Short-Term PV Power Prediction Based on FCM-ISSA-LSTM

Abstract

Accurate prediction of short-term photovoltaic output power is an important technical measure to improve the grid-connected capacity of intermittent photovoltaic power generation. A short-term PV power prediction method is proposed in this paper based on FCM-ISSA-LSTM as to improve the obvious problems of low accuracy and poor applicability. Fuzzy-C-Means clustering method (FCM) is firstly used to identify and classify the input data of three types of typical weather for all-weather historical data, and then improved sparrow search algorithm (ISSA) is presented to optimize the structure and parameters of long short-term memory (LSTM) neural network in the training process for higher optimization efficiency and accuracy. The simulation results show that the proposed method has relatively better prediction accuracy and efficiency in short-term photovoltaic power prediction.