Production manufacturing scheduling problem has been the center of interest for the last decades. As all production systems are integrated into a scheme, the supply of raw material becomes a principle problem in production management system. This problem is more critical in petroleum industry because crude oil has various characteristics and the final product are changed according to crude oil supplied. Petroleum industry consists of a serious of related areas such of oil production, delivery of crude oils, refining operation, and marketing of refined products. Production planning determines aggregated production quantity for each final product while satisfying various firm``s constraints. Production planning is supported by a NLP system called RPMS(Refinery Petrochemical Modeling System). Crude oil daily delivery scheduling problem is tried to solve under the assumption that the monthly total charging quantity of each crude oil is given by the master production plan. In this study, a compound backtrack search algorithm using the existing AI search and intelligent backtracking technique is presented. This heuristic algorithm solved the problem in each subproblems. In advance, the combinations of crude oils that can be delivered simultaneously are composed in order to backtrack at deadend situation easily. The backjumping technique is used for the backtracking and minimizes the occurrence of deadend on the daily delivery scheduling. The quantity of each crude oil at purchasing time is determined at the last subproblem finally. The prototype of crude oil delivery scheduling algorithm is developed and tested.