Forecasting Carbon Emissions in Power Systems Using Autoformer and LSTM: A Case Study of the Spanish Electricity Market

Recent advancements in carbon emission prediction modeling have significantly improved the accuracy of forecasting emission intensity, particularly through the application of artificial intelligence techniques. However, predicting emission intensity within power systems remains challenging due to the complexity of internal variables and the high dimensionality of data. In this study, we introduce a novel Autoformer-LSTM hybrid model designed to predict carbon emission intensity in power markets, and we conduct a comparative analysis with other machine learning models. Empirical results demonstrate that the proposed Autoformer-LSTM model achieves superior predictive performance in the context of the Spanish power market. The Autoformer component effectively captures long-term trends in time series data, while the LSTM component excels at modeling short-term dependencies. This hybrid approach not only delivers accurate predictions but also significantly reduces computational costs. Our findings provide critical theoretical insights and practical guidance for optimizing power market operations and mitigating greenhouse gas emissions.