A Photovoltaic Ultra Short Term Power Prediction Method Considering of Solar Elevation Angle

The paper focuses on how to improve the accuracy of photovoltaic ultra short term power time series prediction by combining of the solar elevation angle. The paper outlines the calculation methods of solar elevation angle, constructs a more stable intermediate variable series combining photovoltaic power and solar elevation angle, clarifies the path of time series prediction method based on random forest algorithm, proposes a photovoltaic ultra short term power time series prediction method combining solar elevation angle, and finally establishes a photovoltaic ultra short term power RF_TSP_SZA prediction model, compares the RF_TSP_R model, the LSTM_TSP_SZA model and the CB_TSP_SZA model, and applies the four models to Zhima PV power station in Yuanmou County, Chuxiong Prefecture, Yunnan Province, China. The research results show that the RF_TSP_SZA prediction model makes more effective use of the solar elevation angle information than the RF_TSP_R model. The RF_TSP_SZA model demonstrates the optimal prediction performance among the four models. The correlation coefficient of the four day daily prediction period from January 28, 2020 to January 31, 2020 exceeds 0.83, and the root mean square error is controlled within 10 MW. The research results can provide guidance for ultra short term time series prediction of photovoltaics and help improve prediction accuracy.