IMPERIAL COLLEGE LONDON
Faculty of Natural Sciences
Centre for Environmental Policy
The Electricity Supply Industry: Past, Present, Future
A report submitted in partial fulfilment of the requirements for the MSc and/or the DIC.
DECLARATION OF OWN WORK
I declare that this thesis
The Electricity Supply Industry: Past, Present, Future
is entirely my own work and that where any material could be construed as the work of others, it is fully cited and referenced, and/or with appropriate acknowledgement given.
Name of student (Please print):.................................................................
Name of supervisor:.....................................................................................
AUTHORISATION TO HOLD ELECTRONIC COPY OF MSc THESIS
The Electricity Supply Industry: Past, Present, Future
Author: Alex Whitney
I hereby assign to Imperial College London, Centre of Environmental Policy the right to hold an electronic copy of the thesis identified above and any supplemental tables, illustrations, appendices or other information submitted therewith (the thesis) in all forms and media, effective when and if the thesis is accepted by the College. This authorisation includes the right to adapt the presentation of the thesis abstract for use in conjunction with computer systems and programs, including reproduction or publication in machine-readable form and incorporation in electronic retrieval systems. Access to the thesis will be limited to ET MSc teaching staff and students and this can be extended to other College staff and students by permission of the ET MSc Course Directors/Examiners Board.
Name printed: ____________________
ContentsIntroduction ......................................................................................................................... 6 Chapter One: A Brief History of the Grid ................................................................................. 7 1.1 Beginnings ...................................................................................................................... 7 1.2 Nationalisation ............................................................................................................... 8 1.3 Privatisation ................................................................................................................. 10 1.4 Re-Integration .............................................................................................................. 11 1.5 Competition ................................................................................................................. 15 1.6 NETA ............................................................................................................................. 17 1.7 In review....................................................................................................................... 20 1.8 Renewables Investment ............................................................................................... 22 1.9 UK Policy ...................................................................................................................... 23 1.10 Electricity Market Reform .......................................................................................... 26 Chapter Two: The Electricity Generation Industry ................................................................ 28 2.1 Model Outline .............................................................................................................. 28 2.2 The Economics of Electricity Generation ..................................................................... 29 2.3 Levelised Cost Model ................................................................................................... 31 2.4 The Grid Today ............................................................................................................. 35 2.5 FPN and MEL Data........................................................................................................ 36 2.6 Balancing Mechanism Data .......................................................................................... 39 2.7 Generation Mix ............................................................................................................ 43 2.8 Prices ............................................................................................................................ 46 2.9 The Big Six .................................................................................................................... 50 Chapter Three: Grid Model .................................................................................................... 53 3.1 Model Design ............................................................................................................... 53 3.2 Demand, Availability, Capacity .................................................................................... 54 3.3 Marginal Cost Curve ..................................................................................................... 54 3.4 Outcome ...................................................................................................................... 57 3.5 Scenarios ...................................................................................................................... 60 3.6 Results .......................................................................................................................... 61 3.7 Oversupply ................................................................................................................... 63 3.8 Storage ......................................................................................................................... 64 3.9 Modelling Storage ........................................................................................................ 65
3.10 Results ........................................................................................................................ 68 Conclusion .............................................................................................................................. 71 References ............................................................................................................................. 72 A description of the main generation technologies .......................................................... 77 UK plant statistics............................................................................................................... 78
List of Figures Page Fig 1.1 Takeovers and mergers of ex-publically-owned enterprises 1995-2011 Fig 1.2 Price support and costs for wind power by country Fig 2.1 Levelised Cost model indicative results by technology Fig 2.2 Interpreted FPN data Fig 2.3 Interpolation rules Fig 2.4 TGSD data and FPN data Fig 2.5 Load factors by plant 2010 Fig 2.6 Load factors by plant, season and settlement period, 2010 Fig 2.7 BMU prices plant and settlement period, 2010 Fig 2.8 BMU prices by cumulative volume and by plant, 2010. Log-log plot Fig 2.9 Detail of Fig 2.5, linear plot Fig 3.1 Marginal cost curve components Fig 3.2a Marginal cost curves by plant Fing 3.2b Overall MCC Fig 3.3a Sample model output Fig 3.3b Sample model prices Fig 3.4a Real-world output Fig 3.4b Real-World prices Fig 3.5 Load factors by year for each scenario Fig 3.6 Oversupply by year Fig 3.7 Modelling Storage 14 26 34 38 39 42 44 46 48 49 49 55 56 56 59 59 59 59 63 64 67
The UKs electricity supply industry (ESI) is poised somewhere between flux and crisis. Our deregulated industry structure has been so lauded and imitated it has become known as the British Model, yet the industry is perpetually under investigation for price gouging and anticompetitive behaviour. The regulator is frequently criticised as toothless and our legislation is often byzantine, self-contradictory, self-defeating or all three. There are no British utilities with anything like the international reach of EdF or E.on and a string of high profile corporate failures have left the majority of our infrastructure in foreign hands. Despite having the best renewable resources in the EU, our generation mix is 80% dependent on coal and gas and investment has been steadily dropping. How did we get into this mess and what happens next? This dissertation is a three-part attempt to answer that question. In the first chapter I trace the history of the grid from its beginnings, paying particular attention to the effects, intended or otherwise, of the 1990 privatisation. I investigate whether the particular structure of the ESI has helped or hindered attempts to kick-start the green revolution. I contrast the UKs approach with the rest of the EU and ask whether the recent Energy Market Reform marks a change in direction. In the second chapter I begin my empirical investigation. I obtain and process data from a variety of sources in order to sketch a detailed picture of both the physical operation of the grid and the underlying economic. I construct a merit order of dispatch and create a model to calculate the levelised cost of various technologies for use in the next chapter. In the third chapter I create a model to estimate the generation mix and costs of electricity through to 2025. Drawing upon section two, I create four possible scenarios for the grid and quantify indicators such as costs, carbon emissions and reliability of supply. I extend the model to investigate the effects of adding energy storage capabilities to the grid and comment upon my results. Finally, I offer some conclusions on the lessons of the past and the challenges of the future. 6
Chapter One: A Brief History of the Grid1.1 BeginningsThe history of the electricity supply industry (ESI) is a fascinating case study of the political history of the UK. Since its inception the ebbs and flows of the industry have mirrored the prevailing political winds, from laissez faire in the 19th and early 20th centuries to the postwar consensus of embedded liberalism, and the post-Thatcher neoliberal turn. This review will draw a sketch history of the grid with particular emphasis upon developments since privatisation. Up until the early 20th century, there was no national grid per se. In the mid-to-late 19th century myriad small networks sprung up, privately owned and operated for profit (the first public utility was a small hydro-electric facility established in Surrey in 1881). As a new technology, electricity had only a few specific uses. The initial motivation came from providing street (and later residential) lighting in competition with town gas. In the absence of common standards relating to voltage, frequency and interconnection, these grids were for the most part non-interoperable (Jamasb & Pollitt, 2007). However it was in the early 20th century, as the domestic and commercial uses of electricity began to multiply, that of the strategic importance of electrical power became evident, and in 1926 the Central Electricity Board was formed to impose order upon the industry. The CEB created operating standards, built high-voltage long-distance interconnects and oversaw the construction of new capacity. The National Grid was created in 1933 to oversee transmission infrastructure. Expansion, integration and technological
improvements began to increase the efficiency and reliability of the supply (Chick, 1995). Yet at this point it is thought that there was still over 600 suppliers operating 400 power stations at any of 19 different voltages (Chesshire, 1996) a proliferation attributed to a lack of a cohesive central government programme for the development of utilities (Byatt, 1979). The essential inefficiency of the existing system coupled with the fact that electricity distribution had monopolistic economies of scale meant that public ownership was (in retrospect) inevitable. This was finally realised in 1947, as part of the radical reinvention of the state engineered by the post-war Labour government who also nationalised coal, transport, gas, iron and steel, healthcare, Cable and Wireless, British Airways and the Bank of England. 7
1.2 NationalisationThrough this process the ESI was formalised into its current structure, which vertical links together four components: generation, transmission, distribution and supply. Point-sources of power generation scattered throughout the nation are interconnected by a high-voltage transmission network, usually overhead pylons. The transmission network feeds into a multitude of (non-interconnected) lower-voltage distribution networks which wind their way capillary-like into every village and street corner in the country. Suppliers contract local demand and oversee metering and billing. The whole ESI operates at 50Hz (threephase) and generators must synchronise at this frequency before they can transmit power. Electricity flows roughly linearly through the system and substations are positioned throughout the network to step the voltages down from 400kV to 275kV, 132kV,33kV, 11kV, 6kV, 450V and finally 240V (single phase) in the home (National Grid Electricity Transmission, 2011) The British Electrical Authority, later the Central Electricity Generating Board (CEGB), was instituted as a vertical monopoly in control of generation and transmission. Distribution, supply and customer services were the responsibilities of 14 regional Area Electricity Boards (AEBs), and this arrangement remained relatively unchanged for the next 43 years. Under nationalisation, integration and standards compliance advanced quickly. During the two decades following the reform, the UK enjoyed unprecedented economic growth, and this was coupled with a huge increase in demand for power as the generation that had never had it so good filled their homes with electrical appliances. As a powerful and highly centralised agency, the CEGB was well placed to meet this growing demand and embarked on a huge plant-building program. Installed generation capacity increased from 16 GW to 65 GW between 1951 and 1971 (or from 0.3 kW to 1.2 kW per capita), whilst total electricity generation grew from 57 to 221 TWh (MacLeay et al., 2011). Indeed a government report from 1969 recommends scaling back the program due to oversupply. (Truly, that was a different era - the same report states that the government expects investment to show a return of at least 8% in real terms and that however careful the forecasting... expenditure generally falls appreciably short of the approved figures.) (Nationalised Industries Review, 1969).
This level of success was in large part due to the fact that the CEGB had the resources to build power plants on a scale that had previously been technically and financially unfeasible. In 1948, the largest power plant was 550 MW, yet by 1965 the average new plant was 1300 MW (ibid). Of the 1960s power plants still in use, five are rated at over 2 GW. The UKs largest (and Europes second largest) power plant, Drax, was commissioned in 1974 and is rated at close to 4 GW (MacLeay et al., 2011). CEGBs preferred technology was the coal-fired steam turbine power plant, coupled with Open-Cycle Gas Turbines (OCGT), and it played an important role in supporting the British coal industry. However the CEGB was also in a position to invest in capital-intensive new technology and from 1970 onwards - at the behest of the government - they constructed a total of ten nuclear power plants. Although it was initially expected that nuclear would prove much cheaper than coal (leading to the now-infamous tagline Too Cheap To Meter), enthusiasm waned after the huge costs and poor reliability of nuclear became apparent (Chesshire, 1996). In spite of this, the CEGB was for many years internationally renowned for its technical and managerial competence. However, in some sense it was too big to last. It was by far the largest of the UK state-owned enterprises and by 1987 its asset base totalled some 27bn (the AEBs owned a further 15bn). The CEGB maintained a duopoly of suppliers for coal plant, awarding contracts on the basis of Buggins turn; as a de facto monopoly (supplying 95% of generation in England and Wales) it was free to spend lavishly on R&D, through which it exercised absolute control over the direction of UK ESI (Chesshire, 1996). From a certain (Thatcherite) point of view it was a lumbering monolith that represented all that was wrong with nationalised industry. In 1987, after two terms into office and fresh from victory in the miners strike, the Tories took aim. Their manifesto from that year proclaims that We will continue the successful programme of privatisation... We will bring forward proposals for privatising the electricity industry subject to proper regulation. In May 1987 Thatcher was re-elected to her third term as Prime Minister and in 1988 the government released the white paper Privatising Electricity. The House of Commons Select Committee on Energy, in its review of governments proposals, commented nervously that Reviews of international experience, particularly of the USA and other European countries, do not reveal any strong, or indeed positive, correlation between, on one hand, utility structure, form of ownership, and the degree of competition, and the level of electricity prices and overall utility performance on the other (Chesshire, 1992).
Nonetheless, the 1989 Electricity Act passed the broad framework for regulation, and so it came to pass that in 1990 the CEGB was dissolved.
1.3 PrivatisationThe neoclassical case against monopolies gives no quarter. A monopoly has no incentive to drive down costs or improve efficiency, and can exploit its position to earn monopoly profits since leaner, more efficient businesses are excluded from entry. Monopolies are not only bad for consumers but bad for economic efficiency - and nationalised monopolies are doubly bad because they gain preferential treatment from the government in place of other worthier investments. The neoliberal school (to which Thatchers advisors subscribed) further argues that the government has no business in business and if at all possible national assets should be privatised and exposed to the unsentimental forces of the market. They add, parenthetically, that this is always possible with appropriate regulation (Harvey, 2005). Despite Thatchers enthusiasm to privatise the system at all costs, the electricity market is peculiarly resistant to marketization. Vickers and Yarrow (1991) highlight some of the unusual economic characteristics of the ESI: 1) The supply chain is tightly vertically integrated 2) Generation technologies are highly capital intensive with long lead times and high sunk costs 3) Most generation technologies cause significant environmental externalities 4) To ensure security of supply, the ESI must operate with excess capacity in most periods 5) Extremely tight technical demands demand and supply must balance exactly at every node across the whole network mean that equilibration will always need some central control regardless of the responsiveness of market mechanisms 6) Transmission and distribution are natural monopolies The Government attempted to deal with the problem of monopolistic transmission and distribution by instituting what became known as unbundling. The four sectors of the industry were to be separated, with the transmission and distribution being privately held but heavily regulated (to ensure that businesses did not abuse their monopoly position) and with incentives to improve efficiency. The other two sectors, generation and supply, were to be sold off atomistically to ensure sufficient competition. A pool system would 10
operate whereby generators would compete amongst themselves for supply and suppliers would compete amongst themselves for customers. A key feature was that no company would be allowed to own both generation and supply assets, thus ensuring that anticompetitive vertical integration did not re-emerge. Through strong competition, efficiency would be increased and consumers would enjoy lower prices. It followed that no capacity payment would be necessary, as inflated pool prices would alert businesses to a capacity shortage and they would respond by investing in further generation (and if existing players refused, a new entrant would step in to take advantage) (Thomas, 1996). Indeed it was supposed that all key planning decisions could be left to the market once it had established itself a feature which would be particularly appealing to weary governments used to shouldering the blame when things go wrong.
