An adaptive single prediction controller

Chemical and petroleum processes generally are nonlinear, and in some cases, time-varying in nature. The mismatch between the process and controller models caused by these characteristics can deteriorate the control performance of the model predictive controllers. GPC (generalized predictive control) is one of the popular adaptive control algorithms that address this problem. However, the mathematical and computational complexity of GPC can be discouraging for industrial applications. In this paper, an adaptive MPC (model predictive control) algorithm is presented that is much simpler computationally. Its performance is presented on three simulation examples where mismatch is considered in dead time, gain, model order, and zeros of the transfer function. Performance is also presented on a laboratory liquid-level system. The results show that the proposed algorithm can handle a fairly large model mismatch. A comparison with the GPC algorithm shows that the control calculations are much simpler, and that the control performance can be somewhat improved.