Research On General Diagnosis Techniques For Overheating Faults Of Power Transformer

Power transformer is the main equipment of power system, it's application is wide, and it's reliable operation is the foundation of system stability. Because of multiple reasons, transformer fault sometimes occur, the main task of power transformer fault diagnosis is to correctly diagnose it's incipient faults.The incipient faults of power transformer mainly consist of overheating and discharge. In them, overheating faults occupy a large proportion, if not handled in time, Transformer will be damaged. This paper systematically studies the general diagnosis methods for overheat faults. Through systematically analyzing the DGA method used to determine whether there is an overheating fault and electrical experiments, infrared temperature monitoring mrthods used to judge the reason and position of the faults, this paper presents mathematical DGA assisted by an expert system to more effectively diagnose overheating faults.Dissolved gas-in-oil analysis (DGA) is one of the very useful methods to detect transformer incipient faults. Among the diverse conventional DGA methods, IEC three ratio method is widely used, but in many cases this method can not accurately diagnose,so it has limitations. This paper put forward two mathematical DGA methods: fuzzy three ratio method and extension set method. The fuzzy three ratio method fuzzified the boundaries of ratio codes. In extension set method: firstly matter-element models are established for three ratio method; secondly the paper divides the intervals in the matter-element models into three kinds, respectively offer the three kinds of extended relation functions of a point to an interval based on the principle of sectional linearization, this is used to solve the limitation of extended relation function of a point to an infinite interval exits in the reference documentations; finally these functions are used to diagnose transformer faults. Simulation proves that these two methods can overcome the drawbacks of the conventional three ratio method (no matching, multiple faults), thus greatly enhanced diagnosing accuracy. The combination of these two methods can effectively determined whether the diagnosing results are credible or doubtful without knowing the actual fault type in advance.As DGA can not determine the cause and position of the faults, this paper presents an overheating expert system based on the expert knowledge of electrical experiments and infrared temperature measurement. In this expert system: knowledge is modularized, rule is based on basic and compound facts, positive and negative direction mixed inference strategy is used, fuzzy mathematics is applied to realize inaccurate knowledge representation and inference, all these are used to optimized the system function. Finally simulation examples show this system has good diagnosing effect for typical overheating faults...