Research On Power Grid Fault Diagnosis Based On Support Vector Machine

With the development of global energy internet,the interconnection can be enhanced between power grids at all levels,especially with the inserting of intermittent new energy and micro power grids,the dynamic behavior of power grid is becoming more and more complicated.The possibility of complex faults remains,accurate and quick power grid fault diagnosis becomes a more urgent requirement.Aiming at complex faults of power grid,this paper studies the deficiency of learning ability of current expert system,and based on historical failure experience,a power grid fault diagnosis method by using support vector machine(SVM)model is proposed.The main work is as follows:By studying on the intrinsic characteristics of statistical learning theory,understanding deeply the VC dimension and structural risk minimization principle,and expounding the mathematical principle of support vector machine model,two practical algorithms of linear SVM and nonlinear SVM are accomplished in the Visual Studio 2010 platform by adopting C++ language,which provides a the theoretical foundation for the application of SVM in power grid fault diagnosis field.According to the demand and characteristics of power grid fault diagnosis,a method is proposed to diagnose complex faults based on SVM.By training and learning from the historical cases,the "hidden" diagnosis knowledge can be obtained from these complex fault cases.These empirical laws can be used to diagnose the current faults,and the SVM model can be re-trained constantly by new fault events.To make the established SVM models more general,based on the information of protection action and circuit breaker tripping operation from the supervisory control and data acquisition system(SCADA),the input features are set up for the SVM models of bus,line and transformer.And the penalty factor and RBF kernel function parameter can be optimized by genetic algorithm.Based on the developed power grid fault diagnosis system,a complex fault diagnosis module is added by adopting the SVM algorithm,and the function design of five components of the module are discussed one by one.In this way the diagnosis efficiency of complex faults in the power grid fault diagnosis system can be improved.Finally the practicality and effectiveness of the module is verified by three typical complex fault cases.