Long-term Electricity Demand Forecasting using THI-ConvNet

Electricity load forecasting is critical for ensuring the stability and efficiency of power supply systems, particularly during holidays when demand patterns tend to deviate from normal trends. However, most existing studies focus on short-term forecasts and lack the capability to effectively incorporate the irregularities caused by long-term holidays. To address this gap, we propose a novel deep learning-based forecasting model called the Temporal Holiday-Integrated (THI) model, which is the first to explicitly integrate both short-term and long-term holiday effects into electricity demand prediction. The model is trained using daily data collected from January 1, 2013, to December 31, 2024 (12 years in total), consisting of 61 meteorological variables from the Korea Meteorological Administration, 1 national holiday indicator, and 1 electricity transaction volume column.

电力负荷预测对于保障供电系统的稳定性与效率至关重要，尤其是在节假日期间——此时需求模式往往偏离正常趋势。然而，现有多数研究聚焦于短期预测，且缺乏有效整合长期节假日所致不规则性的能力。为填补这一空白，我们提出一种基于深度学习的新型预测模型——时间假日整合模型（Temporal Holiday-Integrated, THI），该模型首次将短期与长期假日效应明确整合至电力需求预测中。该模型采用2013年1月1日至2024年12月31日（共计12年）的日度数据进行训练，数据包含来自韩国气象厅的61个气象变量、1个全国假日指标及1个电力交易量列。