Fault diagnosis algorithm for building electrical engineering systems based on BP neural network

Abstract

In view of the fact that the traditional diagnosis technology cannot meet the increasingly complex reliability requirements of modern building electrical systems, a fault diagnosis method of building electrical systems based on hyperparameter back propagation neural network (BPNN) is proposed in this study. By constructing a three-layer BPNN model and comparing various parameter configurations in the experiment, the optimal number of hidden layer nodes, learning rate and momentum factor are determined. The results show that the diagnostic accuracy of the optimized model is 90% and 85% on the training set and the test set respectively, and its performance is significantly better than that of the traditional rule diagnosis and expert system in short circuit, open circuit, overload and grounding fault identification. The diagnosis framework constructed in this study effectively reduces the dependence on manual experience and provides technical support for improving the safety and operation level of building electrical systems. The follow-up work will focus on introducing advanced feature extraction technology and online learning mechanism to further improve the comprehensiveness of system diagnosis.