Cumulative CSP installed capacity grew just over five-fold, globally, between 2010 and 2020, reaching around 6.5 gigawatts (GW) by the end of that period. Breaking the last five years of this down, after modest activity in 2016 and 2017 – with annual additions hovering around 100 megawatts (MW) per year – the global market for Concentrated Solar Power grew in 2018 and 2019. In those years, an increasing number of projects came online in China, Morocco and South Africa. Yet, compared to other renewable power generation technologies, new capacity additions overall remained relatively low, at 860 MW per year in 2018 and 550 MW in 2019.
In 2020, only 150 MW was commissioned globally, with all of this this coming online in China. Hopes for growth in 2021 did not materialise, though 110 MW (all from the Cerro Dominador project) was commissioned during that year in Chile. At the same time, about 265 MW from the Solar Energy Generating Systems (SEGS) plant in the USA – in operation since the late 1980s – was retired. This puts the cumulative global installed capacity of CSP at the end of 2021 at around 6.4 GW.
The sector was optimistic that China’s plans to scale up the technology domestically would provide a boost to the industry and take deployment to new levels. Yet, progress on China’s policy to build-out 20 commercial-scale plants to scale up a variety of technological solutions, develop supply chains and gain operating experience has proved more challenging than anticipated. Developers have struggled and some projects have been lagging, while others have found new developers and some projects appear unlikely to be completed.
The outlook for 2022/2023 is somewhat brighter, however, with the possibility that close to 1.4 GW of new capacity could be commissioned in China and the United Arab Emirates.
Spain has launched an auction that includes 200 MW of CSP capacity, but the results are yet to be announced. The CSP project pipeline includes a 100 MW solar tower project with 12 hours of storage expected to come online by 2024 in South Africa. Botswana’s Ministry of Mineral Resources, Green Technology and Energy Security has initiated a pre-qualification process for participation in a 200 MW CSP tender, while Namibia has announced plans to launch a CSP tender in 2022 for between 50 MW and 130 MW of CSP capacity. In addition to this, a 300 MW project is planned to come online in 2025 in Qinghai, China.
National Energy and Climate Plans (NECPs) of some EU Member States show and indication of the potential development of the CSP project pipeline in the future. For example, Spain plans to add 5 GW and Italy 880 MW of new CSP capacity by 2030.
In the early years of CSP plant development, adding thermal energy storage was often uneconomic and generally unwarranted, so its use was limited.
Since 2015, however, hardly any projects have been built or planned without thermal energy storage. Adding this is now a cost-effective way to raise capacity factors, while it also contributes to a lower LCOE and greater flexibility in dispatch, over the day.
The average thermal storage capacity for PTC plants in the IRENA Renewable Cost Database increased from 3.3 hours between 2010 and 2014 to 6.1 hours between 2015 and 2019 (an 84% increase). For STs, that value increased from 5 hours in the 2010 to 2014 period to 7.7 hours in the 2015 to 2019 period (a 53% increase). In 2020, the 150 MW of newly-commissioned capacity in China had a weighted average storage capacity of 11 hours. Commissioned in 2021, the Cerro Dominador 110 MW ST project, located in Chile’s Atacama Desert, features a storage capacity of 17.5 hours. It is likely that all new CSP projects developed worldwide will include thermal storage.
Total installed costs for both PTC and ST plants are dominated by the cost of the components that make up the solar field. Although data on the total installed cost breakdown for 2010 relies on bottom-up, techno-economic analyses, the data can be paired with IRENA’s project level installed cost to get an understanding of the total installed cost breakdown in 2010/11 and 2019/2020.
In 2010, the solar field of a PTC plant cost an estimated USD 4 209/kW (44% of the total installed cost), but by 2020, this figure had fallen 68% to USD 1 345/kW (30% of the total). With such a dramatic reduction in costs for the solar field, other cost areas with smaller declines saw their share of total installed costs increase. The power block’s share, for example, increased from 15% in 2010 to 19% in 2020, despite its cost falling by 40% over the period, from USD 1 401/kW to USD 834/kW. This was also the case for the heat transfer fluid system, which increased its share from 9% to 11%, despite these costs per kW falling 47% over the period, from USD 886/kW to USD 470/kW.
This also occurred for thermal energy storage. That component’s share of total installed costs increased from 9% in 2010 to 15% in 2020, despite the cost itself falling from USD 815/kW to USD 660/kW. At the same time, during that period, the owner’s costs share rose from 5% to 9%, while it also rose in value, from USD 399/kW to USD 434/kW.
