TCSML-TimesFM: An Exploration for Carbon Trading Prices Prediction with Pre-trained Large Time Series Model

The rapid increase in carbon dioxide emissions is a key driver of global warming, making accurate carbon price forecasting essential for China's climate policy. Despite China's relatively late start in the carbon market, forecasting carbon prices is critical for understanding market fluctuations and supporting government risk warning mechanisms. This study proposes TCSML-TimesFM, a hybrid framework that combines time interpolation and CEEMDAN decomposition for improved forecasting accuracy. The model is first trained using carbon price data from Hubei, and its stability is validated using data from Guangdong. The results show that TCSML-TimesFM significantly outperforms traditional models in accuracy, providing valuable insights for businesses and policymakers navigating China's carbon market.