Bigger Than Nuclear
Bigger Than Nuclear Amory B. Lovins -- perspective 02/11/05
Decentralized electricity generation is the new power technology of choice Decentralized power generation is enjoying impressive progress in the global market despite the many obstacles still in its way. But that progress is often underestimated. Amory Lovins and his Rocky Mountain Institute colleagues Kenneth Davies and Nathan Glasgow present an analysis of the growth of decentralized energy and suggest why its development is outperforming that of centralized plant, including nuclear technology.
Decentralized generation – in The Economist’s apt term, micropower – enjoys an important market share in some countries: in 2004, 52% of the electricity generated in Denmark, 39% in the Netherlands, 37% in Finland, 31% in Russia, 18% in Germany, 16% in Japan, 16 % in Poland, 15% in China, 14% in Portugal, 11% in Canada. Yet it is omitted from many official statistics and projections, underreported in the media, and often dismissed by policymakers as small and slow – a fringe market too trivial to bother with. Surprise! A recent compilation of industry and official data published in June by Rocky Mountain Institute (RMI) found that micropower worldwide has already surpassed nuclear power in both annual output (in 2005) and installed capacity (in 2002), and is growing far faster in absolute terms. In 2004, DG added 2.9 times as much output and 5.9 times as much generating capacity as nuclear power added worldwide. Since nuclear power is often represented as an important technology, providing 20% of US and 16% of world electricity, surpassing this benchmark should at last qualify micropower as a serious competitor. Roughly 65% of micropower’s 2004 capacity and 77% of its output was fossilfuelled CHP; the rest, diverse renewable sources. RMI’s assessment probably understates the actual total for both categories. A separate and similarly detailed compilation in press at Worldwatch Institute by Dr Eric Martinot (formerly at Lawrence Berkeley National Laboratory, now at Tsinghua University) draws similar conclusions but finds more small hydro because it uses China’s definition (up to 30 MWe, not 10 MWe). Neither RMI’s nor Martinot’s assessment includes big hydroelectric projects. In 2003, the US Department of Environment (DOE) estimated that hydroelectricity, very largely in big units and probably omitting many smaller ones, was 20% of world capacity and 17% of world net generation.
Figure 1 shows RMI’s findings about the installed capacity of worldwide micropower in 2004, Figure 2 the annual output. Also shown are the respective industries’ approximate projections to 2010. These are imprecise but qualitatively clear. All data shown are net of increases or decreases in rated capacity and of re- and decommissioned units.
Output lags capacity by three years as CHP has a higher average capacity factor than wind and solar, although some small hydro projects and most geothermal and biomass generation projects also have high capacity factors. Precise data on the fuel mix of CHP are not available, but WADE estimates it at 60%–70% gas-fired. This implies a CHP carbon intensity probably no more than half the global average for centralized fossil-fuelled generation, which is coal-dominated, with a normal range of around 30%–80% depending on relative efficiencies and fuels. 65% of micropower’s 2004 capacity and 77% of its output was fossilfuelled CHP; the rest, diverse renewable sources
LOW OR NO-CARBON EMISSIONS Since the renewable sources shown have zero direct carbon emissions, as does nuclear power if its enrichment energy and any decommissioning or wastedisposal energy are neglected, all the sources of electricity shown in Figures 1 and 2 are low- or no-carbon emitters. Nuclear power’s global capacity and output are both graphed for comparison. Fossil-fuelled, nonCHP generation is several times larger. Micropower’s greater market success compared with nuclear power is all the more impressive because nuclear has been heavily promoted and subsidized by nearly all host governments. In the US, for example, it received 24 times more federal subsidies per kWh in 1985 than did non-hydroelectric renewables.1 In nuclear’s first 15 years of industrial development it was about 30 times more heavily subsidized per kWh than all renewables.2 Moreover, micropower in many countries, including most of the US, is often barred from grid interconnection, on fair terms or completely, while centralized plants are not. Cost comparisons typically burden micropower with both grid expansion and firming or backup costs
(which are often exaggerated) while the corresponding costs of centralized plants, such as reserve margin, are typically socialized as normal overheads of the electricity enterprise. It seems inescapable that fair competition at honest prices would favour micropower over centralized plants even more than the market has been doing lately.
The data and sources behind these graphs are all publicly available3 and use a simple, transparent methodology4 whose main points include: •
•
Fossil-fuelled CHP data are from WADE’s World Survey of Decentralized Energy of March 2005 (www.localpower.org), indicating 282.3 GWe at the end of 2004. This landmark work is based largely on the annual Diesel and Gas Turbine Worldwide sales survey, direct salesdata collection from WADE members that make and sell distributed generating equipment, national market surveys, and standard wind and PV industry sources. WADE has carefully cross-checked its estimates of how much of the equipment sold was used in CHP applications and has conservatively omitted all units less than 1 MWe and all steam turbines outside China. The maximum unit sizes included are 30 MWe for engines and 120 MWe for gas turbines. To prevent double-counting, we subtract photovoltaics, WADE’s estimate of decentralized wind power, and all decentralized biomass CHP (which we count under biomass generation, not fossil-fuelled CHP, using central values from an unpublished WADE estimate that biomass fuels 3%– 5% of CHP in 2004 and could fuel 6%–8% in 2012). We use Michael Brown’s estimate of about 12% of global capacity as realistic in 2012 (below WADE’s 14% target, compared with 2004’s 7.2%) and interpolated intermediate values by curve-fitting. To calculate output, we assume that CHP plants average 7250 h/year of operation (an 82.8% capacity factor) based on his estimate of ‘7000–7500 h/year, possibly more’. Wind data are from the European Wind Energy Association for 1991–2004 and Worldwatch Institute for 1990. The capacity projections shown for 2005–10 range from EWEA’s 2005 to its 2002 Wind Force 12 estimates – the latter adopted by the IEA using a 25% capacity factor. Future values will almost certainly be higher.
