Scholarlink

<p>Accurate day-ahead load forecasting (DALF) is essential for the reliable and economic operation of emerging regional power markets. Jeju Island, operating a pilot electricity market with high renewable energy penetration and non-industrial consumption characteristics, presents unique forecasting challenges due to irregular demand fluctuations and weather-sensitive load behavior. In this study, DALF is operated only for ordinary days, excluding holidays. This study proposes a hybrid ensemble forecasting framework tailored to Jeju’s operational context by integrating similar-day, statistical (ARIMAX), and artificial intelligence (LSTM, XGBoost) models. Each model is designed to capture distinct patterns and dependencies in demand data. The ensemble weights are dynamically optimized using the sequential least squares programming (SLSQP) algorithm based on historical forecasting performance. One year of operational data from Jeju Island is used for validation, showing that the ensemble model achieves substantial improvement in forecasting accuracy compared to standalone approaches. The results confirm the effectiveness of the proposed region-specific DALF framework for renewable-dominant and isolated power systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Practical Framework of Day-Ahead Load Forecast for Jeju Power System Operation Using Ensemble Model

  • 期刊论文

  • 发表日期：

    2026年5月16日

查看全文
  • DongKwan Kim,
  • Woo-Jin Jang,
  • Si-yeon Kim,
  • Sung-Kwan Joo,
  • Young-Min Wi,
  • Jinyeong Lee

摘要

Accurate day-ahead load forecasting (DALF) is essential for the reliable and economic operation of emerging regional power markets. Jeju Island, operating a pilot electricity market with high renewable energy penetration and non-industrial consumption characteristics, presents unique forecasting challenges due to irregular demand fluctuations and weather-sensitive load behavior. In this study, DALF is operated only for ordinary days, excluding holidays. This study proposes a hybrid ensemble forecasting framework tailored to Jeju’s operational context by integrating similar-day, statistical (ARIMAX), and artificial intelligence (LSTM, XGBoost) models. Each model is designed to capture distinct patterns and dependencies in demand data. The ensemble weights are dynamically optimized using the sequential least squares programming (SLSQP) algorithm based on historical forecasting performance. One year of operational data from Jeju Island is used for validation, showing that the ensemble model achieves substantial improvement in forecasting accuracy compared to standalone approaches. The results confirm the effectiveness of the proposed region-specific DALF framework for renewable-dominant and isolated power systems.