Scholarlink

<p>This study proposes a parallel temporal convolutional network (TCN) approach for short-term electricity load forecasting in foundries, designed to significantly enhance predictive accuracy. Beyond forecasting, the method addresses situations where factory electricity loads exceed contracted capacity, generating tangible economic benefits. The model analyzes and predicts power consumption using 2,880 electricity load data records collected over 30 days from a factory. Extensive comparisons with existing forecasting techniques show that TCN-based architectures—particularly the proposed parallel TCN achieve superior accuracy, with a mean absolute error (MAE) of 2.071, a root mean square error (RMSE) of 3.105, and a mean absolute percentage error (MAPE) of only 0.93%. The system contributes to real-time monitoring of electricity load and proactive management strategies, mitigating risks associated with exceeding contracted capacities. It serves as a vital reference for field power scheduling, enabling dynamic production adjustments to reduce operational costs while aligning with intelligent manufacturing frameworks.</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

Parallel TCN-Based Short-Term Electricity Load Forecasting for Foundries

  • 期刊论文

  • 发表日期：

    2026年2月13日

查看全文
  • Min-Sin Liu,
  • Ping-Huan Kuo,
  • Shyh-Leh Chen

摘要

This study proposes a parallel temporal convolutional network (TCN) approach for short-term electricity load forecasting in foundries, designed to significantly enhance predictive accuracy. Beyond forecasting, the method addresses situations where factory electricity loads exceed contracted capacity, generating tangible economic benefits. The model analyzes and predicts power consumption using 2,880 electricity load data records collected over 30 days from a factory. Extensive comparisons with existing forecasting techniques show that TCN-based architectures—particularly the proposed parallel TCN achieve superior accuracy, with a mean absolute error (MAE) of 2.071, a root mean square error (RMSE) of 3.105, and a mean absolute percentage error (MAPE) of only 0.93%. The system contributes to real-time monitoring of electricity load and proactive management strategies, mitigating risks associated with exceeding contracted capacities. It serves as a vital reference for field power scheduling, enabling dynamic production adjustments to reduce operational costs while aligning with intelligent manufacturing frameworks.