Short-Term Photovoltaic Power Forecast Based on Long Short-Term Memory Network

Abstract

The randomness and fluctuation of photovoltaic (PV) power brings new challenges to power system operation. Accurate PV power forecasting is critical to system dispatch. This paper applies long short-term memory network (LSTM) to forecast short-term PV power. First, Pearson correlation analysis is applied to identify features affecting PV power. The features with high correlation coefficient are selected as LSTM inputs. Short-term LSTM PV power forecasting models are then established according to different seasons and weather types. Case study is performed using PV power and numerical weather prediction (NWP) of a practical PV station in northwest China. The results obtained indicate that the forecasting models can effectively improve the forecast accuracy of short-term PV power.