•
•
• • •
Photovoltaic data are from Worldwatch Institute (which uses the industrystandard PV News database) and Solarbuzz’s 2005 Marketbuzz, solving for the lag between shipments and installations in 2004. We assume a 25-year average life, so the first commercial PV panels, made in 1971, started retiring in 1996. That will probably prove conservative since some modern PVs come with a 25-year warranty. We adopt the International Energy Agency’s (IEA’s) 2010 capacity factor. Small hydro’s 2004 capacity, 47 GW (of units up to 10 MW), is from the International Association for Small Hydro (IASH). The 2010 capacity is interpolated from simple exponential growth to the World Energy Council’s favourable scenario for 2020 (75 GW), then pre-2004 capacity is extrapolated backwards. We adopt for all years IASH’s implicit 2010 projection of 45.8% capacity factor. Biomass-fuelled generation: we adopt Navigant Consulting’s incremental capacity additions for 2000–10, renormalize the implied 2010 total installed capacity (55.7 GW) to match the IEA’s projected 55 GW and assume the IEA’s 70% capacity factor. Here, we adopt the IEA’s 2002 capacity, Navigant Consulting’s historical and projected capacity additions (renormalized from 11.8 to the IEA’s 13 GW total for 2010), and the IEA’s capacity factors (72% in 2002, 78% in 2010, scaled linearly). Nuclear capacity data are from the International Atomic Energy Agency (IAEA) through 2004, cross-checked for 2005 against the World Nuclear Association’s database, which we also use for expected 2005–2010 shutdowns. We adjust for upratings (which the IAEA ‘slipstreams’ rather than reporting explicitly) on the assumption that the 3.6 GW of 1991–2004 upratings occurred at a linear rate and can be extrapolated backwards one year to 1990 and forward to 2010 (agreeing within 0.12% with the USDOE’s projection). Nuclear generation data are from the IEA, the most complete historical source. To estimate 2005–2010 nuclear output, we assume a linear increase from 84.1% capacity factor in 2002 to the IEA’s 88.51% in 2010.
DIVERGENCE The divergence between officially favoured and actually achieved market outcomes in recent years became even more striking when RMI examined the first derivative of Figure 1 – the additions of worldwide capacity by technology and year, again including nuclear power as a bench-mark. This comparison, Figure 3, also includes a leading indicator: nuclear construction starts. Those data stopped in 2004 because 2005–2010 construction starts aren’t yet known. But because of this technology’s long lead times, they can’t affect its 2010 installed capacity. Again, an ineluctable conclusion from these data is that even the most favoured and subsidized centralized electric generating technology has been eclipsed by its decentralized competitors. This conclusion would become even stronger if we also considered decentralized electricity savings, i.e. demandside resources. But very few countries and no international organizations properly track demand-side resource additions. Even in the US, where detailed year-byyear evaluation data existed a decade ago, only a few states still measure more than the most aggregated econometric estimates of incremental ‘negawatts’. However, as a very rough indicator of demand-side progress, the 1.98% drop in US electric intensity in 2003 (whatever its causes) would correspond, at constant load factor, to saving 13.8 GWp – 6.3 times US utilities’ declared 2.2 GWp from demandside management5– and the 2004 intensity drop of 2.3% would have saved more than 16 GWp (plus 1 GWp/year from utility load management actually exercised). The US uses only one fourth of the world’s electricity, so it’s hard to imagine that global savings don’t rival or exceed global additions of micropower (24 GW in 2003, 28 GW in 2004). Thus the grand total for global additions of decentralized electrical resources, both demand- and supply-side, must exceed annual nuclear capacity growth by an order of magnitude. That this fact is unknown to most policymakers, commentators and investors is not only remarkable, it is likely to lead to poor decisions and heavy financial losses.
US merchant generating firms, for example, recently lost more than $100 billion of market capitalization, wiping out their equity investors, when they built about 200 GW of combined cycle plants for which there was no demand. The US Energy Policy Act of 8 August 2005 greatly increases subsidies and regulatory aid for supply-side resources, chiefly centralized ones, in apparent disregard of the market’s recent reductions in primary-energy and electric intensity of GDP by roughly 2.5%/year and 2.0%/year respectively. Since about 78% of the increase in US energy services provided in the last decade has been fuelled and powered by unnoticed intensity reductions, and only about 22% by the increases of physical energy supply described in voluminous official statistics, four fifths of what energy markets are doing is invisible. This risks repeating the market crash of the mid-1980s, when the officially promoted and subsidized supply expansions collided with rapid reductions in energy intensity (driven by late-1970s national policies and the 1979 oil price shock), causing glutted energy markets, collapsing prices and bankrupt suppliers. Obviously something other than national policy (in most countries) is causing this virtually unreported but dramatic marginal shift, especially in the electricity sectors, from the most favoured form of centralized supply to decentralized resources.
