Study On Characteristics Between Partial Discharge And Electrical Tree Based On Ultrasonic And Ultra-High Frequency Signals In Polymer

Partial discharge(PD)is one of the main causes of insulation degradation or breakdown in insulation system of power equipment.When local electric field of insulation system is higher than breakdown strength of dielectric,the nonpenetrating PD,and even electrical tree will occur,which seriously limits the service life and reliability of insulation.Therefore,PD and electrical tree have received extensive attention in the field of electrical insulation performance.The current research work of PD mainly focuses on initial voltage,maximum discharge and phase distribution,etc.The defects caused by PD or the development process of electrical tree cannot be effectively estimated.Moreover,the precise location of the partial discharge point needs further discussion,in addition,there is still a lack of in-depth research on estimating insulation life based on the characteristic quantities of electrical tree discharge signals.In the article,based on different discharge types,five types of PD models in liquid/solid composite were built,including needle plate discharge,air gap discharge,suspension discharge,creeping discharge,and sliding flash discharge,and then the electro-acoustic measurement was used to carry out ultrasonic signals and ultra-high frequency signals with the help of ultrasonic sensors and antennas.Then a variety of intelligent algorithms such as wavelet analysis,adaptive multi-stage denoising filtering algorithm,multi-stage wavelet neural network,and quantum genetic algorithm were used to filter ultrasonic signals and ultra-high frequency signals,pattern recognition,and discharge point positioning,etc.And the exploration in this respect was finally systematically verified through theoretical calculation and experimental research.Finally,the characteristics of PD signals generated during the development of electrical tree were used to identify the growth stage of electrical trees in the polymer.It is expected to provide a new method for the location of PD sources and the estimation of insulation operating life.Combined application of multi-intelligent algorithms to filter PD signals,pattern recognition and localization of PD sources.First,an adaptive multi-stage de-noising filtering algorithm was used to filter the ultrasonic and ultra-high frequency signals.Compared with the multi-stage de-noising and horizontal correlation de-noising,the average correlation coefficient is increased by 5.79%and 3.64%,while the mean square The error is reduced by 20.19% and 25.78%;Secondly,in the process of classification and recognition of PD signals,the learning and training of multi-level wavelet neural network was introduced.Through the test of sample space,it is verified that the trained neural network model can realize simple and efficient recognition of discharge patterns;Finally,on the basis of the generalized correlation method,the quantum genetic algorithm is introduced to improve the positioning accuracy of the local discharge source,so that the maximum deviation and the comprehensive error are(0.28±0.16)cm and(0.36±0.19)cm,respectively,and the positioning error less than 4.0%;Therefore,the accuracy of PD signal calibration is improved.Using the precise positioning of the PD source,based on sound pressure simulation and sound wave propagation attenuation characteristics,a quantitative calculation model of ultrasonic signal and apparent discharge was established in the three systems.The systems are that the discharge position and apparent discharge magnitude are known,that the discharge position and apparent discharge magnitude are unknown,and that the unknown discharge position and the known apparent discharge magnitude,respectively.Taking the needle plate discharge ultrasonic frequency of 100 k Hz as an example,it is found that the two calibration curves of the ultrasonic signal and the apparent discharge magnitude of the PD almost overlap when the discharge location is known or unknown,which verifies the accuracy of the quantitative model of the apparent discharge magnitude calibration using the ultrasonic signal.It reveals the correlation between the ultrasonic signal and the apparent discharge magnitude.Multiscale Analysis and BP neural network were used to identify ultra-high frequency signals during the growth of electrical tree,and the length,morphology and conductive properties of electrical trees were effectively characterized.Based on the discharge energy,the identification accuracy of electrical tree characteristics can reach 98.56% after model training by 500 sets of discharge signals by using the characteristic parameters such as mean value,variance,valid value,peak-peak value,skewness,kurtosis and margin,and so on.The identification results of models are in good agreement with the experimental data,which indicates that the established identification model can be applied to the analysis of discharge during electrical tree.And then the relationship between PD signals and the growth process of electrical tree can be discussed.In order to further study the influence of PD on the insulation life,based on the electrical tree dynamics and discharge avalanche theory,combined the apparent discharge magnitude and the electrical tree equivalent circuit,the microscopic characteristic parameters are used to obtain the mathematical model of theory life and running time in polymer,and then its remaining life was calculated.In this paper,the electrical tree-partial discharge detection method was used,and PD characteristics and life estimation of dielectric were calculated and analyzed through combination of theory and experiment,and then the relationship between the apparent discharge magnitude and the life is explained.For polyethylene,when the maximum apparent discharge is 107 p C,compared with the experimental results,the calculation error of the theoretical life of the insulation is 0.29%,and the maximum error of the remaining life is 8.83%,respectively,which are in good agreement,indicating that the life model proposed in this paper can be effectively applied for quantitative calculation and analysis of insulation life.Based on the experimental results of electrical tree discharge,the discharged UHF signal is subjected to threshold denoising,discharge positioning,and pattern recognition.Then,the growth stage and conductivity characteristics of the electrical tree are used as the dynamic health index,combined with the discharge amplitude and frequency,to construct an evaluation model for the dynamic health monitoring of the insulation state.And through the characteristics of the ultrahigh frequency signal of partial discharge in the delayed and rapid growth period,the feasibility analysis of the critical warning before the insulation failure is proposed.So as to provide a new idea for the cost management and safety monitoring of polymer insulation.