Abstract:
Besides the benefits of electric usage and outage monitoring, a practical metering data of customer meters possibly offers electric utilities a high level of operational effectiveness. With nearly real-time measurements, a meter reading data can help indicate feeder conditions and be utilized as a useful input of feeder model represented in power system simulation analysis. Furthermore, it can be used to correct connectivity errors and improve feeder representations quality as well. This study concentrates on correctness improvement of feeder connectivity using 15-minute intervals voltage metering data. Since they are separately constructed by two different techniques, cluster analysis and SVM, the classification procedures with regard to nature of electric radial distribution are proposed to group and identify customer meters connected to the same feeder of distribution primary. The procedures feasibility is demonstrated by two practical case studies, including an industrial estate and a large-scale business district network. Finally, the experimental results present that the procedures are possibly applied in feeder identification and offer adequately accurate results with at least 86.96 percentage correctness. Nevertheless, the supervised procedure with use of SVM classification is more attractive to apply in the real world as it requires a very short-period monitoring and still has many rooms to be improved in order to enhance the correctness performance
Thammasat University. Thammasat University Library