PLAUSIBLE CANDIDATE A plausible candidate would be relative prices. Consistent with the hypothesis that the world is buying more micropower (and electric end-use efficiency) because it costs less than centralized supply, a separate RMI analysis6 offers a simple, transparent, and documented comparison of the delivered cost of one marginal kWh from US centralized and decentralized resources, presented in 2004 dollars on a consistent accounting basis (other than project life, which is appropriate to each technology). Centralized or remote resources incur a delivery cost to the retail meter, conservatively taken as the 1996 embedded historic average (including average grid losses) for US investor-owned utilities – 2.75 cents/kWh. On-site resources avoid this cost because they’re already delivered. That new study’s findings are summarized in Figure 4. The approximately $13 billion of justapproved federal nuclear subsidies – roughly the entire capital cost for the next 6 GW – might, if highly successful, cut the cost of a new delivered nuclear kWh from 9.8 cents/kWh (using a 40 year life and 85% capacity factor) to 7.0 cents/kWh.
Meanwhile, a stiff tax or trading price of $100 per tonne of carbon could raise the nominal cost of a new delivered coal kWh from 7.2 to 9.7 cents/kWh (at $1.33/GJ coal), and that of a new delivered combined cycle, gas-fired kWh from 6.7–8.6 cents/kWh to 7.8–9.8 cents/kWh, assuming a levelized gas price of $3.6–7.6/GJ. That assumption is equivalent to assuming these gas prices initially, escalating at the real discount rate of 5%/year assumed for all the centralized plants. All centralized-plant findings are adopted from the canonical 2003 MIT study The Future of Nuclear Power, which uses a merchant cashflow model.
NUCLEAR’S MARKET FAILURE Thus the National Energy Policy Act of 2005 attempts to reverse the market failure of nuclear power by making it potentially competitive – under very optimistic assumptions – with its coal- and gas-fired central-station competitors. It will become obvious that those are the wrong competitors, because all three types of central plants are uncompetitive with three other options: wind power and some other renewables, CHP and end-use efficiency. Observed market behaviour, which strongly favours these options, implies that even if the policy intent – that yet again, nuclear power cannot be allowed to fail7– were realized, the desired outcome would not be, because the decentralized competitors, being markedly less costly, would continue to win. The documented assumptions behind the decentralized resources’ graphed costs are deliberately chosen to favour centralized plants by: • • •
understating marginal delivery costs counting a generous 0.9 cents/kWh firming and integration cost for wind power but no reserve margin for centralized plants (all types of generators are intermittent, differing only in the size, frequency, duration, cause and predictability of outage) offering the option to back out wind power’s Production Tax Credit, but not the centralized plants’ pre-2005 subsidies (those for nuclear power far exceed the PTC)
• •
using relatively high costs for windpower8 (its assumed basic cost, for 30 year projects including the levelized Production Tax Credit of 0.86 cents/kWh, is 3.0–3.5 cents/kWh – twice that of today’s cheapest projects) using static ‘snapshot’ costs rather than cost trends, which strongly favour micropower and efficiency (only the expected 1 cent/kWh drop in wind power costs to 2012 is shown).
The CHP costs are based on natural-gas prices of $5.4–8.7/GJ, $1/GJ higher than for central stations, and a 10%/year return over 25 years. Industrial CHP costs are based on five actual projects considered representative by Primary Energy LLC, a leading US CHP developer with about 0.9 GW of operating projects. The conventional industrial CHP projects shown are in the 28–64 MWe range. The 60–160 MWe projects using recovered waste heat, which has about 94 GW of untapped US potential9, deliver electricity at negative cost because the heat’s value more than repays capital plus non-fuel operating costs. Gas-fired CHP can yield even lower costs in a suitable building than in an industrial facility if it is highly optimized (typically by trigeneration or better, with system efficiencies equal to or above 0.9) and well integrated with prior demand-side improvements.
WIDE VARIATION Finally, the cost of saving electricity varies widely (as is also true of the supply-side resources), but it’s empirically often below 1 cent/kWh for well designed and well implemented commercialand industrialsector retrofits. It can be higher for suboptimal efforts or for many programmes emphasizing residential shell retrofits but can also often cost less than zero for properly designed new projects and some retrofits in all sectors. This is because they typically downsize, eliminate or greatly simplify HVAC equipment or other parts of whole-system capital cost. The decentralized resources’ compelling cost advantage is robust against even quite implausible improvements in central stations’ technology or regulation. For example, even if some new sort of fission or fusion reactor could provide free steam to the turbine, the balance-of-plant would still cost too much. However, the comparisons presented so far have another major conservatism favouring central stations: they count as zero all but one (thermal integration) of the 207 ‘distributed benefits’ described in RMI’s 2002 Economist book of the year Small Is Profitable: The Hidden Economic Benefits of Making Electrical Resources the Right Size.10 Collectively, counting the documented distributed benefits that generally apply mutatis mutandis to decentralized resources – both supply- and demand-side – makes those resources typically an order of magnitude more valuable. That’s enough to flip almost any investment decision. As the marketplace increasingly recognizes and values distributed benefits, therefore, their already clear economic advantage should become overwhelming. This order-of-magnitude increase in value from distributed benefits has three sources, not including externalities such as environmental and social benefits, which aren’t counted at all. The most important economic benefits come from financial economics, for example: •
• •
Renewable resources avoid the financial risk of volatile fuel prices, adding about 1–2 cents/kWh to the value of a typical wind power project. Conversely, for fair comparison with wind power using the risk-adjusted discount rates that orthodox financial economics requires11, the present value of the gas cost stream for a combined cycle plant must be roughly doubled (assuming that gas prices are only three times as volatile as the stock market, as they were before 2002. Lately they’ve been even more volatile than that). Small, fast modules bear less financial risk than big, slow, lumpy projects. In a typical substation-support application, this factor alone can increase the breakeven capital cost of a distributed resource, like photovoltaics, by a factor of about 2.7. Portable resources can be physically relocated to fit unexpected patterns of utility-system
