Application And Research In Distribution Network Fault Diagnosis By Rough Set Theory And Neural Network

Along with the completion of the networking project of electrical network of our country, the scale of electrical network enlarges and strengthens constantly. When electrical network breaks down, it will be difficult to judge fault areas. So, this paper attempts with the aid of the Rough Set theory, presenting a new attribute reduction arithmetic as well as an improved value reduction arithmetic based on the combination of the Rough Set theory and the Boolean calculation, and using it to handle the reduction progress of the decision table including all kinds of fault cases which is established by considering the signals of protection relays and circuit breakers, and last forming a model of fault diagnosis overall regular knowledge base. Since looked from the example results, it has basically met the practical application design requirements, but it still has a flaw, which is that the fault-tolerance ability of the Rough Set theory is also insufficient ideal. When core attributes are polluted by noise, it may arouse error judgment. Therefore in order to enhance the fault-tolerance ability of Rough Set theory, unify Rough Set theory and Neural Network, construct an intelligent mixture system of Rough Set theory and neural Network, which fully develops the reduction ability of Rough Set theory and the classification ability of Neural Network. First, use Rough Set theory to form a simple rule collection from the reduction of original data, and then neural network carries on the study training through using the simple rule collection. Like this it can both reduce the neural network training time, and enhance the diagnosis accuracy. Finally, carry on the test simulation to 21 test sample data, the simulation results indicate that this algorithm has high fault-tolerance ability and high diagnosis correct rate (100%), which has satisfied the requirement of distribution network fault diagnosis very well.