A Multi-Time-Scale Carbon Emission Forecasting Method for Regional Power Grids Based on an Improved STIRPAT Model

Carbon emission accounting in power grids effectively measures the environmental impact of the system, providing policymakers and operators with a scientific basis for evaluating emission reduction efforts. Carbon emission forecasting further aids in optimizing grid dispatch and enhancing energy utilization efficiency. However, existing forecasting methods rely largely on static assessments, which limits their ability to capture the dynamic characteristics of power systems across multiple time scales and to account for key factors like load fluctuations and renewable energy integration. Although the STIRPAT model can quantitatively analyze emission-driving factors, its dependence on annual data for variables like population and economic indicators introduces latency in regional grid forecasting. In response, this paper proposed a multi-time-scale regional grid carbon emission forecasting model, built on an optimized STIRPAT model with refined input parameters. The model incorporated load, energy structure, and carbon emission factor data across different time scales, enhancing prediction accuracy and addressing the limitations of traditional models. This work first established the rationale for substituting the original STIRPAT parameters with grid load, carbon factor, and energy structure variables. It then introduced a method for acquiring regional grid data based on verified statistical sources. Finally, a simulation using provincial grids’ datasets verified the improved model’s effectiveness in carbon emission forecasting in 15 minutes.