Prediction of Distributed Photovoltaic Users' Electric Energy Data Based on BP Neural Network Algorithm

Abstract

In this paper, a BP neural network (BPN) algorithm model is utilized to forecast the electric energy data of distributed photovoltaic (PV) users. One month's forward active power and voltage data of PV users are collected. The data was collected every hour. So, 24 data were collected every day. Then a BPN algorithm training model are established, First 20 of the days were considered for training data and final 10 days were considered for testing data. Through simulation experiment, the graph of predicted value and actual value of the forward active power and voltage of distributed PV users are obtained. It is concluded that the BPN algorithm model is an accurate model in predicting PV users' data, and the model is more accurate in predicting voltage than in predicting forward active power. The BPN algorithm model could be an effective model for a short-term forecasting of local small distributed PV station output, and has certain significance for the power management department to formulate energy management and dispatching schemes for stability and safekeeping on large grid after PV grid connection.