Photovoltaic Output Prediction Based on CNN-GRU-Attention Algorithm

Aiming at the randomness and fluctuation of power output of photovoltaic power station. In this paper, a neural network model CNN-GRU-Attention based on the combination of Attention Convolutional Neural Network (CNN) and Gated Recurrent Unit (GRU) is used to predict the photovoltaic output of a photovoltaic power station. The model extracts the specific features of photovoltaic data through CNN and sends these spatial features to GRU neural network. GRU learns time sequence features in data, and strengthens the learning of important features by introducing attention mechanism, reduces the information loss caused by too long time sequence, and highlights the influence of strongly correlated features to realize the learning of ultra-long sequences. The experiment shows that compared with ANN network model, GRU network model and CNN-GRU, CNN-GRU-Attention network model is more superior in prediction accuracy.