Photovoltaic output prediction of regional energy Internet based on LSTM algorithm

Abstract

Abstract By the huge development of large scale and modular photovoltaic power generation, accurate photovoltaic (PV) output prediction can help PV power station, scheduling department and power system operate safely and economically. In the process of PV output prediction, the data density is large, and the output data is relatively regular. Therefore, this paper considers the use of long-termed and short-termed memory neural network algorithm to optimize the problem of algorithm gradient vanishment in recurrent neural network, and complete the output prediction of PV power in the regional energy Internet on the basis of historical output data. In this paper, LSTM algorithm is used to analyze the historical output data of PV stations in an industrial zone of a certain city. It can be found that LSTM algorithm has good adaptability for short-term PV output prediction, which can meet the needs of application.