The problem of how to form dynamic cells based on changing production requirements with multiple planning horizons and multiple objectives was discussed.   A nonlinear multi-objective mathematical model of dynamic cell formation was built by weighing the three objectives, including total cost in the process of cell manufacturing and formation, maximum deviation of workload from available capacities of machines, and total number of inter-cell moves.  Using adaptive niche technique, penalty technique, double roulette wheel method, and reserving elite strategy, reserving elite-based random weight multi-objective genetic algorithm was designed for the complicated combination optimization model.   The model and algorithm were analyzed by a numerical example.   The computational results demonstrate that the proposed genetic algorithm is effective.