Research on regional carbon emission scenario simulation based on GA-BP-KDE under uncertain conditions

Abstract

The continuous advancement of the “dual carbon” goal presents a persistent challenge for regions striving to balance rapid economic growth with sustainable development. Consequently, investigating scientifically efficient regional carbon emission reduction pathways has become increasingly significant. This study utilizes carbon emission data and related impact factors from 21 cities in Sichuan province spanning 2011 to 2022 to construct a GA-BP-KDE carbon emission interval prediction model. This model, based on the GA-BP point prediction model, accounts for the inherent uncertainty in carbon emission forecasting. Eight scenarios were employed to predict carbon emission intervals and potential “carbon peak” timeframes for Sichuan province from 2021 to 2040. These scenarios were categorized into three groups: “peak,” “possible peak,” and “unable to peak.” Subsequently, a tailored cascade peak pathway was developed for the 21 cities in Sichuan province. The findings reveal: (1) Pearson coefficient correlation measure results indicate that the overall performance of the genetic algorithm-Back propagation network (GA-BP) model (0.985) surpassed that of the Extreme Learning Machine (ELM) (0.900), Random Forest (RF) (0.981), and support vector machine (SVR) (0.982). The validity of the GA-BP-KDE interval prediction model was evaluated using PICP and PINAW metrics. (2) Under baseline, ideal, and low-carbon scenarios, Sichuan province is projected to definitively reach its carbon peak by 2030. In the weak emission reduction scenario, a 2030 peak remains possible. (3) The 21 cities in Sichuan province were classified into three echelons based on their average peak times, with Chengdu and Mianyang positioned in the first echelon. This research introduces a novel approach to studying “carbon peak” pathways, potentially providing decision-makers with more reliable and comprehensive information for formulating rational and effective carbon emission reduction measures.