Estimation of solar irradiance and solar power based on all-sky images

The ramp event of solar irradiance is prone to occur due to atmospheric conditions, and makes it difficult to integrate solar power into regional power grid. The atmospheric aerosol, cloud occlusion, temperature and weather conditions especially influence the daily solar irradiance, and cause the uncertain of the solar power which would leads to the regional grid overloaded or lower-loaded. To integrate the solar power into the regional grid reliably, the estimation of solar irradiance and solar power output plays an important role. In the study, a system is proposed to perform accurate solar irradiance and solar power estimation. Through the all-sky image feature extractions, the regional and global weights are obtained. The long short-term memory (LSTM) is used as the training model with the weights as inputs to estimate the solar irradiance. The power curve is made by using the estimated solar irradiance and the solar power output to estimate the solar power. Several performance indices are used to evaluate the proposed method, including mean absolute error (MAE), root mean square error (RMSE) and coefficient of determination (R²). The experimental results show that the performance of the solar irradiance estimation of the proposed method is remarkable with the lower MAE and RMSE and the higher R². Furthermore, the performance of the solar power estimation using the proposed method is very close to the actual estimation performance.