Particle Swarm Optimization for Variable Load Rate and Coal Consumption Enhancement in Thermal Power Units

Abstract

The global energy transition requires coal-fired power plants to transition from base-load sources to flexible regulation resources, necessitating a critical balance between rapid load adjustment capabilities and operational economic efficiency. To address the nonlinear, strongly coupled, and multi-constraint characteristics of thermal power units, this study systematically evaluates various intelligent optimization strategies, including gradient descent, reinforcement learning, and evolutionary algorithms. A 660 MW supercritical once-through boiler is used as the research object. The particle swarm optimization (PSO) algorithm is identified as the optimal methodology. Comparative simulation studies reveal that while other algorithms exhibit performance degradation in specific load segments (e.g., momentum gradient descent dropped by 5.32% in the 60–70% load segment), PSO achieved stable positive optimization across the entire 50–100% load range without any negative growth, yielding an average improvement rate of 5.47% in the variable load rate. By collaboratively optimizing total feedwater flow, total air flow, and total coal flow, the PSO strategy significantly enhanced the unit’s dynamic response and economic performance. In terms of flexibility, the variable load rate improvement reached up to 8.7% in the 300 MW–400 MW load interval and peaked at 8.27% in the 400 MW–500 MW interval. Regarding economic operation, the optimization reduced total coal consumption across all load stages by 1.0% to 3.7%, with the most prominent energy-saving effect observed in the 60–70% load stage, where coal consumption decreased from 274.00 t/h to 263.98 t/h (a 3.65% reduction). Furthermore, power generation coal consumption was reduced by 1.12% to 2.62% across all intervals, notably dropping from 641.01 g/kWh to 624.26 g/kWh in the 50–60% load stage. These findings demonstrate that the PSO-based data-driven optimization strategy successfully achieves a dual enhancement of flexible regulation capability and operational economy, providing a feasible technical path for the intelligent upgrading of thermal power units in modern power grids.