Probabilistic carbon price prediction with quantile temporal convolutional network considering uncertain factors.

Area of Science:

• Environmental Economics
• Financial Forecasting
• Machine Learning

Background:

• Accurate carbon price forecasting is crucial for investment and risk management in carbon trading.
• Existing forecast methods face challenges due to escalating uncertain factors.
• Understanding external influences on carbon markets is vital.

Purpose of the Study:

• To develop a novel probabilistic forecast model for precise carbon price fluctuation description.
• To investigate the impact of various external factors on carbon market prices.
• To provide valuable guidelines for carbon market risk management.

Main Methods:

• Development of a quantile temporal convolutional network (QTCN) model.
• Analysis of external factors including energy prices, economic status, international markets, environmental conditions, public concerns, and uncertainties.
• Case study using China's Hubei carbon emissions exchange.

Main Results:

• The QTCN model outperforms classical benchmark models in prediction errors and trading returns.
• Coal prices and EU carbon prices significantly impact Hubei carbon price forecasting.
• Geopolitical risk and economic policy uncertainty substantially contribute to carbon price projections, especially at high quantiles.

Conclusions:

• The QTCN model offers a robust approach to carbon price forecasting amidst uncertainty.
• External factors, particularly energy prices and geopolitical/economic uncertainties, play a significant role in carbon price formation.
• This research provides valuable insights for carbon market risk management and understanding price dynamics.