Interval-based Solar PV Power Forecasting using MLP-NSGA II in Niroo Research Institute of Iran

This research aims to predict PV output power by using different neuro-evolutionary methods. The proposed approach was evaluated by a data set, which was collected at 5-minute intervals in the photovoltaic laboratory of Niroo Research Institute of Iran (Tehran). The data has been divided into three intervals based on the amount of solar irradiation, and different neural networks were used for predicting each interval. NSGA II, a multi-objective optimization algorithm, has been applied to search an appropriate set of weights, which optimized the neural network with two or more conflicting objectives. The MLP-NSGA II algorithm provides better results with the Mean Square Error (MSE) and correlation coefficient (R2) of 0.01 and 0.98, respectively, in comparison with Linear Regression, MLP, and MLP-GA. By the way, obtained results show that the precision of prediction models would be improved by reducing input parameters’ time intervals.