Early fault detection of distribution network based on the ring law of large-dimensional matrix

Abstract

When faults in the distribution network occur, they usually last for a period of time, but if cannot be detected and eliminated in time, the faults can easily expand or even cause systemic collapse. Therefore, early fault detection of distribution grid is very important for ensuring the safe operation of the system. The existing fault diagnosis approaches based on the physical mechanism model have great limitations. This paper presents an early fault detection approach based on large-dimensional matrix ring law. We first construct a spatio-temporal data matrix by using the data from the spatial multi-node of the distribution network. The moving data window method and the ring law are used to track the data behavior, and the mean spectral radius is used to characterize the data behavior. According to the test results of the IEEE standard bus system, our approach is robust to system random fluctuations and measuring errors and can identify abnormal conditions at the earliest stage of failure.