Fault Diagnosis Method Of Power Transformer Base On Artificial Immune And Mind Evolution Algorithm

Power transformer is the key equipment in the power system, the stable and reliable operation of which is an important guarantee for successful operation of power system. Once faulted, it will affect the stability operation of power system, and leading to a greatly inconvenience in the production and people's lives.As a result, power transformer fault diagnosis has been a concern to scholars. The paper briefly introduced the source, background and significance of the topic , analyze the domestic and foreign research status and compared the quality of all kinds of diagnostic methods through reading a large number of documents. It introduces the principle of artificial immune algorithm and the application of transformer fault diagnosis. In order to make up the shortages of traditional Cloning immune algorithm for the transformer fault diagnosis applications, such as the slowly learning speed, non-classified ability and the small antibody space. This paper proposes a classification algorithm of antibody memory based on the principle that the antibodies recognize the antigens in the immune space and antibody memory function. The algorithm, combining the clone selection and the evolutionary algorithm, learns the training antigen and subsidiary type information to build the fault information database which is composed of different types of detection aggregate. The learning speed of this algorithm is very fast while the affinity is high enough. At the same time, it has an expanded immunization search space and improves the antibodies spatial diversity. Therefore, this algorithm can improve the accuracy of fault diagnosis. Besides, it use the artificial recognition ball (ARB) as the basic diagnostic unit. As a result, the speed of diagnosis is greatly improved because of the linear relationship between the ARB and the stimulation level.The artificial immune algorithm has the ability of dynamic learning, adaptive, memory, parallel processing and so on. However, the hypermutation mechanism would lead to the non-functional or poor affinity antibodies from sufficient affinity antibodies and a slower learning speed. In order to make up the shortcomings, this article introduces in mind evolutionary algorithm that the antibody groups will be divided into multiple sub-antibodies in the training process and the antibody affinity is training under the alternately operation of convergence and dissimilation.As a result, it ensures not only a steady increased of antibody affinity but also a faster training Speed.