Grid-Connected PV System Power Forecasting Using Nonlinear Autoregressive Exogenous Model

This paper presents the power forecasting for grid-connected solar photovoltaic (PV) system using artificial intelligence nonlinear autoregressive exogenous (NARX) model. The NARX model consists of fifty hidden layers. The solar irradiation and temperature data generated from public websites for Udaipur, Rajasthan region, are used to train the NARX model. The corresponding power output of the simulated PV system is selected as target data. The Levenberg–Marquardt backpropagation function is used during the training. The complete system is modeled in MATLAB/Simulink environment. A neural network toolbox is utilized for training the system for future prediction. The simulation study is carried out in four cases. Simulation results show that the trained NARX model forecast power output in an effective manner with a root mean square error of 1.4270 for a one-year prediction.