Research On Fault Diagnosis Of Wind Turbine Transmission System Based On Variational Mode Decomposition

Wind power has been the fastest growing renewable energy in recent years. With the gradual commissioning of wind turbines, the equipment which runs for a while has gradually entered the high period of accidents, causing damage of equipment and turbine halting. Wind turbine fault diagnosis can effectively reduce the major accident rate, and provide the necessary time to arrange spare device for maintenance. Therefore, wind turbine fault diagnosis has become an important research direction in the development of wind power. First, the working conditions of doubly-fed induction generator and its effect on the vibration signal is analyzed. The research on fault diagnosis of wind turbine transmission system which is based on variational mode decomposition (VMD) is proceeded with a comparison analysis of empirical mode decomposition (EMD) and local mean decomposition (LMD). The main contents are as follows:An LMD method based on adaptive high frequency harmonic wave and characteristic wave matching is put forward. The proposed endpoint continuation method can effectively restrain non-stationary signal endpoint effect for its consideration of the amount of time on the influence of the characteristic wave selection on the basis of signal characteristic wave matching. On the basis of classification of mode mixing and summary of restraining methods, an adaptive high frequency harmonic LMD(AHFH-LMD) method is proposed. It isolates the high frequency mode from the source signal through a LMD process, and automatically constructs the corresponding frequency and amplitude of high frequency harmonic. It has the characteristics of adaptability and a small amount of calculation. Simulation study and field data detection show that the AHFH-LMD has better inhibition effect for mode mixing than ensemble local mean decomposition and can be used for shaft system fault diagnosis of wind turbine.The variational mode decomposition is introduced into the fault diagnosis of wind turbine transmission system for the first time. In simulation study, it analyses the advantage of VMD in inhibition of two types of mode mixing; Characteristics wave matching method is used to restrain its weak endpoint effect; A determination of the number of mode before decomposition is used based on spectrum, and the impact of penalty factor is analyzed qualitatively. Field data detection shows that unlike the low-pass filtering properties of LMD and EMD, VMD has a similar but different decomposition characteristics compared with wave package analysis, it exhibits a bandpass filtering properties; This feature makes the VMD suitable for shafting fault recognition, and can simultaneously detect the unbalanced fault and incipient wear fault of bearing.In view of the non-stationary characteristics of vibration signal from wind turbine in variable speed operation, we propose a angle domain variational mode decomposition method, which uses computed order tracking technology to make even time sampling signal into the even angle sampling signal for further VMD in angle domain. Research shows that this method is effective to broaden the use scope of variational mode decomposition, and able to handle non-stationary signal over a large area. It not only successfully diagnosed the bearing fault(inner race, outer race and roller) in variable speed condition, but also can diagnose gear slight wearing fault which is easily missing by LMD.A method based on the variational mode decomposition and fuzzy c-means clustering is put forward for state recognition of transmission system. The influence of load variation on the recognition effect is studied. In constant speed operation, the time domain VMD based method can extract the state of bearing characteristics precisely and steadily, and the state characteristic line has good robustness to load change. For variable speed operation condition, angle domain envelope spectrum VMD based method can realize the state recognition of rolling bearing better. At last, a analysis of the influence of load and no-load over characteristics of state line is done, and it points out the importance of the condition identification for state recognition of equipment.Finally, resultant conclusions of the contribution are presented along with promising future researches.