Rearch On Performance Assessment For Hierarchical Structure MPC

With the emerging growth of the scale of industrial processes and the complexity of the plant in recent years, hierarchical multilayer control system composed of scheduling layer,optimization layer,control layer etc.has widespread application in industrial process due to its advantages in system safety,structure flexibility and control robustness,guaranteeing demands in different aspects，including safety,stability,reliability and efficiency of the plant operation. In its control layer, base control layer and constraint control layer are intimately connected,responsible for disturbance restraining,setpoint tracking,economic performance optimization etc.different control task. The performance assessment of all respects is what the engineers care most. The thesis will focus on the following research topics based on current theory and industrial appliation of control system performance assessment mainly in multimode process. This research will investigate the following issues on controller performance benchmark modeling and ecomonic performance assessment in multimode process:1. A method that the sample data is first classified to the coressponding mode then carried out performance assessment is proposed in multimode process in base control layer. Simply employing covariance based benchmark may result in wrong assessment because of the different performance benchmarks of different operation mode. A new method based on conbined shape similarity factor and geometry distance similarty factor classification method is proposed to first determine the mode that real time data belongs to,then carry out performance assessment,which guarantees objectivity and validaty of data driven covariance benchmark performance method.2. A data driven LQG(Linear Quadratic Gaussian) controller performance assessment method is proposed focusing on trasition stage between different operation modes. Trasition stage between different operation mode is characterized as LTV(Linear time varying) property. LQG performance assessment method is first introduced into LTV system to weigh the variance between input and output. Generalized observer Markov parameter for LTV system can be obtained from history benchmark database, through which the system process model and noise model can be identified. The optimal input and output variance for the LTV system stochastic performance part can be obtained by introducing LQG benchmark,constituting joint input and output LTV system performance benchmark based on optimal covariance.3. In the constraint control layer, in order to investigate the effect on the economic performance assessment from the model uncertainty,first a robust linear programming method based on the second order cone programming is introduced to describing model uncertainty, then the strategy of constraint adjustment and variance adjustment is considered to improve economic performance, resulting in carrying out MPC steady economic performance assessment under uncertainty.4. When the MPC steady optimization problem can not be solved feasibly, soft contraint adjustment measure should be take to relax contraints. A multi-parametric programming algorithm is proposed to divide the variation range of coefficient of the soft constraint into different regions. It can determine the most suitable soft constraints adjustment to achieve optimal economic objective by comparing economic objective value in the different region.