The fuel management optimization problem within a light water moderated nuclear reactor is formulated as a multistage decision process, and the amount of fresh fuel in each zone and the shuffling schemes of reloading fuel are simultaneously determined so as to minimize the consumption of fresh fuel throughout plant life. The problem in case of a five-zone core involves about 8000 variables and about 2000 constraints on the variables. However, it is shown that the optimal refueling policy can be obtained in a reasonable length of computing time by using linear programming and, prior to the overall optimization calculation, by solving the subproblem of minimizing the stagewise consumption of fresh fuel. The optimal refueling policy results in reduction of the consumption of fresh fuel by about 10% compared with the conventional policy of uniform partial batch refueling. The essential assumption in the calculation model is that the spatial power distribution in the core does not change significantly with time and the optimal allocation of energy output (integrated power output) for each zone can be attained by optimizing the control rod programming for one core life. Three-dimensional depletion calculations are repeatedly performed to verify feasibility of the model based on that assumption.