Research And Application Of Key Technologies In Electricity Customers Load Pattern Extraction

With the development of smart grids, load monitoring devices are widely used indistribution networks and more and more customers’ time-varying powerconsumption data flow into data warehouse of distribution company, which providingthe base for obtaining customers’ load patterns from their consumption behaviors data.The research on technologies of load pattern extraction is increasingly a hot topic,which is important for improving tariff structures and rates, direct load control andeven for developing new marketing strategies or making demand side response policy,a lot of methods were proposed or introduced to do this task. Cluster technology is apopular measure to extract load pattern and most literature focused on clusteralgorithms, other issues in the cluster procedure are few mentioned in those studies,so they will be studied in this thesis.The influence on clustering result of load curves using different clusteringalgorithms with different data normalization methods is studied and the matchingrelations between data normalization methods and clustering algorithms are obtained,experiments show data normalization methods have different influences on clusteringresults using the same clustering algorithm. Then several distances are used in thek-means algorithm for clustering load curves and their influences on the clusteringresults are analyzed, the result shows distances have different influences on clusteringresults using the same clustering algorithm. The comparison results indicate that thechoice of distances is an important issue in power load pattern extraction usingclustering techniques and a suitable distance may improve the accuracy of miningalgorithms. Finally, the application of load patterns is applied to non-technical lossesdetection, the detection flow is presented and the program is developed.C++and MATLAB mixed programming technology is used to implement thosealgorithms and applications, especially the k-means algorithm. The works show thatthe key technology of load pattern extraction will be helpful for the decision of powersupply companies in operation business.