Study On Fault Diagnosis Method For Power Network Based On Combinational Corss Entropy Algorithm

Power network fault diagnosis method using optimization technique is being taken more seriously because of precise theoretical basis and favorable practicability. However, there are some problems solved necessary in optimization technique: the diagnosis model building distcribing fault adequately; the real-time of optimization algorithm; the tolerance of fault diagnosis system with incomplete and distorted warning messageFor transmission network, paper builds the model of fault diagnosis based on relationship and logic between relay protection and circuit breaker, and weight factor in model is used to avoid the influence for object function when protectors refuse operation. Combinational cross entropy algorithm(CCE) optimizes model to identify fault element(s). After simulating for various fault cases in test power network, proposed method can identify correctly falut elements, and when refusing operation, unwanted operation or multiply faults in protectors and circuit, proposed method also can attain the correct conclusion, which proves that the method has tolerance.For distribution network, the estabilish of fault location model is based on fault current information acquired by FTU in segment swithes. By introducing the direction information of fault current, the traditional model can be improved to solve the fault location problem of distrubtion network with distrubtion generation(s). CCE optimizes model to locate fault section(s). The fault cases extracted from IEEE33 and multi-soures network include single-fault, multiple-fault and little distortion information; method can locate correctly fault section(s) and have tolerance.For solving the problems in power network fault diagnosis using optimization technique, paper does a research, both transmission network and distribution netork, about model building reflecting faults availably. The key point is to propose a new idea that CCE is used to optimize fault diagnosis model. CCE have a fast convergence rate, high precision and good tolerance which indicates that it can also converge to the optimization solution even under incomplete and distorted warning message. By comparing the CCE convergence rate with GA and PSO, paper proves more that CCE satisfies the accuracy and real-time in power network fault diagnosis field.