Bad Load Data Identification And Correction Method Based On Improved FCM Algorithm

With the continuing development of advanced metering infrastructure(AMI),load data in power system are easy to be measured.As one of the basic data of power system,the load data has an important influence on the analysis and operation decision of power system.Due to various reasons,the measurement data from AMI always contains bad data.How to identify and correct these bad data becomes one of the focus in power system data management.Aiming at the processing of power system load measurement data,a method of identifying and correcting bad data based on improved Fuzzy C-Means Clustering(FCM)algorithm was proposed in the thesis,which involves the improvement of FCM algorithm,the bad load data identification and correction methods,the completed work is as follows:When the FCM algorithm is used to detect and identify bad data,the number of load sample clusters needs to be given in advance,and the initial clustering center is often given randomly.In order to overcome this shortcoming of FCM,a hill-climbing method was introduced to figure out the clustering number of the clustering samples,and determine the initial clustering center and initializes the membership degree matrix.The improved FCM algorithm can improve the clustering efficiency,and avoid the objective function into a local minimumIn terms of bad load data identification and correction,aiming at the problem that the data to be detected is sensitive to the feasible region matrix of a particular class and insensitive to the feasible region matrix of the whole sample,an improved method for measured bad data identification and correction was proposed.This method combines the improved FCM algorithm to improve the efficiency of identification.By finding the threshold range of different types of historical normal measurement data,different types of feasible region matrix is formed.On the basis of the improved method,new measurement data could be clustered and identified,the bad data could be corrected by the method based on load characteristic curves.At last,the effectiveness of the proposed method is verified by a practical example based on the actual power grid.