Short-term power forecasting model of floating photovoltaic—A case study of Huainan floating photovoltaic plant, China

Although the development of floating photovoltaic (FPV) in China is rapid, there are few reports on the power generation forecasting of FPV. Based on the monitoring data of the 150MW FPV plant in Huainan, a short-term power forecasting model for FPV plant is proposed in this paper. Firstly, the input variables of deep belief network (DBN) are selected based on the open-source R programming software; then, wavelet transform is used to decompose the power time series into several components with different frequencies, and convolutional neural network (CNN) is used to extract the nonlinear features and invariant structures of each component; finally, the back propagation neural network (BPNN) optimized by genetic algorithm (GA) is used to integrate the outputs of CNN and DBN to obtain the final PV power forecasting value. The simulation results show that the method proposed in this paper can obtain superior forecasting accuracy than other models.