Short-term Prediction of Photovoltaic Power Based on PLSR Improved DBN

Photovoltaic (PV) power prediction is greatly influenced by meteorological factors, so it is of great significance to predict PV power generation accurately and timely. Neural network prediction is widely used in power prediction. Especially, Deep belief networks (DBN) is favored by researchers as it can extract the deep features of data. In the fine-tuning stage of DBN network, Back propagation (BP) algorithm is easy to fall into the local optimal algorithm. This paper proposes a partial least squares regression method instead of BP method to perform parameter tuning method and Cross-verify its rationality. In order to further improve the prediction accuracy, the improved fuzzy C-means clustering (FCM) algorithm was used to cluster the input sample data then the Photovoltaic power prediction model of FCM-DBN-PLSR was established. The result of simulation shows that the prediction model established in this paper can predict photovoltaic power accurately. The prediction error value is 6.01%.