The Fault Diagnosis Of Wind Turbine Gearbox Based On LSSVM Of Parameter Optimization And Improved PSO

The energy crisis has become the most important topic with the development of world economy and the development of new energy is imminent. Wind power is the most abundant natural resources, so it has a wide range of applications and development in the world scope. With the large scale wind generators being put into operation, fault diagnosis problem is carried great weight by people. According to the statistics, gearbox is the part that is easy to break down of wind turbine, which accounted for 65%. In this background, this paper studies the related problems and effective method of gearbox fault diagnosis.The common fault types of wind turbine gearbox are random and are small sample study. SVM theory is a type of machine learning theory for small sample, it is suitable for the characteristics of gearbox fault signal. Therefore, SVM theory is the theoretical basis of fault diagnosis research in this paper.In recent years, SVM theory has become the hot spot in the study of people, SVM classification algorithm is an important application of it. But as a classification algorithm, it has the weakness such as a large amount of calculation and longrunning time, so people put forward LSSVM classification algorithm that is the deformation and improvement of SVM. Experiment results show that LSSVM classifier has better function. We improve the iteration way of inertia weight of PSO, using non-linear decreasing algorithm to get the IPSO. Then we use the concept of black hole and propose the RBH-IPSO algorithm. It can improve the random search ability and balance the global and local searching ability. Finally, we combine RBH-IPSO with SA to get RBH-PSOSA algorithm. It can improve the ability of jumping out of local optimum. Experiments of functions optimization prove the validity and advantage of RBH-PSOSA algorithm. Then we use the new algorithm to optimize the parameters of LSSVM classifier and get an optimized LSSVM classifier.Finally, we use the optimized LSSVM classifier to establish fault diagnosis decision model and apply in fault diagnosis of gearbox. The results show that this method has better accuracy of fault diagnosis compare with the neural network method and not improved LSSVM classifier decision-making model.