Ławryńczuk Maciej
Institute of Control and Computation Engineering, Warsaw University of Technology, ul. Nowowiejska 15/19, 00-665 Warsaw, Poland.
ISA Trans. 2017 Mar;67:476-495. doi: 10.1016/j.isatra.2017.01.016. Epub 2017 Jan 30.
This paper details development of a Model Predictive Control (MPC) algorithm for a boiler-turbine unit, which is a nonlinear multiple-input multiple-output process. The control objective is to follow set-point changes imposed on two state (output) variables and to satisfy constraints imposed on three inputs and one output. In order to obtain a computationally efficient control scheme, the state-space model is successively linearised on-line for the current operating point and used for prediction. In consequence, the future control policy is easily calculated from a quadratic optimisation problem. For state estimation the extended Kalman filter is used. It is demonstrated that the MPC strategy based on constant linear models does not work satisfactorily for the boiler-turbine unit whereas the discussed algorithm with on-line successive model linearisation gives practically the same trajectories as the truly nonlinear MPC controller with nonlinear optimisation repeated at each sampling instant.
本文详细介绍了一种用于锅炉 - 汽轮机单元的模型预测控制（MPC）算法的开发，该单元是一个非线性多输入多输出过程。控制目标是跟踪施加在两个状态（输出）变量上的设定值变化，并满足施加在三个输入和一个输出上的约束。为了获得计算效率高的控制方案，状态空间模型针对当前运行点进行在线连续线性化，并用于预测。因此，未来的控制策略可以通过二次优化问题轻松计算得出。对于状态估计，使用扩展卡尔曼滤波器。结果表明，基于恒定线性模型的MPC策略对于锅炉 - 汽轮机单元不能令人满意地工作，而所讨论的具有在线连续模型线性化的算法给出的轨迹与在每个采样时刻重复进行非线性优化的真正非线性MPC控制器几乎相同。
