Multimodel ensemble projection of photovoltaic power potential in China by the 2060s

Xu Zho ,Xu Yue,＊ ,Chengung Tin ,Ho Zhou ,Bin Wng ,Yuwen Chen ,Yun Zho,Weijie Fu,Yihn Hu

a Jiangsu Key Laboratory of Atmospheric Environment Monitoring and Pollution Control, Collaborative Innovation Center of Atmospheric Environment and Equipment Technology, School of Environmental Science and Engineering, Nanjing University of Information Science &Technology (NUIST), Nanjing, China

b Climate Change Research Center, Institute of Atmospheric Physics (IAP), Chinese Academy of Sciences (CAS), Beijing, China

c University of Chinese Academy of Sciences, Beijing, China

Keywords: Photovoltaic power Climate change Warming extremes CMIP6 Ensemble projection

ABSTRACT China’s demand for solar energy has been growing rapidly to meet energy transformation targets.However,the potential of solar energy is affected by weather conditions and is expected to change under climate warming.Here,the authors project the photovoltaic (PV) power potential over China under low and high emission scenarios by the 2060s,taking advantage of meteorological variables from 24 CMIP6 models and 4 PV models with varied formats.The ensemble mean of these models yields an average PV power of 277.2 KWh m-2 yr-1 during 2004–2014,with a decreasing tendency from the west to east.By 2054–2064,the national average PV power potential is projected to increase by 2.29% under a low emission scenario but decrease by 0.43% under a high emission scenario.The emission control in the former scenario significantly enhances surface solar radiation and promotes PV power in the east.On the contrary,strong warming causes inhibitions to PV power generation under the high emission scenario.Extreme warming events on average decrease the PV power potential by 0.28%under the low emission scenario and 0.44%under the high emission scenario,doubling and tripling the presentday loss,respectively.The projections reveal large benefits of controlling emissions for the future solar energy in China due to both the clean atmosphere and the moderate warming.

1.Introduction

To achieve the goal of carbon neutrality by the year 2060,China has initiated rapid development of non-fossil energy such as hydropower,wind power,solar energy,and biomass energy.The newly installed photovoltaic(PV)power in 2021 was more than 20 times that of in 2011(Liu et al.,2022a).Studies have shown that the full lifecycle carbon emissions of PV power generation are 10–34 g-CO2e/KWh-electricity,much lower than the value of 230–935 g-CO2e/KWh-electricity from coal power generation (Jacobson,2020).Meanwhile,China has abundant solar energy resources,with an estimated potential capacity of 4700 to 39300 GW(He and Kammen,2016),which is 15–128 times the total installed PV power by the year 2021 (National Energy Administration,2022).Therefore,the full utilization of solar energy is one of the effective solutions to reduce the usage of fossil fuels and related carbon emissions.

However,the PV output is affected by various climatic factors such as solar radiation,temperature,and wind speed (Rahman et al.,2015;Razak et al.,2016;AlSkaif et al.,2020).Among these factors,radiation has shown the dominant impact and is predicted to affect PV power in the future climate.For example,Zuluaga et al.(2022)projected the PV power potential in Brazil by 2021–2050 and found the maximum increase of 2.5%–3% in the north due to the decrease in cloudiness and increase in radiation.In Europe,Jerez et al.(2015)estimated an overall reduction of 10%in PV power potential under a high emission scenario,with the larger decline in northern Europe at the end of this century.Globally,Zou et al.(2019) projected a decreasing trend of 0.67 KWh m-2yr-1in PV power following the changes of solar radiation during 2006–2100 based on multiple climate models from phase 5 of the Coupled Model Intercomparison Project(CMIP5).In addition,warming has been found to cause negative impacts on solar energy generation.For example,Danso et al.(2022)found that the increase in temperature could reduce the PV power potential in West Africa by 2%in 2050–2084 based on an ensemble projection using simulated data from multiple CMIP phase 6(CMIP6)models.

While most projections thus far have focused on the changes in the mean state of the climate,only a few studies have explored the significance of impacts of climatic extremes on the PV power potential.Feron et al.(2021) predicted that the number of summer days with very low PV generation would decrease by 50%in southern Europe but increase by over 100% on the Arabian Peninsula by the middle of the century under a medium emission scenario.Using downscaled climate projections from a single regional model,Zhang et al.(2022)found a 36.1%increase in extremely low PV power generation events in China under a 3?C warmer climate.These studies suggest a likely increase in the instability of PV power generation due to the more frequent climatic extremes in the future,posing an unfavorable condition for the widespread usage of PV power in China.

