Research Of Wind Turbine SCADA System’s Fault Classification Method Based On Improved SVM

Currently, wind turbine remote monitoring system adopts SCADAsystem. However, because of SCADA’s hardware structure and externalenvironment, sometimes one fault causes a series of fault at the same time.Now, SCADA system displays only a series of fault alarm signal and notshow the extract fault, so that maintainers can’t find the original faultsource and have to judge the fault source one by one according to theirexperience, which waste large-scale manpower and material resources andadd the downtime. Therefore, it is very important for us to find the truefault rapidly, so as to reduce servicing time and increase effective run time.Based on this background and combined with the SCADA program whichis under development by the wind power company of Shanghai WindEnergy Group, the paper proposes a new method which builds a smartfault expert database through training and learning by algrithm to identifythe fault intelligently. The main research contents are as follows:In view of the above question, we give a lot of fault informationwhich happen at the same time as fault feature for training and treatexperience value as classification categories, then build a better SVMclassifier to realize the automatic classification and identification of thefault and build fault experts database, so as to shorten searching fault timeand increase the benefit.On account of the critical parameterscandgin SVM classifier, weadopt QGA to find the bestcandg. In order to solve the slow convergencespeed and easily trapped in local minima of QGA, dynamic quantum revolving door is introduced to adjust the quantum angle according toprocess dynamics. The strategy adjustment is that setting a larger angle inthe early evolution, then decreasing the angle with the increase ofevolutional generation. Meanwhile, we introduce quantum mutation.Compared the traditional QGA, the improved QGA（IQGA）gives moreefficient result by the typical complex multimodal function test.The improved QGA is used to optimize SVM parameterscand g,which will be applied to fault diagnosis. Through MATLAB simulationexperiment, the improved SVM not only increases accuracy rate thantraditional method, but also reduces elapsed time greatly.We carry out experiments with the SCADA monitoring system in awind field, and program the interface by VB to realize classificationrecognition of the improved SVM. Experiments not only validate thepracticability of the improved method, but also prove the algorithm’sapplication value in engineering.