Research on Early Warning Technology of Lightning Arrester Defects Based on Multi-stage Information and Bayesian Inference

Abstract

The differences in the probability of occurrence of different equipment and defects lead to the small sample characteristics of the defect of the arrester, which makes it difficult to train an accurate prediction model. It is difficult to identify the abnormal state when the arrester monitoring data does not exceed the limit and increase steadily relying on the arrester monitoring index and threshold to judge the defect. Therefore, a lightning arrester defect early warning method based on multi-stage information and Bayesian inference is proposed. The Bayesian inference algorithm is used to calculate the probability of defect cause categories under different feature quantities. According to the new test evidence, the probability of the defect cause category under different feature quantities is updated to identify the defect cause. The algorithm automatically adjusts the prior probability indicators of equipment defects and causes in the model based on the new detection data and annotation conclusions to ensure the accuracy of defect cause classification. The lightning arrester operation and maintenance data and online monitoring system of a power company is used to analyze and verify the effectiveness and correctness of the method proposed in this paper, which provides effective supportfor the lightning arrester operation and maintenance.