Hybrid Deep Learning Architecture Approach for Photovoltaic Power Plant Output Prediction

Abstract

Photovoltaic Power is an interesting type of renewable energy, but the intermittency of solar energy resources makes its prediction an challenging task. This article presents the performance of a Hybrid Convolutional - Long short term memory network (CNN-LSTM) architecture in the prediction of photovoltaic generation. The combination was deemed important, as it can integrate the advantages of both deep learning methodologies: the spatial feature extraction and speed of CNN and the temporal analysis of the LSTM. The developed 4 layer Hybrid CNN-LSTM (HCL) model was applied on a real-world data collection for Photovoltaic Power prediction on which Group Least Square Support Vector Machines (GLSSVM) reported the lowest error in the current state of the art. Alongside the PV output, 4 other predictors are included in the models. The main result obtained from the evaluation metrics reveals that the proposed HCL provides better prediction than the GLSSVM model since the MSE and MAE errors of HCL are significantly lower than the same errors of the GLSSVM. So, the proposed Hybrid CNN-LSTM architecture is a promising approach for increasing the accuracy in Photovoltaic Power Prediction.