We propose a method for predicting the dissolved gas volume fraction in transformer oil based on multi factors to overcome the shortcomings of traditional methods

which use a small amount of parameters

in which online-monitoring data of the dissolved gas in oil

EMS data

and meteorological data are collected

and all kinds of data are cleaned and synchronized by data preprocessing.The gray relational analysis method is applied to determine the key input parameters with high correlation of the dissolved gas and reduce the dimensions of the input data.Thereafter

a multi-factor prediction method based on KNN regression model is established under the normal state of the transformer.By collecting the fault data of the transformer in the failure case database

the relational analysis is applied to select the similar fault samples

and the relational degree results are determined as important weights of each input data.Then a multi-factor prediction method based on weighted mean regression of the fault samples is established under the fault state of the transformer

which overcomes the scarcity of each equipment fault data.The case analysis shows that

compared with the common prediction methods

the prediction accuracy of the dissolved gas volume fraction in oil is obviously improved when multi factors are reasonably incorporated into the prediction model.