AI-driven framework for assessing carbon-reduction costs in power grids: a case study of the Mengdong system

Abstract This work evaluates the systemic costs of carbon emission reduction in the Mengdong power grid and proposes strategies to support renewable energy integration and sustainable development. Using data from the global power plant database, it first analyzes the power grid structure, generation capacity, and emissions. A carbon locking-based cost measurement model is then developed to quantify emission reduction costs. The filtering curve method is introduced to assess the relationship between renewable energy accommodation and system costs. The work proposes a guidance mechanism combining market design, policy support, grid upgrades, and scheduling improvements. Results show the model achieves high prediction accuracy, with errors within 2%, and mean square error and mean absolute error values of 0.025 and 0.153, respectively. The resulting carbon-locking curve indicates a decelerating emission trend, consistent with the model’s forecasts. The carbon locking curve reflects a slowing growth trend in emissions, consistent with model forecasts. The model captures cost increases under high-emission scenarios and accurately evaluates transition costs under low-carbon pathways. Different from previous static or linear cost assessment models, this work innovatively combines the carbon locking theory with the data-driven method based on the global power plant database. It constructs a comprehensive cost assessment and prediction framework that integrates the dynamics of renewable energy consumption, system operation costs, and institutional path dependence. The work provides replicable methodological support for the low-carbon transformation of power systems in high-carbon locking areas.