In this study,we make the first attempt to estimate future PV power potential in China by the 2060s using multiple climate and PV models.The multimodel ensemble method has been widely used in climate projections to reduce the uncertainties from individual models (Kharin and Zwiers,2002;Wu et al.,2020).However,the approach with multiple PV models has not been applied to the projection of PV power potential.We use the predicted future climate from multiple CMIP6 models to drive a group of PV models with different formats.The CMIP6 models have shown better agreement with observed solar radiation compared to CMIP5 models owing to the improvement of radiative processes for clouds and aerosols in the radiation modules(Wild,2020;Feng et al.,2021;Danso et al.,2022),providing a more solid foundation for the projection of PV power potential.In addition to the mean state,we explore how the changes in climate extremes,such as high temperature,affect the efficiency of PV generation.

2.Data and methods

2.1.CMIP6 data

We use the daily output of climate models from CMIP6 under two different Shared Socioeconomic Pathways(SSPs).The SSPs describe the possible future development of society in the absence of climate change or climate policy impacts (O’Neill et al.,2014),which are further combined with the projected radiative forcing by the end of the 21stcentury.For example,the SSP1-2.6 represents a sustainable pathway(SSP1) with a relatively low radiative forcing of 2.6 W m-2at 2100.In contrast,the SSP5-8.5 indicates a business-as-usual development(SSP5)with high radiative forcing of 8.5 W m-2by the end of the century.For this study,we use climate projections under the SSP1-2.6 and SSP5-8.5 scenarios to cap the range of PV power potential in response to climate change.We use daily surface meteorological variables,including downward solar radiation,air temperature,and wind speed,from a total of 24 CMIP6 models (Table S1),which are all the available models we could download from the CMIP6 archive.The climatic data with varied spatial resolutions are interpolated to the same 1?× 1?to derive the ensemble means.Since 2014 is the last year of historical simulations,we select the average of 2004–2014 to represent the present-day climatology.Meanwhile,we choose 2054–2064 as the future period,which are the target years for carbon neutrality in China.A time span of 11 years is selected for both the present day and future to minimize the impact of interannual variability on the predicted climate.

2.2.Observational data

We use observed radiative and meteorological data to validate the CMIP6 models.The surface downward shortwave radiation is adopted from CERES-SYN1deg-Ed4.1 (https://ceres-tool.larc.nasa.gov/),which shows higher accuracy than reanalyses products such as ERA5 and MERRA-2 (Zhou et al.,2019).Surface air temperature is adopted from the Climatic Research Unit(CRU)version 4(https://crudata.uea.ac.uk/cru/data/hrg/).We also use the surface wind speed from the CN05.1 gridded observation dataset (Wu and Gao,2013).The original resolutions for these observations are 1?×1?for CERES,0.5?×0.5?for CRU,and 0.25?×0.25?for CN05.1.All these datasets during 2004–2014 are selected and interpolated to the same 1?× 1?as the CMIP6 models to facilitate the comparisons.

2.3.PV models

Previous studies have proposed several PV models with different combinations of environmental factors(Davis et al.,2001;Mora Segado et al.,2015;Feng et al.,2021;Feron et al.,2021).Despite the different formats of these models,they have demonstrated fair reliability because they take important physical processes,particularly the effects of radiation and air temperature,into account.For this study,we use four different PV models to calculate the PV power potential in China(Table 1).All these models consider the effects of radiation and temperature,and two of the equations additionally consider the effects of wind speed.The ambient air temperature affects cell temperature and further inhibits PV power output with a negative conversion efficiency(-β),which is set to -0.0045?C-1following previous studies.In contrast,an increase in wind speed enhances the ventilation and helps promote PV power generation by decreasing cell temperature.Except for the impacts of meteorology,a constant parameter ofηrefis applied to all PV models to indicate the conversion efficiency from radiation to PV power under standard conditions.

Table 1 Summary of the four PV models used in this study.

3.Results

We evaluate the simulated surface air temperature and solar radiation from 24 CMIP6 models using the Taylor diagram (Fig.S1).For temperature,correlation coefficients of 0.89–0.97 for long-term means and ratios of 0.87–1.34 for standard deviations are derived between observations and simulations from the 24 climate models (Fig.S1(a)).For radiation,the model-to-observation correlation coefficients range from 0.57 to 0.93 with normalized standard deviations of 0.69–1.22 for CMIP6 models.By excluding the models with the lowest correlations(MIROC-ES2L and ISPL-CM6A-LR) or the largest deviations (CNRMCM6-1 and BCC-CSM2-MR),we select a total of 20 models to estimate the future PV power potential.The ensemble means of the selected CMIP6 models yield a high correlation coefficient of 0.96 and a low relative mean bias (RMB) of -0.5% with observed surface air temperature in China (Fig.S2(c)).Meanwhile,these models show a high correlation of 0.92 and a low RMB of 6.5% for solar radiation (Fig.S2(f)),indicating a good capability of the multimodel ensemble in capturing the observed spatial pattern of key climatic parameters for PV power projection.