1.4 Re-IntegrationThere were, however, severe practical problems with implementing these measures. Chief among them, the issue of how to ensure that assets were sold off in such a way that they a) would actually be bought (i.e. would be attractive propositions for investment) and b) would provide the requisite amount of competition. At the time there were really 3 separate grids, two in Scotland and one in England & Wales, with little interconnection between them. The governments plans were compromised immediately when they decided they would split England & Wales generation into just two businesses. Ostensibly this was because the UKs nuclear generation was uneconomically expensive compared to coal and gas and could only survive by being sheltered inside a large company and only another large company would be able to compete with the first. It was therefore conceived that two thirds of the total capacity would go to newly-formed National Power plc (NP), with Powergen plc (PG) taking the rest. However the nuclear assets were so unappealing to investors that they had to be spun out at the last minute into the publicly-owned Nuclear Electric and then subsidised by consumers to the tune of 1bn/year via the fossil fuel levy (which perversely also subsidised power from the French interconnect). The asset sale went ahead anyway. Nuclear Electric operated as a price-taker, meaning that the two remaining generators formed a duopoly which, as we shall see, they werent shy about exploiting (Thomas, 2010). Furthermore, in a move designed to appease furious Scots, the Scottish ESI would remain natively owned in the form of two vertically integrated monopolies, Scottish Power and Scottish Hydro Electric. The England & Wales transmission grid was privatised to become 11
National Grid, who as the designated transmission system operator (TSO) had a very tightly defined remit and was responsible for precisely balancing supply and demand. They soon merged with the privatised gas industry SO to form Transco and have since bought assets in New England. The 12 AEBs in England & Wales were privatised wholesale but since they were both suppliers (regional electricity companies, RECs) and distributors (distribution network operators, DNOs) they were forced to separate the two sides of their businesses. In addition, they were allowed to procure up to 15% of their power from their own plant further violating the principles of de-integration, but at least (in theory) introducing some competition into the generation market as they ordered 10GW of new plant in the early 1990s dash for gas, discussed below. In 1993, in an attempt to stem the market power of the duopoly, the regulator (OFFER, later OFGEM) required that the generators divest 6GW of capacity between them, which was sold to the largest REC, Eastern Electricity (in violation of the 15% rule). Then in 1995 Scottish Power was allowed to take over the REC Manweb, and in 1998 National Power and Powergen were granted permission to take over RECs provided they each divested a further 4GW of plant. Two years later, the RECs were forced to demerge with their respective DNOs, apparently to stop cross-subsidy between them. With the principle of de-integration already scuppered, a mass of acquisitions now took place. Since no one REC or DNO had been allowed to become dominant, most were unable to fend off takeover bids by large foreign utilities (who were able to leverage monopoly positions in their native markets). At this point there is a curious twist in the tale concerning 17 US utilities that suffered a grievous case of groupthink and lost a lot of money as a result. The EU Electricity Directive 1996 (96/92/EC) had mandated the deregulation of EU ESIs, and this sparked a gold rush as US utilities attempted to break into EU markets. Naturally they started in the UK where deregulation was furthest advanced, whence they would launch into mainland Europe. Between 1996 and 2000, seduced by the promise of fast profits on the back of a fast trading market, they made many high-profile acquisitions at heavily inflated prices including 9 RECs and 9 DNOs. But they found that the UK market was far less competitive than imagined and that the rest of the EU was in no hurry to open up their ESI to competition and buy-outs. One by one the utilities lost enthusiasm and dumped their assets at fire-sale prices; by 2003 nearly all had left, nursing estimated losses in excess of $20bn (Haar & Jones, 2008).
In the wake of this fiasco, French, German and Spanish utilities swept in and cleaned up. Figure 1.1 tracks the various fates of the 30 companies brought into existence by privatisation. The apparent endgame of the flurry of mergers and takeovers is that the ESI has contracted to just a handful of mostly foreign-owned corporations (some of which, ironically, are publically held). One imagines that this is probably not what the government had in mind when initiating the great experiment 21 years ago.
Fig 1.1 Takeovers and mergers of ex-publically-owned enterprises 1995-2011Top row colour coding: Red = Generator, Light Green = Distributor Dark Green = Supplier, Mid-Green = Suppler/Distributor, Orange = TSO Else: White box = US Utility, Coloured = Big Six, Grey = Other Note that it is not a full depiction of the UK ESI today as there have been a handful of new entrants in that time Sources: Haar,Laura N, 2008; authors research
1.5 CompetitionThe best laid plans go oft awry whether or not the privatisation played out as intended, the more pertinent question is: does the privatised grid deliver a competitive market for power? And, taking the longer view: was it all worth it? The Pool system implemented in 1990 worked roughly as follows: each day was split into 48 half-hour settlement periods. For each settlement period, each owner of generation placed a number of offers to sell electricity from each of their power plants, priced in /MWh (e.g. in one settlement period, a 1000MW coal-fired plant might offer to sell the first 250 MWhs at 15/MWh and the next 250 MWhs at 20/MWh). The TSO took all of these bids and constructed a marginal price curve for dispatch. The point at which the curve met the predicted power demand was the System Marginal Price (SMP) and all successful bidders were paid this price for their power (MacKerron & Segarra, 1996). (It follows that every REC in fact bought their power from generators at the same price (the SMP), which begs the question of how any REC was supposed to gain a significant cost advantage over the others.) In principle the Pool is quite a pure implementation of marginalist economic theory - but of course it relies heavily on there being sufficient competition to drive down prices for suppliers. Since immediately after privatisation there was effectively only two generators (National Power and Powergen), it is no surprise they were able to manipulate the SMP simply by raising their offer prices. Reviewing the first year of operation, in which the SMP had increased by 29%, OFFER concluded that there is no doubt that the two major generators have recently been able to increase Pool prices significantly (OFFER, 1991). To increase competitiveness the market needed new entrants (termed Independent Power Providers or IPPs). In theory, high Pool prices should have been sufficient to attract them, but entering into competition with two giant incumbents in a new market was extremely risky and no truly independent companies could obtain financing. The RECs, however, were keen to obtain their own generation to avoid being squeezed by NG and PG. The government sought to encourage them by introducing Contracts for Differences (CfDs), which effectively allowed power to be bought from IPPs at fixed prices regardless of the SMP - giving IPPs a guaranteed income. RECs then eliminated all the risk by forming their own (not really independent) IPPs and constructing new Combined Cycle Gas Turbine 15
(CCGT) plant on the basis of back-to-back deals : the gas prices were contracted at fixed prices for 15 years and the power sold for fixed prices for 15 years (Thomas, 2006a). This arrangement sparked the dash for gas leading to a glut of new CCGT, causing the government to place a moratorium on new projects in 1998. The long-term contracts soon became a liability as the price of fuels dropped steadily throughout the 1990s and RECs ended up overpaying for their generation. An REC employee commented at the time that from the RECs point of view it isnt such a good deal at the moment... but as a shareholder of the IPP we are doing very well thank you since companies who buy from the IPP are buying at premium rates (Branston, 2002). Since the domestic market was not opened to competition until 1998, the RECs were in any case able to pass their costs on to their customers. Businesses, however, were able to negotiate and effectively ended up buying the cheaper coal and nuclear power. Studies show that in 1993/1994, 61% of profits were from the domestic sector and 39% from the industrial sector in almost exact reverse to the relative size of the markets (Branston, 2000). CfDs clearly violate the principle of an open and transparent market. The IPPs found that since their generation was already bought and sold, Pool prices were irrelevant - but they still had to ensure their plant was actually accepted for dispatch. IPPs therefore adopted the policy of submitting very low offers to the Pool and effectively taking their generation out of the market. But there were other distortions too: between 1990 and 1998, coal-fired plants were forced to buy quotas of coal from British Coal at above-market prices in order to postpone the collapse of the UK coal industry. RECs were obliged to buy this power at contracted prices, effectively taking this generation off the market too; and at the same time, nuclear power was being heavily subsidised and was also effectively off the market. As Thomas notes, the net result was that for most of the Pools duration, it is clear that more than 95% of RECs needs were supplied from sources that were not required to compete in the Pool (Thomas, 2006a). With such low liquidity, it is no wonder that manipulation was rife. Even while the 1990s saw significant diversification in the generation market (the combined market share of NP and PG fell from 77% in 1990 to 30% in 2000 (MacLeay et al., 2011), it appears that abuse of market power actually increased in that time. A study by Sweeting (2007) estimates that between 1995 and 2000 the SMP was inflated by an average of around 7/MWh translating
into an overpayment of 2.7B per year, most of which inevitably was passed onto domestic consumers; moreover, the peak of exploitation did not occur until Q1 2009. Despite attempting a raft of measures over the years to fix the Pool, OFGEM eventually admitted defeat. In 1999, citing the continuing market power of a number of generators and their willingness to exercise that market power at the expense of customers (OFGEM, 1999), the regulator announced that the Pool was to be scrapped and replaced by the New Energy Trading Arrangements, or NETA. The announcement of NETA amounted to an acknowledgement that the Pool was not a fair and transparent market and that unbundling had been a failure. Whether the Pool could have worked if it had been better implemented is an open question but there is good evidence that the British Model is fundamentally flawed in practice if not necessarily in theory. I shall reflect on the experience of other countries at the end of the chapter.
1.6 NETANETA is a complicated set of agreements but the fundamentals are simple. The most significant innovation is that NETA abolished the mandatory pooling of generation and replaced it with voluntary spot and futures markets. The vast majority of generation is now sold to suppliers over-the-counter in confidential long-term contracts, and each supplier is responsible for contracting its own supply. One hour prior to each settlement period, the TSO calculates the total supply for that period by summing the generation contracted for by each of the suppliers. It also calculates a forecast of total demand using a model. The difference between supply and demand, termed the Net Imbalance Volume (NIV), is settled via the Balancing Mechanism (BM) (OFGEM, 1999). The BM operates like a miniature pool, with generators placing offers to increase generation or bids to decrease generation until demand is met. If the market is short (there is an undersupply) the BM returns the system buy price (SBP); if the market is long (oversupply) it returns the system sell price (SSP). There is one important difference the SBP/SSP is not the marginal price but the average price in the BM. The reverse price (the SSP if the market is short or the SBP if the market is long) is calculated from average spot market price for that settlement period markets close. If the supplier under-forecasts its
demand it pays the SBP for the difference if it over-forecasts it pays the SSP. In this way each supplier has an incentive to calculate its demand as accurately as possible. The SSP and SBP are very volatile and it is doubtful that they give a good indication of the true price of electricity. For example, in December 2010 the SBP ranged from 40.25 to 464 a factor of 11. Even in July (a month of relatively flat demand) prices varied by a factor of 5. The liquidity of the spot market is so low that those prices are equally unreliable. I further explore these issues in chapter 2. The net effect of NETA is to black-box the vast majority of the electricity market. It heavily favours vertically integrated generator/suppliers since the ability to buy energy from oneself greatly reduces ones exposure to the market and therefore risk. An integrated provider will utilise its own generation in the first instance and only enter the market if it can negotiate a particularly favourable deal, thereby squeezing IPPs. This explains in part why the UK ESI underwent marked contraction after 1999 (see Fig 1.1) and is now dominated by the Big Six integrated corporations: EdF, E.on, RWE, Iberdrola, SSE and Centrica. Between them they account for around 70% of generation and over 99% of sales (OFGEM, 2008). I look further into the effects of this oligopoly in section 2.9. How exactly was NETA supposed to deliver economic efficiency and a good deal for consumers? The logic appears to have been that since unbundling had manifestly failed, the regulators decided to go in completely the opposite direction. There are reasons for thinking that vertically integrated suppliers are in fact more efficient due to lower borrowing costs and synergies (i.e. fewer staff). In that case, why not just encourage a vertically integrated market in which the Big 6 compete for consumers on price? Unfortunately there is a rather obvious flaw with this notion: healthy competition relies upon a healthy rate of switching between suppliers but consumers are notoriously sticky (less than 10% consider changing their supplier year-on-year) (OFGEM, 2011) so there is an incentive for suppliers to overcharge and only lower prices when customers threaten to switch. Indeed the back-room costs incurred when a customer switches are so high that if everyone simultaneously decided to switch supplier the cost to consumers would be greater than the realised savings. It appears therefore that the regulator hoped that suppliers would offer competitive prices even without the threat of switching. Irrespective of the theoretical merits (or otherwise) of NETA, what were the actual effects? Initially NETA was widely judged a success for apparently lowering the wholesale and retail
price of electricity after introduction. However, closer inspection reveals that the introduction of NETA coincided with cheaper coal (due to the expiry of expensive Coal Board contracts) and a burst of new generation at the tail end of the dash for gas, so it was likely that wholesale prices would have fallen anyway. In reality the wholesale market decreased in price far more than the retail market, suggesting that the Big Six were enriching themselves at the expense of both IPPs and consumers (Dadeviren, 2009). The fate of British Energy makes an interesting case study in this regard. The UKs nuclear plants were publically owned and supported by the Fossil Fuel Levy up until 1996, but when technical advancements doubled plant availability they became viable businesses. With wholesale prices in the Pool riding high, the government decided to remove the levy and sell the 8 most modern plants under the title of British Energy plc (the privatisation was a disastrous flop, raising only 1.7bn roughly half the construction cost of a single plant) (Thomas, 2010). British Energy prospered for a while but after the introduction of NETA its income crashed by 30% and in 2003 it was bailed out by the public at a total cost of 10bn (European Commission, 2004), much to the chagrin of the National Audit Office who argued that the privatization should never have gone ahead to begin with (National Audit Office, 2004). Yet by the middle of the decade wholesale prices had increased due to a huge spike in the cost of gas, and British Energy was once again a viable business. Finally in 2009 it was acquired by EdF for 12bn meaning that the endgame of this tussle between state and private actors is that the UK nuclear industry is effectively the responsibility of the French public. Apart from this acquisition the ESI has been relatively static for the past few years, which has given researchers a chance to revisit their assessments of NETA and the results have not been good. There was a sharp increase in retail prices around 2005 ostensibly due to wholesale costs, but a study commissioned by the Right to Fuel campaign found that fully half of the rise was simply due to increased supplier margins (Cornwall Energy Associates, 2008). Worse, when wholesale prices fell retail prices refused to follow suit. Last year a comprehensive study was even more damning in its verdict of NETA, concluding that despite NETAs stated intentions of reducing wholesale and thereby retail prices... instead [NETA] merely rearranged where money was made in the system. (Giulietti et al., 2010) page number. This assessment certainly rings true. Judging by the publications of OFGEM, they appear to be locked in a perpetual battle with the utilities - both wholesale and retail markets are 19
currently being investigated for uncompetitive behaviour (the retail investigation has now entered its fourth year) (OFGEM, 2008, 2010). Given that NETA is now ten years old and is looking decidedly creaky, one wonders if another rule-change isnt too far off.
