Syntetic Fault Diagnosis Method Of Power Transformer Based On Rough Set Theory And Improved Artificial Immune Network Classification Algorithm

As a key equipment in power system, power transformer's operation reliability is closely related to the safety and stability of power system. For a long time, the judgement of health state and operation condition of power transformer is carried out by preventive test and periodic maintenance. This kind of maintenance mechanism in time-based manner regardless of power transformer's insulation condition will lead to"over maintenance"or"owing maintenance".As condition-based maintenance is increasingly accepted, it is an inexorable trend to replace periodic maintenance.In this paper, the work includes:1.Collect, collate and summarize the recent domestic and foreign power transformer fault diagnosis in the field of the latest research results and latest progress;2.Use rough set table Reduction technology, to achieve the simplification of knowledge and the compression of fault characteristics, obtain the minimal diagnostic rules;3.Analysis the power transformer insulation fault diagnosis principle, and present artificial immune network classification algorithm in accordance with the power transformer insulation fault diagnosis characteristics ;4.According to complementary strategy, this paper presents a new power transformer fault diagnosis method based on rough sets theory (RST) and improved artificial immune network classification algorithm, combining with IEC ratio method.The representation of Antigens and antibodies in shape space are improved by adding information of fault type to memory antibodies.Analogically to the relationship of antibodies and antigens, antibodies are continuously optimized during learning antigens pattern, thus memory antibodies can preferably learn and memory the features of the same type of fault samples.Then the improved K-Nearest Neighbor method is used to classify the fault samples.The proposed algorithm make use of the distinguishing ability of IEC ratio method,the distilling ability of rough set, the selt-learning and self-memorying ability of immune network. The experiment results demonstrate that this proposed classification algorithm for fault diagnosis of power transformer has remarkable diagnosis accuracy and effective classification for single incipient faults as well as multiple incipient fault.