An integrated framework for STL-LSTM-based energy demand-supply forecasting and ESS operation toward carbon-neutral buildings

초록

Achieving carbon neutrality in the building sector has become an urgent global challenge, driven by the increasing variability of renewable energy and the lack of reliable forecasting–control integration at the building level. In most existing systems, the Energy Storage System (ESS) is either underutilized or manually operated without data-driven intelligence, resulting in inefficiencies and frequent renewable curtailment. To address these limitations, this study proposes an integrated framework that combines Seasonal-Trend decomposition using Loess (STL) with a Long Short-Term Memory (LSTM) network and a rule-based ESS operation strategy. The proposed STL-LSTM-ESS framework explicitly links forecasting accuracy with operational performance within a unified structure, distinguishing it from previous hybrid studies that addressed forecasting and control separately. This integration enables the system to capture nonlinear temporal dynamics while maintaining interpretability and stability under real building conditions. Experimental validation using real building data demonstrates that the proposed method enhances forecasting robustness and substantially improves ESS operational efficiency. More importantly, it provides practical insights for achieving carbon-neutral building operations by transforming predictive intelligence into actionable control strategies.

키워드