development, both increasing the present value of benefits and decreasing the risk of not obtaining them. Together, these and other financialeconomic benefits typically raise decentralized projects’ value by close to an order of magnitude if they’re renewable, or about three- to five-fold if they’re not. Then there are the better-known electrical-engineering benefits, such as avoided grid costs and losses, increased reliability and resilience, more graceful fault management, free reactive power control (from DC sources inverted to AC), and longer distribution equipment life (via reduced heating and tapchanging). Together, these benefits typically increase value by two- to threefold – more if the distribution system is congested, if other circumstances permit deferral or avoidance of new distribution capacity, or if especially reliable or high-quality power is required. Finally, scores of diverse ‘miscellaneous’ benefits typically about double economic value, not counting opportunities for capturing and reusing otherwise wasted heat. One more key advantage distinguishes distributed resources: rapid deployment. This is not just about mass production – building things that are more like cars than cathedrals – and short lead times for siting and installation. It is also about the inherent speed of options that are easily accessible to numerous and diverse market actors, and that can be deployed through routine market transactions rather than requiring highly specialized and ponderous institutions. California illustrated all these effects from 1982 to 1985, when state policymakers encouraged a diverse portfolio of fast technologies installed by varied actors in an open market. At that time, electrical resource acquisition methods and subsidies provided a playing field that was (by historical standards) relatively level between supply- and demand-side investments. In those few years, without today’s concerns about climate change or supply adequacy, the state’s three investorowned utilities’ solicitations elicited (compared with a 37 GW peak load in 1984): • • •
23 GW (62% of load) of contractedfor electricity end-use efficiency to be installed over the following decade 13 GW (35%) of contracted-for new generating capacity, mostly renewable 8 GW (22%) of additional new generating capacity on firm offer, plus a further 9 GW (25%) of new generating offers arriving each year.
EXCEEDING THE FORECAST These contracts and offers totalled 144% of the 1984 peak load, exceeding the forecast load growth. Had bidding not been suspended in April 1985 because of the resulting power glut, another year or so of acquisitions at that pace could have displaced every thermal station in California, which in hindsight could have been a good idea.12 Today’s technologies are much cheaper, more reliable and, in most cases, more readily available than those of 20-odd years ago. And today, we are starting to realize that if issues like climate change are important, we must buy those options that will deliver the most solution both per dollar (ie fewest cents/kWh) and per year (ie fewest years/MW). Centralized plants can seldom if ever satisfy these basic requirements.6 In short, despite the many obstacles still in its way, decentralized power generation is enjoying unheralded but impressive progress in the global market. Other than efficient end-use, which is fortunately its natural partner (together they typically offer better value than their mere sum or than either alone), it has no serious economic competitor. The great disparities in adoption between countries, or even parts of the same country, show how much remains to be done to purge the artificial barriers to true competition between all ways to save or produce energy, regardless of which kind they are, what technology or fuel they use, how big they are or who owns them. But the economic fundamentals of distributed resources promise
an ever faster shift from power plants many orders of magnitude bigger than towards those built at the right size and at the right place for their task. Amory Lovins is co-founder of Rocky Mountain Institute (www.rmi.org), an independent, entrepreneurial, nonprofit, applied research centre in Snowmass, Colorado, US. Fax: +1 970 927 4178 e-mail: [email protected] RMI colleagues Ken Davies and Nathan Glasgow hold masters degrees in engineering and economics respectively.
REFERENCES 1. See the detailed analysis in RMI publications CS85-7 and -22. 1984 federal energy subsidies exceeded $50 billion/year. Per unit of energy or savings delivered, they varied by nearly 200- fold between more and less favoured technologies. Electricity got 65% – 48 times as much per kWh as efficiency. Subsidies may be larger and more lopsided today. 2. Energy Subsidies in the European Union: A brief overview, European Environment Agency (Copenhagen), 2004, at http://reports.eea.eu.int 3. See Excel spreadsheet at www.rmi.org/images/other/Energy/NuclearAndComp.zip, RMI, 21 June 2005. 4. See Methodology memo at www.rmi.org/images/other/Energy/NukeCompMeth.pdf, RMI, 21 June 2005. 5. According to the incremental demand-side resources declared in the US Energy Information Administration’s Electricity Annual 2003, the most recent data available. 6. A.B. Lovins, Nuclear power: economics and climate-protection potential, 11 Sept 2005, www.rmi.org/sitepages/pid171.php#E05-08. 7. Physicist Professor Freeman Dyson in Imagined Worlds, states: ‘There was nothing wrong, and there is still nothing wrong, with using nuclear energy to make electricity. But the rules of the game must be fair, so that nuclear energy competes with other sources of energy and is allowed to fail if it does badly. 8. We don’t explicitly analyse here the costs of other renewables, both because they’re so sitedependent and because from their global market success – much of it outside countries with large subsidies, notably Germany – it’s clear that many renewables besides wind power have favourable value propositions. 9. O. Bailey and E. Worrell, Clean Energy Technologies: A Preliminary Inventory of the Potential for Electricity Generation, http://repositories.cdlib.org/lbnl/LBNL-57451. 10. A. B. Lovins et al, RMI. See website www.smallisprofitable.org. Also available at www.earthscan.co.uk. 11. Versus the single uniform discount rate – normally taken as the weighted-average cost of capital – conventionally used in engineering economics to discount every cost stream of every project. This is as methodologically wrong as, say, comparing junk bonds with Treasury debt by considering only their yield, not also their risk. 12. Similarly, during 1979–1985, the US ordered more new capacity from small hydro and wind power than from coal and nuclear plants, excluding their cancellations, which totalled more than 100 GW – despite nuclear’s approximately 24-fold greater 1984 subsidy per kWh and smaller interconnection obstacles.
