Photovoltaic power prediction by cascade forward artificial neural network

Abstract

Rapid development of Photovoltaic (PV) power and its efficiency is resulting in significant power supply, global warming mitigation and economic benefits. However the power of PV station is highly affected by weather conditions. Therefore forecast of PV power is essential for planning and operation and management of power system. This paper describes the training, validation and application of cascade feed forward back propagation artificial neural network to predict PV power of 3 KW station installed at North China Electric Power University in Beijing. The input parameters for cascade feed forward neural network are meteorological parameters and the target parameter is real PV power to train the proposed network. Solar radiation, Temperature, Humidity, Wind speed are chosen as input parameters to train the model. Photovoltaic power for two days ahead is predicted with a well-trained cascade feed forward neural network. The results show that the proposed network can precisely compute and forecast the PV power of the test data with accuracy.