Research On Intelligent Diagnosis Method Of Vibration Fault In Wind Turbine Transmission System

As the environmental pollution and resource crisis increasingly aggravating, the renewable energy development has been attracted more attention. As a sustainable clean energy, the total installed capacity of wind energy has been rising in recent years, and the relative tertiary industry such as maintenance, monitoring and fault diagnosis will be a new researching direction. Furthermore wind turbines are usually worked in the poor environment, and it is difficult to find and maintenance the fault in time. Therefore the condition monitoring and fault diagnosis of the wind turbine has significant meaning. Some research work shows that the transmission system is the most easily suffered the vibration fault. Due to the vibration signals of the wind turbine transmission system belongs to the non-stationary signal, meanwhile the cibration signal differs in the different parts or different working conditions, so it is hard to realize the accurate diagnosis and in low efficiency, using the traditional spectrum analysis fault diagnosis method. Thus this paper selects the transmission system as the researching objects and based on its characteristics to summarize as following:Firstly, this paper analysis the basic working principle and fault characteristics in wind turbine transmission system, including wheel, low speed shaft, gear box, high speed shaft and generators, and determining gear and bearing damage is the main cause of vibration fault in transmission system, and then choose the measuring point to analysis the vibration signal.Secondly, aimed at the non-stationary and the diagnosis method is constrained by work condition or different components characteristics of the vibration signal, this paper put forward the non-liner state estimate technology (NSET) in the vibration signal diagnosis of wind turbine transmission system. The NSET method belongs to the data statistics model, it means making a memory data in the healthy condition, and then use the NSET model to compare the actual data and healthy data, meanwhile making fault diagnosis. This paper use Matlab simulation software to build NSET model and make fault diagnosis, at last, analyze the diagnosis result, and verify the feasibility of NSET model.Thirdly, aimed at lack of theoretical basis and hard to get the parameters characteristics of the NSET method, this paper propose an improved NSET method, which combine the support vector regression (SVR) and NSET method. This improved method is based on the theory of support vector machine, according the SVR regression function model, comparing the actual data and healthy data, meanwhile making fault diagnosis. Also use Matlab simulation software to build regression function model and make fault diagnosis, at last, analyze the diagnosis result, meanwhile compare with the traditional NSET method and verify to get optimal of improved method in diagnosis result and operational feasibility ways.Fourthly, put the NSET method and improved NSET method into the software of wind turbine monitoring and diagnosis, and apply it to practice, and then verify the feasibility of NSET and improved NSET method in practice engineering.Finally I make a summary of all the study contents of this paper, and make an expectation of improved working in the future and the development prospect of NSET.