Solar PV Power Prediction Using A New Approach Based on Hybrid Deep Neural Network

This paper contributes to the field of deep learning application to power system forecasting problems and presents a novel approach to forecast solar photovoltaic (PV) power output using a Hybrid Deep Neural Network (HDNN) model. The proposed HDNN model, which is the combination of Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) network, considers the historical PV power and other weather parameters as inputs. Several PV power forecasting models are available in the literature with mixed forecasting accuracies. However, the major advantage of the proposed HDNN model is that it extracts the salient features from local raw data to predict the solar PV power output for next several hours and days. The prediction capability of the proposed HDNN model is validated by comparing its performance with other soft computing models. Simulation results demonstrate the suitability of the proposed model to produce a higher degree of short-term PV power forecast accuracy for multiple seasons of the year.