An Electrical Fault Diagnosis Approach Incorporating Volumetric Kalman Filtering and Bayesian Networks

A large amount of multi-source and heterogeneous fault data is generated by different levels of data collection equipment such as metering equipment, low-voltage station monitoring equipment, and camping and distribution fusion system, and in order to improve the accuracy and real-time performance of the electrical fault diagnosis, an electrical fault diagnosis method fusing volumetric Kalman filtering and Bayesian network is proposed. Firstly, a fusion strategy based on timescale alignment is used to construct a dataset within a unified time frame for data from different systems; secondly, volumetric Kalman filtering is used to fuse the dataset to obtain a consistent estimate of the global state of the system, and the abnormal state is identified in time through the prediction and updating steps; then, the basic events are identified through the fault determination results to construct a fault tree and mapped to Bayesian network nodes, and a Bayesian network bi-directional computation function is utilized. Bayesian network bidirectional calculation function to calculate the probability of fault occurrence, reverse inference of the cause of fault occurrence; finally, the experimental results show that, when the ‘Line fault’ occurs, the Bayesian network calculates that the probability of ‘Ground fault’ is 34%, which can be used to determine the fault occurrence of the system. The probability of occurrence is 34%, which can be determined as the cause of fault occurrence, and the F1 score in the comparison experiment reaches 92.27%, which verifies the validity of the fault tree-Bayesian network reasoning about the cause of fault occurrence constructed in this paper, and effectively improves the fault detection and diagnosis capability of the power system, and helps the stable construction of the power system.