Scholarlink

Short-term power load forecasting (STPLF) is essential to the power grid system. Forecasting the power load process can help reduce electricity production costs and optimize power quality. Many studies are in this field, but it is necessary to improve the accuracy of the forecast. For example, forecasting short-term demand power is essential for operating an off-grid system. This article tested three models for forecasting power load in the short term: Decision Tree (DT), Boosting DT, and Bagging DT. The data set was measured in the off-grid labs at the VSB-Technical University of Ostrava, Czech Republic. The data set was clustered using k-means clustering, and the distance between the samples and each cluster was measured, which was fed to the forecast stage. Using three new space features reduced the computation complexity of the training and testing phases. The performance of the designed models was evaluated using the root mean square error (RMSE), mean absolute percentage error (MAPE), and \(R^{2}\) . The Bagging DT achieved the lowest forecast error, followed by Boosting DT and standard DT. In contrast, the standard DT took a shorter time to test one sample than the ensemble DTs.

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

Power Load Forecasting Using Machine Learning Methods

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ibrahim Jahan,
  • Vojtech Blazek,
  • Faisal Mohamed,
  • Abdesselam Mechali,
  • Alaa Alhatab

摘要

Short-term power load forecasting (STPLF) is essential to the power grid system. Forecasting the power load process can help reduce electricity production costs and optimize power quality. Many studies are in this field, but it is necessary to improve the accuracy of the forecast. For example, forecasting short-term demand power is essential for operating an off-grid system. This article tested three models for forecasting power load in the short term: Decision Tree (DT), Boosting DT, and Bagging DT. The data set was measured in the off-grid labs at the VSB-Technical University of Ostrava, Czech Republic. The data set was clustered using k-means clustering, and the distance between the samples and each cluster was measured, which was fed to the forecast stage. Using three new space features reduced the computation complexity of the training and testing phases. The performance of the designed models was evaluated using the root mean square error (RMSE), mean absolute percentage error (MAPE), and \(R^{2}\) . The Bagging DT achieved the lowest forecast error, followed by Boosting DT and standard DT. In contrast, the standard DT took a shorter time to test one sample than the ensemble DTs.