Performance Assessment And Monitoring Of Model Predictive Controller

Model predictive control (MPC) is a new-style algorithm of computer control that emerged in the late 1970's. MPC computes the future control action by minimizing an performance objective function over a finite predictive horizon, including model prediction, receding horizon and feedback control, and can deal with the system with constraint and interaction, etc. It is currently the main vehicle for implementing advanced process control (APC) applications layer. The process and control engineers found that the MPC control systems perform well initially, but many factors can contribute to their abrupt or gradual performance deterioration. Lacking of effective real-time monitoring and maintenance tool, some of the control systems would run poorer and pull down in the end. Most of advanced control systems, such as MPC, need experts diagnose whether the performance is poor or not and offer the measure, but such experts are very little. In order to generate sustained benefit, we need a useful tool to implement and maintain MPC control system. So on the basis of performance assessment and monitoring of control systems, we propose analysis of MPC control systems. In this article we study performance analysis of MPC control system from two aspects, academically and for application, and get some conclusions:1. On the base of conventional performance assessment method we propose integrated performance analysis method of multivariate control system (IPAMMCS). This method can present more information about the existing control system, such as stochastic performance, deterministic performance and oscillation. It not only can assess the generic controller, but also can give effective performance assessment of model predictive controller.2. The performance index based on the idea of designing MPC can give the information of whether the existing control system is running better. If the performance is poor, it can reduce the range of faults.3. Industrial application, simulation and experiment prove that our research work is practical and have high significance.Performance assessment and monitoring of MPC control system is a brand-new field, and it is quality meaningful and practical. With the higher request of controller performance and economy benefit this research field will develop quickly.