Carbon emissions prediction based on the Informer combination forecasting model: a case study of Sichuan

Abstract With the escalating global climate crisis, transitioning to a green, low-carbon economy has become an imperative worldwide. As a key southwestern Chinese province rich in clean energy resources, Sichuan faces significant challenges in balancing economic growth with emission reduction. To overcome the limitations of single-method forecasting approaches, this study develops a novel combination forecasting framework that integrates STIRPAT and Informer models through a Composite Driver Index (CDI). The methodology combines ridge regression-derived elasticity coefficients from an extended STIRPAT model with the temporal forecasting capability of the Informer model, using the CDI as a bridging mechanism. Based on Sichuan's updated emission data (1995–2023), our analysis indicates that the province will achieve carbon peaking by 2030 (280.1 Mt) under the baseline scenario and by 2029 (278.5 Mt) under the low-carbon scenario. The combination model demonstrates superior performance with reasonable error margins when validated against Tsinghua's DPEC model. Our findings highlight industrial restructuring and energy mix optimization as critical pathways, providing targeted policy recommendations for Sichuan's low-carbon transition while offering a transferable methodology for regional emission forecasting.