We project the future changes in temperature and radiation in 2054–2064 relative to 2004–2014.Compared to present day,the average surface temperature by the 2060s increases by 1.62 K in China under the SSP1-2.6 scenario (Fig.1(a)) and 2.86 K under the SSP5-8.5 scenario (Fig.1(b)).For both scenarios,the warming pattern is generally homogeneous,with relatively stronger magnitude in the northern part.In contrast,changes in radiation show a heterogeneous pattern,with an increasing trend over eastern China but a decreasing trend in western and northern parts (Fig.1(c,d)).The enhancement of solar radiation is much stronger under the SSP1-2.6 scenario,with regional hotspots up to 20 W m-2(Fig.1(c)),than under the SSP5-8.5 scenario(Fig.1(d)).On the national scale,solar radiation on average increases by 7.22 W m-2under the SSP1-2.6 scenario,mainly because of the cleaner air due to emission control helps promote downwelling radiation(Tang et al.,2022).By comparison,the national mean solar radiation increases only by 2.89 W m-2under the SSP5-8.5 scenario because of the limited reductions in atmospheric aerosols(Gidden et al.,2019).

In response to the changes in climatic variables,the ensemble mean PV power potential shows contrasting tendencies under the two future scenarios (Fig.2).In the present day,the national total PV power potential is 277.2 KWh m-2yr-1,with high values above 300 KWh m-2yr-1in the west but low values below 240 KWh m-2yr-1in the east(Fig.2(a)),following the spatial pattern of solar radiation in China(Fig.S2(e)).Under the SSP1-2.6 scenario,the national average PV power increases by 6.36 KWh m-2yr-1,with the higher enhancement of 16–24 KWh m-2yr-1over the east (Fig.2(b)).In contrast,the PV power potential on average decreases by 1.19 KWh m-2yr-1(Fig.2(c))under the SSP5-8.5 scenario,even though surface radiation is higher by 4–8 W m-2in the east (Fig.1(d)).These changes are robust since the projections using the four different PV models reveal the same tendencies(Fig.S3).We further compare the uncertainties (indicated by one standard deviation)due to the multiple PV and climate models(Fig.S4).For different PV models,the projection uncertainty is 0.15 KWh m-2yr-1for the SSP1-2.6 scenario and 0.17 KWh m-2yr-1for the SSP5-8.5 scenario.For different climate models,the projection shows uncertainties of 15.52 KWh m-2yr-1for the SSP1-2.6 scenario and 20.14 KWh m-2yr-1for the SSP5-8.5 scenario.As a result,the uncertainties of PV power projection mostly originate from those in climate projections.

We perform sensitivity experiments to isolate the contributions of individual climatic factors to the changes in PV power potential(Fig.3).The warming alone decreases PV power by 0.68%under SSP1-2.6(Fig.3(a))and 1.24%under SSP5-8.5(Fig.3(b)).The larger dampening effect in the latter scenario is attributable to the warmer climate for the same scenario (Fig.1(b)).In contrast,the increase in radiation promotes PV power by 2.99% under SSP1-2.6 (Fig.3(c)) and 0.81% under SSP5-8.5(Fig.3(d)).The net effect of climatic change results in an increase of 2.29% in PV power potential in China under the SSP1-2.6 scenario(Fig.3(e)),because the PV gain from more radiation outweighs the PV loss from warming.On the contrary,the changes in climate lead to a net reduction of-0.43%in PV power potential under the SSP5-8.5 scenario(Fig.3(f)),following the stronger inhibition effect by warming (Fig.3(b)) than the benefit of more radiation (Fig.3(d)).

Fig.3.Relative changes in PV power potential with varied combinations of temperature and radiation compared to present-day values.The combinations include(a,b) present-day radiation but future temperature,(c,d) future radiation but present-day temperature,and (e,f) future radiation and temperature.For future data,either the (a,c,e) SSP1-2.6 or (b,d,f) SSP5-8.5 scenario is employed.

