基于扩张状态观测器的广义预测控制

GENERALIZED PREDICTIVE CONTROL ALGORITHM BASED ON EXTENDED STATE OBSERVER

针对线性对象,提出了一种基于扩张状态观测器的广义预测控制算法.首先,分离出对象中异于标准积分串联型的部分,把它作为扩张状态,再利用观测器对其进行实时估计和补偿,将原始对象近似转化为积分串联型.然后,针对积分串联型这种特殊的对象设计广义预测控制.这既降低了预测控制对于模型的敏感性,同时又极大简化了计算,取消了控制的增量形式以提高系统的鲁棒性但依然保持了闭环无静差特性.通过内模原理得到了闭环特征多项式.仿真结果验证了所提算法对于对象的参数摄动具有很强的动态性能鲁棒性.

A kind of generalized predictive control （GPC） algorithm based on ex- tended state observer for the linear plants is proposed. First of all, the redundant part which makes the plant deviate from the canonical form of the chain of integrators is separated accurately to serve as the extended state. Then a linear observer is utilized to estimate and compensate this component so that an integrators chain can be obtained and the modified GPC is designed for this chain. By using this scheme, both the sensitivity of GPC to the predictive model and the computational complexity can be effectively reduced. The control-increment format result of the existing GPC is eliminated to enhance stability while maintaining zero steady-state capability. The closed-loop characteristic equation can be calculated by using internal model control principle. The simulation results demonstrate that the proposed algorithm exhibits strong dynamic robustness to a wide range of parameter variations.