The Study Of Clustering Method For Power Load Based On Measured Data

Load clustering has become the base of demand side management,load modeling,system planning,etc.it's of great significance to system analysis,operation,planning.With the improvement of power information system,lots of data has been stored,and the information system which is the base of load data mining has been integrated.User load clustering can help to find the pattern of the users' power utilization.Substation load clustering can reflect the status of the power system,it can also effectively solve the problem of time variation and area decentralization of load.Considering the progress that has been made and some difficulties on clustering analysis,the clustering method of user load and substation load is studied respectively.The main works are as follows:(1)Basic procedures of load clustering are summarized,some classic clustering algorithms and usual cluster validity index are introduced,the merits and drawbacks of these algorithms are pointed out.(2)Some suggestions are made by comparing analysis of different kinds of user load clustering algorithms.The effectiveness of common dimension-reducing methods is studied in consideration of the high dimension load data.(3)Traditional substation clustering is based on the industry structure,the short-comings of traditional component analysis method for substations is pointed out,the new component analysis method which is based on clustering of user load is proposed.(4)The model which considering the substation load characteristic and compo-nent is proposed,auto-weighted k-means algorithm and synthetic index is presented.Case analysis based on load data of some power grid is carried out and results show that the proposed method is effective,and strong robustness of the algorithm is con-firmed based the load data that has been added with noise.