Review of Research on Artificial Intelligence-Based Carbon Emission Prediction

Weixian Wang, Lu Zhang, Mingyu Pan, Jie Zhuo, Sirui Lu
Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing, PMLR 278:260-269, 2025.

Abstract

Under the global climate governance framework, carbon emission prediction has emerged as a pivotal technology for low-carbon energy transition. This review systematically examines the advancements in artificial intelligence-based carbon emission forecasting, revealing the evolutionary dynamics between traditional statistical methods and data-driven models. A novel “data-model-scenario" triadic analytical framework is proposed to deconstruct core challenges in this field. The study demonstrates that conventional approaches (e.g., ARIMA, grey models) exhibit structural deficiencies in high renewable energy penetration scenarios, including poor adaptability to abrupt changes and low cross-source data integration efficiency ($<$60%). In contrast, data-driven methods (XGBoost, LSTM, Transformer) achieve significant accuracy improvements through dynamic modeling and feature decoupling. Hybrid paradigms integrating physical constraints and multimodal alignment show promise in bridging the mechanism-data gap, yet face persistent challenges: inefficient multi-source data fusion (feature alignment success rate $<$60%), delayed response to sudden scenarios (recovery time $>$30 minutes), and computational-precision tradeoffs in edge deployment. The paper proposes a “dual-driven" evolutionary path for hybrid modeling and constructs a multi-scale scenario linkage matrix, providing theoretical guidance for next-generation prediction frameworks. Emerging technologies such as digital twins and federated meta-learning are highlighted as critical directions for future research.

Cite this Paper

BibTeX
@InProceedings{pmlr-v278-wang25d, title = {Review of Research on Artificial Intelligence-Based Carbon Emission Prediction}, author = {Wang, Weixian and Zhang, Lu and Pan, Mingyu and Zhuo, Jie and Lu, Sirui}, booktitle = {Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing}, pages = {260--269}, year = {2025}, editor = {Zeng, Nianyin and Pachori, Ram Bilas and Wang, Dongshu}, volume = {278}, series = {Proceedings of Machine Learning Research}, month = {25--27 Apr}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v278/main/assets/wang25d/wang25d.pdf}, url = {https://proceedings.mlr.press/v278/wang25d.html}, abstract = {Under the global climate governance framework, carbon emission prediction has emerged as a pivotal technology for low-carbon energy transition. This review systematically examines the advancements in artificial intelligence-based carbon emission forecasting, revealing the evolutionary dynamics between traditional statistical methods and data-driven models. A novel “data-model-scenario" triadic analytical framework is proposed to deconstruct core challenges in this field. The study demonstrates that conventional approaches (e.g., ARIMA, grey models) exhibit structural deficiencies in high renewable energy penetration scenarios, including poor adaptability to abrupt changes and low cross-source data integration efficiency ($<$60%). In contrast, data-driven methods (XGBoost, LSTM, Transformer) achieve significant accuracy improvements through dynamic modeling and feature decoupling. Hybrid paradigms integrating physical constraints and multimodal alignment show promise in bridging the mechanism-data gap, yet face persistent challenges: inefficient multi-source data fusion (feature alignment success rate $<$60%), delayed response to sudden scenarios (recovery time $>$30 minutes), and computational-precision tradeoffs in edge deployment. The paper proposes a “dual-driven" evolutionary path for hybrid modeling and constructs a multi-scale scenario linkage matrix, providing theoretical guidance for next-generation prediction frameworks. Emerging technologies such as digital twins and federated meta-learning are highlighted as critical directions for future research.} }
Endnote
%0 Conference Paper %T Review of Research on Artificial Intelligence-Based Carbon Emission Prediction %A Weixian Wang %A Lu Zhang %A Mingyu Pan %A Jie Zhuo %A Sirui Lu %B Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing %C Proceedings of Machine Learning Research %D 2025 %E Nianyin Zeng %E Ram Bilas Pachori %E Dongshu Wang %F pmlr-v278-wang25d %I PMLR %P 260--269 %U https://proceedings.mlr.press/v278/wang25d.html %V 278 %X Under the global climate governance framework, carbon emission prediction has emerged as a pivotal technology for low-carbon energy transition. This review systematically examines the advancements in artificial intelligence-based carbon emission forecasting, revealing the evolutionary dynamics between traditional statistical methods and data-driven models. A novel “data-model-scenario" triadic analytical framework is proposed to deconstruct core challenges in this field. The study demonstrates that conventional approaches (e.g., ARIMA, grey models) exhibit structural deficiencies in high renewable energy penetration scenarios, including poor adaptability to abrupt changes and low cross-source data integration efficiency ($<$60%). In contrast, data-driven methods (XGBoost, LSTM, Transformer) achieve significant accuracy improvements through dynamic modeling and feature decoupling. Hybrid paradigms integrating physical constraints and multimodal alignment show promise in bridging the mechanism-data gap, yet face persistent challenges: inefficient multi-source data fusion (feature alignment success rate $<$60%), delayed response to sudden scenarios (recovery time $>$30 minutes), and computational-precision tradeoffs in edge deployment. The paper proposes a “dual-driven" evolutionary path for hybrid modeling and constructs a multi-scale scenario linkage matrix, providing theoretical guidance for next-generation prediction frameworks. Emerging technologies such as digital twins and federated meta-learning are highlighted as critical directions for future research.
APA
Wang, W., Zhang, L., Pan, M., Zhuo, J. & Lu, S.. (2025). Review of Research on Artificial Intelligence-Based Carbon Emission Prediction. Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing, in Proceedings of Machine Learning Research 278:260-269 Available from https://proceedings.mlr.press/v278/wang25d.html.

Related Material