Deep Learning Assisted Prediction for Generation of Power from Solar PV

Abstract

The generation of power from renewable sources is extremely difficult to predict since it is completely dependent over the weather situation. Solar PV-based power generation entirely relies on factors like: solar incident irradiance, PV cell temperature and ambient temperature. The current paper portrays the prediction of power generated from solar PV both in the presence and absence of MPPT controller. In this paper, a linear Deep Neural Network (DNN) model is designed to predict the solar power generated from PV whose performance is compared with state-of-the-art prediction models like Bagged Tree and ARIMA (auto-regressive integrated moving average). Results indicate that DNN is the best model to predict solar power generation both in the presence and absence of MPPT controller having the least RMSE and MSE values for both cases. © 2022 IEEE.