Over the 2010 to 2020 period, the costs of the balance of plant, engineering, and contingencies for PTC plants declined by 60%, 64% and 57% respectively. As a result, the share of balance of plant in total installed costs declined from USD 585/kW (6% of the total) to USD 236/kW (5%) over the same period, while engineering costs fell from USD 473/kW (5% of the total) to USD 168/kW (4%). A measure of how far the weighted average total installed costs for PTC plants have fallen is the fact that the costs of the solar field alone in 2010 were only 5% lower than the weighted average total installed cost in 2020.
For ST plants, this comparison is very similar, with 2010 costs being only 7% lower than the ST weighted average total installed cost value in 2020. Over that decade, the reduction in the cost of the heliostat field was significant, with costs falling 70% between 2011 and 2019, from USD 5 528/kW to USD 1 652/kW. This drove down the field’s share of total installed costs from 31% to 28%.
The cost of the receiver fell by 71% over the 2011 to 2019 period, from USD 2 868/kW to USD 819/kW, with the receiver’s share of total costs falling from 16% to 14%. Balance of plant and engineering saw the largest reduction, however, falling 93% over the same period, from USD 2 804/kW to USD 205/kW. This made this factor’s share of total costs fall from 16% to just 3%.
Contingencies remain an important overall cost component for STs, despite their share falling by 42% between 2011 and 2019, from USD 1 420/kW to USD 820/kW. At 14% of overall costs, in absolute terms, contingencies were still 1.8 times higher for STs than for PTC plants, per kilowatt. This likely reflects the fact that experience with STs remains relatively limited, with the replicability of their development and
construction processes still holding greater uncertainty than for PTC plants.
The latter have a longer commercial track record and a significantly larger number of installed projects. This may also be why owner’s costs have fallen by only 12% over the period, with their share of overall costs increasing to 14% in 2019.
Between 2010 and 2020, the weighted average total installed cost value for CSP plants in IRENA’s Renewable Cost Database fell by one half (50%) to reach USD 4 746/kW.
Total installed costs increased in 2021, to USD 9 090/kW. This trend should be interpreted with care, however, as the 2021 value corresponds to that of the first solar power plant developed in Latin America, which was inaugurated in June that year. Taking that value into account, the total installed cost decline between 2010 and 2021 was 4%. This was despite the fact that the LCOE decline for that period stayed at a similar level to that recorded between 2010 and 2021, given the high capacity factor of the Chilean Cerro Dominador project, which boasts 17.5 hours of storage.
Data from the IRENA Renewable Cost Database shows that total installed costs for CSP plants declined during the last decade, even as the size of these projects’ thermal energy storage systems increased.
During 2018 and 2019, the installed costs of CSP plants with storage were at par or lower than the capital costs of plants without storage commissioned in the 2010 to 2014 period – sometimes dramatically so.
The projects commissioned in 2018 and 2019 and listed in the IRENA Renewable Cost Database had an average of 7 hours of storage. This is 2.1 times more than the average storage value for projects commissioned between 2010 and 2014. Storage continued to grow after that, too. For instance, the weighted average storage level for projects commissioned in 2020 and 2021 was 13.8 hours, which was 70% higher than the level in 2018 and 2019.
The capital costs for CSP projects commissioned in 2020 for which cost data is available in the IRENA Renewable Cost Database ranged between USD 4 449/kW and USD 5 339/kW. With only two projects completed in China in 2020, totalling 150 MW, the data reflect national circumstances, much as the years 2010 to 2012 saw Spain dominate CSP deployment.
The two projects completed in China were part of a programme of 20 pilot projects that were designed to test a range of technology concepts and gain experience in integrating a wide range of technologies and plant configurations into the electricity system. The programme, launched in 2016 and aiming to develop 1.35 GW of capacity, initially targeted completion by 2018, but undoubtedly this timeline was too ambitious. With weighted average total installed costs of USD 4 746/kW in 2020, costs were 31% lower than the weighted average of USD 6 900/kW for projects commissioned in 2019.
During 2018 and 2019, IRENA’s Renewable Cost Database shows a capital cost range of between USD 3 337/kW and USD 9 064/kW for CSP projects with storage capacities of between four and eight hours.
In the same period, the cost range for projects with eight hours or more of thermal storage capacity was narrower – between USD 4 275/kW and USD 7 265/kW. This range also had a lower maximum value.
