Hierarchical Optimization Control Strategy Based On Dynamic Real-time Optimization And Multiple Model GPC

Increasing competition in the market makes a high demands for reducing the cost of consumption and improving economic efficiency, and thus have an increasing demand for the optimal control strategy. The optimal control strategy, which combines real time optimization (RTO) with model predictive control is one of the most common methods. Optimization layer uses real-time optimization to determine the optimal conditions, and the control layer uses model predictive control technology to achieve the optimum adjustment of control process. However, the optimal control strategy faced with the following problems:since the model information of control layer cannot match that of optimization layer, this lead to bad optimization results; the optimization layer based steady-state optimization has limit the flexibility and economic benefits, and so on. Considering these issues, dynamic optimization and multiple model control strategy is introduced in the traditional hierarchical optimal control strategy. Main research works can be summarized as follows:1. Developing the dynamic optimization technology and linear multiple model generalized predictive controller in the hierarchical structure and propose hierarchical optimal control strategy based on linear multiple model generalized predictive controller. The optimization layer depended on the process dynamic model is introduced to deal with the system dynamic behavior and determine the optimal operating conditions. The lower MPC controller with a multi-model generalized predictive controller replaces a single model controller, so that it can effectively eliminate systematic errors, improve system performance and regulation for model parameters transient hopping, and overcome the effects of optimization results from the inconsistence model information between optimization layer and control layer model.2. In the actual industrial process, since the controlled system often has a strong nonlinear characteristic, it is very complicated to model and identify the system with an accurate mathematical model. Linear multiple model lacks of ability to compensate the nonlinearity of the system, so it cannot meet the requirements of complex process. Therefore, hierarchical dynamic optimization control strategy based on the non-linear multi-model GPC is proposed for solving this problem. Dynamic real-time optimization layer determines the optimal condition for realizing economic performance optimization in the production process. Control layer adopts nonlinear multiple model controller which is composed of linear GPC and nonlinear neural network to replace the original linear multiple model controller under the premise of guaranteeing the system stability and improving the control performance of the system. Finally, the feasibility and efficiency of the proposed method is illustrated by a case study.3. In order to improve the low efficiency, low precision and bad control effect of industrial process optimal control strategy when it applies to dynamic optimization, hierarchical dynamic optimization control strategy based on the TLBO algorithm and the nonlinear multi-model GPC (NMGPC) is proposed. Based on dynamic model of process, TLBO algorithm determines the optimal conditions and realizes economic performance optimization in the production process. The NMGPC allows the output of the system asymptotic tracking the reference trajectory and the operation more smoothly under the premise of ensuring the stability of the system. Finally, the simulation shows the feasibility and effectiveness of this hierarchical optimization control strategy.