Summary: On the basis of Holt exponential smoothing model, multi-layer perceptron (MLP) model and support vector machine (SVM) model, an interval optimal combination forecasting model was built by introducing the concept of COWA operator and correlation coefficient, and some properties of the model were discussed. At the same time, the correlation coefficient was used as the contribution index of the single point forecasting method in the combined forecasting, and the Shapley value in the cooperative game was combined to give the approximate solution of the corresponding optimal combined forecasting model. The data of WTI spot price were used to demonstrate the feasibility and effectiveness of the model, and the sensitivity of the attitude parameters on COWA operator was analyzed.