This was due to the fact that the majority of these projects were in China. Between 2018 and 2020, three projects in China were commissioned with greater than 10 hours of storage, with a total installed cost range from USD 4 275/kW to USD 5 339/kW.
For CSP, the quality of the solar resource, along with the technology configuration, are the determining factors in the achievable capacity factor at a given location and technology. CSP is distinctive in that the potential to incorporate low-cost thermal energy storage can increase the capacity factor55 and reduce the LCOE.
This is, however, a complex design optimisation that is driven by the desire to minimise the LCOE and/or meet the operational requirements of grid operators or shareholders in capturing the highest wholesale price.
This optimisation of a CSP plant’s design also requires detailed simulations, which are often aided by technoeconomic optimisation software tools that rely increasingly on advanced algorithms. These simulations must consider the site’s solar resource, the project’s storage capacity and the necessary solar field size to minimise LCOE and ensure optimal utilisation of the heat generated. This is a delicate balance, as smaller than optimal solar field sizes result in under-utilisation of the thermal energy storage system and the selected power block. A larger than optimal solar field size, however, would add additional capital costs, but increase the capacity factor – albeit at the potential risk of heat generation being curtailed at times, due to lack of storage and/or power generation capacity.
Over the last decade, falling costs for thermal energy storage and increased operating temperatures have been important developments in improving the economics of CSP. Increased operating temperatures also lower the cost of storage, as higher heat transfer fluid (HTF) temperatures lower storage costs. For a given DNI level and plant configuration conditions, higher HTF temperatures allow for a larger temperature differential between the ‘hot’ and ‘cold’ storage tanks. This means greater energy (and hence storage duration) can be extracted for a given physical storage size, or less storage medium volume is needed to achieve a given number of storage hours. Combined, since 2010, these factors have increased the optimal level of storage at a given location, helping minimise LCOE.
These drivers have contributed to the global weighted average capacity factor of newly-commissioned plants rising from 30% in 2010 to 42% in 2020 – an increase of 41% over the decade. The 5th and 95th percentiles of the capacity factor values for projects in IRENA’s Renewable Cost Database commissioned in 2019 were 22% and 54%, respectively. In 2020, the range for both projects was from 40% to 46%. The excellent solar resource in Chile’s Atacama Desert, the location of the Cerro Dominador CSP project, meant a very high capacity factor value for 2021, at 80%.
The increasing capacity factors for CSP plants driven by increased storage capacity can clearly be seen. Over time, CSP projects have been commissioned with longer storage durations.
For plants commissioned from 2016 to 2020, inclusive, around four-fifths had at least four hours of storage and 35% had eight hours or more. The impact of the economics of higher energy storage levels is evident in that in 2020, newly commissioned plants had a weighted average capacity factor of 42%, with an average DNI that was lower than for plants commissioned between 2010 and 2013, inclusive. Indeed, during the 2010 to 2013 period, the weighted average capacity factor for newly-commissioned plants was between 27% and 35%.
Both the early period of CSP development in Spain and the more recent one in China have been characterised by small, 50 MW projects. In China’s case, these have predominantly been technology demonstration projects among 20 initial pilot schemes. However, in order to unlock economies of scale – and as competitive procurement has encouraged greater developer choice in plant specifications – average project sizes have risen over time. It is likely that future commercial projects will gravitate towards the 100 MW to 150 MW range, which represents the economic optimum in most locations.
CSP plants are also now routinely being designed to meet evening peaks and overnight demand. CSP with low-cost thermal energy storage can integrate higher shares of variable solar and wind power, meaning that while often underrated, CSP could play an increasingly important role in the future.
The recent increase in storage capacity has also been driven by declining costs of thermal energy storage as the market has matured. This is the result of both declining capital costs and of higher operating temperatures, which allow larger temperature differentials in the molten salt storage systems, increasing the energy stored for the same volume. The result has been an increase in the weighted average number of storage hours through time. This rose more than three-fold between 2010 and 2020, from 3.5 hours to 11 hours. The Cerro Dominador project in Chile that came online in 2021 features the highest known storage capacity in the word, at 17.5 hours.
Although higher direct normal irradiation (DNI) leads to larger capacity factors, all else being equal, there is a much stronger correlation between capacity factors and storage hours. This is, however, only one part of the economics of plants at higher DNI locations. Higher DNIs also reduce the field size needed for a given project capacity – and hence the investment.
