Prediction of daily electricity carbon emission factors for high energy-consuming enterprises based on temporal feature optimization and hybrid intelligent algorithms

With the advancement of China’s “Dual Carbon” strategic goals, carbon emission monitoring and forecasting in the power industry have become increasingly important. The electricity carbon emission factor is a key parameter for measuring the carbon intensity of electricity consumption and is widely used in emission accounting, load dispatching, and electricity trading. However, traditional carbon factor prediction methods are often static and fail to capture temporal variations, making them inadequate for real-time forecasting needs. This paper proposes a carbon factor prediction model based on temporal convolutional networks and attention mechanisms to enhance prediction accuracy and temporal adaptability. The model incorporates a lightweight gated attention mechanism to enhance responsiveness to short-term external disturbances by weighting meteorological features. In addition, a sliding window approach is employed to construct multi-step time series samples, integrating periodicity, lag, and trend features to improve the model’s capability in capturing dynamic patterns. To optimize model performance, Bayesian Optimization and the Marine Predators Algorithm are used for automatic hyperparameter tuning. Experimental results demonstrate that the proposed method outperforms existing models in both prediction accuracy and generalization ability, providing effective technical support for low-carbon scheduling and precise management in high energy-consuming enterprises.