Decentralized electricity generation is the new power technology of choice Decentralized power generation is enjoying impressive progress in the global market despite the many obstacles still in its way. But that progress is often underestimated. Amory Lovins and his Rocky Mountain Institute colleagues Kenneth Davies and Nathan Glasgow present an analysis of the growth of decentralized energy and suggest why its development is outperforming that of centralized plant, including nuclear technology.
Decentralized generation – in The Economist’s apt term, micropower – enjoys an important market share in some countries: in 2004, 52% of the electricity generated in Denmark, 39% in the Netherlands, 37% in Finland, 31% in Russia, 18% in Germany, 16% in Japan, 16 % in Poland, 15% in China, 14% in Portugal, 11% in Canada. Yet it is omitted from many official statistics and projections, underreported in the media, and often dismissed by policymakers as small and slow – a fringe market too trivial to bother with. Surprise! A recent compilation of industry and official data published in June by Rocky Mountain Institute (RMI) found that micropower worldwide has already surpassed nuclear power in both annual output (in 2005) and installed capacity (in 2002), and is growing far faster in absolute terms. In 2004, DG added 2.9 times as much output and 5.9 times as much generating capacity as nuclear power added worldwide. Since nuclear power is often represented as an important technology, providing 20% of US and 16% of world electricity, surpassing this benchmark should at last qualify micropower as a serious competitor. Roughly 65% of micropower’s 2004 capacity and 77% of its output was fossilfuelled CHP; the rest, diverse renewable sources. RMI’s assessment probably understates the actual total for both categories. A separate and similarly detailed compilation in press at Worldwatch Institute by Dr Eric Martinot (formerly at Lawrence Berkeley National Laboratory, now at Tsinghua University) draws similar conclusions but finds more small hydro because it uses China’s definition (up to 30 MWe, not 10 MWe). Neither RMI’s nor Martinot’s assessment includes big hydroelectric projects. In 2003, the US Department of Environment (DOE) estimated that hydroelectricity, very largely in big units and probably omitting many smaller ones, was 20% of world capacity and 17% of world net generation.
Figure 1 shows RMI’s findings about the installed capacity of worldwide micropower in 2004, Figure 2 the annual output. Also shown are the respective industries’ approximate projections to 2010. These are imprecise but qualitatively clear. All data shown are net of increases or decreases in rated capacity and of re- and decommissioned units.
Output lags capacity by three years as CHP has a higher average capacity factor than wind and solar, although some small hydro projects and most geothermal and biomass generation projects also have high capacity factors. Precise data on the fuel mix of CHP are not available, but WADE estimates it at 60%–70% gas-fired. This implies a CHP carbon intensity probably no more than half the global average for centralized fossil-fuelled generation, which is coal-dominated, with a normal range of around 30%–80% depending on relative efficiencies and fuels. 65% of micropower’s 2004 capacity and 77% of its output was fossilfuelled CHP; the rest, diverse renewable sources
LOW OR NO-CARBON EMISSIONS Since the renewable sources shown have zero direct carbon emissions, as does nuclear power if its enrichment energy and any decommissioning or wastedisposal energy are neglected, all the sources of electricity shown in Figures 1 and 2 are low- or no-carbon emitters. Nuclear power’s global capacity and output are both graphed for comparison. Fossil-fuelled, nonCHP generation is several times larger. Micropower’s greater market success compared with nuclear power is all the more impressive because nuclear has been heavily promoted and subsidized by nearly all host governments. In the US, for example, it received 24 times more federal subsidies per kWh in 1985 than did non-hydroelectric renewables.1 In nuclear’s first 15 years of industrial development it was about 30 times more heavily subsidized per kWh than all renewables.2 Moreover, micropower in many countries, including most of the US, is often barred from grid interconnection, on fair terms or completely, while centralized plants are not. Cost comparisons typically burden micropower with both grid expansion and firming or backup costs
(which are often exaggerated) while the corresponding costs of centralized plants, such as reserve margin, are typically socialized as normal overheads of the electricity enterprise. It seems inescapable that fair competition at honest prices would favour micropower over centralized plants even more than the market has been doing lately.
The data and sources behind these graphs are all publicly available3 and use a simple, transparent methodology4 whose main points include: •
•
Fossil-fuelled CHP data are from WADE’s World Survey of Decentralized Energy of March 2005 (www.localpower.org), indicating 282.3 GWe at the end of 2004. This landmark work is based largely on the annual Diesel and Gas Turbine Worldwide sales survey, direct salesdata collection from WADE members that make and sell distributed generating equipment, national market surveys, and standard wind and PV industry sources. WADE has carefully cross-checked its estimates of how much of the equipment sold was used in CHP applications and has conservatively omitted all units less than 1 MWe and all steam turbines outside China. The maximum unit sizes included are 30 MWe for engines and 120 MWe for gas turbines. To prevent double-counting, we subtract photovoltaics, WADE’s estimate of decentralized wind power, and all decentralized biomass CHP (which we count under biomass generation, not fossil-fuelled CHP, using central values from an unpublished WADE estimate that biomass fuels 3%– 5% of CHP in 2004 and could fuel 6%–8% in 2012). We use Michael Brown’s estimate of about 12% of global capacity as realistic in 2012 (below WADE’s 14% target, compared with 2004’s 7.2%) and interpolated intermediate values by curve-fitting. To calculate output, we assume that CHP plants average 7250 h/year of operation (an 82.8% capacity factor) based on his estimate of ‘7000–7500 h/year, possibly more’. Wind data are from the European Wind Energy Association for 1991–2004 and Worldwatch Institute for 1990. The capacity projections shown for 2005–10 range from EWEA’s 2005 to its 2002 Wind Force 12 estimates – the latter adopted by the IEA using a 25% capacity factor. Future values will almost certainly be higher.
