Date of Graduation
Statler College of Engineering and Mineral Resources
Chemical and Biomedical Engineering
Fernando V. Lima
Design and Implementation of Model Predictive Control Strategies for Improved Power Plant Cycling
With the increasing focus on renewable energy sources, traditional power plants such as coal-fired power plants will have to cycle their load to accommodate the penetration of renewables into the power grid. Significant overshooting and oscillatory performance may occur during cycling operations if classical feedback control strategies are employed for plantwide control. To minimize the impact when power plants are operating away from their designed conditions, model-based optimal control strategies would need to be developed for improved power plant performance during cycling.
In this thesis, model predictive control (MPC) strategies are designed and implemented for improved power plant cycling. The MPC strategies addressed correspond to a dynamic matrix control (DMC)-based linear MPC, a classical sequential quadratic programming (SQP)-based nonlinear MPC, a direct transcription-based nonlinear MPC and a proposed modified SQP-based nonlinear MPC. The proposed modified SQP algorithm is based on the backtracking line search framework, which employs a group of relaxed step acceptance conditions for faster convergence. The numerical results for motivating examples, which are selected from literature problem sets, served as proof of concept to verify that the proposed modified SQP has the potential for implementation on high-dimensional systems.
To illustrate the tracking performance and computational efficiency of the developed MPC strategies, three processes of different dimensionalities are addressed. The first process is an integrated gasification combined cycling power plant with a water-gas shift membrane reactor (IGCC-MR), which is represented by a first-principles and simplified systems-level nonlinear model in MATLAB. For this application, a setpoint tracking scenario simulating a step increase in power demand, a disturbance rejection scenario simulating a coal feed quality change, and a trajectory tracking scenario simulating a wind power penetration into the power grid are presented. The second application is an aqueous monoethanolamine (MEA)-based carbon capture process as part of a supercritical pulverized coal-fired (SCPC) power plant, whose model is built in Aspen Plus Dynamics. For this system, disturbance rejection scenarios considering a ramp decrease in the flue gas flow rate as well as wind power penetration, and a scenario considering a combination of disturbance rejection and setpoint tracking are addressed. The third process is the entire SCPC power plant with MEA-based carbon capture (SCPC-MEA), which simulation is also built in Aspen Plus Dynamics. Trajectory tracking and disturbance rejection scenarios associated with wind and solar power penetrations are presented for this process. The MPC implementations on the three processes for the different scenarios addressed are successful. The closed-loop results show that the proposed modified SQP-based nonlinear MPC enhances the tracking performance by up to 96% when compared to the DMC-based linear MPC in terms of integral squared error results. The novel approach also improves the MPC computational efficiency by 20% when compared to classical SQP-based and direct transcription-based nonlinear MPCs.
He, Xin, "Design and Implementation of Model Predictive Control Strategies for Improved Power Plant Cycling" (2019). Graduate Theses, Dissertations, and Problem Reports. 3937.