Life cycle carbon footprint analysis of deep load regulation in coal-fired power plants based on machine learning: A case study of a 1000 MW unit in Hunan province

Abstract

As global climate change intensifies, China has proposed goals of reaching peak emissions before 2030 and carbon neutrality by 2060. Achieving these targets will require coal power plants, the backbone of China's electricity system, to provide more flexible peaking support as renewable penetrations rise. However, the deep peak shaving operation needed to bolster grid flexibility may significantly alter coal plants' carbon emission characteristics. Here we develop a comprehensive, machine learning-based model to analyze the lifecycle carbon footprint of a representative 1000 MW coal unit in Hunan Province under various loading regimes. Our model encompasses emissions from fuel transport, combustion, environmental controls, and other stages, finding that combustion accounts for approximately 60 % of total CO₂. Applying the model to scenarios with deep peak shaving, we show that carbon intensity may rise by 15–30 % compared to baseload conditions. Drawing on these insights, we propose operational strategies such as optimizing combustion and environmental controls at 20–50 % loading and using intelligent dispatch to minimize startup/shutdown cycles below 30 % loading. We demonstrate that such targeted measures could avoid approximately 5 % of CO₂ emissions when the unit operates below 30 % capacity for over 1000 h per year.