Based On Bayesian Networks And Wavelet Algorithms Of The Positioning Grid Fault Diagnosis

Accompanied by the growing power system development, the increasing grid size, and more complex structure, the interconnection between different regions is becoming more tightly in the system, which makes the influence of the system itself also will be expanded. In recent years, the worldwide grid blackout, this large power outage brought a negative impact on the significant loss of social stability and economic development. Thus, the fault after the power failure and fault diagnosis restore is most important to the electricity sector.Diagnostic techniques after years of development and researches are becoming more and more mature. Positioning techniques and methods of fault diagnosis is increasing perfect. Several categories are formed such as:Expert systems, Artificial neural networks, Optimization techniques, Petri network, Rough set theory, Fuzzy set theory, Multi-agent technology, and ways based on the fault recorder information and Bayesian network method. However, with the power system running increasingly complex, the development of information technology and the deepening of the reform of the electricity market, further integration are required to information technology in the traditional grid. The formation of a strong unified smart grid has become the main direction of development of the grid. Most of fault diagnosis methods in the past are based on the information of the switch, not taking full advantage of the presence of electrical quantities.This paper focuses on diagnosis of the power grid fault. Electrical fault information and switch information are compressed to reconstruct by Mallat algorithm of wavelet transform principle, more effective fault characteristics are extracted. Then the fault characteristics are fused by the way of Bayesian parameter estimation. Based on learning and reasoning of Bayesian network structure for fault diagnosis, a diagnostic network model is built by the way of Gibbs sampling and dependency analysis, which is the part of Bayesian network learning methods. The way of belief update joint forecast Bayesian network inference is used to make a fault diagnosis. The simulation study on the numerical example based Bayesian network and wavelet algorithms show that the redundant information of the fault can be effectively reduced. The correlation between the fault information is fully studied to improve the accuracy and efficiency of fault diagnosis.