Projections of future changes in solar radiation in China based on CMIP5 climate models

1 Introduction

Solar radiation is the most important energy source on the earth’s surface.It determines the thermodynamic and dynamic state of the earth’s atmospheric systems and is an important signal reflecting climate change [1, 2].In addition, as global industrialization progresses, human needs for energy continue to grow.Solar photovoltaic power is clean and environmentally friendly energy, and has therefore become the cornerstone of long-term strategies for the future production of energy by human societies [3-5].

Variations in solar radiation and the possible factors affecting that variation in China or in the world have been studied by many scholars using observed data.Variations in surface solar radiation in northern Europe showed a significant decreasing trend from the 1950s to the 1980s,followed by a slight increase, and this study pointed out that cloud cover and atmospheric circulation were the main influencing factors [6].Surface solar radiation in southeastern China decreased significantly from 1961 to 1989, and then increased from 1989 to 2008 [7].Fang Shibo et al.analyzed the variations in total surface solar radiation in eastern and western China from 1961 to 2011, and found that the total surface solar radiation in both eastern and western China showed a decreasing trend, with sunshine percentages and air pollution as the main influencing factors [8].

Current methods for predicting future solar radiation change scenarios mainly involve the extension of statistics and model estimations based on mathematical science [9-11].However, few scholars have previously predicted the surface solar radiation in China or specific regions within China.The coupled model inter-comparison project phase 5(CMIP5), which is a global climate model, can effectively simulate average climate states and change characteristics[12].In this paper, 38 CMIP5 climate models are used to predict changes in surface solar radiation and their effect on photovoltaic power generation in China over the next 40 years.Additionally, several regions that are the key to the development of photovoltaic power stations currently and in the future, are specifically analyzed: Xinjiang, Qinghai,Inner Mongolia, and Shaanxi.This study will be of great significance for understanding future climate change in China, and will help optimize the macro locations of photovoltaic power plants, providing a reference for the future development of solar energy resources.

2 Data and Methods

2.1 Data

Data were selected from the CMIP5 simulation data(http://cmip-pcmdi.llnl.gov/cmip5).In the emission scenario,the CMIP5 trial used four representative concentration pathways (RCPs): RCP2.6, RCP4.5, RCP6.0, and RCP8.5[13].Emissions scenarios are not chosen to accurately predict the future, but to reduce prediction uncertainty and provide a reference for decisions.In this study, the RCP8.5 high radiative forcing scenario is considered.This scenario is improved relative to the CMIP3 model, because it establishes a spatial distribution of atmospheric pollution estimations and strengthens the estimation of land use and land surface changes [14].The RCP8.5 scenario does not include any specific tasks of alleviating climate change pressures.Under the RCP8.5 scenario, radiative forcing will rise to 8.5 W/m2,and the CO2 concentrations in the air will reach 1370 ppm (1 ppm =1 μL/L) by 2100, which is 3–4 times higher than CO2 concentrations prior to the industrial revolution.

The basic information for the 38 CMIP5 climate patterns selected is shown in Table 1.The selected ensemble run for each pattern was rlilp1 (r2ilp1 was selected for CESM1-WACCM and FGOALS-s2).The analyzed variables are surface downward solar radiation (rsds, for all weather conditions), surface downward solar radiation under cloudless conditions (rsdscs; the CMCC-CESM,CMCC-CM and CMCC-CMS models are not available for rsdscs), near-surface air temperature (tas, also known as ambient temperature), and total cloud cover (clt).The time period analyzed in this study ranges from 2006 to 2049(photovoltaic panel lifespans are approximately 40–50 years).The 2006–2015 period is used as the reference period, representing current climatic conditions.We performed a bilinear interpolation of global data in the selected model to a horizontal resolution of 0.75°×0.75°.All data were processed into annual average data.

2.2 Methods

To forecast the effects of climate change on future photovoltaic generation, we use the method proposed by Crook et al.[10], in which photovoltaic generation largely depends on battery materials, sunshine, and ambient temperatures.

