(681h) A Modified SQP Method for MPC of a Supercritical Pulverized Coal-Fired Power Plant during Cycling

In recent years, with rising interest in different sources of energy and focus on greenhouse gas emissions, renewable energy sources have been increasingly penetrated into the grid. As fossil fueled power plants will still play an important role in future power generation [1], optimal operations of such power plants during cycling will become more important. In particular, power plant cycling includes ramping the power generation rate up/down for load-following, as well as shut down or start up, to accommodate the renewable penetrations. To achieve the optimal profiles for cycling the power plant, the development and implementation of model predictive control (MPC) algorithms is critical for the improved plant operation under the challenging scenarios that arise during cycling.

In this work, the design and implementation of linear and nonlinear MPC strategies are considered for power plant cycling scenarios. The proposed MPC strategy is based on a modified sequential quadratic programming (SQP) method, which is designed for better acceptance of long steps and faster convergence. The modified SQP algorithm is built upon a backtracking line search framework and employs a group of relaxed step acceptance conditions. In the proposed algorithm, a step is first tested based on adapted Armijo conditions. If the first condition does not hold, then the second condition is tested, which may still have the step contribution for a sufficient decrease in the objective with a controlled infeasibility. If the second condition fails, then a defined third condition can enforce the step to have global convergence. The proposed MPC strategy is tested and compared to the following other linear and nonlinear MPC controllers developed: (i) a linear MPC control strategy based on dynamic matrix control (DMC); (ii) a nonlinear MPC based on a classical SQP method from the literature as benchmark [2]; and (iii) a nonlinear MPC based on the direct transcription method [3, 4, 5], which employs IPOPT [6] as nonlinear programming solver and ADOL-C [7] for improving derivative calculations.

The application system addressed corresponds to a supercritical pulverized coal-fired power plant with carbon capture. A number of plant-wide control scenarios will be discussed including: (i) power plant load-following, which simulates renewable penetrations into the grid; (ii) disturbance rejection in the coal-feed quality while maintaining the carbon capture rate; and (iii) trajectory tracking for power plant cycling over a relatively long period of operation, exploring the use of industrial power plant data. The closed-loop responses associated with different scenarios and different MPC algorithms will be analyzed. The MPC based on the modified SQP method will be compared to other MPC algorithms in terms of tracking errors and computational time performance.

References

1. Annual Energy Outlook 2015 with projection to 2040. Contract No.: DOE/EIA-0383, 2015.
2. F. Martinsen, L. T. Biegler and B. A. Foss. Application of optimization algorithms to nonlinear MPC. IFAC Proceedings 2002, 35:369-374.
3. F. V. Lima, R. Amrit, M. Tsapatsis, and P. Daoutidis. Nonlinear model predictive control of IGCC plants with membrane reactors for carbon capture. In Proceedings of the American Control Conference. Washington, DC, June 17-19, 2013.
4. F. V. Lima, X. He, R. Amrit and P. Daoutidis. Advanced control strategies for IGCC plants with membrane reactors for CO₂ capture. Process systems and materials for CO₂ capture: modeling, design, control and integration, A.I. Papadopoulos and P. Seferlis (eds.), Wiley, 2017.
5. X. He, Y. Wang, D. Bhattacharyya, F.V. Lima and R. Turton. Dynamic modeling and advanced control of post-combustion CO₂ capture plants. Chemical Engineering Research and Design, 2018, 131:430-439.
6. A. Wächter, L.T. Biegler. On the implementation of an interior-point filter line-search algorithm for large-scale nonlinear programming. Mathematical Programming, 2006,106(1):25-57.
7. A. Wälther, A. Griewank. Getting started with ADOL-C. Combinatorial Scientific Computing, 2012, 181-202.