HMAMRL: Multicriterion Flexible Coordinated Control for Coal-Fired Power Generation Systems under Wide Load Operation

Area of Science:

• Engineering
• Artificial Intelligence
• Energy Systems

Background:

• Flexible and efficient wide-load tracking in coal-fired power generation systems (CPGSs) is essential for integrating renewable energy sources.
• Dynamic characteristics and task distribution differences pose challenges during wide-load operation of thermal power units.

Purpose of the Study:

• To propose a novel hierarchical model-agnostic meta reinforcement learning (HMAMRL) framework for robust wide-load tracking in CPGSs.
• To address the challenges of dynamic characteristics and task distribution in thermal power unit operation.
• To ensure effective generalization across different load conditions.

Main Methods:

• A hierarchical model-agnostic meta reinforcement learning (HMAMRL) framework combining inner and outer meta-learning.
• An adaptive multicriterion reward function to balance load tracking, coal consumption, and input fluctuation costs.
• A truncated proximal policy optimization (TPPO) algorithm for precise load control within physical constraints.

Main Results:

• The proposed HMAMRL framework demonstrated effective and superior performance in wide-load tracking.
• The adaptive reward function successfully balanced multiple cost criteria.
• The TPPO algorithm ensured precise control within operational limits.

Conclusions:

• The HMAMRL framework provides a robust solution for flexible wide-load tracking in CPGSs.
• The study highlights the potential of meta-reinforcement learning for enhancing power system flexibility and renewable energy integration.
• The proposed methods are validated on 160 and 1000 MW CPGSs.