Scholarlink

<p>Flood routing is a critical component of watershed management, requiring accurate predictive models to mitigate risks and enhance disaster preparedness. This study integrates artificial neural networks (ANNs) with metaheuristic optimization algorithms and signal processing techniques to improve flood prediction accuracy. Hybrid models combining ANNs with the genetic algorithm, firefly algorithm, artificial bee colony (ABC), JAYA optimization, and variational mode decomposition (VMD) were developed and tested on hourly flood data from the Kızılırmak and Mera Rivers in Turkey. The results indicate that the ABC-ANN model performed best for the Mera River, achieving R<sup>2</sup> = 0.893, MSE = 0.006, MAE = 0.067, and KGE = 0.916, demonstrating highly accurate flood predictions. For the Kızılırmak River, the VMD-ANN model outperformed the others with R<sup>2</sup> = 0.445, MSE = 285.035, MAE = 14.511, and KGE = 0.239. These findings suggest that integrating optimization algorithms with ANNs enhances prediction reliability, particularly in complex hydrological systems. Furthermore, the study confirms that downstream flood conditions can be effectively predicted based on upstream streamflow variations, contributing to improved flood control strategies. The results underscore the effectiveness of hybrid approaches in flood risk assessment and management, offering valuable insights for early warning systems, infrastructure resilience, and sustainable water resource planning.</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

Hybrid bio-inspired optimization with artificial neural networks for efficient flood routing in watershed management

  • 期刊论文

  • 发表日期：

    2025年6月4日

查看全文
  • N. Elshaboury,
  • O. Bazrafshan,
  • M. Bahrami,
  • S. Berhail,
  • C. B. Pande,
  • M. M. AlAmmar

摘要

Flood routing is a critical component of watershed management, requiring accurate predictive models to mitigate risks and enhance disaster preparedness. This study integrates artificial neural networks (ANNs) with metaheuristic optimization algorithms and signal processing techniques to improve flood prediction accuracy. Hybrid models combining ANNs with the genetic algorithm, firefly algorithm, artificial bee colony (ABC), JAYA optimization, and variational mode decomposition (VMD) were developed and tested on hourly flood data from the Kızılırmak and Mera Rivers in Turkey. The results indicate that the ABC-ANN model performed best for the Mera River, achieving R2 = 0.893, MSE = 0.006, MAE = 0.067, and KGE = 0.916, demonstrating highly accurate flood predictions. For the Kızılırmak River, the VMD-ANN model outperformed the others with R2 = 0.445, MSE = 285.035, MAE = 14.511, and KGE = 0.239. These findings suggest that integrating optimization algorithms with ANNs enhances prediction reliability, particularly in complex hydrological systems. Furthermore, the study confirms that downstream flood conditions can be effectively predicted based on upstream streamflow variations, contributing to improved flood control strategies. The results underscore the effectiveness of hybrid approaches in flood risk assessment and management, offering valuable insights for early warning systems, infrastructure resilience, and sustainable water resource planning.