Model Set Optimization Method for Complex Plant Multi-Model Control

In this paper we present a novel model set optimization method that is applicable to the complex systems with large parameters variations and multiple operating modes. The proposed method is based on the multi-model approach to improve the system performance of a complex control system of linear or nonlinear characteristics when it operates at various operating conditions. The multi-model control scheme depends on the multiple representations of a process using different models that generate the control signal needed to make the system follow a prescribed desired trajectory. It uses a set of models-combining the offline identified models and the dynamic model bank with the iterative learning, which guides the adaptation process. The offline models are obtained from the physics or chemistry principles. Dynamic model bank summarizes the parameters of the models that successfully approximate the plant. The model bank can be automatically created and updated and does not call for an initial set of models. By virtue of the iterative learning, we expect to make the models set consist the models which best represented the character of the plant. It uses a soft switching mechanism that provides a smooth transition from an interpolative to a pure hard switching scheme between the models in the model set. We also demonstrate the advantage of using this approach on an examples considering control of systems with large parameter variations. Simulation results show that the method proposed in this paper has obvious advantage over the only dynamic model bank, and illustrate the potential of the algorithm of the developed method.