Scenario-based Emission Simulations Using Time-Series Generative Adversarial Networks

Integrating carbon emissions accounting with data analytics in electricity generation is essential for achieving sustainability goals and managing risks, supporting global decarbonization efforts. This paper investigates the spatiotemporal correlation of coupling data using a sampling method with a time-variant generative architecture, producing synthetic time-relevant emissions to address the scarcity and inaccuracy of raw data. The proposed evaluation framework, the Coupling Error Ratio Matrix (CERM), offers an intuitive representation of the coupled flow of power and carbon emission scenarios and provides analytical foundations for learning and training optimizations by identifying aggregated features with specific indicators. The evaluation method is initially validated using the IEEE 118-bus system. Additionally, the scalability and efficiency of the framework are confirmed using the power grid of Hainan province in China and photovoltaic generation outputs.