Power Transformer Fault Diagnosis Based On Information Fusion

Power transformer is the core of energy exchange and transmission in power grid, which is the key equipment in the power grid security defense system. So it is very important to prevent and reduce the occurrence of transformer fault. Because of the information and knowledge in power transformer fault diagnosis being characterized by randomness and uncertainty, the idea of transformer fault diagnosis based on information fusion is designed.Due to the complexity of transformer fault mechanism, transformer fault diagnosis method in this paper is studied on the use of content data of dissolved gas in transformer oil and combined with the theory of information fusion. In this paper, based on the extreme learning machines algorithm, the use of information fusion technology to improve extreme learning machines thinking through always, to solve the limitations of a single diagnosis caused the slow training speed and low diagnostic accuracy.Firstly, it introduces the related knowledge of information fusion technology, including the development history and research status of information fusion technology, the basic principle and hierarchy structure, and compares the advantages and disadvantages of the common information fusion algorithm, and discusses the application of information fusion technology in the military and civil fields.Secondly, it introduces the basic idea, learning algorithm and classification of ELM. According to the characteristic of transformer fault, the transformer fault diagnosis based on ELM is designed. The ELM fault diagnosis model is established. The input feature vector and output fault classification vector are determined. The influence of activation function and the number of hidden layer nodes on diagnostic performance is analyzed by simulation experiment. The advantages and disadvantages of ELM and SVM in transformer fault diagnosis performance are compared and analyzed. The effectiveness of ELM in transformer fault diagnosis is analyzed through an example. At the same time, it is found that there is a deficiency of fitting.Finally, based on the principle of information fusion technology, transformer fault diagnosis based on fusion ELM and FELM is proposed, which can achieve decision level fusion, improve the performance of fault diagnosis based on ELM, and retain the advantages of ELM training speed and good generalization performance. Fusion ELM is the fusion product of ELM, probability theory and the weight. The traditional output of ELM is mapped to the probability output, and the fusion result is obtained after the weighting process. The fusion ELM can effectively solve the problem of the output inconsistency caused by the randomness. FELM is a combination of ELM and fuzzy theory. The traditional output of ELM is transformed into fuzzy vector output, which effectively solve the weighted classification problem, and show a clear diagnostic advantage in the case of less samples.