Short-term photovoltaic power forecasting method based on convolutional neural network
Yutong He, Qingzhong Gao, Yuanyuan Jin, Fang Liu
Abstract
This research proposes a hybrid model that combines the convolutional neural network (CNN) and the bidirectional long short-term memory network (BiLSTM) to accurately estimate the energy output of a short-term photovoltaic system. Firstly, Pearson correlation analysis is introduced to screen out meteorological factors with high correlation with photovoltaic (PV) power output. Then, a convolutional neural network-bidirectional long short-term memory network (CNN-BiLSTM) combined algorithm is used to extract the characteristics of influencing factors by CNN, and BiLSTM is used for timing prediction. Last but not least, using simulation analysis of data from a particular region in China over the previous two years, the results show that this model reduces training time, improves prediction accuracy, and outperforms the conventional prediction model in terms of the effectiveness of forecasting results, which could also satisfy the demands of the practical application of PV energy generation prediction.