•
•
• • •
Photovoltaic data are from Worldwatch Institute (which uses the industrystandard PV News database) and Solarbuzz’s 2005 Marketbuzz, solving for the lag between shipments and installations in 2004. We assume a 25-year average life, so the first commercial PV panels, made in 1971, started retiring in 1996. That will probably prove conservative since some modern PVs come with a 25-year warranty. We adopt the International Energy Agency’s (IEA’s) 2010 capacity factor. Small hydro’s 2004 capacity, 47 GW (of units up to 10 MW), is from the International Association for Small Hydro (IASH). The 2010 capacity is interpolated from simple exponential growth to the World Energy Council’s favourable scenario for 2020 (75 GW), then pre-2004 capacity is extrapolated backwards. We adopt for all years IASH’s implicit 2010 projection of 45.8% capacity factor. Biomass-fuelled generation: we adopt Navigant Consulting’s incremental capacity additions for 2000–10, renormalize the implied 2010 total installed capacity (55.7 GW) to match the IEA’s projected 55 GW and assume the IEA’s 70% capacity factor. Here, we adopt the IEA’s 2002 capacity, Navigant Consulting’s historical and projected capacity additions (renormalized from 11.8 to the IEA’s 13 GW total for 2010), and the IEA’s capacity factors (72% in 2002, 78% in 2010, scaled linearly). Nuclear capacity data are from the International Atomic Energy Agency (IAEA) through 2004, cross-checked for 2005 against the World Nuclear Association’s database, which we also use for expected 2005–2010 shutdowns. We adjust for upratings (which the IAEA ‘slipstreams’ rather than reporting explicitly) on the assumption that the 3.6 GW of 1991–2004 upratings occurred at a linear rate and can be extrapolated backwards one year to 1990 and forward to 2010 (agreeing within 0.12% with the USDOE’s projection). Nuclear generation data are from the IEA, the most complete historical source. To estimate 2005–2010 nuclear output, we assume a linear increase from 84.1% capacity factor in 2002 to the IEA’s 88.51% in 2010.
DIVERGENCE The divergence between officially favoured and actually achieved market outcomes in recent years became even more striking when RMI examined the first derivative of Figure 1 – the additions of worldwide capacity by technology and year, again including nuclear power as a bench-mark. This comparison, Figure 3, also includes a leading indicator: nuclear construction starts. Those data stopped in 2004 because 2005–2010 construction starts aren’t yet known. But because of this technology’s long lead times, they can’t affect its 2010 installed capacity. Again, an ineluctable conclusion from these data is that even the most favoured and subsidized centralized electric generating technology has been eclipsed by its decentralized competitors. This conclusion would become even stronger if we also considered decentralized electricity savings, i.e. demandside resources. But very few countries and no international organizations properly track demand-side resource additions. Even in the US, where detailed year-byyear evaluation data existed a decade ago, only a few states still measure more than the most aggregated econometric estimates of incremental ‘negawatts’. However, as a very rough indicator of demand-side progress, the 1.98% drop in US electric intensity in 2003 (whatever its causes) would correspond, at constant load factor, to saving 13.8 GWp – 6.3 times US utilities’ declared 2.2 GWp from demandside management5– and the 2004 intensity drop of 2.3% would have saved more than 16 GWp (plus 1 GWp/year from utility load management actually exercised). The US uses only one fourth of the world’s electricity, so it’s hard to imagine that global savings don’t rival or exceed global additions of micropower (24 GW in 2003, 28 GW in 2004). Thus the grand total for global additions of decentralized electrical resources, both demand- and supply-side, must exceed annual nuclear capacity growth by an order of magnitude. That this fact is unknown to most policymakers, commentators and investors is not only remarkable, it is likely to lead to poor decisions and heavy financial losses.
US merchant generating firms, for example, recently lost more than $100 billion of market capitalization, wiping out their equity investors, when they built about 200 GW of combined cycle plants for which there was no demand. The US Energy Policy Act of 8 August 2005 greatly increases subsidies and regulatory aid for supply-side resources, chiefly centralized ones, in apparent disregard of the market’s recent reductions in primary-energy and electric intensity of GDP by roughly 2.5%/year and 2.0%/year respectively. Since about 78% of the increase in US energy services provided in the last decade has been fuelled and powered by unnoticed intensity reductions, and only about 22% by the increases of physical energy supply described in voluminous official statistics, four fifths of what energy markets are doing is invisible. This risks repeating the market crash of the mid-1980s, when the officially promoted and subsidized supply expansions collided with rapid reductions in energy intensity (driven by late-1970s national policies and the 1979 oil price shock), causing glutted energy markets, collapsing prices and bankrupt suppliers. Obviously something other than national policy (in most countries) is causing this virtually unreported but dramatic marginal shift, especially in the electricity sectors, from the most favoured form of centralized supply to decentralized resources.
PLAUSIBLE CANDIDATE A plausible candidate would be relative prices. Consistent with the hypothesis that the world is buying more micropower (and electric end-use efficiency) because it costs less than centralized supply, a separate RMI analysis6 offers a simple, transparent, and documented comparison of the delivered cost of one marginal kWh from US centralized and decentralized resources, presented in 2004 dollars on a consistent accounting basis (other than project life, which is appropriate to each technology). Centralized or remote resources incur a delivery cost to the retail meter, conservatively taken as the 1996 embedded historic average (including average grid losses) for US investor-owned utilities – 2.75 cents/kWh. On-site resources avoid this cost because they’re already delivered. That new study’s findings are summarized in Figure 4. The approximately $13 billion of justapproved federal nuclear subsidies – roughly the entire capital cost for the next 6 GW – might, if highly successful, cut the cost of a new delivered nuclear kWh from 9.8 cents/kWh (using a 40 year life and 85% capacity factor) to 7.0 cents/kWh.
