A Two-Stage LSTM Optimization Method for Ultrashort Term PV Power Prediction Considering Major Meteorological Factors

Abstract

Ultrashort term photovoltaic (PV) power prediction is one of the important tasks for the intraday scheduling of PV power station integrated into the power grid system. To tackle the deficiency problem of conventional prediction methods, a novel two-stage long short-term memory network (LSTM) optimization method considering major meteorological factors is proposed for ultrashort term PV power prediction. In the first stage, an input data optimization method is developed to improve the accuracy and efficiency of LSTM, which combines major meteorological factors extraction based on factor analysis, similar pattern clustering using fuzzy c-means algorithm, and maximum similar pattern recognition based on grey correlation analysis and cosine similarity. In the second stage, a LSTM optimization method using an improved sparrow search algorithm is proposed to further improve prediction accuracy. Finally, comprehensive experiment results indicate that compared with other methods, the proposed method has higher accuracy and faster computational efficiency in ultrashort term PV power prediction.