Study On Fault Diagnosis For Power Transformer Based On Support Vector Machine Of Artificial Immune Algorithm

Power transformer is the pivot equipment in the power system, its operating reliability is closely related to the safe operation of the power system. Therefore, the strengthening the detection of the transformer operating state, early detecting and handling the latent failure of the transformer, and improving the operation reliability of transformer, are of great significance to the safe and reliable operation of the grid.Transformer fault diagnosis is based on the fault features to judge the failure type.These provide foundation for transformer maintenance. The practice shows that dissolved gases analysis plays a major role in the transformer fault diagnosis. On the basis of the analysis of transformer failure characteristics, according to the dissolved gas in oil, this paper summed up the relationship between the gases dissolved in transformer oil and the fault types of transform. Due to that traditional transformer fault diagnosis method is not high in detection accuracy, and need a lot of experience, therefore it is urgent to introduce some new transformer fault diagnosis methods to make up for the defects of the current transformer fault diagnosis. Support vector machine is a new machine learning method,based on statistical learning theory. it is not only better able to solve high dimensional problems and local minima problems in the neural network algorithm, but also deal with the small sample data effectively.Therefore, it is highly advantageous to the power transformer fault diagnosis where the fault sample data is small in quantity. However, due to that the support vector machine kernel function and its parameters have a great influence on the classification performance of support vector machine, the traditional methods of parameter selection have large amount of calculation and are time-consuming, therefore they can not meet the requirements of the transformer fault diagnosis. Relying on positive selection and clone of excellent antibody, Artificial immune algorithm is able to ensure that the algorithm has a strong global optimization features, and a fast convergence speed in finding global optimal solution, and have a great advantage in parameter optimization. On the basis of analysising a variety of transformer fault diagnosis methods, this paper presents the method which is to optimize the support vector machine kernel parameter using immune evolutionary algorithm. Sample simulation results show that the method has the good effect of parameter optimization and effectively solves the problem when choosing support vector machine parameters.This paper applied classification and identification capability of support vector machine into fault diagnosis of power transformers, and presented a new method of using the immune evolutionary algorithm to optimize the support vector machine for diagnosising transformer fault type. Through the simulation and test of the actual fault data, the results show that the optimized support vector machine’s generalization ability and diagnosis accuracy has been greatly improved, thus the correctness and validity of the method is and verified.Compared to support vector machine optimized by cross validation and support vector machine optimized by genetic algorithm in transformer fault diagnosis, this method is better than the other two methods in fault diagnosis accuracy and convergence, and can substantially improve the accuracy and reliability of diagnosis.It plays a larger role in the power transformer fault diagnosis.