Scholarlink

<p>Water demand prediction is crucial for ensuring water security and plays a key role in achieving the goal of “ensuring access to water and sanitation for all”. However, due to the complex interactions between socio-economic, climatic, and hydrologic factors, existing long-term water demand prediction methods often face challenges such as slow convergence and inadequate handling of nonlinear dynamics, which limit the accuracy of predictions. This study incorporates innovative optimization Sine cosine and Cauchy variation Sparrow Search Algorithm (SCSSA) and Adjustable Self-feedback Gain Elman Neural Networks (ASGENN) to an integrated algorithm, named SCSSA-ASGENN, to tackle these challenges effectively. The ASGENN incorporates an adjustable self-feedback gain to enhance the network’s adaptability, while SCSSA integrates refractive back learning, Sine-cosine algorithms, and Cauchy mutation strategies to accelerate convergence and prevent entrapment in local optima. The approach is applied for projecting water demand for Beijing using the data from 1978 to 2024. The results show that SCSSA-ASGENN outperforms other benchmark models, achieving a mean absolute error (MAE) of 1.62&#xa0;billion m<sup>3</sup>, root mean square error (RMSE) of 1.73&#xa0;billion m<sup>3</sup>, and mean absolute percentage error (MAPE) of 4.22%. Compared to SSA-ASGENN, it improves mean square error (MSE) by 18.02%. Notably, the number of convergence iterations is reduced by 80% compared to models such as SLSSA-ASGENN. The robustness of this model in capturing nonlinear trends and extreme values highlights its potential for sustainable water resource planning in regions facing water scarcity. This approach provides a reliable framework for addressing the challenges of complex long-term water demand prediction.</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 New Water Demand Prediction Approach using Adjustable Self-Feedback Gain Elman Neural Networks Coupled with Enhanced Sparrow Search Algorithm

  • 期刊论文

  • 发表日期：

    2026年2月18日

查看全文
  • Liying Wang,
  • Qiang Ge,
  • Zhenxing Zhang,
  • Shixiang Gu,
  • Nima Khodadadi,
  • Cheng-I Hsieh,
  • Weiguo Zhao

摘要

Water demand prediction is crucial for ensuring water security and plays a key role in achieving the goal of “ensuring access to water and sanitation for all”. However, due to the complex interactions between socio-economic, climatic, and hydrologic factors, existing long-term water demand prediction methods often face challenges such as slow convergence and inadequate handling of nonlinear dynamics, which limit the accuracy of predictions. This study incorporates innovative optimization Sine cosine and Cauchy variation Sparrow Search Algorithm (SCSSA) and Adjustable Self-feedback Gain Elman Neural Networks (ASGENN) to an integrated algorithm, named SCSSA-ASGENN, to tackle these challenges effectively. The ASGENN incorporates an adjustable self-feedback gain to enhance the network’s adaptability, while SCSSA integrates refractive back learning, Sine-cosine algorithms, and Cauchy mutation strategies to accelerate convergence and prevent entrapment in local optima. The approach is applied for projecting water demand for Beijing using the data from 1978 to 2024. The results show that SCSSA-ASGENN outperforms other benchmark models, achieving a mean absolute error (MAE) of 1.62 billion m3, root mean square error (RMSE) of 1.73 billion m3, and mean absolute percentage error (MAPE) of 4.22%. Compared to SSA-ASGENN, it improves mean square error (MSE) by 18.02%. Notably, the number of convergence iterations is reduced by 80% compared to models such as SLSSA-ASGENN. The robustness of this model in capturing nonlinear trends and extreme values highlights its potential for sustainable water resource planning in regions facing water scarcity. This approach provides a reliable framework for addressing the challenges of complex long-term water demand prediction.