This paper analyzes the basic principle of support vector machine, determines the sample inputs for oil consumption prediction model of aviation troops. The prediction function for aviation troops oil consumption is constructed, LibSVM Matlab toolbox is used to solve the model, and three indexes are selected to evaluated the prediction results. The oil consumption of one aviation troop is taken as an example. This aviation troops oil consumption prediction model is used to forecast the oil consumption in 2009 based on SVM, and the prediction results are compared with the actual value. It shows the high prediction accuracy, which provides scientific quantitative analytical method for prediction battlefield oil consumption.