Development of Memory-Based Control Systems with a Concurrent Structure of Process Assessment and Process Control

近年来，过程工业对提高生产质量和降低生产成本的要求越来越高，控制系统在这种情况下发挥着重要的作用。特别是，过程系统的性质经常由于操作条件的变化和/或系统的非线性而发生变化。因此，需要根据系统特性的变化，重新调整相应的控制参数，以保持期望的控制性能。作为其中一种策略，自整定控制被提出。然而，递推最小二乘法在非线性系统中并不有效。同时，为了保证工业过程的安全或生产率，近十年来对控制性能评价（CPA）的研究受到了广泛的关注。在过程系统中，如何统一CPA和控制器的设计是工业过程系统中最重要的问题。本课题提出了一种性能自适应PID控制器的设计方案。根据所提出的控制方案，首先对系统的建模性能进行了评价，并利用数据驱动系统辨识方法对非线性系统进行了参数辨识，该方法在建模性能不佳的情况下是有效的。此外，根据预估值调整PID参数，使控制性能满足预先确定的期望性能。具体采用基于广义预测控制方案的自整定PID控制对PID参数进行整定，并选择用户指定的参数，即表示期望控制性能的控制权重因子。通过数值模拟算例和实际系统的应用验证了该方案的性能。

Recently, the improvement of production quality and the reduction of the production cost have been further advanced in process industries, and the control systems play an important role in such a situation. In particular, the properties of process systems have been frequently changed due to the change of operating condition and/or nonlinearity of the systems. Therefore, it is necessary to readjust the control parameters corresponding to the change of system properties in order to maintain the desired control performance. As one of such strategies, the self-tuning control has been proposed. However, it is said that the recursive least squares method is not effective in nonlinear systems. By the way, for the purpose of maintaining the safety or the productivity of industrial process, the researches on control performance assessment(CPA) have been paid to attention in last decade. In process systems, it is most important problem to unify the CPA and the controller design in industries. In this research project, a design scheme of performance-adaptive PID controllers has been newly proposed. According to the proposed control scheme, the modeling performance is firstly evaluated, and system parameters are identified using the data-driven system identification which is effective in nonlinear systems, if the modeling performance is not good. In addition, PID parameters are adjusted based on the estimates so that the control performance satisfies the desired performance determined in advance. Concretely, the self-tuning PID control based on generalized predictive control scheme is adopted in order to tune PID parameters, and the user-specified parameter, that is, the control weighting factor which indicates the desired control performance, is chosen. The behavior of the proposed scheme has been examined by some numerical simulation examples and application for a real system.