Research For Aging Characterization Method Of Transformer Oil-immersed Paper Insulation Based On Ternary Chemical Indicators In Oil

Oil-paper insulation(composed of insulating oil and paper insulation),as the main insulation form within the transformer,gradually ages under the action of thermal stress during the long-term service of the transformer.As a result,the electrical insulation performance and mechanical support performance of the insulation system continue to decline,which can easily induce partial discharge.The insulating oil can restore its insulating properties through filtering/changing oil.However,the performance degradation of paper insulation is irreversible.Therefore,the industry generally believes that the aging state of paper insulation is the key to determining the insulation performance of transformers.Since it is difficult to directly measure the intrinsic parameters of paper insulation of on-site transformers(usually requiring sampling from the hood during a power outage),the aging state of transformer paper insulation is usually assessed by indirect methods such as chemical indicators method.Furfural,methanol and ethanol chemical indicators are produced by the aging of paper insulation and are dissolved in the insulating oil.Moreover,this method has the advantages of online sampling and in-situ rapidity,and has greater potential for practical engineering applications.However,different chemical indicators have different generation characteristics,and their sensitivity to reflect the aging degree of transformer paper insulation is not the same.It is impossible to comprehensively,accurately and reliably evaluate the aging state of paper insulation throughout the life cycle of a transformer by using only a single chemical indicator.Therefore,it is more scientific and reasonable to use multiple chemical indicators in oil to jointly characterize the aging state of transformer paper insulation.In view of this,this study combines simulation and experimental analysis to explore the generation mechanism of ternary chemical indicators(furfural,methanol and ethanol).Moreover,its concentration change rules and loss characteristics are analyzed.In addition,the interpretable method based on subjective and objective weighting and deep neural network-SHAP(SHapley Additive ex Planations)provides technical support for the accuracy of ternary chemical indicators to determine the aging state of paper insulation.The specific innovation work is as follows:1)A high-precision cellulose molecular simulation model is constructed,and the microscopic mechanism of the aging of transformer paper insulation to generate ternary chemical indicators is revealed by simulation.Based on macroscopic experiments,the influence mechanism of different aging factors(temperature,moisture content and aging degree)on the concentration of ternary aging indicators is clarified.In addition,reaction molecular dynamics simulations are carried out based on simulation software such as LAMMPS,and the main generation pathways and reaction types of ternary chemical indicators during the pyrolysis of paper insulation are accurately analyzed.On this basis,oil-paper insulation samples under the action of multiple aging factors are prepared through an accelerated thermal aging test platform.The evolution rules on the degree of polymerization of paper insulation and the ternary chemical indicators content are clarified.The multiple regression equation for the cracking rate of paper insulation under the influence of temperature,moisture content and aging degree is established.Furthermore,the influence for aging factors on the production of ternary chemical indicators is explored.2)This study revealed the concentration changes of ternary chemical indicators in oil under different temperatures,moisture contents and oil-to-paper ratios,and clarified the mass loss mechanism of low-molecular alcohols(methanol and ethanol)in the middle and late stages of oil-paper insulation aging.Based on accelerated thermal aging experiments,the evolution rules of ternary aging indicators in oil under the combined effect of temperature and moisture are obtained.Besides,the distribution characteristics of ternary aging indicators between oil-paper under different oil-to-paper ratios are analyzed.Based on the above,in view of the concentration loss phenomenon of low molecular alcohols in the middle and late stages of oil-paper insulation aging,the kinetic parameters and thermodynamic characteristics of the reaction between low molecular alcohols and organic acids are calculated though Gaussian quantum chemical calculations and macroscopic experimental analysis.Furthermore,the main types of organic acids that consume low-molecular alcohols are clarified,and the loss mechanism of methanol/ethanol in the middle and late stages of oil-paper insulation aging is clarified.3)The weight proportion of ternary chemical indicators in oil to evaluate the aging state of paper insulation is determined,and a correlation model between the concentration of ternary chemical indicators in oil and the aging state of paper insulation is established.Through the analytic hierarchy process and the entropy weight method,the subjective/objective weight ratio of the ternary chemical indicators that characterizes the aging state of paper insulation is solved.Furthermore,the comprehensive weight rate of ternary chemical indicators concentration in oil to evaluate the degree of polymerization of paper insulation during the entire aging cycle is analyzed.Based on the generation rate and concentration change characteristics of ternary aging indicators under the different aging factors,the equation for ternary aging indicators concentration in oil and the degree of polymerization of paper insulation is established.Then,an aging state assessment model of transformer oil-immersed paper insulation based on ternary chemical indicators in oil is proposed.4)The interpretable analysis model of the weight proportion of ternary chemical indicators based on deep neural network is proposed.The SHAP interpretable analysis method is used to conduct a rational analysis for the comprehensive weight rate of the ternary chemical indicators.The ternary chemical indicator experiment/test data set is obtained through the accelerated thermal aging experiment of transformer oil-paper insulation,and the evaluation model is established using a deep neural network.The validity is verified from the aspects of error analysis,correlation and coefficient of determination.On this basis,a SHAP-based model interpretable analysis is carried out,which provided a reasonable explanation and verification for the weight distribution of ternary chemical indicators in the characterization of the aging state of paper insulation.It provides scientific experimental guidance and theoretical support to accurately determine the status of paper insulation for old transformers.