Scholarlink

<p>The secondary network supply water temperature is critical to heat transmission and distribution from heating systems to end users, and its prediction accuracy directly determines heating quality and energy utilization efficiency. At present, most heating systems rely solely on single-source meteorological data for regulation and suffer from pronounced prediction lag, creating an urgent need for intelligent prediction methods that integrate multi-dimensional features. Therefore, this paper adopts Ensemble Empirical Mode Decomposition (EEMD) for multi-scale time series decomposition, employs Convolutional Neural Network (CNN) for spatial feature extraction and Bidirectional Long Short-Term Memory (BiLSTM) for temporal dependency mining, and introduces Dynamic Weight Allocation (DWA) to realize adaptive fusion of dual-model outputs. On this basis, an EEMD-DWA-CNN-BiLSTM multi-scale feature fusion prediction framework is constructed. We trained and validated the proposed model using 1776 sets of secondary network supply water temperature data collected from a heat exchange station in Qiqihar City, Heilongjiang Province, spanning from December 2024 to February 2025. The proposed EEMD-DWA-CNN-BiLSTM combined model achieved excellent prediction performance with a Mean Absolute Error (MAE) of 0.31, a Mean Absolute Percentage Error (MAPE) of 0.68, a Root Mean Square Error (RMSE) of 0.38, and a coefficient of determination (R<sup>2</sup>) of 88.54. This algorithm exhibits good prediction accuracy and reliable generalization ability, which provides solid technical support for heating enterprises to achieve energy conservation, emission reduction and precise heating regulation.</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

Prediction of secondary network supply water temperature based on EEMD-DWA-CNN-BiLSTM model for energy conservation

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月30日

查看全文
  • Ye Wang,
  • Yang Song,
  • Xinning Wang,
  • Shiming Feng,
  • Guangjun Pan,
  • Jian Xing

摘要

The secondary network supply water temperature is critical to heat transmission and distribution from heating systems to end users, and its prediction accuracy directly determines heating quality and energy utilization efficiency. At present, most heating systems rely solely on single-source meteorological data for regulation and suffer from pronounced prediction lag, creating an urgent need for intelligent prediction methods that integrate multi-dimensional features. Therefore, this paper adopts Ensemble Empirical Mode Decomposition (EEMD) for multi-scale time series decomposition, employs Convolutional Neural Network (CNN) for spatial feature extraction and Bidirectional Long Short-Term Memory (BiLSTM) for temporal dependency mining, and introduces Dynamic Weight Allocation (DWA) to realize adaptive fusion of dual-model outputs. On this basis, an EEMD-DWA-CNN-BiLSTM multi-scale feature fusion prediction framework is constructed. We trained and validated the proposed model using 1776 sets of secondary network supply water temperature data collected from a heat exchange station in Qiqihar City, Heilongjiang Province, spanning from December 2024 to February 2025. The proposed EEMD-DWA-CNN-BiLSTM combined model achieved excellent prediction performance with a Mean Absolute Error (MAE) of 0.31, a Mean Absolute Percentage Error (MAPE) of 0.68, a Root Mean Square Error (RMSE) of 0.38, and a coefficient of determination (R2) of 88.54. This algorithm exhibits good prediction accuracy and reliable generalization ability, which provides solid technical support for heating enterprises to achieve energy conservation, emission reduction and precise heating regulation.