An Improved Photovoltaic Power Generation Forecast Model Based on NARX Neural Network

A simple photovoltaic power system tends to show stochastic behaviors considering complex environmental factors. A photovoltaic power generation forecast model with great accuracy can help decrease the intermittence and uncertainty. With the expansion of the installed capacity of the photovoltaic system, traditional forecast models can no longer satisfy our needs. To address that, we analyze defects of the current forecast model based on the back propagation neural network and propose a novel forecast model based on the nonlinear autoregressive exogenous neural network. Then we use an improved particle swarm optimization algorithm to optimize our forecast model. In addition, we also provide a new normalization method for time series. In the end, we evaluate the performance of the proposed model. Good predictions are achieved regardless of what the season is. Furthermore, the forecasting model based on the nonlinear autoregressive exogenous neural network has proved capable of predicting photovoltaic power generation. The proposed model could be a useful tool for the management of a power system.