The invention discloses a transformer oil chromatographic data forecasting method based on the D-S evidence theory. The transformer oil chromatographic data forecasting method comprises the following steps: (1) acquiring historical concentration data of various gases in transformer oil chromatographic data; (2) performing training through various intelligent forecasting algorithms by taking the concentrations of the multiple gases at a certain moment as inputs, the concentration of a gas to be forecast in the multiple gases at the next moment as an output and the acquired historical concentration data of the multiple gases as a training sample set, and calculating a training error; (3) performing weight extraction and fusion on forecast results obtained by a plurality of forecasting models in the step (2) according to the D-S evidence theory to obtain the weight of the forecasting model for the gas to be forecast, and finally obtaining a forecast value; (4) finishing the forecasting. According to the transformer oil chromatographic data forecasting method, the gas concentrations are forecast through the fusion performance of the D-S evidence theory on the forecast results obtained by multiple methods, so that the precision and the generalization property of a transformer oil chromatographic data forecasting model are improved, and the transformer oil chromatographic data forecasting model is higher in forecasting precision.