Research On Fault Diagnosis Method Of Wind Turbine Based On Photoelectric Detection And Information Processing

As the capacity of single-turbine wind turbines increases and wind power generation technology continues to improve,the complexity of wind turbines continues to increase,and wind turbines can generally be used in harsh natural environments.Therefore,wind turbines are prone to various failures,causing additional economic losses to wind farms.Therefore,fault analysis and diagnosis of wind turbines is very important,which can maximize the efficiency of production,eliminate safety risks and avoid dangerous situations.In this context,conventional fault diagnosis methods do not fully explore the spatiotemporal distribution characteristics of wind turbine operation data and can not diagnose all the faults of wind turbine at the same time.An intelligent fault diagnosis method based on image texture analysis is proposed.The time-domain signal generated by wind turbine operation is sampled as a two-dimensional grayscale image.Then the image pattern is recognized and the corresponding fault is diagnosed.Due to the influence of wind shear effect and unbalanced blade load,there is usually a lot of measurement noise in the vibration acceleration signal of wind turbine tower.Therefore,the vibration acceleration signals at the top,middle and bottom of the tower are measured by the fiber grating vibration acceleration sensor,and the comprehensive vibration information of the tower is obtained,and the temperature compensation is carried out.In order to reduce the measurement noise caused by unbalanced blade load and wind shear effect,an empirical mode soft threshold decomposition method was designed to de-noise the measured tower vibration acceleration signals,which effectively reduced the influence of measurement noise.Then the processed vibration acceleration signal is introduced into the designed fault diagnosis method to improve the effectiveness of feature extraction and fault diagnosis rate.In this paper,three texture features are used: statistical feature,wavelet feature and Gabor feature,and four classification algorithms are designed: k-nearest neighbor classification model(KNN),linear discriminant analysis classification model(LDA),decision tree classification model(ID3)and ensemble classification tree model(Bagging).Through 10 layered cross validation method to select the relevant characteristics and classification of training algorithm,through the analysis of the precision of different classification algorithm to select the best classification algorithm,analysis the best classification accuracy of the algorithm and the relationship between the characteristic number,the selection of optimal characteristic number and then merge the same type of failure in order to enhance the effect of the diagnosis of classification algorithm.Finally,the simulation experiment is designed under the condition of multiple fault conditions of wind turbine,and the correctness and rationality of the proposed method are proved.Although the above methods are able to make a comprehensive diagnosis of the malfunction.However,it is difficult to solve the problem of diagnosing wind turbine actuator rapid malfunction and obtaining accurate pneumatic torque in actual use.A non-singular terminal sliding adjustable observer is offered.The predicted non-singular terminal sliding surface can effectively reduce the observer shake in the traditional sliding mode,avoiding the problem of the vibration phenomenon that causes undiagnosed.Accurate and without system failure diagnosis,and effectively improve the efficiency of system failure diagnosis.The adaptation rules were implemented in the observer’s design to ensure that the sliding motion mode is not affected by any unknown interference of the system.By entering the fault indicator parameter,the hydraulic pressure reduction of the stepper actuator is changed to an additional fault.Two stacked scroll observers are then used to observe the pitch system,and state estimation and error generation are provided for a limited time.Achieving the results of a quick fault diagnosis demonstrating the validity and feasibility of the proposed method.