Design of a Database-Driven Nonlinear Generalized Predictive Controller

Abstract

This paper addresses a regulation problem of non-linear systems via database-driven nonlinear generalized predictive controller without model information. In industrial processes, lots of controlled systems with unknown time-delay and strong nonlinearity, are difficult to be handled in terms of control performance. Advanced controllers are considered to be established to deal with those nonlinear systems. In several design methods, advanced controllers are designed based on model information. However, it is time- and cost-consuming to identify the model of controlled systems, and requires regular maintenance to maintain acceptable performance. The database-driven approach has been attracted attentions to tackle those issues without model information. The controller can be designed and tuned only based on data, which is the main feature of this approach. Besides, the database-driven approach can deal with strong nonlinear systems. Additionally, the Generalized Predictive Control (GPC) is one of predictive controllers and widely applied in industrial processes. The GPC controller is developed based on multi-step prediction, therefore, it is effective to those systems subject to unknown or time-delay. As a result, a nonlinear GPC controller in the proposed scheme inherits the advantage of GPC, and is also tuned by the database-driven approach. The effectiveness and benefits of the proposed scheme are demonstrated through a numerical simulation and a comparative study.