Solar Photovoltaic Forecasting Using ANN Network for Central and Southern Regions of Jordan

Abstract

The use of renewable energy has increased during the last several decades. The most popular renewable energy source is photovoltaic (PV) technology, which uses solar radiation to create electricity. However, a number of variables, such as position, weather, etc., have an impact on the production of PV electricity. It is crucial to control the inherent changeability of PV plants as they expand and contribute significantly to the production of grid power. Predicting solar PV is therefore essential for ensuring efficient and dependable grid functioning. The forecasting model's inputs were historic PV power output data from two solar power installations in Jordan's central and southern regions. The prediction of PV power production in this research takes into account a stacked long short-term memory network (LSTM), a crucial part of the deep recurrent neural network. This model and Nonlinear Autoregressive NARX have been contrasted (Dynamic Neural Network). The outcomes demonstrated equivalent, admirable performance for both the dynamic NARX ANN and the LSTM, with NARX being better. The dynamic ANN can be claimed to be superior to the deep neural network (DNN) for time-based performance modeling of PV systems with varying data.