Fault Intelligent Classification Method And Applications Of Wind Turbines Based On Support Vector Machine

Recently, wind power has been rapid development in our country. Although wind turbine design and manufacturing has been gradually improved, the failure rate of the wind turbine is still high because of the operating load complex, bad condition and other causes. So it is important for safety and reliability of wind power equipment by research and promotion in wind turbine condition monitoring and fault diagnosis.For this situation, this paper researches intelligent classification of wind turbine transmission chain fault as to improving its automatic recognition capability and accuracy, the equipment maintenance technology and management level. The main research is as follows:First, this paper summarized the structure,the failure mechanism and types of wind turbine,and introduced three fault feature extraction methods based on vibration signal analysis, including time domain, frequency domain and wavelet packet feature extraction method and analyzed their classification effect with many practical examples. At the same time, the paper introduced the principle of principal component analysis method, combining the principal component analysis method and wavelet packet feature extraction method to transform several characteristic values into a few comprehensive indexes in order to obtain the most important information.Second, fault classification method of support vector machine is applied to the classification of wind power gear box. The effectiveness of the two classification and multi classification support vector machine is verified by practical cases. This paper designed a fault diagnosis model called WP-PSO-SVM for the transmission chain of the wind turbine based on wavelet packet feature extraction, particle swarm optimization algorithm and support vector machine based on problems of large numbers of training samples and the local extreme value in the neural network with an aim at improving the classification accuracy,and verified its validity by the real data. The thesis also discussed the influence of the three aspects on classification accuracy, including the band-pass filtering, feature extraction and principal component analysis method.Finally,this thesis worked out semi-supervised result of support vector machine based on high dimensional feature space support vector point clustering and similarity threshold discrimination for the restriction in practical applications of the support vector machine,which is a supervised learning method,to achieve the identification of unknown samples and apply it to the intelligent fault diagnosis of the wind power plants.