Research On The System Of Online Monitoring And Intelligent Fault Diagnosis For Power Transformer

As central equipment in the power distribution system,the transformer plays a key role in the stable and safe operation of the whole power grid.Therefore,it has long been an important task to build and improve the monitoring and fault diagnosis system of the transformer.The state-of-the-art research on the transformer monitoring and diagnosis system is mainly about improving the system real-time performance and the diagnosis accuracy,and the trend is to transform the single-function and patterned systems into integrated-function and intelligent ones.The research of this thesis focuses on the oil-immersed power transformer.On the basis of dissolved gas analysis,an intelligent algorithm considering multiple transformer parameters is adopted.With the help of modern sensor technologies and communication methods,the whole online monitoring and intelligent fault diagnosis system is researched.After studying fault types and the gas generation mechanism in the transformer,existing judging methods based on dissolved gas analysis are analyzed.Despite of advantages in different aspects,shortages including incompleteness of judgement criteria and not being able to covering all types of fault are common to these methods.An intelligent fault diagnosis strategy based on wavelet neural network optimized by improved Artificial Bee Colony algorithm is proposed.By introducing the Artificial Bee Colony algorithm featuring efficient search and clear objective,and constructing the searching mechanism instructed by elitist bee colony,both the convergence speed and convergence accuracy are improved.The chaos strategy is adopted to ensure the variety of initial colony distribution,effectively avoiding localized optimization and improving the accuracy simultaneously.The adaptive value and the follower selection method are improved to guarantee the convergence speed of the algorithm.To avoid the ambiguity of conventional BP neural network,the wavelet neural-network-based model is selected with parameters are optimized by improved artificial bee colony algorithm.The simulation results of test functions show fast and accurate convergence together with capability of avoiding localized optimization of the proposed method.Test results after applying the optimized wavelet neural network to the transformer fault diagnosis verify the fast convergence and high accuracy of the proposed diagnosis strategy.To achieve function integrity and completeness,five monitoring subsystems,including those designed for monitoring dissolved gas in oil and micro-water,are constructed,all of which are equipped with precise sensors and data from which can be integrated and transmit to host computers in real time.The diagnosis solutions integrate independent judgements of each monitoring subsystem with both traditional and intelligent diagnosis strategies and support all kinds of combination patterns,thus effectively overcoming drawbacks of dissolved gas analysis-based methods in fault points determination.The QT 5.3 software and Microsoft SQL server 2014 data access are utilized to construct the software environment of the host computer for online monitoring and intelligent fault diagnosis of the transformer,thus meeting the real-time performance and accuracy requirements of the online monitoring and fault diagnosis system sufficiently.