Scholarlink

<p>Intelligent water supply relies heavily on water demand forecasting. Due to nonlinear changes in the water consumption process, traditional prediction models have significant limitations regarding precision and consistency. Water wastage is monitored and the water distribution system is wirelessly monitored via the Internet of Things (IoT) technology. This study introduces a Hierarchical Hybrid Deep-Net Framework for water demand forecasting, combining advanced deep learning architectures and the Osprey-Pelican Hybrid Optimization (OPHO) algorithm to enhance prediction accuracy and efficiency. The framework integrates RNN-LSTM to capture temporal patterns, Caps Net to understand hierarchical relationships, and MobileNetV2 for efficient data processing. The methodology begins with comprehensive data preprocessing, including data cleaning to handle inconsistencies, Min–Max normalization for feature scaling, and feature extraction through statistical measures such as mean, median, standard deviation, and skewness and correlation-based features like Spearman rank correlation. High-level abstract features are extracted using Inception V3, a pre-trained deep learning model. Feature selection is performed using the OPHO algorithm, which combines the precision of Osprey optimization and the local search capabilities of Pelican optimization to identify the most informative features while reducing dimensionality. The framework is validated on a real-world dataset sourced from United Utilities, a water and wastewater management firm in the North West of England, obtained from Kaggle. The proposed model is evaluated for both 70% and 80% of training data for the metrics MAE of 0.0300, 0.0100, MAPE of 0.0119, 0.0223. Also, the MSE, NMSE, RMSE, and IoA have superior performance when compared to the other existing methods.</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

A Hierarchical Hybrid Deep-Net Framework for Water Demand Forecasting Using OPHO Optimization and Deep Learning Techniques

  • 期刊论文

  • 发表日期：

    2025年1月24日

查看全文
  • Subha Jeyavelu,
  • Kowsigan Mohan

摘要

Intelligent water supply relies heavily on water demand forecasting. Due to nonlinear changes in the water consumption process, traditional prediction models have significant limitations regarding precision and consistency. Water wastage is monitored and the water distribution system is wirelessly monitored via the Internet of Things (IoT) technology. This study introduces a Hierarchical Hybrid Deep-Net Framework for water demand forecasting, combining advanced deep learning architectures and the Osprey-Pelican Hybrid Optimization (OPHO) algorithm to enhance prediction accuracy and efficiency. The framework integrates RNN-LSTM to capture temporal patterns, Caps Net to understand hierarchical relationships, and MobileNetV2 for efficient data processing. The methodology begins with comprehensive data preprocessing, including data cleaning to handle inconsistencies, Min–Max normalization for feature scaling, and feature extraction through statistical measures such as mean, median, standard deviation, and skewness and correlation-based features like Spearman rank correlation. High-level abstract features are extracted using Inception V3, a pre-trained deep learning model. Feature selection is performed using the OPHO algorithm, which combines the precision of Osprey optimization and the local search capabilities of Pelican optimization to identify the most informative features while reducing dimensionality. The framework is validated on a real-world dataset sourced from United Utilities, a water and wastewater management firm in the North West of England, obtained from Kaggle. The proposed model is evaluated for both 70% and 80% of training data for the metrics MAE of 0.0300, 0.0100, MAPE of 0.0119, 0.0223. Also, the MSE, NMSE, RMSE, and IoA have superior performance when compared to the other existing methods.