Without considering other loss factors and system components such as inverters, the power output of a photovoltaic panel is

Where Gtot is the downward solar radiation received by the earth’s surface under all weather conditions and is the efficiency of the photovoltaic panel (Photovoltaic conversion rate).

Where is the reference efficiency and βand γ depend on the material and structure of the battery.Parida et al.[15]showed that the monocrystalline silicon cell dominates production, that β=0.0045, and that γ=0.1.Tref = 25 ℃ is the reference temperature and Tcell is the panel temperature,which is determined as follows [10]:

where T is the ambient air temperature (℃).c1, c2 and c3 depend on the properties of the battery and c1= -3.75 ℃,c2= 1.14, c3= 0.0175 ℃m2W-1 for a monocrystalline silicon cell.

We calculate the change (∆P/ P) in future photovoltaic power generation relative to that under current climatic conditions, because the value ofis unknown.In addition,we consider only horizontal irradiance, disregarding the fact that slant panels or tracing ray panels could receive more radiation.

Table 1 Basic information on the climate models in CMIP5

Model Unit, country Resolution Model Unit, country Resolution ACCESS1-0 CSIRO-BOM, Australia 192 × 145 GFDL-ESM2G NOAA GFDL, America 144 × 90 ACCESS1-3 CSIRO-BOM, Australia 192 × 145 GFDL-ESM2M NOAA GFDL, America 144 × 90 BCC-CSM1-1 BCC, China 128 × 64 GISS-E2-H NASA GISS, America 144 × 90 BCC-CSM1-1-m BCC, China 320 × 160 GISS-E2-R NASA GISS, America 144 × 90 BNU-ESM GCESS, China 128 × 64 HadGEM2-AO NIMR/KMA, Korea/England 192 ×145 CanESM2 CCCMA, Canada 128 × 64 HadGEM2-CC MOHC, England 192 ×145 CCSM4 NCAR, America 288× 192 HadGEM2-ES MOHC, England 192 ×145 CESM1-BGC NSF-DOE-NCAR, America 288 × 192 INMCM4 INM, Russia 180 × 120 CESM1-CAM5 NSF-DOE-NCAR, America 288 × 192 IPSL-CM5A-LR IPSL, France 96 × 96 CESM1-WACCM NSF-DOE-NCAR, America 144 × 96 IPSL-CM5A-MR IPSL, France 144 × 143 CMCC-CESM CMCC, Italy 96 × 48 IPSL-CM5B-LR IPSL, France 96 × 96 CMCC-CM CMCC, Italy 480 × 240 MIROC5 MIROC, Japan 256 × 128 CMCC-CMS CMCC, Italy 192 × 96 MIROC-ESM MIROC, Japan 128 × 64 CNRM-CM5 CNRM-CERFACS, France 256 × 128 MIROC-ESM-CHEM MIROC, Japan 128 × 64 CSIRO-Mk3-6-0 CSIRO-QCCCE, Australia 192 × 96 MPI-ESM-LR MPI-M, Germany 192 × 96 FGOALS-g2 LASG-CESS, China 128 × 60 MPI-ESM-MR MPI-M, Germany 192 × 96 FGOALS-s2 LASG-IAP, China 128 × 108 MRI-CGCM3 MRI, Japan 320 × 160 FIO-ESM FIO, China 128 × 64 NorESM1-M NCC, Norway 144 × 96 GFDL-CM3 NOAA GFDL, America 144 × 90 NorESM1-ME NCC, Germany 144 × 96

3 Results

3.1 Estimated projections of radiation, temperature, and cloud cover in China by CMIP5 model

We use linear regression to analyze the evolution of climate elements on a long-term scale [16].Fig.1 shows a spatial distribution of the linear trends of various climates elements from 2006–2049: (a) downward solar radiation received by the earth’s surface; (b) downward solar radiation received by the earth’s surface under cloudless conditions; (c)near-surface air temperatures; (d) total cloud fractions.The regression coefficients are confirmed by a test of significance for the linear regression equation (t-test, α = 0.05).We used a median collection scheme (that is, the median of the 38 pattern data values at each grid point for each variable); Fig.1 shows the synthesis of the 38 pattern data.

