Output Power Prediction of Photovoltaic Module Using Nonlinear Autoregressive Neural Network

Abstract

Precise prediction of generated output power plays an essential aspect in many sectors of power system like in solar energy sources which is the current topic being discussed on. It is of great role in every system but the prediction of output power for solar energy system is a tough task due to the influence of numerous parameters and fluctuations. Photovoltaic module being main part of the solar power system has many factors which can influence its performance where temperature is paramount. In this paper, the output power of a certain photovoltaic module was estimated under change of temperature and prediction of its future output power was done referring to the estimated power by nonlinear neural network. Both monthly and annual predictions were done through training, validation and test processes. The best monthly performance was achieved equal to 0.9743 at epoch 3 with regression values for training, test and validation all equal to 0.74274, 0.7166, 0.83388 and 0.75604 respectively. While the best annual best performance was achieved equal to 0.10284 at epoch 6 with regression values for training, test, validation and all equal to 0.76576, 0.73665, 0.71678 and 0.75386 respectively. Finally, results showed that nonlinear autoregressive neural network was good and effective for prediction of the photovoltaic module output power.