The economics of solar power in the UK

Lecture to the Institution of Mechanical Engineers Manchester, 12th November 2024

In this lecture I will cover three aspects of the economics of solar power. The first aspect is what I will call the microeconomic issues – what does it cost and how does it perform? This will draw on material that I published in a paper which was published by the Renewable Energy Foundation in 2023.[1] It focuses on utility-scale solar generation, which I define as solar plants with a minimum capacity of 1 MW but with the primary focus on plants of at least 5 MW. Readers should refer to that paper for full details of my analysis as there is only time to provide a summary here.

The second aspect concerns the macro impact on the power market in the UK of a large expansion in the amount of solar capacity by 2030 or 2035. Government plans envisage an increase in total solar solar generating capacity from about 16 GW at the end of 2023 to at least 50 GW by the early 2030s. Notionally this is the target for 2030, but that implies a rate of new construction at least twice the level achieved during the previous boom from 2013 to 2018 so it is very likely that the target will slip. My analysis will examine what difference should be expected between the performance of GB electricity systems with 30, 50 and 70 GW of solar capacity holding other things constant.

The third aspect brings the micro and macro perspectives together to examine the viability of investment in solar power under different scenarios. The most important question is what level of subsidies will be required to achieve the targets for a rapid increase in the solar generating capacity within the next decade?

At the outset I want to emphasize an obvious but often forgotten point about solar power. The UK is a long way (literally) from being the most favourable location for solar generation in Europe, let alone the world. It is at least 15° north of the states in the US where most solar development is concentrated. Add in a maritime climate with lots of cloud and stormy weather. The performance of solar plants in Britain will never match those in the south of Spain, let alone the deserts of Mexico and Chile where the lowest costs are claimed. Hence, the development of solar plants in Britain is entirely the product of subsidies for renewable and low carbon generation.

Costs and performance of solar plants

Capex costs. It is well-known that the cost of solar panels fell sharply during the 2010s. Many have assumed that the overall cost of building solar plants has fallen similarly and, even more important, will continue to fall in future. The data show that there was a 15% decline in the average capex cost per MW of capacity from 2011-13 to 2014-16 and a 10% decline from 2014-16 to 2017-20 – see Table 1. The average capex cost per MW was £0.95 million at 2018 prices, equivalent to £1.16 million at 2023 prices. The trend in capex costs is consistent with the fall in the costs of solar panels and inverters, but other costs have increased over the period and appear to be affected by a scarcity of equipment and skilled labour. Further falls in the cost of solar panels will only have a limited impact on total capex costs. 

Opex costs. The average level of opex costs for solar plants is 3 to 4 times the official assumptions. The best estimate is about £44,500 per MW per year at 2023 price for a plant in the size category of 10-20 MW. Opex costs are highly variable over time and across plants because of equipment failures and other factors, but the pooled data suggests that they tend to increase with the age of the plant. The estimated rate of increase over time was about 5% per year in real terms. That rate of increase may fall as the industry matures but it would be prudent to assume that opex costs will increase by 2.5% to 3% per year in real terms.

Output. There is extremely strong evidence from both the UK and the US that the output of solar plants falls at 1% to 2% per year after age 3 after controlling for the level of solar radiation. The rate of decline in output is higher in the US than in the UK which may reflect differences in maintenance practices or the greater length of experience in the US. If the US pattern prevails in the UK, solar plants reaching the end of their period of eligibility for ROCs will have an expected output for standard weather conditions which is 30% lower than in their early years of operation.

Economic life. The combination of rising opex costs and declining performance means that existing solar plants will not cover their operating costs once their period of eligibility for ROCs comes to an end after 20 years and they move to operating as merchant generators. Recently, many of the SPVs which own and operate solar plants have changed their accounting assumptions to increase the economic life of their assets from 25 to 35 years. This modification is ill-judged and potentially damaging to investors as the evidence suggests that the economic life of solar assets is unlikely to be significantly greater than 20 years.

Financial prospects. The breakeven price of electricity for new investment in solar plants is £132 per MWh at 2023 prices over a 25-year life under the most optimistic assumptions about opex costs and performance and it is £150 per MWh under more realistic assumptions. These breakeven prices are significantly higher than for onshore wind but comparable with breakeven prices for offshore wind.   

