Study On Health Management System Of Oil-immersed Power Transformer Based On Data And Model

Oil-immersed power transformer is widely used in power grid,which makes its safety and reliability become one of the key factors affecting the quality and reliability of power supply,and the health management of transformer has been widely concerned.The health management of oil-immersed transformer has always occupied a large number of resources of power grid enterprises due to the diversification of manufacturer,process,voltage level,capacity,and the complexity of operation environment.Under the background of smart grid,the rapid development of modern science and technology promote the building of smart substation,which makes the real-time online monitoring of transformer operation status possible,thus laying a physical foundation for transformer health management.Based on the application scenario of the prognostics and health management(PHM)of transformer,this dissertation researches on the key theory and technology of the transformer health management system.And on this basis,through making full use of advanced computer,communication and other information technology,a transformer PHM cloud platform is developed for effectively improving the management level and efficiency of power grid enterprises on transformer assets.The main research contents of the dissertation are following.(1)Aiming at the problem that single transformer attribute is difficult to effectively and accurately realize the state evaluation of transformer,the multi-attribute characteristics of transformer are analyzed,and the appropriate weight is assigned to every attribute.On this basis,a multi-attribute model based on fuzzy logic,to be used to evaluate the transformer state,is proposed.The proposed model has simple fuzzy rules and less input parameters,and overcomes the shortcomings of previous fuzzy logic models and traditional transformer health assessment methods.The model performance,including the correctness and reliability of the proposed model,are verified by field transformer testing.(2)In order to settle thematter that the analysis and value of heat dissipation resistance in the empirical formula for calculating the hot spot temperature of transformer is relatively simple,which can not effectively reflect the influence of environmental factors on temperature,resulting in relatively large error of calculation results,the internal heat transfer mode and mechanism of transformer under different load current,different cooling mode and different internal temperature are studied.The methods on the structure and parameters of transformer thermal circuit model under various working circumstances is studied,and then the calculation method of top oil and hot spot temperature based on the improved model is given,and the accuracy of calculation results is evaluated.The testconclusionsprove that the difference between the temperature of top oil and hot spot calculated by the improved model and its actual value is less than 2.2k,that is to say,the calculation method of heat dissipation resistance can effectively improve the accuracy of thermal circuit model.(3)Three common transformer fault diagnosis methods based on dissolved gas in oil are analyzed.The basic principle of genetic algorithm is studied.For solving the shortage of the basic genetic algorithm which is easy to be trapped to local optimization,an improved method is proposed to adjust the crossover and mutation probability as well as the individual reproduction number according to the moderate value.The simulation results show that the improved method significantly improves the global search ability of the algorithm;by optimizing the initial weights of BP neural network using the improved genetic algorithm,a transformer fault diagnosis model based on improved genetic algorithm is established,which effectively solves the problems of slow convergence speed and poor accuracy of BP neural network.Compared with three common fault diagnosis methods,the conclusions show that the proposed diagnosis model has better diagnosis speed and accuracy.(4)Based on the proposed thermal circuit model of transformer under different working conditions,the insulation life evaluation model of transformer using hot spot temperature analysis is established.An aging test of oil paper insulation under thermal factors is designed,and four statistical maps based on pulse phase distribution mode are extracted,then 10 principal component factors are obtained through the factor analysis method,and thereby a method for evaluating the oil paper life of transformer with genetic algorithm optimizing BP network is advanced,and the test results show that the proposed method has a credible diagnosis effect.Finally,a power transformer life prediction model based on the Weibull distribution is established by analyzing the correlation between the Weibull distribution and the law of electrical equipment life,and the simulation conclusions prove that the prediction results are in good accordance with with the actual situation.(5)Based on the need of online and real-time management of the transformer PHM,the platform of the transformer PHM system is developed by using advanced information network technology.The key technology theories and design principles included in the platform are described.According to the functional requirements of the PHM and the status of technical resources,the overall constructure and functional modules of the PHM platform are planned.Through the field operation of the developed PHM platform to a transformers substation,the operation conclusions prove that the developed platform can satisfy the transformer PHM.