Research On Transformer Fault State Prediction Method Based On Oil Chromatogram Data

At present,with the development and popularization of smart grid,the relevant parameter data of power system are integrated and shared on a unified platform.The generation of a large number of data makes the traditional theoretical analysis model difficult to deal with this high-dimensional,large amount of data information.Therefore,the data-driven modeling method opens up a new way for power transformer fault prediction.In order to solve the problem of inconsistent data magnitude in transformer oil chromatogram data,it is necessary to normalize the data before transformer condition evaluation and prediction.In this paper,combined with IEC three ratio method[1],the standardized formula of benefit index is improved.The above method can normalize the transformer data well,and provide data support for subsequent transformer condition evaluation and prediction.In view of the traditional three ratio method can not systematically and intuitively evaluate the potential fault degree of the current transformer,this paper improves the three ratio method,using the three ratio data of the failed transformer and the fault data of the normal operation state of the transformer,using Entropy TOPSIS and grey correlation degree to calculate the distance between the three ratio data of the current transformer and the two states The proportion of fault distance.The above method can accurately evaluate the depth of the transformer fault,and according to the transformer fault proximity,the transformer fault degree can be divided into four levels from shallow to deep,so as to facilitate the maintenance personnel to observe.Aiming at the problem of transformer state prediction,a data-driven transformer symbolic regression prediction scheme is proposed.The prediction method in this paper is based on GEP Algorithm,and takes the time stamp of historical transformer fault proximity as the input parameter to train the prediction model.The system predicts the depth of transformer fault,which can accurately predict the fault time of transformer,reduce the subjectivity and experience in traditional transformer fault prediction,and provide reference for the reasonable arrangement of transformer maintenance,maintenance and maintenance It provides a scientific basis for the safe operation of transformer.Aiming at the problem of low diversity of GEP population,this paper introduces group climbing algorithm to ensure the diversity of individuals and the gene expression of each individual is not the same,so as to improve the diversity of the population and improve the efficiency of the algorithm.To sum up,based on the transformer oil chromatogram data,this paper studies and diagnoses the potential fault state of the transformer from the perspective of the system,judges the fault depth of the transformer according to the evaluated potential fault degree,and takes it as the input of the prediction algorithm to predict and evaluate the future fault state of the transformer.In order to improve the prediction accuracy and modeling efficiency of the prediction algorithm,this paper improves the original GEP by adding group climbing algorithm.Experiments show that this algorithm can better diagnose the potential fault degree of transformer,and can accurately predict the occurrence of transformer fault.