Advancing Long-Term Natural Gas Flow Forecasting with Transformer Architecture

Accurate natural gas flow forecasting is crucial for energy supply chain management. Traditional time series models often struggle with the complex non-linear dynamics and long-term dependencies inherent in such data. This paper proposes a novel Transformer-based deep learning model specifically designed for natural gas flow forecasting. We utilize an encoder-only Transformer architecture, leveraging its self-attention mechanism to capture long-range dependencies effectively. The model is trained on historical natural gas node flow data, preprocessed using standardization and a sliding window approach. Experimental results demonstrate that the proposed model achieves high prediction accuracy, with an average accuracy of 94.7% for one-year ahead forecasting. The model successfully captures trends, seasonality, and fluctuations in gas flow, indicating its robustness and suitability for practical applications in energy forecasting.