1.7 In reviewNow that we are up to date, it would be useful to put this remarkable history in context. There is something very striking about the story of the UK ESI; though it has its own quirks, curiosities, heroes and villains, in fact it gives the complete history of British capitalism in a distilled form. In particular the reforms since 1990 are an excellent case study of actually existing neoliberalism: the dispassionate hidden hand of the market shall be unleashed upon the lumbering, monolithic state enterprises and deliver efficiency, growth and prosperity for all or at the very least send large profits to monopolistic private corporations (Harvey, 2005). In fact, there are good reasons why the ESI is not as amenable to competition to other industries even in theory (adapted from Thomas (2006b)):
1) Inabiliy to store power: Storage in other industries allows one to balance out fluctuations between supply and demand. In the absence of storage a free market suffers huge price volatility 2) Supply and demand must balance perfectly: Some central control will always be necessary to co-ordinate supply and demand. The free-market ideal of free entry and exit is clearly impossible 3) No substitutes: Most products are readily substitutable which effectively increases competition in the marketplace and acts as a check on suppliers. No such check exists for electricity suppliers 4) Vital to modern society: Unlike other (substitutable) products, a constant and reliable power supply is essential to modern living. Under no circumstance would a government allow the ESI to collapse it is the ultimate too big to fail 5) Lack of investment: For security of supply it is necessary that there is excess capacity in the system. A free market will underinvest in capacity because it will not be able to turn a profit on plant that is only operated a few times a year 6) Unsustainable price structure: The power supply is by design completely standardised, meaning that competition is based entirely on price. But if 20
competition pushes prices down to the marginal costs, generators will make a loss on their capital investment (see section 2.2) 7) Environmental impacts: The environmental impacts of electricity generation are substantial and regulation is necessary to ensure they are accounted for Lohmanns principle of frame overflow (2009) provides a useful metaphor to explain the difficulties that the UK ESI has faced since privatisation. The government attempted to put the ESI into a market frame, but the ESI is inherently resistant to such framing for the reasons outlined above, so it generated frame overflows (e.g. collusion, underinvestment). The government has tried to get the overflow back in frame by introducing new rules and regulations (e.g. divestment of assets, vertical integration) but this very process has inevitably generated more overflows therefore re-framing is a never-ending process. In this context the new regulations to encourage renewable generation are another attempt at reframing due to an environmental overflow. This is the topic of the next section. An objective assessment of privatisation must surely deliver the verdict that it simply wasnt worth all the effort (Dadeviren, 2009); nonetheless it is surprising the extent to which the British Model has been championed worldwide (Joskow, 2008) (though it is perhaps less surprising when one recalls that the neoliberal mantra is TINA: There Is No Alternative). In the mid-1990s the World Bank and the European Commission played leading roles in exporting the British Model to countries around the world, both developing and developed. The World Bank in particular has often made utility privatisation a prerequisite of aid packages, most notoriously under the guise of Structural Adjustment Programmes (Stiglitz, 2002). The twin principles of privatisation and deregulation form the very lifeblood of the neoliberal project, so it has been necessary to tout the British Model as a success in the face of all evidence, as an act of paradigm maintenance (Wade, 1996). While the spread of the British Model is a story of its own, suffice to say that when applying the template to developing countries the results have tended to be far more devastating than in the UK. This was an entirely predictable outcome given that such countries generally have much weaker regulatory regimes (Dadeviren, 2007) . Developed countries, in contrast, are better able to resist and nations such as France, Germany and Japan have effective control their ESIs even if there has been a degree of deregulation. In the last decade the focus of energy policy has shifted from competition and markets to cutting carbon emissions. I will now consider how government policy has influenced the
deployment of renewable energy sources. As we shall see, despite the notably collectivist context (saving the planet), governments will always find a role for the market.
1.8 Renewables InvestmentIn the last decade successive British governments have pledged to various CO2 emission reduction targets in an attempt to halt climate change. The Climate Change Act 2008 set a legally-binding target of at least 80% cut in emissions by 2050 (against a 1990 baseline). Compared to other sectors of the economy it is relatively straightforward to decarbonise the ESI so it is expected to make a large contribution to the cuts. The EU Renewable Directive 2009 set the 20 20 20 target that by 2020, renewable energy (RE) will source 20% of total energy consumption (15% in the UKs case). This amounts to RE making up some 40% of the UK electricity supply within the next 10 years (European Parliament and the Council of the European Union, 2009). The fundamental difficulty with RE - which must eventually be confronted by governments - is that renewables are more expensive than conventional generation. (Economists have clever ways of proving that actually they are cheaper if you take into account the environmental benefits, but there is no escaping the impact on the bottom line.) In a nationalised industry this would not present a problem: the state would just build the things and pass the extra costs onto the consumer. But in a competitive market, a RE generator has to be sure that a supplier will buy their expensive electricity; and a supplier has to be sure that its competitors will also endure higher costs, or else it will lose market share. Given that the ESI is inherently conservative in nature, it clearly needs a big incentive to undertake investment in RE. (I should note at this point that in northern Europe, renewable energy essentially means wind energy although sometimes one includes nuclear too.) In response, governments throughout the EU have introduced a raft of taxes, trading schemes, tariffs and incentives to send the correct price signals to the market. Two schemes have been most influential Renewable Energy Feed-In Tariffs (REFITs) and Tradable Green Certificates (TGCs). REFITs simply give RE a guaranteed price in the wholesale electricity market; this price may be fixed or may be pegged at a certain rate above the basic wholesale price. Energy suppliers are forced to buy RE at this price whenever it becomes available (though the precise rules vary). The benefit of this system (besides its simplicity) is that by guaranteeing 22
a fixed return in investment it eliminates capital risk - which is very important for speculative new technologies (where even the perception of risk can push up the cost of finance beyond what is affordable). Because the government sets the tariff level, the REFIT affords the state a great deal of control over the deployment of RE. While some see this as an advantage, others see it as a weakness. Since the price has not been set by the market, it is not efficient i.e. the price does not represent the marginal cost of generation, therefore REFITs in theory enable RE generators to make profits at the expense of consumers. TGCs are mooted as a solution to this problem. Under this scheme, RE generators are given certificates (TGCs) for the energy they supply to the grid and suppliers are obligated to purchase a certain number of TGCs per year. This sets up a separate market for RE on top of the normal electricity market and, so the argument goes, that leads to efficient market outcomes (though at the cost of the state surrendering control). In fact, some have argued that the above logic is flawed and the TGC is no more efficient than the REFIT. As the EC commented in 2005, both instruments are equally market-based in that the regulatory body sets either the price or the quantity and leaves the determination of the other to the market (Commission of the European Communities, 2005).
1.9 UK PolicyAlmost every state in the EU has chosen one of the above two mechanisms, or some hybrid of the two. It is interesting to note that the choice of policy appears to have been strongly influenced by the ideological disposition of the state. The UKs first announcement of RE policy arrived in 1999 after years of political wrangling; unsurprisingly, given their neoliberal bent, the New Labour government opted for a variant on TGCs, the Renewables Obligation (RO), which was formally implemented in 2002 (Toke & Lauber, 2007). It will be instructive to trace UK RE policy and compare the outcomes to those of other EU nations. The RO policy set an obligation on suppliers to purchase Renewable Obligation Certificates (ROCs) from RE generators equal to 10% of output by 2010, 15% by 2015 and 20% by 2020. The alternative was to pay a buy-out price of 30/MWh. Normally this would act as a cap on the market price of ROCs, but there was an extra twist: the fines were recycled back to the owners of ROCs so that the value of each ROC increases in proportion to the ROC shortfall. Moreover, by design there was always at least 10% fewer ROCs available than
necessary for suppliers to meet their obligations. The net result of this is that the ROC price had a price floor of 30 (Toke, 2010). At the time there were complaints that the buy-out was too low to stimulate investment in large scale wind, especially offshore wind, and yet too generous to other better-established technologies. There were arguments that ROCs be banded so that, for example, offshore wind ROCs would be worth more than hydroelectric ROCs. The government defended its position by stating that it is no longer Governments job to pick winners... the future role of [Government] will be one of action but not direct intervention (Department of Trade and Industry, 2007). It soon became clear that investment was indeed being directed away from large-scale offshore wind projects and into small-scale onshore wind and landfill gas projects indeed landfill gas accounted for 44% of all ROCs issued between 2002 and 2005 and it appeared that the UK was set to miss its 2010 RE targets. Table 1.1 shows the volumes and worth of ROCs over time, indicating that by 2006 RE generation was at barely two thirds the target amount. A 2006 government review paradoxically concluded that the Obligation is operating largely as anticipated and therefore needed amending: the conclusion was that the RO was to be banded after all (OFGEM, 2006). Thereafter the rules were changed so that (e.g.) offshore wind was awarded 2 ROCs per MWh whereas landfill gas received only 0.5 an amendment that was labelled a quasi-feed-in-tariff by an irate British Wind Energy Association (2009).
Worth () 45.9 53.4 45.1 42.5 49.3 53.0 54.4 52.4
2002 2003 2004 2005 2006 2007 2008 2009
8,393,972.00 12,387,720.00 14,315,784.00 16,175,906.00 19,390,016.00 22,857,584.00 25,944,763.00 26,971,916.00
4,973,091.00 6,914,524.00 9,971,851.00
59.2 55.8 69.7
15.9 22.9 13.7 10.2 16.0 18.7 18.6 15.2
12,232,153.00 75.6 12,868,408.00 66.4 14,562,876.00 63.7 16,813,731.00 64.8 18,747,129.00 69.5
Table 1.1 Value of ROCs 2002-2009 Source: Ofgem (2007, 2011)
While this move did stimulate further investment in offshore wind, it was clear by now that the RO was in any case forcing consumers to overpay massively for RE. The average auction price for ROCs from 2002 to present was between 40 and 50 (e-ROC, 2011) which would boost the revenue of onshore wind by at least 100%. That represents a huge premium for a technology thought to be only around a third more expensive than CCGT (see section 2.3). Whats more, the price of ROCs was not coming down over time as had been anticipated (see Table 1.1). By the middle of the decade there was very good evidence that REFIT systems, despite their alleged inefficiency, were achieving higher RE penetration at a lower cost. Initially the EC were very pro-TGC schemes (Commission of the European Communities, 1999) by 2005 their own data was showing that (for wind) the countries with TGC schemes were the least efficient and had the worst penetration rates - with the UK the worst performer of all (see figure 1.2). They put this down to higher risk premium requested by investors, the administrative costs and the still immature green certificate market (Commission of the European Communities, 2005) page no).
Fig 1.2 Price support and costs for wind power by country From Commission of the European Communities, 2005
1.10 Electricity Market ReformIt seems probable that OFGEMs continued promotion of markets at all costs even thoroughly dysfunctional ones has imposed huge net costs on consumers. However when even the World Bank is (supposedly) rethinking its position (Thomas, 2006b) there is reason to think that the healing has begun. In 2010 the Energy Minister Mike OBrien stated that in order to ensure that we were able to make an energy revolution ... we had to get OFGEM to stop being so pedantically market driven (quoted in (Toke, 2011)) and the regulator has become notably more interventionist in recent years. The latest White Paper, entitled Energy Market Reform, was released by DECC in July 2011 (Department of Energy and Climate Change, July 2011). As the title hints, it calls for a rethink of RE policy. Among its proposals are: phasing out the RO; replacing it with contracts for differences feed-intariffs (CfD FITs);a carbon price floor; an emissions standard; capacity payments. The mechanism for capacity payments has yet to be announced but will be geared towards keeping the lights on rather than encouraging RE investment. The emissions can basically be thought of as a ban on new coal-fired (but not gas-fired) plant. From the point of view of RE, CfD FITs represent the most significant change. From 2014 RE generators will be able to sign long-term contracts with an as-yet unspecified counterparty (possibly the government) 26
to pay them a fixed price per MWh, the strike price. The energy will still be sold on the wholesale market at the market price - but the counterparty will make up the difference to the strike price. The logic is that this will guarantee a return on investment, thereby reducing the risk-premium for RE developers, without distorting the rest of the market. Many details are still to be settled, however, most notably the level of the strike price. There are also concerns that the liquidity of the wholesale market is too low to deliver a stable price, as is necessary for such a mechanism. OFGEM is conducting a separate review on how to increase liquidity in the wholesale markets (OFGEM, 2010). The carbon price floor is designed to reduce uncertainty surrounding the EUETS carbon price, which is notoriously volatile. The floor has been set at 16/tonne CO2 in 2012 rising to 30/tonne by 2020 and 70/tonne by 2030. Though the price floor will increase the cost of gas and (particularly) coal plants, it alone is unlikely to increase prices enough to support RE (and it will in any case be nullified by CfD FITs). For this reason the price floor has been widely interpreted as a subsidy for nuclear power: according to the treasury, the price floor will results in the nuclear industry (read: EdF) benefitting by an average of 50m/year to 2030. Industry reactions to the EMR have been mixed. Unsurprisingly EdF were delighted, stating that This is good news for customers, policy makers and investors (EDF Energy, 2011), whereas most environmental groups expressed disappointment. Greenpeace criticized the EMR for lacking ambition and ignoring the structural problems of the industry, commenting there are six winners from today's white paper and millions of losers. Overall the response was muted with most commentators suspending judgement until further details are released. This was typified by the Association of Electricity Producers who responded that they had some concerns but that there is a great deal of detail to be agreed before all this takes effect. (Carrington, 2011). Undoubtedly the industry is right to be cautious; nonetheless it is certainly possible to speculate on how the EMR will affect the ESI in the future. That topic forms the basis of the remainder of this paper.
