Feature Extraction And Fault Analysis Of Hydropower Units Based On Improved Ensemble Empirical Mode Decomposition

Under the development trend of complex and changeable grid structure,largescale hydropower units,and complicated operating conditions,higher requirements are put forward for the safety and stability of hydropower units.Studies have shown that the probability of vibration fault of hydropower units is greater.The vibration fault of hydropower generators is complex,gradual and irregular,and the vibration signal is non-stationary and non-linear.It is of great significance to realize the fault diagnosis of the unit and provides a new idea for the fault diagnosis method,which is to separate the signals containing a large amount of fault information from the complex operating environment,carry out effective feature extraction,adopt a certain pattern recognition method to distinguish the state of the unit,build fault early warning indicators on this basis,and establish fault reasoning models based on fault early warning indicators and hydropower unit fault mechanisms.The main research contents of this paper are as follows:(1)Aiming at the end effect problem in ensemble empirical mode decomposition(EEMD),the reasons for the end effect are analyzed,and an empirical mode decomposition end effect suppression method based on support vector regression(SVR)optimized by grid search(GS)and window methods is proposed.This method uses an optimized algorithm to obtain SVR parameters to ensure the validity of the continuation waveform.The combination of SVR and window methods ensures that the end point of the signal is converged after SVR extension,while avoiding the waveform change caused by directly windowing the signal.Compare the end effect suppression effect with original EEMD,EEMD based on GS-SVR continuation,and EEMD based on window methods.Introduce index of orthogonality and regional equalization index to compare the decomposition effect of the simulated signal and the measured vibration signal of the hydropower unit.The analysis result shows that the method proposed in this paper has the best end effect suppression effect and has a good application effect in actual engineering.(2)Aiming at the problem that the noise environment of the vibration signal of the hydropower unit interferes with the essential characteristics of the signal,a denoising algorithm for the vibration signal of the hydropower unit based on improved EEMD and singular value decomposition(SVD)is proposed.After the signal is processed by the improved EEMD,the high-frequency noise components are separated and eliminated,and the remaining components are denoised by Hankel-SVD.Finally,the remaining components are accumulated to achieve denoising.The simulated signal and the measured signal are analyzed and verified respectively,and the results show that the method proposed in this paper has a better denoising effect.(3)Aiming at the problem of the non-stationary and non-linear characteristics of the vibration signal of the hydropower units and the difficulty of intuitive identification of the signal characteristics,this paper proposes a feature extraction of vibration signal of hydropower units based on the improved EEMD and approximate entropy,and on this basis,a warning method for vibration faults of the units is proposed.In the process of signal feature extraction,the kurtosis-standard correlation coefficient index is used to screen the effective IMF components,which improves the efficiency of the algorithm.In the fault early warning research,the automatic working condition clustering algorithm based on the maximum and minimum distance method is introduced to improve the reliability of feature recognition in multiple working conditions.Calculate the mean value of the eigenvectors under different working conditions to obtain the health reference value under the working conditions.The distance between the characteristic value of the sample to be tested and the corresponding health reference value(JS divergence)is the fault early warning indicator,and when the alarm threshold is reached,the system sends out an alarm signal.The simulated vibration signal of the rotor test bench and the measured vibration signal of the hydropower unit are analyzed to verify the effectiveness of the proposed method.(4)This paper provides a new idea of fault analysis,which effectively combines signal feature extraction with fault mechanism analysis,and establish a typical fault reasoning process for hydropower units.Start with the analysis of the fault component structure,analyze the fault mechanism,summarize the fault characteristics,and determine the associated early warning indicators of the fault,finally combine the early warning indicators and other feature combinations to determine whether the unit has a corresponding fault.