Research On Short-term Load Forecasting Based On Improved Fuzzy Clustering Algorithm

Power system short-term load forecasting is an important part ofpower dispatching.Accurate and quick completion of short-term load forecasting can make the power system run smoothly and economically.Therefore,how to improve the accuracy of short-term load forecasting in power systems has always been the focus of scholars at home and abroad.The traditional short-term load forecast is mainly to forecast the overall load of multiple users,but the load change is closely related to the power consumption characteristics of each user.Predicting the total load will lose the information ofu ser's power consumption behavior,resulting in a decline in the prediction accuracy.At this time,the traditional forecasting method is no longer applicable.Therefore,this parper presents a short-term load forecasting method based on improved fuzzy clustering algorithm.First,the FCM algorithm is used to cluster the overall user load.The traditional FCM algorithm needs to know the number of clusters in advance,and depends on the initial value of the cluster center.Therefore,this paper obtains the optimal number of clusters through the Calinski-Harabasz index,introduces the magnetic optimization algorithm to obtain the initial value of the cluster center,and solves the problem of poor stability of the FCM algorithm.Furthermore,differential evolution and global memory are used to improve the stability and rapidity of magnetic optimization.Then,the load curve clustering is performed under the FCM algorithm that determines the number of clusters and the initial value of the cluster center to obtain load groups with similar load characteristics.Based on several user groups with similar load changes obtained by clustering,the GRU network is used for short-term load forecasting.For each user group,GRU is used as a meta-model to establish a GRU short-term load forecasting model.Finally,the load forecast values of each user group are superimposed to obtain an overall load forecast value.The simulation results show that the GRU network load prediction accuracy using cluster analysis is higher than the prediction accuracy without cluster analysis of user groups.At the same time,the prediction results of GRU compared with BP and Elman neural networks show that GRU is better for load prediction.