Meanwhile, a stiff tax or trading price of $100 per tonne of carbon could raise the nominal cost of a new delivered coal kWh from 7.2 to 9.7 cents/kWh (at $1.33/GJ coal), and that of a new delivered combined cycle, gas-fired kWh from 6.7–8.6 cents/kWh to 7.8–9.8 cents/kWh, assuming a levelized gas price of $3.6–7.6/GJ. That assumption is equivalent to assuming these gas prices initially, escalating at the real discount rate of 5%/year assumed for all the centralized plants. All centralized-plant findings are adopted from the canonical 2003 MIT study The Future of Nuclear Power, which uses a merchant cashflow model.
NUCLEAR’S MARKET FAILURE Thus the National Energy Policy Act of 2005 attempts to reverse the market failure of nuclear power by making it potentially competitive – under very optimistic assumptions – with its coal- and gas-fired central-station competitors. It will become obvious that those are the wrong competitors, because all three types of central plants are uncompetitive with three other options: wind power and some other renewables, CHP and end-use efficiency. Observed market behaviour, which strongly favours these options, implies that even if the policy intent – that yet again, nuclear power cannot be allowed to fail7– were realized, the desired outcome would not be, because the decentralized competitors, being markedly less costly, would continue to win. The documented assumptions behind the decentralized resources’ graphed costs are deliberately chosen to favour centralized plants by: • • •
understating marginal delivery costs counting a generous 0.9 cents/kWh firming and integration cost for wind power but no reserve margin for centralized plants (all types of generators are intermittent, differing only in the size, frequency, duration, cause and predictability of outage) offering the option to back out wind power’s Production Tax Credit, but not the centralized plants’ pre-2005 subsidies (those for nuclear power far exceed the PTC)
• •
using relatively high costs for windpower8 (its assumed basic cost, for 30 year projects including the levelized Production Tax Credit of 0.86 cents/kWh, is 3.0–3.5 cents/kWh – twice that of today’s cheapest projects) using static ‘snapshot’ costs rather than cost trends, which strongly favour micropower and efficiency (only the expected 1 cent/kWh drop in wind power costs to 2012 is shown).
The CHP costs are based on natural-gas prices of $5.4–8.7/GJ, $1/GJ higher than for central stations, and a 10%/year return over 25 years. Industrial CHP costs are based on five actual projects considered representative by Primary Energy LLC, a leading US CHP developer with about 0.9 GW of operating projects. The conventional industrial CHP projects shown are in the 28–64 MWe range. The 60–160 MWe projects using recovered waste heat, which has about 94 GW of untapped US potential9, deliver electricity at negative cost because the heat’s value more than repays capital plus non-fuel operating costs. Gas-fired CHP can yield even lower costs in a suitable building than in an industrial facility if it is highly optimized (typically by trigeneration or better, with system efficiencies equal to or above 0.9) and well integrated with prior demand-side improvements.
WIDE VARIATION Finally, the cost of saving electricity varies widely (as is also true of the supply-side resources), but it’s empirically often below 1 cent/kWh for well designed and well implemented commercialand industrialsector retrofits. It can be higher for suboptimal efforts or for many programmes emphasizing residential shell retrofits but can also often cost less than zero for properly designed new projects and some retrofits in all sectors. This is because they typically downsize, eliminate or greatly simplify HVAC equipment or other parts of whole-system capital cost. The decentralized resources’ compelling cost advantage is robust against even quite implausible improvements in central stations’ technology or regulation. For example, even if some new sort of fission or fusion reactor could provide free steam to the turbine, the balance-of-plant would still cost too much. However, the comparisons presented so far have another major conservatism favouring central stations: they count as zero all but one (thermal integration) of the 207 ‘distributed benefits’ described in RMI’s 2002 Economist book of the year Small Is Profitable: The Hidden Economic Benefits of Making Electrical Resources the Right Size.10 Collectively, counting the documented distributed benefits that generally apply mutatis mutandis to decentralized resources – both supply- and demand-side – makes those resources typically an order of magnitude more valuable. That’s enough to flip almost any investment decision. As the marketplace increasingly recognizes and values distributed benefits, therefore, their already clear economic advantage should become overwhelming. This order-of-magnitude increase in value from distributed benefits has three sources, not including externalities such as environmental and social benefits, which aren’t counted at all. The most important economic benefits come from financial economics, for example: •
• •
Renewable resources avoid the financial risk of volatile fuel prices, adding about 1–2 cents/kWh to the value of a typical wind power project. Conversely, for fair comparison with wind power using the risk-adjusted discount rates that orthodox financial economics requires11, the present value of the gas cost stream for a combined cycle plant must be roughly doubled (assuming that gas prices are only three times as volatile as the stock market, as they were before 2002. Lately they’ve been even more volatile than that). Small, fast modules bear less financial risk than big, slow, lumpy projects. In a typical substation-support application, this factor alone can increase the breakeven capital cost of a distributed resource, like photovoltaics, by a factor of about 2.7. Portable resources can be physically relocated to fit unexpected patterns of utility-system
development, both increasing the present value of benefits and decreasing the risk of not obtaining them. Together, these and other financialeconomic benefits typically raise decentralized projects’ value by close to an order of magnitude if they’re renewable, or about three- to five-fold if they’re not. Then there are the better-known electrical-engineering benefits, such as avoided grid costs and losses, increased reliability and resilience, more graceful fault management, free reactive power control (from DC sources inverted to AC), and longer distribution equipment life (via reduced heating and tapchanging). Together, these benefits typically increase value by two- to threefold – more if the distribution system is congested, if other circumstances permit deferral or avoidance of new distribution capacity, or if especially reliable or high-quality power is required. Finally, scores of diverse ‘miscellaneous’ benefits typically about double economic value, not counting opportunities for capturing and reusing otherwise wasted heat. One more key advantage distinguishes distributed resources: rapid deployment. This is not just about mass production – building things that are more like cars than cathedrals – and short lead times for siting and installation. It is also about the inherent speed of options that are easily accessible to numerous and diverse market actors, and that can be deployed through routine market transactions rather than requiring highly specialized and ponderous institutions. California illustrated all these effects from 1982 to 1985, when state policymakers encouraged a diverse portfolio of fast technologies installed by varied actors in an open market. At that time, electrical resource acquisition methods and subsidies provided a playing field that was (by historical standards) relatively level between supply- and demand-side investments. In those few years, without today’s concerns about climate change or supply adequacy, the state’s three investorowned utilities’ solicitations elicited (compared with a 37 GW peak load in 1984): • • •
23 GW (62% of load) of contractedfor electricity end-use efficiency to be installed over the following decade 13 GW (35%) of contracted-for new generating capacity, mostly renewable 8 GW (22%) of additional new generating capacity on firm offer, plus a further 9 GW (25%) of new generating offers arriving each year.
