Doubly Fed Induction Wind Generator Fault Diagnosis Based on Support Vector Machine
|Keywords||doubly fed induction wind generator fault diagnosiscondition monitoring principal component analysissupport vector machin|
The wind energy is a clean renewable energy which can be widely developed andutilized by most of the counties in the world. Along with this trend, the technologyof wind power has become the focus of current domestic and foreign research. Withthe increasing installed capacity of the wind power generation and the developmentof offshore wind power, a new thorny issue which is to improve and optimize windgenerator fault diagnosis should be solved as soon as possible.This thesis has mainly researched on three aspects which refer to signalacquisition, feature extraction and fault diagnosis, all of which are concerning thedoubly fed induction wind generator on line monitoring and the intelligent diagnosisin early stage of the fault.First, extracting accurately fault characteristics of the doubly fed induction windgenerator will directly affect the early fault diagnosis. Bases on the doubly fedinduction wind generator, this thesis analyzes phenomenon and reasons of inter-turnfault of stator phase winding, the rotor phase fault and the bearing failure, thenobtain the basic equation of generator internal failure. By observing theelectromagnetic characteristics and mechanical characteristics during the operationof brush and slip ring system, the causes of abnormal operation of brush areanalyzed. In order to simulate the different operating conditions, the laboratorialplatform of the doubly fed induction wind generator is built. The real-time currentsignals of doubly fed induction wind generator such as in the states of its normalstate, short circuit between the stator turns, or the rotor fault, in different the outputpower are gathered.Secondly, by using the fast fourier transform (FFT), all the current signalsgathered from the doubly fed induction wind generator can be analyzed in the timeand frequency domains. The theoretic analysis on the doubly fed induction windgenerator reveals that the failure mechanism of doubly fed induction wind generatoris more complex. Therefore it has to apply principal component analysis (PCA) algorithm based on FFT processing to reduce current signal dimensions and removenoises, extracting the frequency features of the doubly fed induction generator underdifferent operating states.Finally, the algorithm of support vector machine (SVM) is applied to analyze thefault intelligent diagnosis of double fed induction wind generator. Regarding themotor fault intelligent diagnosis as a pattern classification decision problem, thefault of intelligent identification can be achieved by the use of the support vectormachine. As the later studies are conducted, the number of the training data becomesmuch larger with the decreasing of the experimental data. Since the large number ofthe training data, another algorithm, namely the vector of the incremental algorithmbased on ball ring is introduced in this thesis. And an incremental support vectormachine model using for the doubly fed induction wind generator fault diagnosis isdeveloped. According to the features of frequency components that extracted in thisthesis, the training and testing result showed that the method is efficacious fordoubly-fed induction wind generator fault diagnosis.