Dynamic Optimization Algorithms for Baseload Power Plant Cycling under Variable Renewable Energy

The growing deployment of variable renewable energy (VRE) sources, such as wind and solar, is mainly due to the decline in the cost of renewable technologies and the increase of societal and cultural pressures. Solar and wind power generation are also known to have zero marginal costs and fuel emissions during dispatch. Thereby, the VRE from these sources should be prioritized when available. However, the rapid deployment of VRE has heightened concerns regarding the challenges in the integration between fossil-fueled and renewable energy systems. The high variability introduced by the VRE as well as the limited alignment between demand and wind/solar power generation led to the increased need of dispatchable energy sources such as baseload natural gas- and coal-fired power plants to cycle their power outputs more often to reliably supply the net load. The increasing power plant cycling can introduce unexpected inefficiencies into the system that potentially incur higher costs, emissions, and wear-and-tear, as the power plants are no longer operating at their optimal design points. In this dissertation, dynamic optimization algorithms are developed and implemented for baseload power plant cycling under VRE penetration. Specifically, two different dynamic optimization strategies are developed for the minute and hourly time scales of grid operation. The minute-level strategy is based on a mixed-integer linear programming (MILP) formulation for dynamic dispatch of energy systems, such as natural gas- and coal-fired power plants and sodium sulfur batteries, under VRE while considering power plant equipment health-related constraints. The hourly-level strategy is based on a Nonlinear Multi-objective dynamic real-time Predictive Optimization (NMPO) implemented in a supercritical pulverized coal-fired (SCPC) power plant with a postcombustion carbon capture system (CCS), considering economic and environmental objectives. Different strategies are employed and explored to improve computational tractability, such as mathematical reformulations, automatic differentiation (AD), and parallelization of a metaheuristic particle swarm optimization (PSO) component. The MILP-based dynamic dispatch framework is used to simulate case studies considering different loads and renewable penetration levels for a suite of energy systems. The results show that grid flexibility is mostly provided by the natural gas power plant, while the batteries are used sparingly. Additionally, considering the post-optimization equivalent carbon analysis, the environmental performance is intrinsically connected to grid flexibility and the level of VRE penetration. The stress results reinforce the necessity of further considering and including equipment health-related constraints during dispatch. The results of the NMPO successfully implemented for a large-scale SCPC-CCS show that