时滞网络预测PI控制算法鲁棒稳定性分析

Robust stability of predictive PI control algorithm for delay network

针对网络中存在的时滞给主动队列管理算法性能带来的不利影响,将Sm ith预估补偿器与达林算法相结合,提出了一种预测PI控制算法,既利用Sm ith预估器克服了大时滞给系统性能带来的影响,也减少了控制器参数整定数量.利用控制理论分析了该算法中Sm ith预估模型与实际对象模型分别存在增益、时间常数和时延失配情况下系统的鲁棒稳定性,并分别给出了保持系统稳定的失配参数条件.理论分析表明,通过适当选择期望闭环传递函数时间常数和预估模型参数,可以使系统获得较强鲁棒性和较快响应速度.仿真试验验证了理论分析的正确性.

A predictive proportional integral （PPI） control algorithm is proposed, which combines the Smith predictive compensator with Dahlin algorithm to alleviate the time-delay influence on the performance of active queue management （AQM） in network. In PPI algorithm, large delays are overcome with Smith predictor and the number of running parameters is reduced. Robust stability of system is analyzed when gain, time constant and time delay parameters of both Smith predictive model and actual plant model are mismatched respectively, and the conditions of mismatch parameters used to maintain system stability are given. The theoretic analysis shows that the tradeoff between robustness and response is achieved by selecting the time constant of the system closed loop desired transfer function and parameters of the predictive model properly. The simulations validate the theoretic analysis results.