According to Fig.1c, near-surface layer air temperatures are warming throughout China at a rate of 0.05 K/year.Panel efficiency decreases with the increasing panel temperatures induced by increased air temperatures,assuming the same amount of radiation [17].Fig.1b shows that surface downward solar radiation under cloudless conditions has a linear trend close to zero or a small negative trend over western China, and a maximum positive trend that is approximately 0.3 W/m2/year over southeastern China.Previous studies have shown that atmospheric turbidity dominates surface solar radiation in eastern China [8].Therefore, the maximum positive trend in southeastern China may result from a decrease in aerosol concentrations.Positive and negative trends in the spatial distribution of surface downward solar radiation under all weather conditions (rsds), which mainly includes the effects of clouds, is consistent with that under cloudless conditions(rsdscs), as shown in Fig.1a.Cloud cover exhibits a negative trend throughout China; in the southeast, that negative trend is up to -0.1%/year.However, rsds increases,with a trend of 0.4 W/m2/year (Fig.1d).The results are consistent with those of Wild et al.[9].

Fig.1 Distribution of median trends of all considered models between 2006 and 2049 in China

3.2 Estimates of selected areas of radiation,temperature and cloud cover by CMIP5 model

Combined with the National Energy Administration’s 12th Five-Year Plan and 13th Five-Year Plan for developing solar power, the eastern part of Xinjiang (86°E ～ 95°E,42°N ～ 44°N), Qinghai (91°E ～ 101°E, 34°N ～ 38°N), Inner Mongolia Municipal (101°E ～ 112°E, 39°N ～ 42°N) and the northern part of Shaanxi (108°E ～ 110.7°E, 36°N ～ 39°N)are selected to analyze, because these regions will develop photovoltaic rapidly in the future (Fig.2).

The average annual variations in climatic factors in the selected area (deviation from the average of ten years from 2006 to 2015) are shown in Fig.3.The black lines in the figure represent annual variations in the deviation values of the 38 models (rsds, rsdscs, tas, clt).The red lines in the figure show a linear trend fit of the median of 38 model deviation values.As shown in Fig.3, the linear temperature trends (tas) increase in selected regions.The increase rates are only slightly different, and all approximately 0.49 K/year.Downward solar radiation under cloudless conditions(rsdscs) exhibits negative trends in Xinjiang, Qinghai, and Inner Mongolia.The maximum negative trend (-0.051 W/m2/year) occurs in Xinjiang.In Shaanxi, the rsdscs trends before 2025 are not significant, and increase after 2025.The surface downward solar radiation (rsds) in Xinjiang,Qinghai and Inner Mongolia exhibit negative trends.The maximum negative trend (-0.080 W/m2/year) appears in Qinghai.In Shaanxi, the rsds increases at a rate of 0.030 W/m2/year.The amount of clouds selected in the figure is negative, and the maximum negative trend (-0.025 %/year)appears in Qinghai.Previous studies have shown that total cloud cover has an irregular correlation with surface solar radiation in western China [8].The surface downward solar radiation in the western region is mainly characterized by a decreasing trend, which may be related to atmospheric turbidity and the percentage of sunshine.In short, surface downward solar radiation in the northwestern region of China mainly exhibits a negative trend.The temperature is warming in China.Those all negatively affect power generation by photovoltaic power plants in northwest China.

Fig.2 Four regions of China selected in this study

Fig.3 Annual variation of rsds [W/m2], rsdscs [W/m2], tas [K]and clt [%]for the four focal regions (m is the regression coefficient, units are per year)

3.3 Potential impact of climate elements on solar photovoltaic power plants

Power generation in photovoltaic systems is greatly affected by natural environmental conditions.The study of the prediction on PV power generation by using meteorological factors needs to be further improved [18].Lv et al.[19]found that radiation and temperature are the two main factors affecting photovoltaic power generation.As is known from formulas (1, 2, and 3) in section 1.2, in this study we analyze the effects of surface solar radiation and near-surface temperatures on photovoltaic power generation.We hope this will provide a reference for predicting the output of photovoltaic power stations.