Chapter Two: The Electricity Generation Industry
I will now begin the empirical part of the investigation, the eventual objective of which is to create a simple model of electricity generation in the UK. From there I will attempt to project into the future taking into account the UKs renewables policy as described in section 1.10. Why am I focussing on generation? Put simply, because it is where the action is; it is the sector which accounts for most investment and the sector which is mooted to undergo a green revolution (indeed in 2009 the Minister for Energy encouraged a roomful of offshore wind developers to imagine you are pin-striped revolutionaries in the spirit of Che Guevara on the Sierra Madre (quoted in (Toke, 2011)). This is in contrast with the rest of the supply chain - essentially just the means by which electricity is delivered to the plug which will remain relatively static for the foreseeable future. However, we should not overestimate the significance of generation to consumers (provided the lights dont go out): a recent Ofgem report reveals that on average consumers pay 13.4p/kWh or 134/MWh for electricity (c.f. 39/KWh for gas) (Ofgem, 22 June 2011), of which generation accounts for only 42% (see Table 2.1) (although we should expect this fraction to increase in the future). Contribution Generation Operating costs VAT & other costs Supplier Margin Total /MWh 56.25 16.25 52.5 8.75 133.75
Table 2.2 Electricity costs by source
2.1 Model OutlineThe model I will use is simple but still flexible enough to investigate a range of future scenarios. It will be useful to outline it now, though I will elaborate in Section 3.1. My guiding principle in this investigation is that the UK ESI is driven by cost that the cost
determines which power plants get built in the first place and which plants are given priority for dispatch (the so-called merit order). Suppose that one knows both the total power required at a given moment in time and the costs of power generation for each power plant on the grid. The model is a ladder model: it will meet the power demand simply by activating the plants one-by-one in merit order (i.e. starting with the cheapest) until the electricity demand is met. This will deliver the least-cost way of meeting electricity demand for a given set of power plants. By adjusting various parameters e.g. carbon prices, fuel prices, fixed costs, electricity demand one can then explore a number of possible scenarios and speculate on what DECC likes to call Our Energy Future. Clearly for such a model to be accurate it is crucial to get the costs right - therefore the first step will be to gain an understanding the economics of the industry.
2.2 The Economics of Electricity GenerationBroadly speaking, the economics of the power industry are not that dissimilar to any other industry: a product (electricity) is made a factory (power plant) and revenue is generated through the sale of said product. The costs of plant can be broken down into capital expenditure (CAPEX, the up-front costs) and operating expenditure (OPEX, the running costs) (Berrie, 1983). The subtlety is that there are numerous plant types each with a different balance of capex and opex making them suitable for different roles within the ESI. Roughly speaking, if you were to build a plant tomorrow your choice would be between an expensive machine with cheap fuel or a cheap machine with expensive fuel. The former includes renewables such as nuclear, wind, solar and tidal. The latter includes fossil-fuel burning plants such as coal, CCGT, OCGT, diesel and gas CHP. A brief profile of the most common plant types is given in Appendix A. For any new plant, the main contributor to the capex is Engineering, Procurement and Construction (EPC): the actual building of the structure. To this we add related costs such as ancillary equipment, land purchase, planning, legal fees and network connection. Arguably we should include decommissioning costs, though these occur at the end of plant lifespan. Since all these costs are one-offs it is possible to subsume capex into a single representative figure for each type of plant, commonly quoted in /kW capacity (this assumes that economies of scale are already taken into account). It is important to note, however, that capex is an inherently uncertain metric and cost escalations are common. Commodity prices have shot up in the past few years to the extent that the real capex of a
coal or nuclear plant has more than doubled. It remains to be seen whether this price shift is permanent. Opex can be broken down into costs that are constant year-on-year (fixed costs) and costs that vary depending upon mode of operation (variable costs). The main drivers of fixed costs are labour, business rates, insurance, network charges and financing. The main variable costs are fuel purchase, carbon taxes, fuel disposal and handling of by-products. The distinction between fixed and variable costs is not always clear; maintenance will have both a fixed and variable components. The fixed opex can be subsumed into a headline figure similarly to capex it quoted in /KW/year. The variable opex can be quoted in /MWh. This figure is particularly important as it is the marginal cost (MC) of electricity for a given plant. I will discuss the implications of this figure later in this section. Assuming that for a given plant one can calculate these three figures capex, fixed opex, variable opex with reasonable accuracy, how can we combine them into a single indicative cost of electricity generation? The standard way to do so is to introduce the Levelised Cost (LC). The LC can be thought of as the (constant) price at which a generator would have to sell electricity if it were to exactly break even on its complete lifetime investment. It is defined as the net present value of all costs (in s) divided by the net present value of energy generated (in MWhs). It is calculated by summing the expected costs for each year for the lifetime of the plant, applying a discount rate to each years costs, summing the discounted costs across all years and dividing by discounted lifetime energy generated to end up with a levelised cost of electricity in /MWh. Expressed algebraically, where T is the lifespan of the plant, Ct, Ft and Vt are the capex, fixed opex and variable opex in year t, K is the plant capacity, r is the discount rate and Et is energy generated in year t. The LC provides a common metric against which one can compare power plants with wildly different cost structures - e.g. heavily front-loaded investment versus uniform investment. However it does not necessarily denote a lower bound on the price which the plant will sell electricity for that is the MC. To elaborate: suppose that a nuclear power plant has a LC of 50/MWh, but this is mostly due to very high up-front capital costs. The cost of actually 30
operating the plant, including fuel - the MC - is just 15/MWh (uranium is cheap). Now suppose that due an excess of cheap CCGT, the wholesale price of electricity is 35/MWh. Should the nuclear plant operate? Clearly it should at the moment it is losing (LC MC) = 35 every hour for every MW of capacity, just in capital and fixed costs (so a 1000 MW plant will be losing 35,000/hour). If it generates at full capacity it is now losing (LC MC + MC price) = 15 per MW capacity per hour so it is better off generating even if it is still losing money. Therefore we should not expect the LC to necessarily reflect the sale price that any given plant achieves. It is worth noting that this situation in itself is not that different to any other market: if I manufacture a trinket for 30 but the market price is 20, Im still better off selling it and taking a 10 loss rather than not and losing 30. If this continued for long in a free market, I would go bust and the world would be better off. But remember that the power supply is unique: it must never fail and it cannot be stored. Therefore it is always necessary to have excess capacity in the system, so even non-profit-making plants cannot be allowed to close. This is in fact precisely what happened in the British Energy debacle (section 1.6), resulting in a 10bn public bailout. It does appear that there is a fundamental contradiction between free electricity markets and the need for capacity. At the moment the UKs oligopolistic market is just about unfree enough to deliver stability in that regard (the integrated suppliers essentially subsidise their own unprofitable peaking plant), but as mentioned in section 1.10, DECC have recently acknowledged the need to introduce a new capacity mechanism to address this.
2.3 Levelised Cost ModelLet us look more closely at formula (1). Though it is by no means simple, behind the symbols hides considerable further complexity. The lifespan of a plant can be anything from 15 to 50 years the levelised cost calls for one to know or calculate the opex and energy output for every year in that time (capex is less problematic since it is heavily frontloaded). This may require, for example, predicting the price of fuel 20 years hence including carbon taxes. Et is a function of plant load therefore depends upon the total energy demand, generation mix and merit order of the plant in the tth year. In the future there may be technical advances, government interventions or changes in market structure which could work in favour of or against any particular plant. Moreover the results are sensitive to the discount rate, but discount rate does not have a well-defined value, rather it requires some hybrid of financial and political judgement (low discount rates imply we 31
care more about the future and favours front-loaded investment; conversely high discount rates favour uniform or tail-loaded investment). Nonetheless it will be very useful to calculate approximate LC values to use as a basis for ascribing prices to technologies in the grid model. Rather than try to account for all of these details explicitly, my approach was to reconstruct the model used in a Mott MacDonald in a recent study for the UK government (Mott MacDonald, June 2010). The LC model works according the principles outlined above: taking each technology in turn, one inputs a host of parameters such as lifespan, efficiency, construction time, EPC cost around 40 in all and the model calculates the costs for each year of operation broken down by capex, opex, fuel and carbon costs. A sample list of parameters for the base CCGT case are reproduced in Table 2.2. It also calculates total energy generated per year. Each cost was discounted by the appropriate amount and then summed over all years and divided by summed discounted output to give the LC. My objective was firstly to program my own version of the model in MATLAB and use the supplied parameters to replicate the results; secondly to adjust the parameters to explore a range of different scenarios. Unfortunately the paper does not disclose the Mott MacDonald model in enough detail for the results to be replicated with perfect accuracy; however I was able to reverse-engineer a model that produced results within 1-2 % of the original findings, which is good enough for our purposes. From this point on all references to the LC model will refer to my own.
Parameter Pre-develop period Construction period Plant Lifespan
Unit years years years MW % % % % % % /kw m /kw m /kw m /kw m % /kw /MW/year m/year /MWh m/year m/year /MW/year m/year /MW/year m/year /MWh m/year /MW/year m/year
Value 2 2.5 30 830 59 3.5 91.2 90 2.3 0 25 20.8 25 20.8 656.3 544.7 12 10 7.6 718.3 15000 12.5 2.2 13.1 25.6 5000 4.2 6000 5 0 0 26000 34.7 Table 2.2 Example Input Parameters for Levelised Cost model Taken from Mott MacDonald June 2010 Note not all parameters are shown e.g. Carbon price, Load Profile
Gross power Gross Efficiency Ave Degradation Ave Availability Ave Load Factor Aux Power CO2 Removal
Pre-license costs Pre-license costs Reg/license/enquiry Reg/license/enquiry EPC EPC Infrastructure Infrastructure Dev as share of EPC Total CAPEX
O&M fixed O&M fixed O&M variable O&M variable Total O&M Insurance Insurance Connection/UoS Connection/UoS CO2 trans storage CO2 trans storage Total fixed/year Total Opex
The central results from the LC model are shown in Figure 2.1, split into fixed and variable costs (including a carbon tax). We can have reasonable confidence in these results as they are in agreement not only with the original paper but with other similar studies, for example (UK Energy Research Centre (UKERC), 2007), (Arup, 2011), (PB Power, 2004).
250.0 200.0 150.0 100.0 50.0 0.0 Var Costs Fixed Costs
Fig 2.1 Levelised Cost model indicative results by technology
One can see that offshore wind in particular is very expensive relative to traditional technologies, although onshore wind is roughly competitive. We can also see, for example, that with this modelled carbon price CCS (carbon capture and storage) is actually more expensive to install than the savings it would deliver. I will discuss the findings further at the modelling stage, but for now it is important to note, however, that these are LCs for new-build plants and since EPC costs have spiked in recent years they very likely overestimate the LC of any existing plant and it would be unwise to rely uniquely on modelled LCs to determine ESI pricings. And, of course, LCs tell us little about how power
plants are actually utilised on a day to day basis. To draw a more complete picture of the industry, it will be necessary to analyse real-world data.
2.4 The Grid TodayThis brings us to our next section, a close look at how the various generation assets are operated and traded in todays ESI - information which will be invaluable when formulating the grid model. I found that when attempting such an analysis one hits an immediate roadblock - since most intimate dealings within the ESI (i.e. who buys what from whom for how much) are now trade secrets, it is very resistant to outside scrutiny. Indeed just working out who owns what is non-trivial since assets change hands frequently. Much of the time I have relied upon government reports and independent analyses - the Digest of UK Energy Statistics (DUKES) proved particularly useful (though the 2010 data was not published until early August (MacLeay et al., 2011)). However my main source has been the mass of raw data available on the website www.bmreports.com, from where it is possible to download data relating to each Balancing Mechanism Unit (BMU) in the UK going back several years in half-hour intervals. (A Balancing Mechanism Unit is National Grids term for any piece of infrastructure which puts power into or takes power out of the grid here I use it to refer to any generation asset i.e. one that puts power into the grid.) The main dataset of interest was the power output of each BMU at a given moment in time, known as the Final Physical Notification (FPN). I was also interested in a sister dataset, the Maximum Export Limit (MEL). The MEL is the maximum power that a BMU is capable of generating in a given time period, and by comparison with the actual output (FPN), allows one to calculate the load factor of the BMU. One might expect the MEL to be a constant equal to the capacity of the power plant, and much of the time it is, but there are plenty of occasions when a BMU might not be able to operate at maximum capacity, for example due to planned maintenance. Finally, I was interested in the full results of the balancing mechanism- which means the Net Imbalance Volume (NIV) and all the offers and bids put in by each BMU to increase or decrease their output to meet the NIV. This data can be used to infer the marginal cost of generation for each BMU (with caveats). I decided that my data range would be the entirety of 2010, which amounts to some 365*48 = 17520 settlement periods. In terms of sheer quantity of data this is perhaps overkill, but any shorter time period would risk missing out seasonal variations.
