Generally, the oil support demand is forecasted through the method of historical data regression analysis, these methods become not so effective when confront with the situations of sudden force action style which are short of the historical oil consumption data. In order to solve this problem, case-based reasoning techniques is introduced ,then the oil support demand cases property weights are calculated by the analytic hierarchy process; Aiming at a universal problem that missing of some case properties, a similarity calculating method called ''twice-calculating'' is proposed which base on the structure comparable degree and the similarity degree. Finally the theory and the method proposed above are simulated and verified by a case of oil support demand forecasting of a tank regiment's military action. The result shows that, without the historical data support, the Case-based Reasoning can still forecast the oil demand accurately.