Ensemble system for short term carbon dioxide emissions forecasting based on multi-objective tangent search...

Area of Science:

• Environmental Science
• Climate Science
• Data Science

Background:

• Global warming and climate change are significantly driven by carbon emissions.
• Accurate carbon emissions forecasting is essential for effective emissions reduction strategies and achieving carbon neutrality.
• Existing forecasting methods show limited performance due to a lack of objective model selection and interval forecasting.

Purpose of the Study:

• To propose a novel ensemble prediction system for carbon emissions forecasting.
• To improve the accuracy and stability of carbon emissions predictions.
• To provide both point and interval predictions for carbon emissions.

Main Methods:

• A novel ensemble prediction system integrating data decomposition, model selection, phase space reconstruction, ensemble point prediction, and interval prediction.
• Application of the proposed system to carbon emissions datasets (Dataset A and Dataset B).
• Evaluation of forecasting performance using Mean Absolute Percentage Error (MAPE) for point prediction and Conditional Weighted Coverage (CWC) for interval prediction.

Main Results:

• The proposed forecasting strategy achieved mean MAPE values of 1.1102% (Dataset A) and 1.1382% (Dataset B) for point predictions.
• The system demonstrated strong interval forecasting capabilities with mean CWC values of 0.3512 (Dataset A) and 0.1572 (Dataset B).
• The ensemble system significantly outperformed other models in terms of forecasting accuracy and stability.

Conclusions:

• The developed ensemble prediction system offers superior accuracy and stability for carbon emissions forecasting.
• The system's ability to provide both point and interval predictions is crucial for informed policymaking.
• This research provides valuable insights and references for policymakers aiming to address climate change and achieve carbon neutrality.