Abstract: A mathematical model and an optimization algorithm, which is based on PBIL (population-based incremental learning) and network maximum flow, were proposed for the charge design problem of steel-making. The algorithm first finds an upper bound of the number of charges, which serves as the baseline for designing PBIL with 0-1 chromosome encoding, through a heuristic rule. Each chromosome represents a selection scheme of charges, and the network maximum flow theory is used to calculate the fitness value for chromosome. The optimal order-furnace matching strategy could be obtained after several iterations. Simulations on real production data indicated that the proposed algorithm can obtain an optimized matching solution within reasonable time, and can provide enough decision support for planners.