A short-term prediction model to forecast power of photovoltaic based on MFA-Elman

Abstract

Due to the low accuracy and reliability as well as slow prediction speed of photovoltaic (PV) power short-term prediction, a short-term PV power prediction method based on MFA-Elman neural network is proposed. K-means mean algorithm is employed firstly in this method to cluster the weather type, and then EEDM is used to decompose the PV output power on the various data obtained by clustering. Finally, each decomposition subsequence is inputted into the MFA-optimized Elman neural network to predicting the PV output power, which solves the randomness of initial weight and threshold and slow training speed of Elman neural network. The simulation results show that the MFA-Elman model has smaller prediction error, higher prediction accuracy and reliability, faster prediction speed, which can reach the requirements of short-term prediction of PV power.