These results reflect averages for all GB utility-scale solar plants. There are large differences between plants. Some enjoy favourable locations with good solar resources. Others have low capital costs or relatively high commercial power purchase prices, but the general pattern is that GB solar plants require high levels of subsidies to be viable. ROC supported plants receive an average of 1.3 ROCs per MWh worth £76.70 per MWh at 2023 prices on top of the market price. Current CfD strike prices for solar plants are much lower than breakeven price - £50-£66 per MWh. These seem to follow the usual pattern for CfD contracts – a combination of extreme optimism about future costs, hopes for high market prices in future, and treating CfD contracts as options.

Impact on the electricity system and markets

Projecting the impact of a much higher level of solar capacity on the electricity market is complicated by one critical feature. Almost all solar plants are embedded, that is they are connected to distribution networks rather than the grid. Most embedded plants are neither monitored nor controlled by the grid. Their output is delivered directly to consumers and, thus, constitute negative demand from the point of view of the system operator. While such output may be sold through the wholesale market, most of it falls outside the balancing market and associated arrangements such as constraint payments. This is important when considering how the system might respond to excess generation from renewable sources.

                                                                       Figure 1

                                                        Source: Author’s estimates

Backup generation. Figure 1 shows the distribution of backup generation requirements by expected hours of operation in 2035 for three scenarios of total solar capacity – 30 GW (blue), 50 GW (purple) and 70 GW (green). The projections are derived a detailed dispatch model of the GB electricity system which is based on demand, weather conditions, and market behaviour over the period from 2015 to 2024. The model takes account of planned changes in onshore and offshore wind, biomass and other bioenergy, hydro, and nuclear power. It treats embedded and grid-connection generation separately. Further, it takes account of the expiry of ROC-eligibility and Cfd contracts for renewable generators. All forms of subsidised generation are treated as non-market generators with priority access to dispatch, while backup demand refers to the gap between total demand and non-market demand while will be supplied primarily from gas plants.

There are two critical conclusions from Figure 1. First, demand for backup generation will exceed 35 GW for an average of at least 50 hour per year. If we were to allow a conventional reserve margin of 15% of expected peak demand (58 GW), the requirement for flexible generation would be 44 GW. Of that, between 8.5 GW and 12.5 GW would be expected to operate for sufficient hours per year to warrant use of CCGTs, while the remainder would be provided by gas turbines, reciprocating engines and similar generators. The peak requirement is the same in all scenarios. Building additional solar capacity makes no difference, because solar plants don’t generate in the periods when backup demand is at its peak.

                                                                      Figure 2

                                                            Source: Author’s estimates

Second, adding 20GW of solar has a relatively small impact on the number of hours per year that backup generators would expect to run. For example, the increase of 40GW of solar from the low to the high scenario only reduces the backup capacity that will operate for at least 3,000 hour per year – the threshold between CCGTs and GTs – by 4GW or 10% of the increase in solar capacity.

Figure 2 examines what would happen if 20 GW of 4-hour battery storage were added to the system. It turns out that this makes no difference to the amount of backup capacity required and minimal difference to the amount of CCGT backup. In this scenario, there would be small reductions in the expected numbers of hours that CCGT plants would operate but the effect on the expected numbers of operating hours for GTs would be insignificant.

Surplus generation. There is another issue which will be very important for investment decisions. How will surplus generation be resolved in 2035? Figure 3 shows the distribution of surplus generation from non-market generators by capacity and expected hours. The calculation assumes that (a) all market generation and imports are zero, and (b) exports are equal to the willingness of inter-connected markets to buy power, bearing in mind that those markets may have surplus generation during periods of surplus in the GB market.

                                                                      Figure 3

                                                            Source: Author’s estimates

Because of the subsidies received by non-market generators the probable outcome is that market prices will be zero or negative during periods of surplus generation. The most recent CfD contracts contain provisions that are supposed to ensure that new plants do not operate during periods of negative market prices. It is far certain that they will respond to negative prices in this way. In any case, the many generators covered by ROCs or early-CfD contracts have no incentive to cut output.

Consider what Figure 3 implies. Because there is surplus generation for about 2,500 hours in the year in the 30GW scenario, market prices can be expected to be zero or negative for more than 28% of hour in the year. That proportion will increase to more than 37% of hours in the year in the 70GW scenario. Adding 20 GW, as show in Figure 4, would reduce the expected number of hours in which market prices would be zero or negative, but it would still be about 24% of hours in a year.

