Research Of Rough Set And RBF Neural Network Fault Diagnosis Method About Distribution

With the constant expansion of the grid and raise the level of integrated automation,distribution network after a failure, more and more information into the dispatch center. Sincea huge amount of information,as well as their uncertainties and inaccuracies,and even someimportant information is missing,may lead to inaccurate diagnosis or even wrong.To solve the above problems,this paper rough set theory and the combination of RBFneural network fault diagnosis method of the grid. First,the use of rough set theory isincomplete and inaccurate knowledge of data processing methods,the dispatch center into alarge number of uncertain data table and make editorial decisions about Jane, has beenminimal reduction of nuclear properties and attributes,then these properties as RBF neuralnetwork training sample input, the network training,the use of trained neural network module,the power system fault diagnosis, determine the fault location.In this paper, the use of RBF neural network training, because it is compared with thetraditional BP neural network: The BP do not have the best approximation and the globaloptimal performance characteristics; and simple structure, training speed, suitable forlarge-scale grid. Paper, rough sets and neural networks to optimize the combination of theadvantages of using two methods, which greatly improves system fault tolerance.Articles using vs2005development environment, using vc++language part of rough setreduction program, and then in vc++environment called matlab neural network toolbox tocreate a simple system module to realize the fault diagnosis process.Finally, this paper, a numerical example, be diagnosed by step, to determine the faulttype and fault diagnosis with traditional methods of power compared to verify the correctnessand practicality.