(545g) Distributed and Multiple Model Predictive Control for Rapid Load-Following Operation of Supercritical Pulverized Coal Power Plants

With the increasing penetration of renewable energy sources, many fossil-fuel-based thermal power plants are forced to cycle their load rapidly and frequently over a wide range leading to increased emissions, decreased thermal efficiency, and increased equipment damage [1-3]. Even the newer generation supercritical pulverized coal (SCPC) power plants that were base loaded earlier due to their high efficiency are being subjected to extensive cycling [4]. As the load of the SCPC plants is reduced, different locations inside the waterwall section transitions to subcritical regime at varying load. This is partly due to the sliding pressure operation of the SCPC plants; thus, the boiler pressure decreases with the decrease in load and vice versa. Transition from/to supercritical to/from subcritical range is associated with high nonlinearity and large change in the thermo-physical properties of steam. Furthermore, the resulting change in the hydraulics also affect the heat transfer characteristics. Therefore, rapid load-following operation of these plants in a wide range of load is challenging without unacceptable thermal excursions. Traditional coordinated control systems in existing power plants are inadequate to address these challenging control problems. In this work, two advanced control, multi-model predictive control (MMPC) and distributed model predictive control (DMPC), are developed and investigated for large range of operating scenarios.

A high-fidelity model of an SCPC plant that is validated with the industrial data is used for developing and evaluating the control algorithms. The MMPC approach partitions the desired operating range of 40% to 100% into multiple regions optimally selected based on the nonlinearity of the region and operating conditions. Linear or simple nonlinear models are optimally identified for each of these ranges. Model weights for an operating region are estimated from Bayesian statistics, updated recursively, based on the model residuals [5]. In the DMPC approach, the main control objectives such as the load control, steam temperature control, pressure control, etc. are split into multiple sub-MPC’s with an efficient communication strategy of their respective states and inputs at desired intervals. This approach is not only computationally easier to solve and implement [6], but also facilitates difference in execution time of the underlying sub-MPCS.

Results are presented to validate the performance of the controllers for disturbance rejection at the nominal load, followed by studies of their relative performance under load-following at different ramp rates across the whole range. During rapid load-following, the magnitude of temperature excursion constrains the rate of change as large temperature excursions can lead to low efficiencies and equipment damage. In this work, the adverse effect of temperature is quantified in terms of evolving thermal stress on boiler components, available through the high-fidelity boiler model that is incorporated into the dynamic model [7]. One of the issues with the DMPC is that they can lead to suboptimal performance. Therefore, in this work, we also investigate synergistic and coordinated control using both MMPC and DMPC to take advantage of both control approaches. The efficacy of the proposed approaches is evaluated by comparing their performances in terms of optimality, constraint satisfaction, convergence characteristics, and computational expense with the industry-standard coordinated control systems.

References:

[1] R. K. Smith, “Analysis of hourly generation patterns at large coal-fired units and implications of transitioning from baseload to load-following electricity supplier,” Journal of Modern Power Systems and Clean Energy, (2018).

[2] S. Hesler, “Mitigating the Effects of Flexible Operations on Coal-Fired Power Plants,” Power Magazine, August 1 (2011).

[3] A. Shibli and J. Ford, “Damage to coal power plants due to cyclic operation,” Coal Power Plant Materials and Life Assessment: Developments and Applications, Elsevier Inc., (2014).

[4] P. Sarda, E. Hedrick, K. Reynolds, D. Bhattacharyya, S.E. Zitney, B. Omell, “Development of a Dynamic Model and Control System for Load-Following Studies of Supercritical Pulverized Coal Power Plants”, Processes, Vol. 6, Pages 226, (2018).

[5] M. Kuure-Kinsey, B. Bequette, “Multiple Model Predictive Control Strategy for Disturbance Rejection,” Industrial and Engineering Chemistry Research, Vol. 49, Pages 7983-7989, (2010).

[6] P.D. Christofides, R. Scattolini, D. Muñoz de la Peña, J. Liu, “Distributed Model Predictive Control: A Tutorial Review and Future Research Directions,” Computer & Chemical Engineering, Vol. 51, Pages 21-41, (2013).

[7] K. Hedrick, E. Hedrick, B. Omell, S.E. Zitney, and D. Bhattacharyya, “Dynamic Modeling, Parameter Estimation, and Data Reconciliation of a Supercritical Pulverized Coal-Fired Boiler,” Manuscript under Review