模型预测控制的性能监视

Performance Monitoring for MPC Systems

由于线性系统的参数变化是随时间变化的,使得评估控制系统性能变得困难。模型预测控制有一个优点就是可以直接地描述约束背离的情况,而约束违背的时变特性与模型预测控制系统的性能监视问题是直接相关的。该文将用一种对约束的全新算法来描述这种相关性,通过检测真实控制性能与基准控制性能的统计学差异来监视控制系统的退化。

Actually the constraints make the process nonlinear, more precisely, it makes the process time -varying even for a linear system. Under this situation, it is becoming difficult to evaluate the performance of an actual control system. Model predictive control (MPC) systems have a merit that he can describe the violation of constraints directly, such as actuator saturations and limitation of values. And the constraints are explicitly involved in the performance monitoring problem for the MPC systems. In this paper, time -varying nature of the violation of the constraints are captured by a new expression, and the degradation of the control performance is monitored by detecting the statistical variation of the real control performance away from a benchmark control performance.