2.5 FPN and MEL DataThe bmreports website, administered by National Grid subsidiary Elexon, allows one to download raw system data via the TIBCO relay service. TIBCO relay data is released daily in the form of a ~50Mb comma separated value (CSV) text file. Buried within this file is the FPN, MEL and balancing mechanism information for each BMU for each settlement period. Helpfully the website also pre-extracts the FPN and MEL data from the relay file ready for download. The date and data type are specified entirely within the URL for example to download FPN data for 1st January 2010, one queries the following URL:http://www.bmreports.com/tibcodata/2010-01-01/tib_messages_FPN.2010-0101.gz
The format of the URL lends itself to scripting. To obtain the data for all of 2010 I wrote a script which generates each date of the year in turn, downloads from the corresponding URL and extracts the file, resulting in 365 CSV files for each data type (FPN and MEL). The FPN and MEL data were in the same format. A sample of raw FPN data (to which I have added headers) is shown in Table 2.3:
Data Type PN PN PN PN
Settlement Start Date/Time Period
T_DRAXX-1 T_DRAXX-2 T_DRAXX-2 T_DRAXX-2
3 3 3 3
20100101010000 645 20100101010000 645 20100101010100 635 20100101010200 631 Table 2.3 Sample FPN data
20100101013000 645 20100101010100 635 20100101010200 631 20100101013000 631
Taking the columns in turn Data Type: PN stands for Physical Notification BMU: indentifies the balancing mechanism unit via a unique ID. Here the data refers to Drax Power Station units one and two. Some large power stations, such as Drax, have multiple turbines which can be operated independently; hence they have more than one BMU (Drax in fact has nine). 36
Settlement period: tells us which half-hour period the data concerns. 3 refers to the third half-hour period of the day i.e. 01:00 01:30. Start Date/Time: The start date and time of this particular relay entry a numerical string in the format yyyymmddhhmm Output: The MW output from the BMU at the start time End Date/Time: the end date and time of the entry Output: The MW output at the end time
It is good practise that each BMU reports its output at the start and the end of each settlement period regardless of whether its output changes in that time. If a particular entry gives different start and end outputs (i.e. if output changes), one assumes a constant rate of change between the two times. If there is a gap in the time series one linearly interprets between known data points. The full interpretation of Table 2.3 is that on the 1st Jan 2010, Drax unit 1 output 645MW continuously between 01:00 and 01:30 whereas Drax unit 2 output went from 645MW at 01:00 to 635 MW at 01:01 to 631 MW at 01:02, where it stayed until 01:30. This is shown schematically in Fig X. MEL data would be interpreted in the same way except that it would represent the maximum power that a BMU was capable of outputting as opposed to the power actually output. This could plausibly be constrained by fuel or staffing availability, maintenance, planned and unplanned downtime or other circumstances.
Fig 2.2 Interpreted FPN data Numbers indicate corresponding row entry in Table 2.3
A number of steps were required to put the data into a usable form. I wrote a series of scripts to perform the following operations: Scan each days data into MATLAB Output the data from each individual settlement period into its own .mat file Download a list of all generator BMUs from the bmreports website and scan into MATLAB Take each BMU from the list in turn, search each of the 17520 settlement period files for the corresponding BMU entry and collate the data for each BMU into a new output file The net result was a file for each generator BMU (266 in number) containing the FPNs for the whole year in other words a complete record of the power generated by each power station in the country. Similar data was aggregated for the BMU MELs. Having so processed the data, a degree of cleaning was necessary. There are a number of ways that the data for a given BMU can be inconsistent: 1) Consecutive records show a discontinuity of output 2) Two records overlap in time but with the same output 3) Two records overlap in time with different outputs 4) Data missing i.e. a temporal gap 38
All these inconsistencies were indeed found to occur with various frequencies. I opted for the following solutions: 1) The second record is assumed to take precedence. The first record is shortened to achieve consistency with the second 2) As in 1), the first record is shortened. This leaves us with a situation as in 3) 3) No change. It is assumed that the BMU changes its output very quickly in a way that I will not attempt to quantify 4) A third record is created which interpolates across the gap The inconsistencies and their solutions are shown schematically in Fig 2.3. A series of scripts were run to carry out the modifications to the data for both FPNs and MELs.
Fig 2.3 Interpolation rules see text
2.6 Balancing Mechanism DataThe BM data is a little different. There are a large number of bids and offers that are put in for each settlement period, but a quick perusal of the data shows that many of them are not competitive. That is, since it costs nothing to make a bid/offer, BMUs often place unrealistic ones on the off chance that they get accepted due to some miscalculation or system failure. (This has happened on a handful of occasions, most famously on 10th Dec 2002 when two large power plants failed at short notice causing the marginal system buy price to top out at 9,999/MWh (ERI, 2004).) 39
The uncompetitive bids/offers are not particularly relevant to my work, so to simplify things I opted to download just the bids and offers selected by the TSO. In effect this means just the cheapest 10-20 such bids and offers, depending upon the imbalance volume. To download such data one queries the soapserver via the following URL:http://www.bmreports.com/bsp/additional/soapfunctions.php?output=CSV&dT=YYY Y-MM-DD&SP=#N&element=DETSYSPRICE&submit=Invoke
replacing YYYY-MM-DD with the date and #N with the settlement period of interest. A script was written to cycle through the 17520 different datasets. The format of the datasets is a little more complicated but an edited snapshot is reproduced in Table 2.4 (once again with my own headers). Bid/Offer Date SP Index BMU ID Offer Price BID BID OFFER OFFER 20100101 20100101 20100101 20100101 2 2 2 2 1 2 1 2 T_RATS-4 T_RATS-3 T_CDCL-1 T_COSO-1 24.7 24.65 35 36.58 Offer Volume -0.283 -0.283 2.15 87.5 Imbalance Volume 0 0 2.15 87.5
Table 2.4 Sample BMU data
Bid/Offer: Indicates whether it is a bid (to reduce a BMUs output) or an offer (to increase output) Date: in the format YYYYMMDD SP: Settlement period number Index: The merit order of the bid/offer in the stack. i.e. if the system is short (undersupplied) then the TSO will accept offers in order of their index, whereas if it is long the TSO will accept bids in index order. The index is found by sorting the bids and offers in ascending price order
BMU ID: As before. Here the bids come from Ratcliffe-on-Soar coal-fired plant units 4 and 3. The offers come from Cottam Development Centre CCGT plant and Coryton CCGT plant
Offer price: The price the plant will pay (per MWh) to decrease (bid) or increase (offer) their power output. It may seem strange that a plant would pay to reduce
their output but if they have already contracted that power for (say) 30/MWh then it makes sense to pay 24 to lower output and pocket the difference Offer Volume: Offer/bid volume in MWh. Since it is for a half-hour period, double this to work out the actual MW output Imbalance Volume: The volume that was accepted i.e. did in fact end up getting used. This information is calculated retroactively once metering is completed. In this case the system turned out to be short so the bids went unused and the two offers were contracted for their full respective volumes (if they had been lower in the stack they may have had a smaller or indeed zero volume accepted). Once the imbalance volumes have been measured, the SBP (if short) or SSP (if long) is calculated by where Pi is the ith offer price and Vi is the ith imbalance volume. The BM data was analysed in much the same way as the FPN and MEL data to produce a separate file for each BMU detailing all the offers and bids it made during 2010. Each entry that was ultimately accepted by the system operator was flagged. I also undertook a statistical analysis of the BM data for each SP, calculating the maximum, minimum, mean, median and standard deviation for each set of bids and offers, performing separate calculations for the set of all bids/offers and the subset of only those which were accepted. To summarise then, my main data sets are FPNs, MELs and BM bids and offers for every generator BMU on National Grids system from 1st Jan 2010 to 31st Dec 2010. How representative is this dataset? It is important to note that all data comes from the BM, and that not every power source is a participant. There are many small power sources (for example, local CHP or wind power schemes) which are either not connected to the grid or are too small to be significant for system balancing. However the vast majority of larger plants are participants, including every plant of capacity greater than 50MW. It is probably fair to say that no one knows exactly how much generation capacity there is in the UK - the 2011 DUKES attempts to identify every generator over 1 MW capacity and lists a total of 370 plants with a total capacity of 85GW. Of that, the BM accounts for 80GW with just 114
plants (266 BMUs), showing that the vast majority of UK capacity is concentrated within the subset for which I possess data. I was also able to check the data against the NG Total Gross System Demand (TGSD) data, which is declared separately. TGSD is simply the sum of power output of every station connected to the grid and is reported for each settlement period. I calculated the sum of all BMU FPNs and compared them with the TGSD for each settlement period and found that they differed by an average of 1.3GW or about 3.5%. Since the FPNs are issued before final load balancing (via the BM), this is more or less what we would expect. Therefore from here on I shall assume that the dataset is complete even if that is not strictly the case. For comparison, Fig 2.4 shows the two demands shown side by side for a typical week in March (chosen at random). Interestingly we can see that the BMU FPNs tend to overestimate peak demand and underestimate trough demand this may be due to the effect of the French interconnect which tends to import at the former and export at the latter time (thereby requiring the BMUs to lower and increase their output respectively).
Fig 2.4 TGSD data (red) and FPN data (blue)
2.7 Generation MixAfter cleaning the data, each BMU was assigned a fuel type from the categories of Coal, CCGT, OGCT, Nuclear, Hydroelectric, Pumped Storage, Wind and Oil, allowing similar BMUs to be grouped together. I also assigned the ownership of each BMU from the options of Centrica, EdF, EoN, SSE, Iberdrola, RWE and Independent. Two particularly important plant statistics are the load factor (current power output as a fraction of possible output) and availability (possible output as a fraction of rated capacity). A high load means a plant is being well-utilised; a high availability means a plant is very reliable. I was able to calculate the load for each power-plant at each moment in time by dividing each FPN entry by its corresponding MEL entry. I calculated the BMU availability at each moment in time by dividing the MEL by the maximum MEL reported during the year. I was now in a position to answer a very wide range of queries. Average CCGT BMU size? 514MW. Average load factor for coal plants? 58%. Correlation coefficient between BMU capacity and load factor? 0.19. Hours of downtime for Drax turbine 1? 91. Hours of operation for Pembroke power station? Zero (it isnt built yet). Obviously there are a huge number of angles from which one could attack the dataset, so it is worth outlining my objectives in a little more detail. The overall aim is to ascertain the merit order of plant in the UK. In the Section 2.2 it was explained that the costs of generating electricity can be roughly broken down into fixed and variable costs. The variable costs are equal to the marginal costs of production (MC) and (I argued) it is this value that should determine the merit order of dispatch. If there was a strict merit order by plant type e.g. Nuclear < CCGT < Coal < Hydro < OCGT then one would expect the load factors to be something like this: Nuclear 100%, CCGT 100%, Coal 45%, Hydro 0%, OCGT 0% i.e. certain plant types would never get used as they would be too low down the merit order. In reality, as any economist will tell you, MC curves are not flat- particularly in this case where each plant type is made up of many different plants each with their own characteristics. One would expect the curves from plants types to overlap such that it might be cheaper e.g to increase Coal load from 0% to 10% rather than increase CCGT load from 80% to 90%.
Fig 2.5 shows the load factors for each plant type throughout 2010, calculated as 4-week moving averages. This allows one to follow the seasonal changes in demand while ignoring the short-term variations that depend on time-of-day and day-of-week. The dotted line shows the moving average of power demand, normalised by dividing by the mean demand for the year; it indicates how the total load varies within a year.
Fig 2.5 Load factors by plant 2010 The graph allows us to comment on the overall merit order. Taking each plant type in turn: Nuclear: consistently operated at full load irrespective of other factors, implying that it is top of the merit order. This agrees with what we know about nuclear i.e. that it has very high capital costs and low marginal costs. Coal: operates at between 40% and 85% and appears to follow the shape of the normalised demand curve. This implies that most of the time coal is the marginal plant in the merit order. OCGT and Oil: practically zero load throughout the year. This is because they are right at the bottom of the merit order and are only used as peaking plant at times of exceptional demand. Wind: achieves average loads of between 20% and 40%, but the output fluctuates wildly and at random. No surprises there wind of course is not dispatchable and indeed is notoriously unpredictable, and the grid must absorb it whenever it is available. The results
are in line with other studies of wind which shows that you can expect an average capacity factor of roughly 30% (depending upon location). Hydro: appears to be closely correlated with wind. Here then is proof that when it is windy, it is rainy! Though hydro plants have a limited ability to choose when to dispatch power over the course of the day/week, those variations are not visible on the graph so it appears to follow the load of wind. In fact if you look closely you can see that hydro load seems to lag behind wind by a few days; this could be evidence that the hydro plants wait a little before choosing the optimum time to empty their reservoirs. CCGT: a flattish load curve implying that it is not strongly influenced by total demand. However it does not operate at close to maximum, rather between about 60% and 75% - it appears that the majority of CCGT operates at baseload (always-on) but the rest is rarely used. The reasons for this are a little more complicated. A plausible explanation is that a lot of CCGT is still tied to long-term baseload contracts dating from the mid-1990s (see Section 1.5) and otherwise would be lower down the merit-order than coal plant. This is supported by the fact that the fuel cost of coal is currently much lower than gas so one would expect the MC of coal power to be lower than that of CCGT (IMF, 2011). Figure 2.6 shows the data in a slightly different way, breaking down average load factors for each plant type by time of day. The data is shown in 3-month blocks to allow seasonal comparisons. The dotted line shows the system load profile. One can see that system load shares many features across all seasons: a low load at night, a peak in the morning, a slight tailing off and plateau followed by a second peak in the evening. The chief difference between seasons (besides the absolute magnitude of output) is in the relative prominence of the peaks and the plateau.
Here the differences in load profiles between plant types are starker. Nuclear and wind are flat; hydro shows very pronounced peaks coinciding with demand peaks, indicating that dispatch is being controlled to maximise profit; oil and OCGT are again barely visible. CCGT load looks like a more flattened version of the system load, whereas Coal follows the system load in an exaggerated fashion. While there are many more features which one could comment upon, the important point is that the data presented in this way basically supports the conclusions made above.
