Fault Electricity Metering Detection using A Rule-based Model Tuned by Particle Swarm Optimization

摘要

High electricity loss in distribution systems is one of major concerns of most electricity utilities. The electricity loss can be categorized as technical and non-technical losses. The fault metering which includes electricity theft and meter defect mainly contributes to the non-technical losses. They are difficult to detect and tackled. An Automatic Meter Reading (AMR) system records electric energy usages of individual customer along with voltages and currents at the meter. These records provide insight how the meter operates and can be used to detect fault metering. This paper proposes a rule-based model which is tuned by the particle swarm optimization algorithm to detect fault electricity metering, using the AMR data. The data used in this study were from 500 AMR meters installed in a 22 kV distribution system. The model has 100% fault metering detection rate with fault positive rate (FPR) at 1.09% and Matthew Correlation Coefficient (MCC) at 94.01%. The benefits of this research include reducing the workload expense for on-site meter inspection and non-technical loss reduction.