| At present,wind power generation has been widely used in the actual power system.Due to the impact of harsh environment,wind turbine faults occur frequently.Among them,the misalignment fault is a kind of fault type that often occurs in the transmission system and it has a certain potential,which will affect the quality of power generation and lead to the failure of internal parts in the unit.Therefore,it is very meaningful for the safe and stable operation of wind turbines to use reasonable and effective fault prediction method to forecast the status of equipment and give timely early warning.The thesis analyses the misalignment fault mechanism of the drive system for wind turbines at first.Considering the limitation of single prediction models,a combined prediction model for the misalignment fault is proposed in this thesis.The details are as follows:Based on the 3D model in SolidWorks and dynamic simulation model in Adams for wind turbine drive system,the vibration signal of misaligned fault is extracted.Based on the electromechanical joint simulation established by MATLAB / Simulink combining with Adams,the stator current signal of misalignment fault is extracted.Then,the vibration and stator current signals are simultaneously measured on the misalignment fault test bench.After that,the time domain,frequency domain and time-frequency domain features of fault signals are extracted.Among them,the time-frequency domain features of the vibration signal are extracted by the Improved Empirical Mode Decomposition with the image extension(IEMD).The time-frequency domain features of stator current signal are extracted by Dual-tree Complex Wavelet Transform(DTCWT).The single prediction methods which are respectively Improved Multivariate Grey Model(IMGM(1,n))and Improved Artificial Fish Swarm Algorithm(IAFSA)optimized Least Square Support Vector Machine(LSSVM)are adopted to predict the vibration and stator current signals of misaligned faults.The LSSVM optimized by IAFSA(LSSVM-IAFSA)is regarded as the combiner of the combined forecasting model,the predictive values of single prediction models are taken as the input of the combiner,and the actual kurtosis index is used as the output.The simulation and experimental results show that the combined prediction model has higher prediction accuracy than the single prediction models for the vibration and stator current signals of misalignment faults.In order to reduce the influence of noise in the experimental signal on the prediction effect,the noise reduction method of Singular Value Decomposition(SVD)is determined through the comparison of the noise reduction effect,and the noise reduction order is determined by singular value energy standard spectrum.The experimental vibration and stator current signals are denoised by SVD respectively.After the feature parameters are extracted,the fault signals are predicted by IMGM(1,n)and LSSVM-IAFSA,and the predicted values are used as the inputs of the combiner to get the predicted value of the combined prediction model after noise reduction.The experimental results show that the noise reduction method of SVD improves the prediction accuracy and robustness of the combined forecasting model.Finally,according to the 3σ principle of the normal distribution,the early warning line is set up to realize the early warning of misalignment fault. |
