Application of Improved Grey GM(1, m) Model in Fault Prediction of Transformer

In order to solve problem of big error in predicting data sequence with fluctuation while using grey model to predict transformer fault, an improvement scheme of gray GM(1, m) predicting model was proposed. In the scheme, original data sequence is processed for better exponential rule to meet smoothness requirement of the predicting model, and the processed data sequence is analyzed by grey relational degree method to get relationship between variables. Background value of the predicting model is optimized and used for establishing the model, and equal-dimension new-information model is used to predict data. The improved GM(1, m) model was used to predict volume fraction of seven kinds of characteristic gases in some transformer oil and both average differences and posteriori relative error of predicted data were smaller than the ones of GM(1, 1) model and traditional GM(1, m) model, which showed the improved grey GM(1, m) has better prediction accuracy.