EXCEEDING THE FORECAST These contracts and offers totalled 144% of the 1984 peak load, exceeding the forecast load growth. Had bidding not been suspended in April 1985 because of the resulting power glut, another year or so of acquisitions at that pace could have displaced every thermal station in California, which in hindsight could have been a good idea.12 Today’s technologies are much cheaper, more reliable and, in most cases, more readily available than those of 20-odd years ago. And today, we are starting to realize that if issues like climate change are important, we must buy those options that will deliver the most solution both per dollar (ie fewest cents/kWh) and per year (ie fewest years/MW). Centralized plants can seldom if ever satisfy these basic requirements.6 In short, despite the many obstacles still in its way, decentralized power generation is enjoying unheralded but impressive progress in the global market. Other than efficient end-use, which is fortunately its natural partner (together they typically offer better value than their mere sum or than either alone), it has no serious economic competitor. The great disparities in adoption between countries, or even parts of the same country, show how much remains to be done to purge the artificial barriers to true competition between all ways to save or produce energy, regardless of which kind they are, what technology or fuel they use, how big they are or who owns them. But the economic fundamentals of distributed resources promise
an ever faster shift from power plants many orders of magnitude bigger than towards those built at the right size and at the right place for their task. Amory Lovins is co-founder of Rocky Mountain Institute (www.rmi.org), an independent, entrepreneurial, nonprofit, applied research centre in Snowmass, Colorado, US. Fax: +1 970 927 4178 e-mail: [email protected] RMI colleagues Ken Davies and Nathan Glasgow hold masters degrees in engineering and economics respectively.
REFERENCES 1. See the detailed analysis in RMI publications CS85-7 and -22. 1984 federal energy subsidies exceeded $50 billion/year. Per unit of energy or savings delivered, they varied by nearly 200- fold between more and less favoured technologies. Electricity got 65% – 48 times as much per kWh as efficiency. Subsidies may be larger and more lopsided today. 2. Energy Subsidies in the European Union: A brief overview, European Environment Agency (Copenhagen), 2004, at http://reports.eea.eu.int 3. See Excel spreadsheet at www.rmi.org/images/other/Energy/NuclearAndComp.zip, RMI, 21 June 2005. 4. See Methodology memo at www.rmi.org/images/other/Energy/NukeCompMeth.pdf, RMI, 21 June 2005. 5. According to the incremental demand-side resources declared in the US Energy Information Administration’s Electricity Annual 2003, the most recent data available. 6. A.B. Lovins, Nuclear power: economics and climate-protection potential, 11 Sept 2005, www.rmi.org/sitepages/pid171.php#E05-08. 7. Physicist Professor Freeman Dyson in Imagined Worlds, states: ‘There was nothing wrong, and there is still nothing wrong, with using nuclear energy to make electricity. But the rules of the game must be fair, so that nuclear energy competes with other sources of energy and is allowed to fail if it does badly. 8. We don’t explicitly analyse here the costs of other renewables, both because they’re so sitedependent and because from their global market success – much of it outside countries with large subsidies, notably Germany – it’s clear that many renewables besides wind power have favourable value propositions. 9. O. Bailey and E. Worrell, Clean Energy Technologies: A Preliminary Inventory of the Potential for Electricity Generation, http://repositories.cdlib.org/lbnl/LBNL-57451. 10. A. B. Lovins et al, RMI. See website www.smallisprofitable.org. Also available at www.earthscan.co.uk. 11. Versus the single uniform discount rate – normally taken as the weighted-average cost of capital – conventionally used in engineering economics to discount every cost stream of every project. This is as methodologically wrong as, say, comparing junk bonds with Treasury debt by considering only their yield, not also their risk. 12. Similarly, during 1979–1985, the US ordered more new capacity from small hydro and wind power than from coal and nuclear plants, excluding their cancellations, which totalled more than 100 GW – despite nuclear’s approximately 24-fold greater 1984 subsidy per kWh and smaller interconnection obstacles.
