主管:中国科学院
主办:中国优选法统筹法与经济数学研究会
   中国科学院科技战略咨询研究院

中国管理科学 ›› 2025, Vol. 33 ›› Issue (3): 339-350.doi: 10.16381/j.cnki.issn1003-207x.2022.0097cstr: 32146.14/j.cnki.issn1003-207x.2022.0097

• • 上一篇    下一篇

基于LSTM神经网络的中国省级碳达峰路径分析

章高敏1,2,3, 王腾1,2, 娄渊雨1,2, 关忠诚1,2, 郑海军2, 李强2, 武佳倩4()   

  1. 1.中国科学院大学公共政策与管理学院,北京 100049
    2.中国科学院科技战略咨询研究院,北京 100190
    3.香港科技大学(广州)碳中和与气候变化学域,广东 广州 511458
    4.郑州大学管理学院,河南 郑州 450001
  • 收稿日期:2022-01-14 修回日期:2022-04-30 出版日期:2025-03-25 发布日期:2025-04-07
  • 通讯作者: 武佳倩 E-mail:wujiaqian@zzu.edu.cn
  • 基金资助:
    国家自然科学基金青年项目(72004206);河南省哲学社会科学规划项目(2020CJJ097)

Research on China's Provincial Carbon Emission Peak Path Based on a LSTM Neural Network Approach

Gaomin Zhang1,2,3, Teng Wang1,2, Yuanyu Lou1,2, Zhongcheng Guan1,2, Haijun Zheng2, Qiang Li2, Jiaqian Wu4()   

  1. 1.School of Public Policy and Management,University of Chinese Academy of Sciences,Beijing 100049,China
    2.Institutes of Science and Development,Chinese Academy of Sciences,Beijing 100190,China
    3.Carbon Neutrality and Climate Chang Thrust,Society Hub,The Hong Kong University of Science and Technology (Guangzhou),Guangzhou,511458,China
    4.School of Management,Zhengzhou University,Zhengzhou 450001,China
  • Received:2022-01-14 Revised:2022-04-30 Online:2025-03-25 Published:2025-04-07
  • Contact: Jiaqian Wu E-mail:wujiaqian@zzu.edu.cn

摘要:

作为世界最大的碳排放国家和第二大经济体,中国承诺二氧化碳排放量将于2030年左右达到峰值,届时二氧化碳排放强度较2005年下降60%~65%。然而,中国各省份之间存在较大的异质性,其碳达峰路径不能“一刀切”。基于各省份的“十四五”规划,本文设计了基准、绿色发展、高碳发展三种情境,采用LSTM神经网络动态预测2020-2040年中国及各省份的碳达峰路径,并根据各省份碳排放强度、累积碳排放量和达峰时间三因素分析了其适合的达峰路径。结果表明:在不同的情境下中国都将在2030年之前实现碳达峰目标,峰值水平在10884-11792百万吨;24个省份能够在至少一种情境下实现2030年之前碳达峰目标,且大部分省份都呈现绿色发展情境下达峰时间早、峰值低,高碳情境达峰时间晚、峰值高的特点;北京、上海、福建、浙江等省份有望在2035年之后实现负碳。研究结果对中国合理制定2030年碳达峰路径措施,统筹分配减排任务具有重要参考价值。

关键词: 碳达峰路径, LSTM, 神经网络, 情境分析

Abstract:

As the world’s largest carbon emitter and second largest economy, China has pledged that its carbon emissions will peak before 2030. Meanwhile, the intensity of carbon emissions will drop by 60%~65% compared to that of 2005. However, due to the varying carbon emissions trajectories of individual provinces, their carbon emissions paths cannot be one-size-fits-all. Three scenarios, i.e., benchmark, green development, and high-carbon, are designed based on the “14th Five-Year Plan” of each province. Based on LSTM, the carbon emissions of each province from 2020—2040 are predicted under different scenarios. Finally, the appropriate peak paths for individual provinces are analyzed based simultaneously on carbon emission intensity, cumulative carbon emissions and peak time. The results show that the higher the growths rate is, the later the peak time will arrive; China will achieve its carbon emissions peak before 2030, with a peak level of 11884~11792Mt; 24 provinces can achieve the carbon emissions peak before 2030 under at least one scenario; Beijing, Shanghai, Zhejiang, etc. can achieve negative carbon emissions after 2035; An important reference is provided for national policy makers to allocate emission reduction tasks and optimize emission reduction policies.

Key words: carbon emissions peak path, LSTM, neural network, scenario analysis

中图分类号: