Optimal Transition Pathways Toward a Low-Carbon Energy System in China: a Data-Driven Optimization Prediction Model in Machine Learning.

Abstract

The rapid expansion of renewable energy and the imperative for carbon reduction have prompted significant coal phase-outs. Coal is the largest contributor to energy-related carbon emissions globally, accounting for over one-third of the total. Coal-fired electric generating units (EGUs) play a significant role in these emissions, with over 5,000 units in China contributing to around 15% of global carbon emissions. These units, relatively young with an average age of less than 15 years, are facing challenges such as the absence of power purchase agreements, the prospect of early retirement, amid renewable energy growth, and ongoing retrofits for energy efficiency and carbon reduction. The transition pathway of the coal-fired power sector is crucial for its evolution and the integration of renewable energy. Hence, a data-driven optimization prediction model is introduced in this study, aiming to delineate an optimal transition pathway for the coal-fired power sector under different scenarios, guiding its evolution towards a low-carbon energy system. The model comprises two modules: the phase-out module and the retrofit optimization prediction module. A unified unit-level database, encompassing operational data from over 5000 coal-fired EGUs in China, as well as techno-economic information associated with 21 types of carbon reduction retrofits, serves as the foundation for the most cost-effective pathway towards a low-carbon transition in the power sector. The phase-out module predicts the phase-out and remaining capacities, including the potential portion replaced by renewable energy. The phase-out determination involves assessing the cost of replacing coal-fired power with renewable power generation, along with considerations of the economics and carbon emissions associated with units under normal operation before retirement. This laterally furnishes valuable information for comprehending the potential capacity for renewable generation, ensuring that the transition pathways in the coal-fired sector are realized in a manner that safeguards the stability and reliability of the future power system. The optimization prediction model employs machine learning algorithms consisting of the predictor and the optimizer. The predictor provides estimates for overall carbon reduction potential (CRP) for the coal-fired power sector, even for the power sector, as well as near-term levelized costs of carbon emissions reduction (LCOC) and electricity (LCOE), approached from the unit-level perspective. The optimizer identifies portfolios that maximize carbon emission potential while minimizing costs. This study ultimately provides a comprehensive analysis of the low-carbon transition pathway for the primary source of emissions in the energy sector, namely the coal-fired power sector, conducted from both techno-economic and environmental (specifically carbon reduction) standpoints, employing an optimization prediction model.