Research On Model Predictive Control Strategies In Thermal Process

The generator units are developing towards large capacity and high parameters. The characters of thermal process become more and more complex and the demand to quality of control becomes more and more strict. The performances of general PID control strategy which is widely used now can not satisfy this demand. The reasearch of advanced control strategy applicated in thermal process has attached increasing attention in recent years. Based on pioneer's fruits, some model predictive control strategies of thermal plant control are researched in this dissertation, and the main research works are as follows:(1) The main steam temperature possesses non-linear features under load changes. Aim to that, a nonlinear model predictive control strategy is proposed. A linear parameter varying (LPV) model is used to represent this nonlinear feature, which is identified using test data of selected working-points and transient processdata. In each sample time, the model is linearized through the scheduling variable, and the linear predictive controller is computed. This method keeps better control performance than traditional PID and linear MPC, and use low cost test. The effectiveness is verified in simulation platform of real 200MW thermal power plant units.(2) Model predictive control strategies for cascade system are researched focusing on the two-stage spray desuperheating process of superheated steam. Using the actual data acquired form 300MW units of Wen Zhou power plant, tracking and rejection disturbance performance of three diferent model predictive control strategies for cascade system are compared. A single controler strategy using predictive error as feedfoward variable is proposed. Single controler are used to control the cascade system incorporating multiple segments. The predictive error of output from the prior segment are selected as the feedforward variable of the latter segment. This method rejected the effect of unmeasured disturbance, and enlarged the feasible region of controler.(3) Aiming to satisfy the control requirements of fast and smooth in Coordinated control system of the turbine -generator units, the performances of model predictive control and bang-bang control, which was considered a classical time optimal control strategy are analyzed first. Then, two time-optimal predivtive control strategies are proposed. In the first method, an objective funtion is constructed, which incorporate both quadratic performance index which reflected the smooth needs and time length index which reflected the minimum time demand and terminal equality constraints with infinite horizon are added. The proof of algorithm stability assured that a sequence of control action can be acquired to make closed-loop system asymptotically stable in each sample time. By that the system can be state stable quickly through this method. In the second method, the first time-optimal model predictive control strategy without terminal constraints is used to make the system close to the target as soon as possible. Then, traditional MPC method is used instead. It weaks the limit of terminal constraints and further reduce the time to reach the control target. Simulation results verified that these two methods both keep better performance in speedability than traditional MPC and satisfy the smooth requirements.