| 基于机器学习的山西省资源型城市碳排放影响因素分析与达峰预测 |
| 摘要点击 453 全文点击 42 投稿时间：2025-03-26 修订日期：2025-06-10 |
| 查看HTML全文
查看全文 查看/发表评论 下载PDF阅读器 |
| 中文关键词 碳达峰 机器学习 XGBoost 山西省 资源型城市 |
| 英文关键词 carbon peaking machine learning XGBoost Shanxi Province resource-based cities |
| DOI 10.13227/j.hjkx.202503299 |
|
| 中文摘要 |
| 作为国家资源型经济转型综合配套改革试验区，预测山西省资源型城市碳排放情况对“双碳”目标有序推进具有重要意义. 以全球大气研究排放数据库（EDGAR）2000~2023年山西省11 个地市的碳排放量为研究样本，基于随机森林算法识别碳排放的关键影响因素，构建PSO-LSTM-XGBoost混合模型预测山西省资源型城市碳达峰趋势. 结果表明：①2000~2023年山西省碳排放总体上呈现出波动上升的趋势，除临汾市近年来有下降趋势外，其余各地市皆波动上升. ②能源消费总量、城市化水平、第二产业GDP、总GDP、公路客运量、R&D 内部支出、第一产业结构以及第一产业GDP 是影响山西省碳排放的关键影响因素. ③PSO-LSTM-XGBoost模型在多种优化模型中预测误差最低，拟合优度最高，为山西省碳排放情景预测提供了相对精准的工具. ④基准情景下，山西省于2031年达到峰值7.05×108 t；低碳情景较基准情景提前3a实现碳达峰且峰值降低7.94%；高碳情景下，达峰将延迟至2034年且峰值近8×108 t. 基于此，研究提出具体到市、优化产能结构和跨区协同治理的低碳发展建议，可为精准施策和有序降碳提供参考. |
| 英文摘要 |
| As a national resource-based economy transformation comprehensive supporting reform pilot area， predicting the carbon emissions of resource-based cities in Shanxi Province is of great significance for the orderly promotion of the dual-carbon target. The study takes the carbon emissions of 11 cities in Shanxi Province from 2000 to 2023 provided by the Emissions Database for Global Atmospheric Research （EDGAR） as a sample， identifies the key influencing factors of carbon emissions based on the random forest algorithm， and constructs a PSO-LSTM-XGBoost hybrid model to predict the carbon peak trend of resource-based cities in Shanxi Province. The results showed that： ① From 2000 to 2023， the carbon emission in Shanxi Province demonstrated a fluctuating upward trend in general， except for Linfen City， which had a downward trend in recent years， and the remaining cities had a fluctuating upward trend. ② Total energy consumption， urbanization level， secondary industry GDP， total GDP， highway passenger traffic， R&D internal expenditure， primary industry structure， and primary industry GDP were the key influencing factors of carbon emission in Shanxi Province. ③ The PSO-LSTM-XGBoost model had the lowest prediction error and the highest goodness of fit among multiple optimization models， providing a relatively accurate tool for the subsequent prediction of carbon emission scenarios in Shanxi Province. ④ Under the baseline scenario， Shanxi Province reaches a peak of 705 million tons in 2031， and the low-carbon scenario achieves peak carbon three years earlier than in the baseline scenario and reduces the peak value by 7.94%， while under the high-carbon scenario， the peak value is delayed until 2034， and the peak value is nearly 800 million tons. Based on this， the study proposes city-specific， optimized capacity structure and cross-region synergistic governance for low-carbon development， which provides a reference for city-specific policymaking and precise carbon reduction. |