Fig 2.6 Load factors by plant, season and settlement period, 2010
2.8 PricesSo what can we say about that actual prices that suppliers pay for wholesale electricity? This part is tricky because of aforementioned trade secrets. In the vast majority of cases, suppliers (i.e. the Big Six) buy their electricity from themselves or else from IPPs through long-term bilateral contracts. In each case the sale price is confidential. There are spot and futures markets in electricity but liquidity is very low. I calculated that in 2010 the average trading volume per settlement period was just 554MWh or roughly 3% of all electricity (in a healthy market this number would be at least 200%). In such a situation the spot prices are likely to be more volatile and probably systematically biased relative to
the true system price (since only desperate firms participate). Moreover the actual bid and offer prices are, once again, confidential: the exchange (AP ENDEX) only releases the average price per settlement period, which is unhelpful for anyone trying to distinguish between different power sources. There are other snatches of information available. OFGEM issues occasional reports on wholesale and retail prices (for example Table 2.1) and forces the Big Six to publish yearly segmental accounts which purport to break down the costs of generation and supply (see the next section) though it seems to me that some of the numbers are dubious: does Scottish Power really spend three times as much as Scottish and Southern to generate 1 MWh? However our main source of pricing information remains the Balancing Mechanism bids and offers. To reiterate: these are bids/offers that BMUs make every settlement period to balance supply and demand - each entry consists of BMU ID, bid/offer volume, bid/offer price and a flag (accepted/declined). In theory each bid/offer should represent the marginal cost of electricity for each power plant at that moment in time. However the dataset, while very large (10-20 entries for each of the 17520 settlement periods), was never really supposed to ascertain the true marginal cost of electricity so should be used with care. Firstly, not every BMU chooses to make bids and offers so it is an unrepresentative (self-selecting) dataset. Secondly, the balancing mechanism only settles relatively small volumes in 2010 the average (absolute) NIV was just 294 MWh and tells us very little about the other 98% of generation. Finally there is the problem of bids/offers themselves. Bids are always lower than offers, so which price best represents the marginal cost of generation? Suppose that a power plant is on a long-term contract whereby it sells electricity at its marginal cost of 30/MWh. It may then choose to use the BM to place an offer of 35/MWh and a bid of 25/MWh and bag itself a 5/MWh profit if either is accepted (otherwise it may as well not bother). So we expect that the true marginal cost for a plant actually lies somewhere between the average bid and average offer price. With these caveats in mind, let us look at the data. Fig 2.7 shows the average lowest, highest and mean accepted bids and offers for each settlement period. Unsurprisingly the offer price profiles follow the load profile through the day, featuring the same twin-peaked characteristics. The bid price profiles are very flat indicating that it is always relatively cheap to lower ones electricity production. Note also that there is a consistent gap
between the bid and offer prices. The mean offer is always at least 10/MWh above the mean bid, rising to 40+/MWh during periods of peak demand.
Fig 2.7 BMU prices plant and settlement period, 2010
Fig 2.8 is a little more complicated. It shows a plot of cumulative offer volume versus offer price for each plant type, sorted ascending; the axes are logarithmic in order to show all data at once. The bid data has much the same shape but with slightly lower prices overall. There is no nuclear or wind data as they do not participate in the balancing mechanism. We can make a few observations. The coal and CCGT curves are near-identical, suggesting they have very similar costs or alternatively that some sort of gaming behaviour is occurring, given coal marginal costs are thought to be lower. The other plant types are priced in the order predicted in the previous section. Every curve has a prominent flick in its tail (even on a logarithmic scale): these are indicative of speculative bids that were accepted at a time when the grid was operating very close to full capacity; hence they are not marginal offers, indeed they likely deliver windfall profits to generators. Fig 2.9 shows a section of Fig 2.8 with linear scale allowing one to better make out the gradient of CCGT and coal curves.
Fig 2.8 BMU prices by cumulative volume and by plant, 2010. Log-log plot Fig 2.9 Detail of Fig 2.8, linear plot
It is important to recognise that Figs 2.5 and 2.6 are not marginal cost curves nonetheless they obviously tell us something about electricity prices. Indicative statistics are shown in Table 2.4. I shall use these results in the next section when choosing prices for the grid model.
Plant CCGT Coal Hydro
Min Mean Median Stand. Dev. 27 30 45 60.7 61.9 136.9 147.6 339.2 277.7 57.5 58 125 144 345 270 19.5 21.2 58.5 34.8 141.7 41.5
Pumped Storage 45 Oil OCGT Nuclear Wind 75 180 N/A N/A
Table 2.4 BMU Statistics by plant type
2.9 The Big SixAs a slight digression before moving on to the final section, it is worth looking briefly at wholesale market structure. As mentioned in section 1.6, the Big Six integrated suppliers have achieved very dominant positions in the ESI, accounting for 70% of UK capacity and 99% of supply. I have summarised key corporate indicators for their UK and international operations in Tables 2.5a and 2.5b. One can see that the UK-based companies are notably smaller by asset base and it would not be surprising at some point to see them being bought out or perhaps even merging. The other four are truly corporate behemoths - EdF is the worlds largest utility. Section 1.6 mentioned that the Big Six are currently under review for abusing their oligopolistic market positions. There is clear evidence of this in the retail market where high margins are indicative of market power, and NETA would seem to lend itself to abuse
at the wholesale end too. I was interested to see if the Big Six leverage their positions to squeeze out IPPs.
Generation Capacity Company EdF E.on RWE Iberdrola SSE Centrica Total (MW) 14087 10170 11751 5889 9270 4363 55530
Gen Revenue (m) 3574 1575 733 1643.5 841 1075 9441.5
Gen costs (m) 2433 1330 392 1233 424 893 6705
Gen 2010 (TWh) 71.6 29.8 32.6 26.9 33 22.8 216.7
Sales 2010 (TWh) 63.6 48.3 49.8 23.1 60 45.1 289.9 Gen Costs (/MWh) 49.92 52.85 22.48 61.10 25.48 47.15 43.57 Gen Margin (/MWh) 15.94 8.22 10.46 15.26 12.64 7.98 12.63
UK revenue Company Ownership French (85% EdF E.on RWE Iberdrola SSE Centrica State) German German Spanish UK UK 5.36 4.48 4.49 2.32 5.78 4.63 (m)
UK profits (m)
Group revenue (m)
Group profits (bn)
Group Assets (bn)
Market Cap (bn)
-0.18 0.21 -0.09 0.00 0.27 0.23
57.36 81.75 46.94 26.79 28.33 22.40
0.90 5.15 2.91 2.53 1.42 1.30
211.27 133.80 81.87 82.48 21.45 19.28
33.54 24.82 12.06 29.05 11.30 14.91
Table 2.5a Big Six UK statistics Table 2.5b Big Six Group statistics
One way to do this would be for the Big Six to put a downward pressure on wholesale prices (while cross-subsidising their own wholesale businesses from retail), thus making IPPs less profitable and therefore vulnerable to takeovers. The number of IPPs that have
gone bankrupt and/or been bought out in the last decade suggests this may well have occurred (c.f. the fate of British Energy, Section 1.6). However in the absence of reliable wholesale prices, this behaviour is difficult to prove; instead, I decided to look at plant utilisation. If, for example, it turned out that IPPs were achieving lower average plant loads than the Big 6, that might be evidence that the Big 6 have a preference for using their own plant i.e. IPPs are getting cut out of the market. To that end, I calculated the average load for CCGT and coal plants owned by the Big Six and IPPs respectively. The results (shown in Table 2.6) actually suggest the opposite that IPPs achieve considerably higher loads overall. Could this be evidence that the IPPs are superefficient, proof that market forces are working their magic? A more likely explanation is that since IPPs are by nature financially precarious it is very important for them to achieve a high utilisation. Whereas the Big Six can afford to shop around for their wholesale power, IPPs will to keep the turbines turning 'at any cost to bring in revenue. One would expect this to result in lower per-MWh income for IPPs i.e. a price squeeze. However in the absence of price data, this is just speculation. It occurred to me that if IPPs are being squeezed, they might make a more aggressive effort to exploit the BM to top-up their earnings. This would manifest itself as IPPs making a disproportionately large number of offers and bids and taking a disproportionate amount of BM revenue. However I found that IPPs accounted for 27% of BM bids/offers (by volume) and took 27% of BM revenue, which is close proportion to their ~30% market share. Overall, then, the dataset gives evidence that IPPs and the Big Six do have different modes of operation, but in the absence of detailed price information it is not possible to prove abuse of market power. Big6 CCGT Coal 0.43 0.35 IPPs 0.60 0.60
Table 2.6 Load factors by plant and owner
Chapter Three: Grid Model
In this final chapter I develop a (relatively) simple model of electricity supply which enables me to investigate various possible futures of the Grid. Whilst a fully realistic model would be an unfeasibly immense undertaking, I believe that as long as I capture the key features of the industry I can still contribute a useful analysis. Moreover, I am aware that this work is not original and many others have attacked the problem with far greater resources, for example Poyry (2010); therefore I see the modelling in part as an intellectual exercise how far can you get with just a computer, a coffee machine and a deadline? However, I believe that my method and results do offer some novelty, particularly in the final section.
3.1 Model DesignThe model treats all UK power as originating from just 6 generalised sources: Nuclear, CCGT, Coal, OCGT, Onshore Wind and Offshore Wind. Other sources are ignored as they make up an insignificant fraction of generation (e.g. hydro) or are similar enough to be subsumed into another source (e.g Oil). In outline, the model works as follows: Taking each settlement period one-by-one and a given set of input parameters (in particular, the power demand to be met), it generates a marginal cost curve of electricity. From this curve the model estimates the order of dispatch, returning, amongst other things, the power output and revenue for each type of power plant. From this one can calculate, for example, average electricity cost, average plant loads, or total carbon emissions. By looping through several days or weeks one can simulate the output and costs over a period of time. The model is constructed so that every parameter can be specified at every point in the time period or just a handful of times, or just once. By changing the input parameters (for example, the balance of nuclear and wind energy, or the cost of CCGT) one can explore a range of hypothetical scenarios. Costs are calculated separately for baseload and variable generation. For each plant type one specifies the fraction that is baseload and fraction that is variable; baseload generation operates at a fixed price and output regardless of power demand or marginal price. The power generated through baseload is subtracted from system demand before the marginal cost curve is calculated, then added back in when calculating that final outcomes. For each plant I chose the cost/MWh assigned to baseload generation to be
equal to the median cost of variable generation, or else (if there is no variable generation) equal to coefficient c1 (see below).
3.2 Demand, Availability, CapacityThe demand profile is an important input. I used the profile from 2010 as the starting point. The demand profile is then modified for future years by multiplying throughout by a constant in proportion to the expected change in demand (which may be positive or negative). This approach succeeds in changing overall demand while maintaining the same generic shape for each year (important to capture daily, weekly and seasonal variations). The plant capacity the total possible output of a plant if it is fully functional also draws upon 2010 data as a starting point (see appendix A). By modifying the inputs one can map alternative scenarios: for example one might wish to steadily increase the amount of wind power and decrease coal as time progresses. The availability the fraction of plant which is operable at a given moment in time can arguably be ignored in some circumstances (i.e. simply set to 100%, or 85% or some other constant). However, analysis of the MEL data (section X) showed that availability does tend to go through a seasonal cycle, decreasing at times of lower demand - presumably because plants are taken offline and/or scheduled for maintenance at times when they are less likely to be needed. Therefore I created a profile for each plant based on the MEL data (smoothed by taking the 4-week rolling average). The profile was duplicated for each year of simulation. The exception was wind power, for which I assigned a randomly-chosen weeks worth of availability data from 2010 to each week of simulation data one-by-one. This is a simple way of capturing the extreme variability inherent in wind power.
3.3 Marginal Cost CurveThe marginal cost curve (MCC) is the key component of the model as it determines the variable outputs. It is generated first by creating a load-price curve for each power plant (running from 0 to 1). The load curve is a sum of linear and exponential curves, obeying the equation:
where p is the price, x is the load and c1 c5 are coefficients to be specified. I chose this form as it allows one to emulate the shape of the curves in Fig 2.8, i.e. smoothly increasing 54
from a constant but with a sharp flick at the tail. Fig 3.1 shows a such a curve achieved by setting c1 c5 = 15, 10, 5, 0.03 and 12 respectively .
Fig 3.1 Marginal cost curve components
Next each load curve is scaled by multiplying by the available capacity of its respective power plant yielding a series of six curves of varying ranges. The curves are combined together and sorted by price to give a single marginal cost curve. Getting the shape of the MCC is important but also a matter of judgement. The intention is to get the load curves the correct shape to start with and then explore different scenarios mainly by changing the constant c1, i.e. changing only the intercept. The logic behind the shape of each plant is as follows: Nuclear: Completely flat. Will be run at 100% baseload so the only important coefficient is c1.
CCGT: Curving gradually upwards according to the principles of increasing marginal cost. Sharply rising at near to 100% load to model the fact that the system prefers to have reserve capacity, and in circumstances where supply is stretched, marginal prices go 55
through the roof. CCGT will be run at 40% baseload in the present day scenario as per Section 2.7 but will be run at 0% baseload in most other cases. Coal: Similar to CCGT; the main point of differentiation between them is the intercept c1 with opportunity to increase c2 to reflect the ratio different ramping ratios. 0% baseload. OCGT: Linear load curve with high intercept and sharp gradient. A get out valve which stops the price rising far beyond a certain point (200-300/MWh) when supply is severely stretched. Wind: A special case. Will be run at 100% baseload, to reflect that it is always absorbed by the grid when operational. The variability of wind will be modelled by varying the availability (above), not the load factor. Therefore as with Nuclear, c1 is the only relevant coefficient; it should be set to equal the price paid for wind by the new REFIT mechanism. Onshore and Offshore Wind are differentiated to allow for two-tier REFITs. Fig 3.2a which needs labelling shows example load curves for each of the plants and Fig3.2b shows the corresponding cost curves along with the overall marginal cost curve.
