Research On Vibration Fault Diagnosis Of Hydropower Units Based On Adaptive Successive Variational Mode Decomposition

With the implementation of the "double carbon" strategy,it has become a consensus to vigorously develop hydropower and actively promote the construction of hydropower bases.The safe and stable operation of Hydropower Units is crucial to ensure the performance of power generation system,power generation efficiency and plant safety.With the increasing demand for electric energy,the output requirements of hydropower stations are gradually increasing.However,hydropower units are in complex and harsh environments such as high temperature and high pressure,variable load and variable speed for a long-time during operation,and their safety problems are increasingly prominent.Research shows that 80 % of faults in hydropower units are reflected in vibration signals.Therefore,the accurate and efficient fault diagnosis method based on vibration signal is of great significance for timely understanding the operation state of hydropower units,ensuring the safe operation of hydropower units and giving full play to the comprehensive benefits of hydropower stations.In view of this,this paper studies the three aspects of vibration fault diagnosis of hydropower units:data preprocessing,feature extraction and fault recognition,in order to improve the accuracy of vibration fault of hydropower units.The specific research contents are as follows:(1)Hydropower units are in harsh operating environment for a long time.If the collected vibration signals are analyzed directly,the noise contained in the vibration signals will seriously affect the accuracy of fault diagnosis.The wavelet denoising method is widely used and has good denoising effect.However,the wavelet denoising method contains many parameters and lacks an effective basis for parameter selection.It is difficult to realize the optimal denoising of vibration signals of hydropower units.In order to reduce the noise interference,this paper proposes a denoising method for vibration fault signal of hydropower units based on adaptive wavelet denoising.The minimum value of Permutation Entropy(PE)is used as the fitness function,and the Whale Optimization Algorithm(WOA)is used to optimize the wavelet denoising parameters.The experimental results show that the proposed method can adaptively obtain the optimal parameters of wavelet noise reduction and improve the noise reduction performance of vibration signals of hydropower units.(2)In order to improve the sensitivity of vibration fault feature extraction results of hydropower units,the Successive Variational Mode Decomposition(SVMD)method is introduced into the feature extraction process of vibration fault of hydropower units,and a feature extraction method of vibration fault of hydropower units based on SVMD is proposed.Considering that the penalty parameter α in SVMD method has an important influence on the decomposition results,a new fitness function is constructed by combining kurtosis index and permutation entropy theory,and the penalty parameter α of SVMD is optimized by using the Salp Swarm Algorithm(SSA).The effective extraction of vibration fault characteristics of hydropower units is realized by combining the Synthetic Detection Index(SDI)and the T-distributed Stochastic Neighborhood Embedding(T-SNE)methods.The experimental results show that the vibration fault feature extraction method proposed in this paper can effectively extract the sensitive characteristics of vibration fault signals.(3)In order to improve the accuracy of fault diagnosis of hydropower units,an adaptive parameter optimization Support Vector Machine(SVM)fault classification method is proposed.The maximum classification accuracy is used as the fitness function,and the whale optimization algorithm is used to optimize the parameters of SVM.The recognition accuracy of four SVM models with different kernel functions is compared and analyzed.The experimental results show that the proposed method can effectively improve the fault identification accuracy of SVM,and the linear kernel and polynomial kernel SVM can obtain higher fault diagnosis accuracy in the vibration fault identification of hydropower units.