Research On Propagation Characteristics And Denoising Of Partial Discharge Signal For Hydro Generator

With the development of the power industry and the advancement of technology,the generator is developing in the direction of high voltage and large capacity.The 24 kV,1000MW giant hydro generator is about to put into operation,and hydropower engineering in China has entered the key period of transition from the project construction to operation management.At the same time,as Energy Internet construction and energy restructuring,the capacity of intermittent new energy has grown rapidly,such as wind power and photovoltaic power generation,thus the task of peaking and frequency modulation of hydroelectric generating units is more arduous,while the operation mode and operation condition of the units is getting worse.Generally,the safe and stable operation of large hydro generators is becoming increasingly important,which poses unprecedented challenges for condition monitoring.Partial discharge(PD)monitoring is one of the important means to improve the condition monitoring level and operational safety of hydro generator,since it can detect insulation defects and fault signatures by detecting PD pulses originating from stator windings,and diagnosis defects and prevent faults of stator winding insulation by adopting pattern recognition and fault diagnosis technology.At present,a lot of research has been carried out on PD monitoring of high-voltage equipment at home and abroad,and has achieved fruitful application practices and results.However,with the deepening of engineering practice,the problem of PD monitoring system has become increasingly prominent,including measurement incomprehensive,unable to locate the PD source and difficult to calibrate the PD charge,and PD monitoring faces new challenges.Therefore,in-depth research and exploration should be conducted on current general concerns,and new scientific and technical problems that need to be solved should be tackled.Focuses on the problems and challenges exposed in engineering practice,this paper conducted in-depth research on several technologies in PD monitoring of large hydro generators.Firstly,the PD monitoring should be carried out based on the in-depth study of the insulation structure and failure mechanism of the stator winding,combined with the study of the PD mechanism and characteristics.To this end,the characteristics of the stator winding are analyzed in depth,and an identification method for the winding connection style is proposed.After that,combined the damage procedures of stator winding insulation caused by thermal,mechanical,environmental and electrical,the stator failure mechanisms are studied systematically.Then,combined with engineering practice requirements and the latest research progress,this paper extends the concept and connotation of PD and presents a new definition of PD,and presents the statistics of mechanism,location,pulse spectrum and hazard of each type of PDs as well.Finally,the correspondence between insulation fault and PD is summarized.Secondly,the realization of accurate monitoring of PD is based on obtain PD signals comprehensively and effectively.The stator winding of large hydro generator is a complex distributed structure,and different degrees of attenuation and deformation will occur during the propagation of PD signal from the discharge point to the detection point.The research on the propagation characteristics of PD signal in the stator winding can provide a theoretical basis for reasonable selection of detection points and detection frequency bands,and for effective suppression of noise and interference,and for localization of PD source location and charge quantity calibration as well.According to the PD mechanism and characteristics summarized in this paper,an experimental research scheme of stator winding propagation characteristics is designed.Through the destructive experiment performed on hydro generator onsite,the propagation mode and cross-coupling phenomenon of the PD pulse in the stator winding are analyzed in detail,combined with stator bar conduction,slot winding coupling and end winding coupling experiment,the analysis of the shortcomings and contradictions in the current research with detailed experimental results is performed.Finally,toward general modeling method research,an insulation parameter identification method based on conventional measurement is proposed.Thirdly,the location of the PD source and the calibration of PD charge are all dependent on an accurate and effective stator winding model.Therefore,it is necessary to establish a wide-frequency model of the stator winding that is accurate enough to cover the spectrum of PD signal.This paper proposes a multi-segment multi-conductor transmission line cascade model based on multi-conductor transmission line theory.From the theory of multi-conductor transmission line,the scientific problems and technical problems faced in the modeling of stator bars/windings are studied.Aiming at the non-uniformity of the stator winding structure,a multi-segment cascade model is proposed to homogenize the model,and the corresponding segmentation model cascade rules and methods are studied.A method of solving the wide-frequency parameters is proposed by combining finite element method and reluctance network method,to deal with the frequence-dependent of the model parameters caused by the skin effect and the proximity effect.The equivalent permeability is introduced to describle the effect of the core lamination structure on model parameters.Further,the problem in simulation of frequency-dependent parameter model is analyzed,and the spectrum response method is proposed to reduce the time complexity of model reuse.On this basis,the design and development of general modeling software based on stator winding connection diagram is presented.Afterward,the variation characteristics of polarity of PD pulse is studied in view of theoretical.Finally,since the existing of serious noise and interference during online PD monitoring,if fast and accurate denoising cannot be achieved,it will not only lead to false positive or false negative indications,but also reduce the credibility of the monitoring results.Moreover,the pattern recognition theory based on the discharge distribution will be useless,thus the users can not identify the type of PDs accurately.To this end,based on the existing research results of the research group,the research on a rapid denoising method is carried out.Based on the characteristics of noise and interference,this paper proposes a hierarchical step-by-step signal denoising method.Firstly,a time domain reduction method combining mathematical morphological filter and spectrum correction is proposed to deal with discrete spectrum interference(DSI)with large amplitude and long duration.As to white noise suppression,inheriting the wavelet threshold denoising used by the research group,a modified method is proposed to deal with the problems of the uncertainty of the number of decomposition level and the sensitivity of the threshold to sampling parameters.Since the number of decomposition level depends on the signal frequency distribution,an effective lower limit frequency identification method based on the signal effective cumulative energy distribution is proposed.To solved threshold randomness problem,an adaptive threshold calculation method based on iterative filtering is proposed.In this method,the iterative filtering idea,the modified sliding energy window method and the hypothesis test are introduced and combined to perform threshold estimation automatically and adaptively.The generator neutral PD monitoring system developed by our research group has been successfully applied to more than forty large hydro generators in the power plants of the Three Gorges,Gezhouba,Shuibuya,Geheyan and Sanbanxi.The on-site monitoring results verify the effectiveness and practicability of the denoising method proposed in this paper.And the system effectively promotes the condition-based maintenance and the construction of intelligent power stations.