Yet, technology improvements and cost reductions for thermal energy storage also mean that higher capacity factors can be achieved even in areas without world class DNI. The 2020 data show the impact of higher storage levels, with newly commissioned plants recording a weighted average capacity factor of 42% that year, even though the average DNI in 2020 was lower than for plants commissioned between 2010 and 2013, inclusive. During that earlier period, the weighted average capacity factor was between 27% and 35% for newly commissioned plants.
OPERATION AND MAINTENANCE COSTS
For CSP plants, all-in O&M costs, which include insurance and other asset management costs, are substantial compared to solar PV and onshore wind. They also vary from location to location, depending on differences in irradiation, plant design, technology, labour costs and individual market component pricing, which is linked to local cost differences.
Historically, the largest individual O&M cost for CSP plants has been expenditure on receiver and mirror replacements. As the market has matured, experience, as well as new designs and improved technology, have helped reduce failure rates for receivers and mirrors, however, driving down these costs.
In addition, personnel costs represent a significant component of O&M, with the mechanical and electrical complexity of CSP plants relative to solar PV, in particular, driving this. Insurance charges continue to be an important further contributor to O&M costs, and typically range between 0.5% and 1% of the initial capital outlay (a figure that is lower than the total installed cost)
With some exceptions, typical O&M costs for early CSP plants still in operation today range from USD 0.02/kWh to USD 0.04/kWh. This is likely a good approximation for the current levels of O&M in relevant markets for projects built in and around 2010, globally, even if this is based on an analysis relying on a mix of bottom-up engineering estimates and best-available reported project data.
Analysis by IRENA undertaken in collaboration with the Institute of Solar Research (Das Institut für Solarforschung des Deutschen Zentrums für Luft- und Raumfahrt [DLR]) shows, however, that more competitive O&M costs are possible in a range of markets where projects achieved financial closure in 2019 and 2020.
The O&M costs per kWh in many of these markets are high in absolute terms, compared to solar PV and many onshore wind farms, but are about 18% to 20% of the LCOE for projects in G20 countries. Taking this into account, the LCOE calculations in the following section reflect O&M costs in the IRENA Renewable Cost Database that declined from a capacity weighted average of USD 0.037/kWh in 2010 to USD 0.015/kWh in 2020 (a 59% decline). The corresponding 2021 value is USD 0.022/kWh (40% lower than in 2010).
LEVELISED COST OF ELECTRICITY
With total installed costs, O&M costs and financing costs all falling, while capacity factors rose, the LCOE for CSP fell significantly between 2010 and 2020. Indeed, Over that period, the global weighted average LCOE of newly commissioned CSP plants fell by 70%, from USD 0.361/kWh to USD 0.107/kWh.
With deployment during the period 2010 to 2012 inclusive being dominated by Spain – and mostly comprised of PTC plant – the global weighted average LCOE by project declined only slightly, albeit within a widening range, as new projects came online. This changed in 2013, when a clear downward trend in the LCOE of projects emerged as the market broadened, experience was gained and more competitive procurement started to have an impact. Rather than technology-learning effects alone driving lower project LCOEs from 2013 onward, the shift in deployment to areas with higher DNIs during the period 2013 to 2015 also played a role.
In the period 2016 to 2019, costs continued to fall and the commissioning of projects in China became evident, with projects commissioned there in 2018 and beyond achieving estimated LCOEs of between USD 0.08/kWh and USD 0.14/kWh. At the same time, projects commissioned in 2018 and 2019 in Morocco and South Africa tended to have higher costs than this.
For projects commissioned between 2014 and 2017, their location in places with higher DNIs was a major contributor to increased capacity factors (and therefore lower LCOE values). The weighted average DNI of projects commissioned during that period, at around 2 600 kWh/square metre (m2)/year, was 28% higher than in the period 2010 to 2013. As already noted, however, this was not the only driver of LCOE trends, as technological improvements saw a move towards plant configurations with higher storage capacities.
CSP with low-cost thermal energy storage has shown it can play an important role in integrating higher shares of variable renewables in areas with good DNI.
In 2016 and 2017, only a handful of plants were completed, with around 100 MW added in each year. The results for these two years are therefore volatile and driven by specific plant costs. In 2016, the increase in LCOE was driven by the higher costs of the early projects in South Africa and Morocco commissioned that year. In 2017, the global weighted average LCOE fell back to the level set in 2014 and 2015.
New capacity additions then rebounded in 2018 and 2019, with at least 600 MW added in each year. In 2018, plants were commissioned in China, Morocco and South Africa, with LCOEs ranging from a low of USD 0.076/kWh in China, to a high of USD 0.234/kWh in South Africa. In contrast, 2019 saw higher LCOEs, as two delayed Israeli projects came online. Costs that year ranged from USD 0.107/kWh for a project in China to USD 0.404/kWh for the Israeli PTC project.