Previous studies have shown that the surface solar radiation and temperature values simulated by climate models deviate somewhat from observed values, mainly because of the difficulty of cloud parameterization [20].Therefore, some studies [21]have corrected the temperature and radiation before calculating the amount of photovoltaic power generation.To evaluate the effects of systematic deviations in surface solar radiation and temperature on the calculation of photovoltaic power generation, Wild et al.[9]performed a sensitivity experiment.However, the researchers did not find that the correction of their data has a significant impact on their estimations of photovoltaic power generation trends.Therefore, we have not corrected the surface solar radiation and temperature data.In this study, we used irradiance data for the horizontal plane.If the panel is slanted or is a chasing photovoltaic panel,the effects of radiation and temperature changes on photovoltaic power generation will be greater.For example,chasing panels are affected twice as much as horizontal panels [9, 22].

We calculated the relative deviations in the annual photovoltaic power generation for 2006–2049 according to 38 models, and calculated the average photovoltaic power generation from 2006–2015 at each grid point.Then we fit linear trends for the relative deviation data for the 38 models.Finally, the median of 38 model calculation results(regression coefficients) is selected for each grid point(Fig.4).Fig.4 is very similar to Fig.1a, because the trends in temperature changes throughout China are relatively consistent.Formula (1) indicates that temperature increases will lead to decreases in photovoltaic power generation.Additionally, the decline in surface solar radiation (rsds)in western China will lead to a decrease in photovoltaic power generation.Compared with the average photovoltaic power generation from 2006–2015, photovoltaic power generation in western China will decrease by -0.04 %/year.In southeastern China, photovoltaic power generation will increase by 0.06-0.1%/year.The effect of surface solar radiation on power generation is dominant.

Comparing the relative changes in photovoltaic power generation in the eastern and western regions of China over the next 40 years, we find that the eastern region shows a larger change than the western region, mainly because the variation of surface solar radiation in eastern China is significantly greater than that in the west.This is partly because of the effects of cloud cover.In addition, smog and air pollution are the main factors affecting surface solar radiation in the eastern region.Over the past 50 years, the frequency of haze in eastern China has increased significantly, while the frequency in the west has not changed significantly.Therefore, in the next 40 years, smog and other polluted weather in eastern China will decrease.Additionally, wind speeds are negatively correlated with aerosol thickness.Therefore, whether the wind speeds will increase in the future need further research.Although the western part of China shows a negative trend, the changes are relatively small, and the solar energy resources are abundant [23, 24].The development of photovoltaic power stations is also relatively stable.

Fig.4 Spatial distribution of median trends of the relative changes in photovoltaic generation capacity in China from 2006–2049 (Unit: %/year)

4 Conclusions

In this study, CMIP5 global climate model data are used to predict surface solar radiation, temperature, and cloud cover in China over the next 40 years.Additionally, the effects of surface solar radiation and temperature changes on future photovoltaic power generation are analyzed.The main conclusions are as follows:

(1) For surface temperatures, the median trends of all considered models show a warming trend in China, at a rate of about 0.05 K/year.The maximum positive trend for allsky radiation appears in the southeast of China, reaching about 0.4 W/m2/year.Cloud cover mainly decreases throughout China.

(2) The all-sky radiation for most selected regions shows a decreasing trend.The maximum negative value (-0.08 W/m2/year) appears in Qinghai.

(3) Compared with the average photovoltaic power output from 2006 to 2015, the photovoltaic power output in western China will decrease by -0.04 %/year, while the photovoltaic power output in southeastern China will increase by 0.06–0.1%/year.

The new-generation CMIP5 model represents a significant improvement on past climate models.However,there are still some problems associated with the simulation of solar radiation in current climate models, which makes long-term solar radiation predictions uncertain, especially in local regions.The models should be further improved to address these issues.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (No.41805085, 41475066 and 41575112), the State Grid Science and Technology Project (SGGSKY00FJJS1700304), and the Foundation for Excellent Youth Scholars of Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences.

Thanks for the simulation test data to the CMIP5 model groups shown in Table 1.Thanks to Doris Folini for his guidance.