(394g) Domain-Compliant AI Adoption Enhances the Operation Excellence of Power Systems

The slow artificial intelligence (AI) adoption in industrial operations of energy and chemical process complexes is a missed opportunity to exploit AI potential for operation excellence of industrial power systems that supports industrial decarbonisation. A key issue in the adoption of the AI based decision support system for power system’ operation is the domain inconsistent decisions made by AI based analytics which can not be verified or implemented in industrial environment. To this potential issue barring the AI adoption in industry, we present Machine Learning and OPTimisation (MLOPT) framework that is complimented with human-in-the-loop (HITL) to derive domain-consistent optimal solutions for the operation excellence of industrial systems. We illustrate the effectiveness of our proposed framework through a case study focused on improving the thermal efficiency and lowering the turbine heat rate of a 660 MW supercritical coal-fired power plant [1]. By incorporating domain knowledge into our optimization process, our approach generates solutions that harmonize with the plant's operational behaviors and seamlessly integrate into its control systems. Empirical validation shows an average enhancement in thermal efficiency by 0.64% and an average decrease in turbine heat rate by 93 kJ/kWh. Scaling our analysis to 59 global coal power plants of similar capacity and fuel type suggests a cumulative lifetime reduction of 156.4 million tons of carbon emissions. These findings highlight the transformative potential of our HITL-MLOPT framework in delivering practical, domain-compliant solutions for industrial decarbonization, offering a scalable approach to mitigating the environmental impact of coal-based power generation worldwide.

Reference

[1] Ashraf, W. M., Dua, V., and Debnath, R., 2025, "Domain Consistent Industrial Decarbonisation of Global Coal Power Plants," arXiv preprint arXiv:2503.03571.