Fig 3.2a Marginal cost curves by plant Fig 3.2b Overall MCC (blue)
3.4 OutcomeHow realistic is this model? Clearly it is highly stylised, a toy model. Some of the most glaring simplifications and omissions include: No geographical model of the Grid therefore no accounting for transmission bottlenecks and losses No model of market structure assumes perfectly efficient market, always choosing least-cost first No model of demand side demand is a model input Time resolution limited to every 30 minutes no treatment of fine variations in demand Similarly, no treatment of plant ramp rates assume plants can increase and decrease output without penalty All variations between plants subsumed into a single cost curve. Several technologies omitted completely, most notably hydro Monotonically increasing MCC in reality running a plant at a lower load factor is often less efficient, therefore sometimes reducing output would increase marginal cost In spite of this I believe the model captures enough detail to make it useful to look at some loosely-sketched future scenarios, provided the input parameters are chosen wisely. Let us look at a sample set of outputs. Fig 3.3a shows a set of results from a week picked more-orless at random, the week starting 25th Nov 2010. The parameters are close to real-life except that the amount of Wind has been doubled to provide a little interest. One can see that Wind fluctuates semi-randomly contributing occasional bursts of energy to the grid, whereas Nuclear is completely flat operating completely in baseload. CCGT and coal make up the rest of the supply in roughly equal proportion; CCGT is running at 40% baseload but is slightly more expensive than Coal, so Coal shows greater variability. At times of maximum demand, OCGT steps in to contribute the final few GWs of supply. Fig3.3b shows the corresponding prices calculated by the model. The peak variable price (i.e. the marginal price) usually follows the supply curve closely but we can see large peaks at times of 57
exceptionally high demand. The average variable price is lower and much less variable but shows the same basic shape, as one would expect. The overall price takes into account the payments made to wind and nuclear too; in this example Nuclear, Onshore Wind and Offshore Wind are receiving payments of 30/MWh, 70/MWh and 140/MWh respectively. This means that at periods of high wind the overall price is pushed up somewhat; otherwise it is pushed slightly down. Comparing the model outputs with the real-world data in Fig 3.4a, the similarities are encouraging. The shapes are slightly different because the model uses the TSGD as its demand input which does not match the FPN data exactly (Section 2.7 ); nonetheless the generation mix is a close match and could be further improved by tweaking parameters. Comparing prices (Fig 3.4b), it appears the model marginal price is a good proxy for market spot price. Of course the whole point is that the market spot price is a very flawed metric and we do not know real prices very accurately, so we neednt be too concerned with discrepancies as long as the model results are robust.
Fig 3.3a Sample model output Fig 3.3b Sample model prices
Fig 3.4a and 3.4b Corresponding real-world data 59
3.5 ScenariosHaving outlined the model and established that the results are plausible, let us put it to use. Obviously there are potentially thousands of input parameters 37 for each cycle so to impose some logic on the process I have developed four scenarios for the future. Each scenario sketches out a possible path for development of the UK ESI. Each scenario runs from 2010 until 2025 and the simulation will run for the entirety of that period (a total of 48*365*15 = 262800 cycles - taking roughly 15 minutes to execute). I have assumed that the basic costs of building plant remains the same in all scenarios - now is finally the time to confront what prices to choose. As mentioned above, the load curves have been chosen such that I need only pick the minimum price c1. However there is a rather glaring inconsistency between my data sources. The BM prices (Table 2.4) suggest that some CCGT and Coal plants have a marginal price as low as 30/MWh, and the OFGEM data gives an overall average of 56/MWh (Table 2.1). However the levelised cost data suggests that new-builds will have to charge at least 80/MWh to break even with Offshore Wind costing a boggling 190/MWh. To quote a consultant at Poyry (2011), electricity costs are going up... way up! How to reconcile the data sources? As a test I ran the simulation with todays parameters and values of c1 equal to the minimum values in Table 2.4, and found that the overall average price came out at 57/MWh, very close to OFGEMs value. Therefore I will use these as the 2011 prices. My strategy is then to increase prices linearly until they reach the LC levels by 2020, after which they will remain flat. This is not the most elegant solution but clearly increased capital costs cannot just be ignored. However, (here it gets a little tricky), the LC levels will not necessarily be the ones calculated in Section 2.3. The LC model has several input parameters which should for consistency be obtained from the grid model for example, plant load factor but the output from the LC model also affects the grid model, i.e. there is a feedback loop between the LC model and the grid model. However there should be a stable equilibrium where the price gives the output that gives the price, so I constructed a script to continually change the parameters until this equilibrium is found. With the input prices settled, lets take a look at the scenarios: Base Case: Based on government projections from 2010 before the EMR was announced. Low case taken on the assumption that business-as-usual policies Total demand to increase by 6%; capacity to increase by 14% 60
Wind to increase to 30% of capacity; CCGT to 40%; others to decrease Carbon prices to increase from 16 to 49 EUR/tonne Wind energy payments same as today: 50/MWh onshore, 100/MWh offshore (eroc auction) Wind 50/50 onshore and offshore
Energy Market Reform: As above but implementing new EMR policies. Lower payments for Wind through REFIT : 40/MWh onshore, 80/MWh offshore Nevertheless, more Wind built increasing to 40% capacity Higher Coal and CCGT prices due to carbon price floor More Nuclear built: capacity decreases then increases at end of decade, 10% capacity by 2025 Go For Green: An alternative which attempts to aggressively cuts carbon by introducing wind power at an accelerated rate. Demand-side reduction: demand falls by 10% Higher carbon price, increasing faster and further from 16 to 100 EUR/tonne Coal and Nuclear phased out CCS on new CCGT plants reduces emissions by 30% but increases costs by 40% Higher offshore REFIT: 110/MWh More Wind built: increasing to 50% of capacity Greater fraction of Wind built onshore (66%)
Too Cheap To Meter: An alternative which brings lots of Nuclear online, backed up by CCGT Nuclear construction increases linearly to 35% of capacity by 2025 Coal phased out Slow Wind uptake, increasing to 20% of capacity CCS on new CCGT plants reduces emissions by 30% but increases costs by 40%
3.6 ResultsSummary results (across 15 years) are shown in Table 3.1.
This result does not appear to vary much between scenarios the difference between in price highest and lowest is only 15%. This is likely due to the fact that all scenarios have the same start point and there is a limit to how far they can evolve for that point in 15 years. In all scenarios the fractional price increase over today is substantial ranging roughly from 20-40%. The EMR price is notably higher than others, which I believe is due mainly to the effect of the carbon price floor pushing up baseload prices - and at least it does achieve a 10% reduction in CO2 emissions versus BC. How is it that the GFG and TCTM scenarios achieve lower prices? Simply because they reduce both CO2 by a substantial amount, and carbon prices are expected to be a significant cost in the future, adding around 30/MWh to the price of Coal and 12.50 to CCGT by 2025. The GFG also benefits from the enlightened populace putting up with lots of onshore wind which is substantially cheaper than offshore. However the intermittency of Wind means that CCGT still has a large role to play in the GFG scenario. This is less of an issue with the TCTM scenario between Wind and substantial Nuclear generation, it achieves the highest carbon reductions of all and at relatively low cost (perhaps this is something for environmentalists consider?).
Average supply Revenue (GW) 39.37 39.60 37.12 40.31 (bn) 626.27 727.21 642.85 663.69
CO2 Emssions Average (Mtonnes) Error (%) 0.30 0.87 1.12 2.62
BC EMR GFG TCTM
5173.62 5203.34 4877.75 5297.34
60.53 69.88 65.90 62.64
1766.62 1560.40 1380.36 950.67
Table 3.1 Model summary results by scenario
Fig 3.5 shows the generation mix by year for each scenario. The most striking thing is how similar they all look, especially wind. The GFG scenario ends up with a huge 47GW of Wind capacity by 2025, enough in theory to deliver the vast majority of our electricity. But the load factor averages only around 30%, so even in the end it is still only providing 40% of energy (and a fraction of this will be unwanted see below). The overall view is that CCGT dominates generation at present and will continue to do so in the future perhaps even more so as Coal plants are retired. Consequently the most important developments from a consumers point of view is anything which impacts the price of CCGT, be it gas prices, a carbon tax or mandatory CCS. 62
Fig 3.5 Load factors by year for each scenario
3.7 OversupplyThe GFG and TCTM scenarios may seem appealing but they have a distinct problem, as we shall see. The final column of Table 2.1 shows the Average Error; this is the average percentage difference between supply and demand, calculated as where S is
total supply and D is total demand. In a well-functioning grid supply and demand are in perfect balance at all times (otherwise the system frequency drops and customers may experience brownouts or blackout) but in the future this may not be the case unless there is careful planning. If there is too much wind and not enough dispatchable generation then when wind drops to near zero (it happens on a UK-wide basis several times a year) there will be insufficient power supply. Equally, if there is a burst of wind the grid will struggle to absorb the excess power - electricity prices may even go negative as suppliers have pay businesses to increase power consumption or else risk damaging the grid infrastructure. Nuclear is also a liability because for technical reasons it can only change its output the course of weeks, not intra-day as load balancing requires. I have included average error as a measure of how much of a problem this is likely to be. Given that the supply and demand are in balance most of the time, an average error of 163
2% conceals some quite severe individual imbalances. As we can see, this problem is bad enough with lots of wind (GFG scenario) but with potentially disastrous with wind and nuclear and not much else (TCTM). Figure 3.6 illustrates this point further. It shows the oversupply each year i.e the number of GWh which were generated beyond what was required (the undersupply is tiny by comparison). It shows, unsurprisingly, that the more wind and/or nuclear, the more likely you are to have a oversupply problems, whereas CCGT alleviates the problem - it is no coincidence that the TCTM scenario peaks at the point where there is maximum nuclear and minimum CCGT. Though the GFG and TCTM scenarios are unrealistic in that they call for rapid deployment of new technologies that (to put it bluntly) simply wont happen, they highlight the point that sometime soon intermittency will become a serious problem. I will spend the remainder of this chapter with a brief investigation of the mooted solution to these problems: storage.
Fig 3.6 Oversupply by year
3.8 StorageIn the past five or so years there has been increasing interest in the energy storage as a possible solution to these problems (MacKay, 2008). Actually there has always been a small subset of the ESI concerned with energy storage, even before the prospect of significant non-dispatchable loads. The rationale is simple: The UK power supply can vary in magnitude by up to 100% within 24 hours. At peak periods this means activating the less efficient, more expensive and more polluting peaking plants. However, if one could store 64
energy from baseload sources at times of low demand and offload it at peak periods, the need for peaking plant could be reduced. Not only is this potentially cheaper, it puts less stress on the grid (since the supply curve is smoothed) and reduces carbon emissions. Moreover, by buying low and selling high, storage operators can make a tidy profit. When factors in the need to level loads from wind and other intermittent sources, one can see why storage has become a very important - and potentially very valuable commodity. (I should add that there are also many ancillary ways that storage can add value e.g. load balancing, black start, backup power - a recent report identified 17 separate uses (Eyer & Corey, 2010). However here I will focus on the arbitrage aspect.) The real stumbling block to energy storage is technology. The only viable large-scale technology is pumped hydroelectric, where energy is stored by pumping water uphill and released by sending it back through turbines. Storage in this way is relatively cheap and achieves round-trip efficiencies of 70-80%; the problem is that it requires a spare mountain (with planning permission). Therefore pumped-storage projects tend to be few and far between. In anticipation of future need, there are a host of other technologies vying to be the next big thing in storage, for example Compressed Air, Pumped Heat, Flow Battery and Flywheel technologies, but as yet nothing competitive with Pumped Hydro (Walawalkar, 2008). The UK grid has four pumped-hydro plants but capacity is dominated by Dinorwig, an amazing structure set inside a mountain in North Wales. It was initiated in 1974, supposedly for the nuclear revolution which never arrived, has a peak output of 1.8 GW and a maximum capacity of 9.4 GWh (MacLeay et al., 2011). There are currently no plans for new pumped hydro to be constructed in the future, which begs the question do we have enough?
3.9 Modelling StorageThis is a complicated question (storage is a complicated topic) so I will aim to provide an answer in a fairly narrow way. I will use my model to see a) how the introduction of storage would alter the above scenarios and b) how much money a storage provider could make in the process. I chose to model storage as follows: prior to running the grid model, the power demand profile and baseload output are supplied to a separate script. Given a store capacity (MWh) and efficiency (%), this script will optimise the demand so that the store exactly fills and 65
exactly empties once every day, or else flattens the load profile whichever would create least overall demand. In the case where there is too much baseload power to be fully absorbed, the script simply optimises as best it can (some energy will still have to be dumped). The model is then executed as before with the new demand profile. The marginal price for each settlement period is taken to be the price at which storage provider buys/sells power. By multiplying the list of marginal prices by storage electricity purchases/sales and summing, we obtain the total profit (or loss) made by the storage provider. By comparing the overall simulation indicators (e.g. revenue, CO2 emissions etc) with the zero-storage case we can see what kind of effect the introduction of storage has had. This method is summarised schematically in Fig 3.7.
Fig 3.7 Modelling Storage
3.10 ResultsI chose to model 3 storage scenarios on top of the four previous scenarios. The first scenario has 10,000 MWh storage (Store10K), the second has 100,000 MWh storage(Store100K) and the third has infinite storage (StoreInf). The third one is also slightly different in that it balances the loads across four days rather than just one. Efficiency is set equal to 80% in all cases. Store10K is a realistic scenario (c.f Dinorwig), Store100K represents the limits of what could plausibly happen, and StoreInf is the best possible scenario to test the limits of the usefulness of the concept. The results are shown in Table 3.2 alongside the original simulation results for comparison. Three columns have been added. Store profit is straightforward, but note that it excludes the costs of building and operating the storage. Benefit to Grid is the reduction in revenue (i.e. reduced cost) to the ESI as a whole due to using fewer peaking plants etc. Total benefit is the sum of these two measures and is equal to the total welfare benefit to society.
