An Improved Model for Power Prediction of PV System Based on Elman Neural Networks

High penetration of Photovoltaic (PV) systems is variable resource as challenges to the stability and power quality of electrical grids. Accurate prediction of PV power has been recognized as a way to solve this problem. An improved model for power prediction of PV system based on Elman neural networks is proposed in this paper. Comparing with traditional BP network, the context layer is added in Elman, which make it fewer iterations and less computation. Except from irradiance and temperature, four physical quantities, wind speed, the direction of the wind, humidity, air pressure, are u] sed as inputs in the proposed model. The influence of different weather factors on the accuracy of the proposed model is discussed. The proposed model is applied to experimental data of PV station and compared with BP neural networks. The results indicate that the Elman model with 6 inputs has best accuracy in performance prediction. Moreover, the accuracy is improved especially in the region of low solar irradiance by adding more weather factors. The proposed model can effectively and accurately predict the output power of PV system, which is useful to enhance the safety and stability of the electrical grid.