Short-term Prediction Of Photovoltaic Power Based On Improved BP Neural Network

Because of its advantages of clean,safe,reliable and low cost photovoltaic solar power generation has become one of the main green and clean energy in China.Photovoltaic power is affected by solar irradiance,temperature,humidity,and other factors,resulting in high random fluctuation of power size.When it is connected to the grid system,its callability will be reduced.The security and stability of the grid will be affected.Aiming at the problem of photovoltaic power prediction,association rules(Apriori)algorithm,and gray Wolf(GWO)algorithm are analyzed and studied in depth.The improved Apriori algorithm and GWO algorithm are used to optimize the BP neural network,and the improved BP neural network algorithm is used to make a short-term prediction of photovoltaic power.The main research contents are as follows:As the BP neural network is more sensitive to the input feature values,it will lead to bias in the prediction results when the input feature values are less correlated with the prediction results.To address this problem,the prediction feature value extraction based on the Apriori algorithm is proposed.The Apriori algorithm is less efficient,and in order to obtain the prediction feature values quickly,the Apriori algorithm is improved by converting each item set in the algorithm into a binary code,and compressing the transaction set binary code,and reducing the time complexity of the algorithm by replacing the set operations between item sets with bit operations.The Apriori algorithm’s operational efficiency is improved.The improved Apriori algorithm is used to mine the association rules of various meteorological factors in PV data,and only the irradiance,temperature and humidity,which are more correlated with PV output power,are selected as the input feature values for the prediction model of PV power prediction.This provides a new idea of input parameter screening.To address the problem that BP neural networks are prone to fall into local extremes,the GWO algorithm is introduced to optimize the parameters of BP neural networks,and the GWO-BP neural network PV prediction model is proposed,based on the GWO algorithm’s better able to find optimal solutions,which solves the problem that BP neural networks are prone to fall into local extremes in training and lead to reduced prediction accuracy.Then combined with the improved the proposed GWO algorithm can find the optimal solution to solve the problem that the BP neural network is prone to fall into local extremes during training,which may lead to the reduction of prediction accuracy.The results of the experimental analysis show that the combined GWO-BP neural network PV short-term prediction model has higher accuracy and more reliable prediction results,and has broad application prospects.