Study On Condition Assessment And Fault Diagnosis For Pawer Transformers Based On Big Data Analytics

Oil-immersed power transformer is applied widely and the most critical power equipment in the transmission and conversion of electricity, whose performance relate with the safe and reliable operation of power system, so grasping the running state of the power transformer accurately and finding the transformer latent fault timely and effectively, can guide the power transformer state maintenance work and reduce the power transformer faults and accident probability effectively. This paper adopts the date analysis method, based on the transformer operation data, selects effective characterization of influence the characteristics of transformer running state as evaluation indexes, set up power transformer state comprehensive evaluation system, and combining with the intelligent optimization algorithm to build evaluation model and do transformer state evaluation research; Characterized by the ratio of content of dissolved gas, combined with intelligent optimization algorithm to build fault diagnosis model and do transformer fault diagnosis. The main outcome of our research includes:Study of existing research results and collected a large number of standards, procedures and expert experience, to analyze and select the all possible indexes that can affect the transformer running state, combined with the principle of the establishment of index system for power transformer, establishes the comprehensive evaluation index system of transformer operation.The extreme learning machine (ELM) is applied to transformer condition assessment, use the Self-adaptive Evolutionary algorithm (SaE) optimize the ELM, build a power transformer running state comprehensive evaluation model based on the SaE-ELM. Study of the effect of the transformer inherent differences factors to the accuracy of the transformer condition assessment, and compared with SVM model, ELM model. Test results show that increase transformer inherent difference factors as condition assessment index can improve the accuracy of assessment; The SaE-ELM model can be used for transformer condition assessment and evaluation accuracy is higher than the SVM model and ELM model.To limit the randomness of parameters in ELM algorithm in practical application, the Artificial Fish-swarm (AFSA) is applied to optimize the weights and thresholds of ELM. In order to overcome the influence of the fault diagnosis results of the transformer fault sample data, in this paper, a new method of fault diagnosis for the fault diagnosis method of artificial fish swarm is constructed based on rough set of DGA. Use of power transformer fault diagnosis data set for training and testing, case study analysis shows that the method is good at transformer fault diagnosis and is put forward a kind of extremely value of new ideas for transformer fault diagnosis.