| With the rapid development of China’s economic in recent years, the electric power system has been towards the changes of larger capacity, high voltage, and high level of automation. Meanwhile, the power system security requirements also become higher too. Power transformer is an important equipment of power system, the price of the transformer manufacturing is high and it undertakes an important task. The transformer can run safe and stable or not directly related to the power system security. However, according to statistics, transformer is one of the frequent accident power equipments, and because of its importance, failures often result in serious consequences. Faced with this situation, the power transformer fault diagnosis technology increasingly has valued by researchers. Traditional transformer fault diagnosis is generally based on transformer oil dissolved gas components, and according to different levels of the gas, classified into different types of faults. This paper focuses on the modified three-ratio, a method recommended by China’s current national standard.Traditional fault diagnosis method has some drawbacks, such as classification fuzzy, difficult to distinguish the case of multiple faults, etc. Therefore, this paper introduces the BP network. Design a BP network based on the modified three-ratio method, and train the fault diagnosis model through the study of sample. But BP network is based on the gradient method to determine the weights, and the gradient descents are inherently vulnerable to the effects of local minima. Therefore, we introduce the genetic algorithms to make up for this lack of BP network. This paper attempts to improve the traditional genetic, the operational efficiency is improved, and to achieve functional of improve the fault diagnosis system.At last, we train the improved GA-BP network by30groups of typical sample, and select the20groups of test data to test the network to verify the accuracy of fault diagnosis. |
PDF Full Text Request