The Short-Term Load Forecasting for Special Days Based on Bagged Regression Trees in Qingdao, China

Area of Science:

• Electrical Engineering
• Data Science
• Time Series Analysis

Background:

• Short-term load forecasting is crucial for power system operation.
• Existing models often overlook the impact of special days like holidays.
• Accurate forecasting is essential for grid stability and economic efficiency.

Purpose of the Study:

• To investigate the effectiveness of the bagged regression trees (BRT) model for short-term load forecasting.
• To develop an improved forecasting model that accounts for special days.
• To enhance the accuracy of load prediction in Qingdao.

Main Methods:

• Applied the bagged regression trees (BRT) model with eight variables.
• Compared BRT performance against the artificial neural network (ANN) model.
• Introduced a novel indicator variable to capture data anomalies on special days (holidays, bridging, proximity days).
• Tested the enhanced BRT model on 2018 load data.

Main Results:

• Bagged regression trees (BRT) showed improved accuracy and speed over artificial neural networks (ANN).
• The proposed indicator variable significantly enhanced predictive accuracy for special days.
• The improved BRT model outperformed the standard BRT model in forecasting special day loads.

Conclusions:

• The BRT model is a viable and effective tool for short-term load forecasting.
• Incorporating an indicator variable for special days substantially improves forecasting accuracy.
• The enhanced model provides more reliable electricity demand predictions, especially during periods with abnormal load patterns.