Total Supply (TWh) 5173.62 5179.29 5087.70 5067.22 5203.34 5205.04 5105.94 5074.34 4877.75 4871.41 4784.64 4750.01 5297.34 5283.20 5166.79 5124.39
Ave supply (GW) 39.37 39.42 38.72 38.67 39.60 39.61 38.86 38.72 37.12 37.07 36.41 36.25 40.31 40.21 39.32 39.11
Average Error (%) 0.30 0.11 0.06 0.05 0.87 0.20 0.05 0.05 1.12 0.5 0.06 0.06 2.62 0.11 0.06 0.06
Store Profit (bn) 0.00 0.28 4.55 4.75 0.00 0.68 5.62 6.35 0.00 0.34 4.94 6.24 0.00 1.24 3.77 4.38
Benefit to Grid (bn) 0.00 2.92 17.51 20.36 0.00 1.64 21.81 27.35 0.00 3.57 16.50 22.74 0.00 2.64 19.34 25.16
Total Benefit (bn)
BC BCS10k BCS100k BCSInf EMR EMRS10k EMRS100k EMRSInf GFG GFGS10k GFGS100k GFGSInf TCTM TCTMS10k TCTMS100k TCTMSInf
626.27 623.35 608.76 605.91 727.21 725.57 705.40 699.86 642.85 639.28 626.35 620.11 663.69 661.06 644.36 638.53
60.53 60.18 59.83 59.79 69.88 69.70 69.08 68.96 65.90 65.62 65.45 65.28 62.64 62.56 62.36 62.30
1766.62 1757.13 1686.31 1675.74 1560.40 1554.91 1478.62 1463.33 1380.36 1379.37 1337.68 1325.59 950.67 945.23 893.70 877.36
0.00 3.20 22.07 25.11 0.00 2.32 27.43 33.71 0.00 3.92 21.44 28.98 0.00 3.88 23.11 29.54
Table 3.2 Storage model summary results by scenario
There is a wealth of information here. The simulations confirm most of the trends one would expect: more storage means lower costs, less CO2 and lower errors (i.e. better grid reliability). We can see plenty of interesting features : for example, it appears that the two grid scenarios with most nuclear (EMR and TCTM) also benefit the most from storage. This suggests, somewhat contrary to received wisdom, that storage is even more important for nuclear than for wind perhaps simply because nuclear loads are consistently higher than wind and so when they do cause problems, they cause big problems. I dont want to dwell on all the features of the results. The general message is what is important: storage is going to be big. If we can increase our storage tenfold from where we are today (roughly, Store10K to Store100K), the potential benefit is enormous in most scenarios (including EMR) averaging well over 1bn/year. The overall conclusion is that it is essential to invest in storage, the more, and the sooner, the better. I would add one point: the storage profits tend make up only around a fifth of the total welfare benefit. Is a large enough slice of the pie to encourage investment? There has been some work that suggests it is (Sioshansi et al., 2009), but I am not so sure. Is this set to be the big next failure of the free market in the UK ESI? At the very least, we can say that it would have good pedigree.
In this dissertation I have drawn to sketch history of the UK ESI, detailled the current state of play of the industry, commented on the latest incentive scheme and speculated on the future direction. Given the vastness of the subject it has been necessary to gloss over and simplify many important points. Nevertheless, I will now draw some overall conclusions from the study. Firstly, the UK has not been well served by privatisation. The state has lost control of a vital economic and strategic asset, customers have been overcharged, many businesses have gone bust, and been the cause of who-knows how many headaches for the regulator. The state should not be afraid to become more interventionist, perhaps even creating a new state-owned integrated supplier, in an attempt to wrest back control of the industry. Secondly, the state faces a huge challenge in the next decade to marshal the expertise and investment necessary for the Green Revolution (while also keeping the lights on). Recently it has made some progress in this area but I am still not convinced it understands the scale of the task it has set itself. The EMR is a good start, but now it needs to think bigger. Finally, energy prices are going to rise in the future, of that there is no doubt but with careful planning, and particularly with shrewd strategic investment in storage R&D and assets, the costs to the consumer will be minimised and we might even create a few green jobs in the process. Overall, the ESI has had an interesting couple of decades - and it is about to have at least a couple more.
ReferencesConservative Party General Election Manifesto (1987) [Online] Available from: http://www.conservativemanifesto.com/1987/1987-conservative-manifesto.shtml [Accessed July 2011]. Nationalised Industries Investment Review. (1969), National Archives. [Online] Available from: http://filestore.nationalarchives.gov.uk/pdfs/large/cab-129-143.pdf [Accessed July 2011]. Arup (2011) Review of the generation costs and deployment potential of renewable electricity technologies in the UK. Department of Energy and Climate Change. Report number: REP001. Berrie, T. W. (1983) Power System Economics. Institution of Engineering and Technology. Branston, J. R. (2000) A counterfactual price analysis of British electricity privatisation. Utilities Policy, 9 (1), 31-46. Branston, J. R. (2002) The price of independents: an analysis of the independent power sector in England and Wales. Energy Policy, 30 (15), 1313-1325. British Wind Energy Association. (2009) Consultation on Renewable Electricity Financial Incentives: BWEA response Byatt, I. C. R. (1979) The British Electrical Industry 18751914: The Economic Returns to a New Technology. Carrington, Damian. (12th July 2011) Live blogged: The UK's new energy future. [Online] Available from: http://www.guardian.co.uk/environment/damian-carringtonblog/2011/jul/12/electricity-reform-energy-nuclear-carbon . Chesshire, John. (1996) UK Electricity Supply Under Public Ownership. In: Surrey, John (ed.) The British Electricity Experiment. Privatisation: the Record, the Issues, the Lessons. Chesshire, John. (1992) Why nuclear power failed the market test in the UK. Energy Policy, 20 (8), 744-754. 72
Chick, M. (1995) The political economy of nationalisation: the electricity industry. In: Milward, R. & Singleton, J. (eds.) The Political Economy of Nationalisation in Britain 19201950. Commission of the European Communities. (1999) Electricity from renewable energy sources and the internal electricity market. Commission Working Document. Report number: SEC(1999) 470. Commission of the European Communities. (2005) The support of electricity from renewable energy sources. Brussels, Commission of the European Communities. Report number: COM(2005) 627. Cornwall Energy Associates. (2008) Gas and electricity costs to consumers. National Right to Fuel Campaign. Dadeviren, Hlya. (2009) Limits to Competition and Regulation in Privatised Electricity Markets. Annals of Public and Cooperative Economics, 80 (4), 641-664. Dadeviren, Hlya. (2007) University of Manchester. Privatisation of electricity and water Is it still worthwhile? Global Poverty Research Group and Brooks World Poverty Institute Conference. Poverty and Capital. pp. 1-20. Department of Energy and Climate Change. (July 2011) Planning our electric future: a White Paper for secure, affordable and low-carbon electricity. Department of Trade and Industry. (2007) Renewable Energy: Reform of the Renewables Obligation. EDF Energy. (2011) Press Release: Tuesday 12 July 2011. e-ROC. (2011) e-ROC Track Record Data. [Online] Available from: http://www.eroc.co.uk/trackrecord.htm [Accessed September 2011]. European Commission. (2004) Commission decision of 22 September 2004 on the State aid which the United Kingdom is planning to implement for British Energy plc. Report number: 2005/407/EC.
European Parliament and the Council of the European Union. (2009) Directive 2009/28/EC of the European Parliament and of the Council of 23 April 2009 on the promotion of the use of energy from renewable sources and amending and subsequently repealing Directives 2001/77/EC and2003/30/EC. Official Journal of the European Union, I (140), 16-62. Eyer, Jim & Corey, Garth. (2010) Energy Storage for the Electricity Grid: Benefits and Market Potential Assessment Guide. A Study for the DOE Energy Storage Systems Program. Giulietti, Monica, Grossi, Luigi & Waterson, Michael. (2010) Price transmission in the UK electricity market: Was NETA beneficial? Energy Economics, 32 (5), 1165-1174. Haar, Laura N. & Jones, Trefor. (2008) Misreading liberalisation and privatisation: The case of the US energy utilities in Europe. Energy Policy, 36 (7), 2610-2619. Harvey, D. (2005) A Brief History of Neoliberalism. Oxford, Oxford University Press. IMF. (2011) Primary Commodities Prices Tables 1980-2011. [Online] Available from: http://www.imf.org/external/np/res/commod/index.aspx [Accessed July 2011]. Jamasb, Tooraj & Pollitt, Michael. (2007) Incentive regulation of electricity distribution networks: Lessons of experience from Britain. Energy Policy, 35 (12), 6163-6187. Joskow, Paul L. (2008) Lessons Learned From Electricity Market Liberalization. Energy Journal, 9-42. Lohmann, Larry. (2009) Toward a different debate in environmental accounting: The cases of carbon and costbenefit. Accounting, Organizations and Society, 34 (3-4), 499-534. MacKay, David. (2008) Sustainable Energy - Without the Hot Air. MacKerron, Gordon & Segarra, Isabel Boira. (1996) Regulation. In: Surrey, John (ed.) The British Electricity Experiment. Privatisation: the Record, the Issues, the Lessons. pp. 95-119. MacLeay, Iain, Harris, Kevin & Annut, Anwar. (2011) Digest of United Kingdom Energy Statistics 2011. Mott MacDonald. (June 2010) UK Electricity Generation Costs Update.
National Audit Office (NAO). (2004) Risk Management: The Nuclear Liabilities of British Energy plc. Report number: HC 264. National Grid Electricity Transmission plc. (May 2011) National Electricity Transmission System Seven Year Statement. OFFER. (1991) Report on Pool Price Inquiry. Ofgem. (22 June 2011) Electricity and Gas Supply Market Report. Report number: 81/11. Ofgem. (6 October 2008) Energy Supply Probe: Summary of initial findings and remedies. Ofgem. (2010) Liquidity Proposals for the GB wholesale electricity market. Report number: 22/10. Ofgem. (1999) The New Electricity Trading Arrangements: Ofgem/DTI Conclusions Document. Ofgem. (2011) The Retail Market Review: Findings and initial proposals. Report number: 34/11. Ofgem. (1999) Rises in Pool Prices in July: A Decision Document. PB Power. (2004) The Cost of Generating Electricity. Pyry Energy Consulting. (2010) Demand side reponse: conflict between supply and network driven optimisation Sioshansi, Ramteen, Denholm, Paul, Jenkin, Thomas & Weiss, Jurgen. (2009) Estimating the value of electricity storage in PJM: Arbitrage and some welfare effects. Energy Economics, 31 (2), 269-277. Stiglitz, J. (2002) Globalization and it Discontents. Sweeting, Andrew. (2007) Market power in the England and Wales wholesale electricity market 1995-2000. Economic Journal, 117 (520), 654-685.
Thomas, S. (1996) The development of competition. In: Surrey, John (ed.) The British Electricity Experiment. Privatisation: the Record, the Issues, the Lessons. Earthscan Publications Limited, pp. 67-94. Thomas, S. (2006a) The British Model in Britain: Failing slowly. Energy Policy, 34 (5), 583600. Thomas, S. (2006b) The grin of the Cheshire cat. Energy Policy, 34 (15), 1974-1983. Thomas, Steve. (2010) Competitive energy markets and nuclear power: Can we have both, do we want either? Energy Policy, 38 (9), 4903-4908. Toke, David. (2010) Politics by Heuristics: Policy Networks with a Focus on Actor Resources, as Illustrated by the Case of Renewable Energy Policy Under New Labour. Public Administration, 88 (3), 764-781. Toke, David. (2011) The UK offshore wind power programme: A sea-change in UK energy policy? Energy Policy, 39 (2), 526-534. Toke, David & Lauber, Volkmar. (2007) Anglo-Saxon and German approaches to neoliberalism and environmental policy: The case of financing renewable energy. Geoforum, 38 (4), 677-687. UK Energy Research Centre (UKERC). (2007) A Review of Electricity Unit Cost Estimates. Report number: UKERC/WP/TPA/2007/006. Vickers, John & Yarrow, George. (1991) Economic Perspectives on Privatization. The Journal of Economic Perspectives, 5 (2), pp. 111-132. Wade, R. (1996) Japan, the World Bank, and the Art of Paradigm Maintenance: The East Asian Miracle in Political Perspective. New Left Review, I (217), 3-36. Walawalkar, Rahul S. (2008) Economics of Emerging Electric Energy Storage Technologies and Demand Response in Deregulated Electricity Markets. DPhil. Carnegie Mellon University.
Appendix AA description of the main generation technologies
Description Coal-fired steam boilers are the old 'work horses' of generation. The standard configuration is to pulverise the coal, burn it in a boiler and use heat exchangers to create high-pressure superheated steam (around 500C and 150 bar). The steam generates power by passing through a series of turbines before being released into a cooling tower. All but one of the UKs coal was built in the CEGB era and most of it is large, between 2 and 4GW. Coal is by far the most polluting energy source and many plants are scheduled for closure - others have installed expensive flue-gas desulpherisers. Coal designs are flexible and some can also accept gas, oil and more recently biomass as a feedstock. Open Cycle Gas Turbines are a relatively old technology that have mostly been superseded by CCGTs (below). In their simplest implementation they consists of a compressor, combusion chamber, expander and turbine on a single shaft. Air is passed through the compressor, mixed with a natural gas or vapourised oil feedstock, ignited and expanded to drive the turbine. Since OCGT are cheap, small and inefficient they nowadays exist mainly as 'peaking' plants, providing power only in times of exceptional demand. Some operate only a handful of times every year - others are mothballed for years at a time. Combined Cycle Gas Turbines are the 'state of the art' in conventional fossil fuel generation. Based on the OCGT, the hot gas leaving the expander is passed through a heat-exchanger to create steam which is passed through a second expander on the same shaft, greatly enhancing efficiency. Compared to coal, CCGT has a smaller footprint, lower capital costs, is more flexible and emits around 60% less CO2 per MWh (however it requires more maintenance and fuel costs are significantly higher). Since privatisation almost all new plants have been CCGT. The UK is one of the few countries that embraced nuclear technology, building the world's first commercial station in 1954 and amassing a stock of 16 by 1990, though enthusiasm has waned due to cost and safety concerns. There are many different implementations and the UK's are particularly idiosyncratic, but the basic principle is to initiate the fission of Uranium-235, causing a chain reaction which gives off large quantities of heat. The heat is used to run steam turbines similarly to a coal boiler. Unlike other technologies, nuclear is very inflexible and is run at a constant load for weeks or months at a time. Though once labeled "too cheap to meter", it is currently the most expensive type of generation by some margin - but also the only renewable technology with proven capacity. The preferred power source of environmentalists, being non-hazardous and zero-carbon, wind is perhaps the UKs best bet for a 'green revolution'. Essentially just a windmill joined to a turbine, wind power is expensive relative to conventional plants at present but this may change if/when carbon taxes are introduced. The biggest problem is land - of all technologies discussed here, wind farms have by far the largest footprint for a given power capacity and planning permission can be a problem. Offshore wind may be the answer, but it is even more expensive.
UK plant statistics
Total Capacity (MW) 28766
Indicative Thermal Efficiency 38%