Research On Vibration Signal Denoising And Feature Extraction Of Hydropower Units

With the constant adjustment of the energy structure and the need of social and economic development,the demand for electric energy in China is constantly increasing,and the role of hydroelectric power in the power system is also increasingly prominent.In the process of hydroelectric power generation,the hydroelectric unit is at the core position,so it plays a vital role in ensuring its safe and stable operation.Through the investigation that more than 70% of the common faults of hydraulic turbines are reflected through the form of vibration,so the analysis and processing of fault signals is the key to the maintenance and overhaul of hydropower units.However,the running conditions of the hydropower unit are increasingly complex,and the fault feature extraction of the hydropower unit is very challenging.Therefore,the research on the method of denoising and feature extraction of the Hydropower unit under the complex working conditions is put forward.First of all,because the vibration fault characteristics of hydropower units are different from those of general power machinery,it is necessary to have a basic understanding of the vibration fault of the sets and the arrangement of measuring points in order to select a more targeted method for feature extraction of the signals.Therefore,the vibration fault characteristics of the unit are first understood,and then the common faults of the unit are further studied from three aspects of hydraulic,mechanical and electrical.At last,the basic basis of the determination of the vibration monitoring position and the selection of the sensor of the hydroelectric unit is mastered.Then,in order to solve the problem that the original signal and the noise signal are mixed with each other during the acquisition of the hydropower unit,a method to remove the background noise of the hydro-generator set by combining FDM and permutation entropy is proposed.Based on Fourier decomposition,the signal to be analyzed is decomposed into several natural frequency band functions.Then,the permutation entropy is calculated,and the natural frequency band function obtained is screened by the permutation entropy’s sensitivity to noise,and the function satisfying the conditions is reconstructed to achieve the effect of signal denoising.Finally,this method is used for simulation and example verification,and compared with EMD permutation entropy method.The results show that the method based on FDM and permutation entropy has good denoising effect and is suitable for the unit to remove background noise.Then,in order to realize the problem that it is difficult to extract the characteristics of the vibration signal of the unit effectively under the noise interference,this paper proposes a method combining the adaptive stochastic resonance based on genetic optimization algorithm and FDM.Firstly,the genetic algorithm is used to obtain the optimal resonance parameters and improve the signal-to-noise ratio.Secondly,FDM is used to process signals,and the frequency band functions satisfying the conditions are screened and reconstructed.Finally,the fault feature frequencies to be analyzed were read from the Hilbert spectrum to extract the feature.In order to further demonstrate the effectiveness of this method compared with other methods,a comparative experiment was designed with the simulated signal and the measured swing signal of a hydropower unit as the research object.The results show that this method can reduce the interference of noise in feature extraction to a certain extent,and improve the signal to noise ratio to a certain extent in the process of signal processing,and extract the signal characteristic frequency,which can provide a certain value in the actual engineering application.Finally,in order to solve the problem that the weak fault signals of the turbine are easily affected by the noise and the mutual interference between the signals,a method combining CEEMDAN-ICA and singular value differential spectrum is proposed to extract the fault features.Firstly,the signal to be analyzed was decomposed by CEEMDAN to obtain a certain number of modal functions,and the signal was reconstructed with the correlation coefficient as the judgment index.Then,the reconstructed signal and the original signal are constructed into a virtual channel for ICA separation,and the characteristic signal and noise signal are further separated.Finally,the signal obtained from ICA separation is analyzed by singular value differential spectrum,and the singular value difference between the noise signal and the characteristic signal is used to extract the characteristic signal accurately.The simulation and example verification show that the method can suppress the interaction between noise interference and characteristic signal,and is more comprehensive and accurate to extract the early fault characteristic signal of hydropower units,which can meet the needs of practical engineering.