We further explore the effects of extreme warming on PV power generation.For most PV panels,the power conversion efficiency decreases significantly if the temperature is higher than 25?C(Davis et al.,2001;Mora Segado et al.,2015;Feng et al.,2021;Feron et al.,2021).Here,we calculate the number of warming days with average temperature of>25?C for different periods.In the present day,an ensemble mean of 48.5 warming days is predicted in China,with 40.58%of grids experiencing more than 60 warming days(Fig.4(a)).By the 2060s,the warming days increase to 76.8 days under SSP1-2.6(Fig.4(b))and 94.7 days under SSP5-8.5(Fig.4(c))on the national scale.The percentage of grids with more than 60 warming days increases to 52.82%and 59.67%for these two scenarios,respectively.In response to such warming,PV power generation reduces by 0.13% in the 2010s (Fig.4(d))due to the negative impacts of high temperatures.Such inhibiting effects become stronger by the 2060s,with 0.28%for SSP1-2.6(Fig.4(e))and 0.44%for SSP5-8.5(Fig.4(f)).As a result,the warming-induced loss of PV power is doubled for a low-warming scenario but tripled for a high-warming scenario.

Fig.4.Ensemble projected(a–c)number of warming days(>25?C)and(d–f)loss of PV power potential in the(a,d)present day and future climate under the(b,e)SSP1-2.6 and (c,f) SSP5-8.5 scenarios.Results in (d–f) are calculated as the differences in PV power potential if all warming days are set to 25?C,indicating the percentage reduction of PV power potential due to the increased frequency of high temperatures.

4.Discussion and conclusions

We project the future changes of PV power potential in China using an ensemble of 24 climate models and 4 PV models.Based on the evaluations,20 climate models are selected to reproduce the observed distribution of surface air temperature and solar radiation.In the present day,the multi-model mean PV power potential is 277.2 KWh m-2yr-1in China,similar to the estimates of 276 KWh m-2yr-1by Liu et al.(2022b).By the middle of the century,the PV power potential is projected to increase by 6.7 KWh m-2yr-1(2.29%) under the SSP1-2.6 scenario,because the increase in solar radiation outweighs the negative effects of increasing temperature.In contrast,the PV power potential is estimated to decrease by 1.19 KWh m-2yr-1(-0.43%)owing to the stronger inhibiting effect by enhanced warming under the SSP5-8.5 scenario.For both scenarios,the PV power potential is projected to increase in the east,following the regional enhancement of solar radiation by air pollution regulation.The predicted PV power potential is also modulated by warming events,which decrease the PV power by 0.28%under SSP1-2.6 and 0.44%under SSP5-8.5,suggesting increasingly negative impacts of extreme warmings on future PV power generation.

There are,however,some limitations to our study.First,we apply daily meteorological variables in the PV calculation due to the availability of data.Both solar radiation and temperature reach their maximums at around noontime.The strong warming-induced inhibition may dampen the conversion efficiency of PV panels even though the radiation is at high levels.Such an association cannot be resolved from daily data,leading to an overestimation in the PV prediction (Chen et al.,2023).Second,we omit the effects of other meteorological factors such as wind speed,which is considered in two out of the four PV models(Table 1).Evaluations against CN05.1 show that the multimodel ensemble in general captures the spatial pattern of wind speed but overestimates the magnitude by 51% (Fig.S5).In a sensitivity test,we scale the wind speed from CMIP6 models with a uniform factor of 0.7 and yield a low relative mean bias of 3%.With the adjusted wind speed,the derived PV power potential is lower by only 0.5 KWh m-2yr-1(0.2%) for both present-day and future scenarios (Fig.S6).Sensitivity experiments also show identical PV power with the original(Fig.3)and adjusted (Fig.S7) wind speed.Moreover,projections using CMIP6 models reveal very limited changes in surface wind speed over China by the middle of the century (Wu et al.,2020),suggesting that our main conclusions are unlikely to be affected by the biases in wind speed.Third,we do not consider the negative effects of other extreme climatic events,such as cold surges,floods,and sleet.There are very limited records of PV power generation during these types of extreme events,meaning there is a low level of understanding regarding their impacts.Furthermore,projections of these extreme events are subject to uncertainties in climate models (Luo et al.,2020).Despite these limitations,this study reveals the contrasting tendencies in future PV power generation under different emission scenarios.We suggest that the low-emission pathway will be beneficial for the PV power industry in China because of the large enhancement in solar radiation with moderate warming inhibition.

Funding

This research was jointly supported by the National Natural Science Foundation of China[grant number 42275128]and the Natural Science Foundation of Jiangsu Province[grant number BK20220031].

Supplementary materials

Supplementary material associated with this article can be found,in the online version,at doi:10.1016/j.aosl.2023.100403.