Deployment in 2020 did not exceed 150 MW, though low capital costs for the projects occurring in China pushed down the weighted average LCOE for that year to USD 0.107/kWh. In 2021, the LCOE value was 7% higher than in 2020, at USD 0.114/kWh – although this was still 68% lower than in 2010. The 2021 figure was, however, based on a very thin market.
At 64%, the largest share of the reduction was taken by the decline in the total installed cost of CSP plants over the period. Improvements in technology and cost reductions in thermal energy storage – which led to projects with longer storage duration being commissioned in 2020 – led to an improvement in capacity factors. This, in turn, accounted for 17% of the reduction in LCOE over the 2010 to 2020 period. Lower O&M costs accounted for 10% of the total decline in LCOE during that time, while the reduction in the weighted average cost of capital accounted for the remaining 9%.
This same analysis yields quite different results for the period 2010 to 2021, given the high total installed costs/high capacity factor structure of the 2021 project in Chile. Accounting for this results in the capacity factor being the major contributor (77%) to cost reduction between 2010 and 2021. Lower O&M costs account for a tenth of the reduction, while reductions in the global weighted average total installed costs of newly commissioned CSP plants accounted for 7%. Improvements in the weighted average cost of capital account for 6% of the total decline in LCOE over the period.
In the absence of strong policy support for CSP, the market remains small and the pipeline for new projects unambitious. This is disappointing, given the remarkable success in reducing costs since 2010, despite just 6.4 GW being deployed globally by the end of 2021. Given the growth in variable renewables competitiveness since 2010, the value of CSP’s ability to provide dispatchable power 24/7 in areas with high DNI at reasonable cost is only set to grow. Greater policy support would be instrumental in bringing costs down even further – and in reducing overall electricity system costs – by providing firm, renewable capacity and flexibility services to integrate very high shares of renewables.
For PTCs, cost reductions have been pursued by trying to reduce the costs of the parabolic troughs themselves and by improving their performance. Essentially, the challenge has been to raise absorption of solar heat and reduce heat losses in the HTF conveyed to the power block, while at the same time, reducing the capital cost of the components.
Improvements in special coatings on the absorber tube and insulation measures for the receiver have helped reduce thermal losses. To reduce capital costs, efforts have focused on reducing materials costs relative to heat generation.
To the extent possible, given the loads on the structure, light-weighting of the mirrors and supporting frameworks has been pursued. Aperture widths have also been increased to allow for greater solar radiation to be focused.
Between 2010 and the 2018-2020 period, the weighted average aperture width of the parabolic troughs used in projects increased from around 5.7 metres (m) to around 7 m. In 2010, Spanish projects were dominant, using troughs with widths in the relatively narrow range of 5.5 m to 5.8 m. In the period 2018 to 2020, although deployment had slowed, it was more geographically diverse and used a wider range of troughs. These went from 5.8 m widths – not dissimilar to those used in two projects in 2010 – to larger, 8.2 m ‘space tube’ troughs.
With an increased share of STs in deployment, the increased operating temperatures made possible by the use of molten salt HTFs or direct steam generation saw weighted average receiver outlet temperatures increase. These rose from 396°C in 2010, when PTC plants represented all capacity added for which there is data, to 485°C in 2019, as STs with receiver outlet temperatures ranging from 560°C to 565°C were commissioned.
Higher temperature differentials in the hot and cold tanks allow greater energy to be stored for a given volume. Yet, the benefit of higher operating temperatures is not just lower cost thermal energy storage, but also that they allow more efficient steam cycles to recover more electricity from the available resource. With the increasing share of STs, the weighted average turbine efficiency for projects where data is available rose from 38% in 2010 to 44% in 2019.
Efforts continue to commercialise molten salts as an HTF for PTC plants, since it can lead to higher HTF temperatures (530°C) than the currently prevalent thermal oil (393°C). Silicon-based HTF have also been proposed as an alternative and can achieve 425°C (Jung et al., 2015). However, for the moment, the largest efficiency gains and greatest potential for longer storage remains with ST plants that can already operate at higher temperatures and efficiencies. Greater scale in the deployment of STs would also help to narrow their installed cost premium over PTC plants, potentially allowing STs a decisive advantage over PTCs in LCOE terms, in locations where the air is clear.