Study Of Building Electrical Fault Diagnosis Based On Neural Network

Building electrical system is one of the most important part in modern architecture. As more attentionhas been paid in buiding safety, people now has an increasingly need in the reliabilty of building electricalsystem. Some tradional testing methods can not dispose the complex electrical system, because both thediagnosis speed and diagnosis precision can not meet the requiements.In this paper, a new neural network control theory has been proposed, which is based on the feature ofbuilding electrical fault; that is the mapping of sign set and fault set. At the same time, the spike neuron andgenetic algorithm are used to optimize the neural network, which can further up to strengthen the faultpattern recognition capability of Radial Basic Function Neural Network and Ellipsoid Basic FunctionNeural Network. The simulation experiment shows that boht algorithms are effective in fault diagnosis.Based on the simulation experiment, in this paper, one physical real research has been held in buildingelectricity fault text platform MA2067. This platform can simulate some traditonal electrical accidents,such as earth fault, lightning protection, insulation fault. In the end, the proposed fault diagnosis methodshave proved by simulation result and experimental result.