전력 수요 예측에서의 산업별 지표 활용 및 변수 영향력 분석

초록

This study quantitatively analyzes the impact of industry-specific economic activity indices on electricity demand forecasting and proposes a predictive model that incorporates these effects. For this purpose, daily data from 2010 to 2024 were utilized, including electricity demand, temperature, the KOSPI index, exchange rates, and 17 sector-specific stock indices. The empirical results indicate that the Reg-SARIMA-GARCH model incorporating industry-specific indices outperforms other models in terms of both predictive accuracy and explanatory power. This finding suggests that the level of industrial activity exerts a significant influence on electricity demand. In addition, a comparative analysis with the LGBM model—advantageous in interpreting SHAP values—was conducted. The SHAP-based evaluation of variable importance and directional impact revealed that the Machinery and Equipment and Energy and Chemicals indices exhibit the highest levels of importance. These results provide empirical evidence identifying key industrial sectors that substantially affect electricity demand. As one of the first domestic studies to apply industry-level economic indicators to short-term electricity demand forecasting, this research provides a valuable foundation for the future development of dynamic forecasting models that reflect time-varying variable influence.

키워드