本文针对输入要求平稳的系统设备提出了一种预测控制新算法。该算法通过对输入增量引入柔化系数矩阵从而进行约束，避免了矩阵求逆的计算，大大减少了计算量，确保了系统的快速性，同时保留了广义预测控制的基本特征。其次，控制算法中改善了传统广义预测控制算法的单一性，充分利用预测信息的补偿作用进行加权平均，很好得抑制了超调的出现。通过挤出机的压力系统做仿真试验，结果验证了该算法的有效性。关键词： 广义预测控制；快速性；无超调；逆矩阵Abstract : Considering some industry system equipments are ordered smoothing input signalsagainst some sharp signals damages equipments, A kind of new control algorithm is presented, which not only makes input increase constrainted by softing coefficient matrix to obtain smoothing input, overcoming the disadvantage of calculating inverse matrix to increase responding speed of the system, but also makes fully use of compensations of predictive information to form weighted average to restrain overshoot. The simulation results have proved the validity of the algorithm by comparative study for extruder pressure system.Key words Generalized predictive control; A fast speed; Non-overshoot; inverse matrix