                                                                    Figure 4

                                                              Source: Author’s estimates

Such extended periods of zero or negative prices would be deeply disruptive to the functioning of the power market. No one has produced a convincing story about what happens if the market price is negative or zero for more than a quarter of hours in the year. Economists may blithely assume there is no conceptual difference between negative and positive prices. Still, I would place a large bet that the public doesn’t see things in that way, especially for generators who are receiving large subsidies financed by customers.

Since almost all solar plants are embedded there are additional complications about what prices they receive if/when market prices are negative. Most have power purchase agreements (PPAs) with intermediary traders or corporate buyers. However, as the plants are not registered as Balancing Market Units (BMUs) there is almost no transparency about what they are paid and the SPV company accounts provide no details. In the medium term, it is unlikely that traders and corporations will protect solar plants from negative market prices, so the average offtake price for their output is likely be substantially below historic average market prices. That will affect the guaranteed payments or subsidies required to achieve the levels of investment in new plants required in each scenario.

Investment decisions. Table 2 summarises the core investment metrics for new solar plants at 2023 prices. The first two rows are based on actual prices from 2017 to 2023 with the second row excluding the very high prices during the 2022 price spike. The operating margins allow for average opex costs discussed above. The payback period – the overnight capital cost divided by the annual operating margin – is 35 years based on the average prices from 2017 to 2023 and double that if 2022 is excluded. A reasonable maximum value for the payback period is 10 years if a project is to be commercially viable once allowance is made for price uncertainty and a base cost of capital of 6%.

A related metric is the guaranteed annual payment per MW that is required to bring the payback period down to 10 years, thus making the project investible in commercial terms. This is over £82,000 if we use the average price for 2017-23 or nearly £100,000 if 2022 is excluded.

The second part of the table adjusts the expected values of revenues and investment metrics for the effect of periods of surplus generation on market prices. This treats the market price as zero during such periods, as generators have the option of ceasing to generate. The base prices, before allowing for surplus generation, are those for the full period 2017-23. Under the different scenarios of solar capacity by 2035 the payback period increases from 35 to between 40 and 47 years. The guaranteed annual payments required for a payback period of 10 years increase to between £86,000 and £91,000. These increases are less than might be expected because market prices were already relatively low under the conditions that are likely to lead to surplus generation in future.

Conclusion

In the introduction I emphasized that the UK is a notably poor location for the large-scale development of solar power. Not only are solar resources poor, but the value of suitable land is high as are grid and other operating costs. Hence, solar power in Britain is entirely dependent on subsidies, not only the visible payments via feed-in tariffs, ROCs, and CfDs but also less visible benefits such as the indulgent operating regime that applies to embedded generators.

Both policymakers and investors have convinced themselves that the capital costs of building new solar plants have fallen and will continue to fall much faster than the evidence from company accounts demonstrates. There is no doubt that the cost of solar PV modules has fallen, and this trend may continue in future. However, solar PV modules account for a diminishing share of total capital costs and experience suggests that the costs of other equipment, civil works and other items are very sensitive to the rate of new construction. The current government’s plans will almost certainly push up such costs by a significant margin, which means that overall capital costs may fall by much less than expected and may even rise.

Similarly, operating costs in the UK are far higher than in more favourable locations for solar power. The combination of limited solar resources plus high capital and operating costs imply that solar plants are and will continue to be a poor investment in purely commercial terms. What is equally important is that very few plants will be able to earn an adequate margin over operating costs once their eligibility for subsidies expire. This means that the economic life of British solar assets is no more than 20 years. Investment funds which claim that solar plants will have an economic life of 35 years are misleading themselves and their investors.

The UK has a solar industry based on subsidies. Still, we should remember the costs. Based on the figures discussed above, a commitment to support total solar capacity of 50 GW within a decade implies a minimum level of subsidies of £4.5 billion per year at 2023 prices. Directly or indirectly that cost must fall on UK households, so the eventual cost will be about £160 per household.  Bear in mind that the average electricity bill for a 3-bedroom house is £868 per year. Subsidising solar generation will add between 15% and 20% to that bill.

The goal in subsidising the development of solar plants is to reduce emissions of CO2 by substituting solar generation for gas-fired power generation. The results from the model discussed above imply that adding 40 GW of solar capacity will reduce expected CO2 emissions by 5.79 million tonnes per year at a cost of £3.6 billion per year. That is an average cost of £622 per tonne of CO2 saved. By any conventional standards that is an extraordinarily expensive strategy.

[1] https://www.ref.org.uk/attachments/article/374/Economic-Solar-Generation.pdf