Deep Learning Models in Photovoltaic Power Forecasting: A Review

This paper presents a systematic and comprehensive review on application of deep learning (DL) models for direct photovoltaic (PV) power forecasting. Several factors which influenced PV power forecasting, namely forecasting horizon, forecasting input, data preprocessing and optimization of models, are discussed. DL models, including, gated recurrent unit (GRU), convolutional neural network (CNN), deep belief network (DBN), and long short term memory (LSTM), along with their utilization and contribution in direct PV power forecasting are presented. Further, this review also highlights the superiority of hybrid and optimized models in providing more accurate forecasts compared to stand alone models over the recent years. The research presented here provides fundamental guidelines for researchers and solar power forecasters on how DL techniques can improve the PV power prediction accuracy.