为了解决网络控制系统中不确定时滞、尤其是随机的长时延对控制性能造成的影响,提出了基于模型匹配和多步预测输出补偿的预测控制思想来改善控制性能。该算法通过采集到的传感器端至控制器端的时滞来估算控制器端至执行器端的时滞,并给出了传感器端和执行器端数据处理的算法,建立起与实际网络结构匹配的预测控制模型,采用多步预测输出来补偿控制量的传输滞后。该模型不依赖于网络参数和时滞分布特性,可根据不同对象选取合适的预测控制算法,适用于实时系统。通过基于倒立摆对象的仿真和控制实例验证了该算法能有效地改善控制性能。 In order to solve the influence of uncertain delay, especially random long delay, on the control performance, a predictive control idea based on model matching and multi-step predictive output compensation is proposed to improve the control performance. The algorithm estimates the time delay from the sensor end to the actuator end by collecting the delay from the sensor end to the controller end and gives the algorithm of data processing on the sensor end and the actuator end to establish the prediction that matches the actual network structure Control model, using multi-step prediction output to compensate for the control of the transmission lag. The model does not depend on the characteristics of network parameters and delay distribution, and can select the appropriate predictive control algorithm according to different objects and is suitable for real-time system. The simulation and control examples based on the inverted pendulum object show that the algorithm can effectively improve the control performance.