Research On Fault Diagnosis Of Direct Drive Wind Generator Bearing Based On Multiclass Cost-Sensitive Support Vector Machine

With the rapid development of wind power industry, wind power generation technology has become a research hotspot at home and abroad. Owning to the operation of the wind turbine of high accident rate, there is a need for high effective fault diagnosis method in order to minimize downtime and maintenance cost while increasing energy availability and life time service of the wind farms, which has a great significance for promoting the development of wind power industry.The basic principle of fault diagnosis is introduced and the common faults and fault mechanism of wind power generator are summarized. The improved wavelet packet feature extraction method is proposed based on analysis of the basic principle of the existing feature extraction, first, analysis of the signal spectrum analysis to determine the length and the position analysis, second, wavelet packet decompose. Comparative analysis of extraction and improved wavelet packet feature extraction by wavelet packet feature extraction, multi feature, show the effectiveness of the improved feature extraction method.Analysis of the standard support vector machine, multi class support vector machine, for the direct drive wind generator bearing sample class imbalance problem, combined with the research and development direction of wind turbine fault diagnosis, a fault diagnosis approach is proposed for cost sensitive support vector. Analysis of mesh optimization algorithm, particle swarm optimization algorithm and genetic algorithm, on this basis, the improved particle swarm optimization algorithm with three parameters for cost sensitive support vector machine optimization. After comparison with the traditional optimization algorithm analysis, proved that the improved algorithm is faster.In the foundation of the contents, construction of direct drive wind turbine bearing fault diagnosis model, through the simulation of the fault samples, analysis ability to identify fault sensitivity, the robustness of the models, the new type sample, demonstrate the excellent performance of cost sensitive support vector machine fault diagnosis model.The need to improve and further study of direct drive wind generator bearing fault diagnosis is summarized. The work of the direct drive permanent magnet synchronous wind generator fault diagnosis is of important reference value.