跳转至

迈向可持续的人工智能：深度学习模型碳足迹的综合综述与比较分析

文章背景与核心概要

随着人工智能（AI）和机器学习（ML）技术的迅猛发展，其带来的高能耗和碳排放问题引发了学界和业界对环境可持续性的广泛关注。大规模深度学习（DL）架构在提供卓越预测能力的同时，也消耗了巨大的计算资源。为了应对这一挑战，本文系统综述了绿色AI、绿色DL以及旨在减少环境影响的优化技术，并评估了多种碳排放测量工具。

通过基于CPU的实验配置，研究团队针对多标签分类任务对六种深度学习模型进行了实证评估，旨在量化并比较它们的碳足迹，找出模型生命周期中碳排放最多的阶段。研究结果表明，训练阶段是碳排放的主要来源；同时，更高的架构复杂度并不总是能带来成比例的准确率提升。这一发现强调了在模型选择和系统设计中，必须平衡预测性能与环境可持续性。

文章元数据

  • 作者： Samar Garrab, Sarra Boughriou, Manel BenSassi
  • 主要学科： 人工智能 (cs.AI)
  • 次要学科： 计算机与社会 (cs.CY)、机器学习 (cs.LG)、软件工程 (cs.SE)
  • 提交日期： 2026年8月7日
  • 期刊引用： Applied Intelligence, vol. 56, Article 173, 2026
  • 标识符：
  • arXiv: 2608.09998 [cs.AI]
  • DOI: 10.1007/s10489-026-07208-y

摘要

人工智能（AI）和机器学习（ML）已成为支持和自动化人类复杂任务的强大工具。尽管它们带来了诸多益处，但人们对其环境影响的关注日益增加，这主要是由于它们的高能耗和伴随而来的碳排放。随着大规模模型（特别是深度学习（DL）架构）的不断部署，这种担忧显得尤为重要，因为这些架构虽然具备高级预测能力，但需要大量的计算资源。

本文对旨在减少AI模型环境影响的绿色AI、绿色DL和优化技术的研究进行了系统性综述。此外，我们审查并比较了几种用于估算AI算法产生碳排放的碳测量工具。为了对综述进行补充，我们使用基于CPU的实验装置进行了实证评估，在多标签分类任务中实现了六个DL模型。其目的是量体并比较它们的总体碳排放，并确定DL生命周期的哪些阶段对总碳足迹的贡献最大。

结果表明，训练阶段是碳排放的主要来源。此外，研究发现架构复杂度的增加并不系统性地转化为成比例的准确率提升，这凸显了仔细平衡预测性能与环境成本的重要性。这些结果进一步证明了将可持续性考量整合到模型选择和AI系统设计中的必要性。

Artificial Intelligence (AI) and Machine Learning (ML) have become powerful tools for supporting and automating complex human tasks. Despite their benefits, growing attention has been directed toward their environmental implications, primarily due to their high energy demands and associated carbon emissions. This concern is particularly relevant in light of the increasing deployment of large-scale models, especially Deep Learning (DL) architectures, which provide advanced predictive capabilities but require substantial computational resources.

This paper presents a systematic review of research on Green AI, Green DL, and optimization techniques aimed at reducing the environmental impact of AI models. In addition, we examine and compare several carbon measurement tools for estimating emissions generated by AI algorithms. To complement the review, we conducted an empirical evaluation using a CPU-based experimental setup, in which six DL models were implemented for a multi-label classification task. The objective was to quantify and compare their overall carbon emissions and to determine which stages of the DL lifecycle contribute most significantly to the total footprint.

The results show that the training phase is the primary source of emissions. Moreover, the findings reveal that increased architectural complexity does not systematically translate into proportional accuracy gains, highlighting the importance of carefully balancing predictive performance and environmental cost. These results reinforce the need to integrate sustainability considerations into model selection and AI system design.

核心要点

  1. 训练阶段的影响： 实证评估证实，与推理和其他生命周期阶段相比，训练深度学习模型占据了绝大部分的碳排放。
  2. 复杂度与准确率的权衡： 更高的架构复杂度并不能保证成比例的性能提升，这表明市场迫切需要更高效的模型架构。
  3. 绿色AI的整合： 可持续性指标和优化技术必须成为软件工程、机器学习研究以及部署流水线中的标准实践。
  1. Training Phase Impact: The empirical evaluation confirms that training deep learning models accounts for the vast majority of carbon emissions compared to inference and other lifecycle stages.
  2. Complexity vs. Accuracy Trade-off: Greater architectural complexity does not guarantee proportional performance improvements, signaling a need for more efficient model architectures.
  3. Green AI Integration: Sustainability metrics and optimization techniques must become standard practices in software engineering, machine learning research, and deployment pipelines.