Scholarlink

<p>Hourly streamflow prediction is a critical component in various fields, including water resource management, energy production, network optimization, and flood risk mitigation. In flood management, accurate short-term discharge forecasts are essential for anticipating river surges and implementing preventive or mitigation strategies. This study proposes a forecasting approach for hourly streamflow at multiple time horizons (t, 1&#xa0;h, 3&#xa0;h, 6&#xa0;h, 12&#xa0;h, 24&#xa0;h) using Extreme Learning Machine (ELM) models optimized by five bio-inspired algorithms: Grey Wolf Optimizer (GWO), Bat Algorithm (BAT), Differential Evolution (DE), Whale Optimization Algorithm (WOA), and Genetic Algorithm (GA). The models leverage temporal dependencies from past discharge values (t−1 to t−6) to improve predictive performance. Experimental results show that model performance varies significantly across time horizons. For immediate forecasts (t + 1h), the ELM-GWO1, ELM-BAT1, and ELM-DE1 models achieved the best accuracy, with R = 0.991, 0.990, and 0.988, respectively, and NSE values above 0.97. At longer horizons (t + 24h), predictive accuracy decreases, with models such as ELM-GA6 and ELM-WOA6 showing the lowest performance (R = 0.415 and 0.435, NSE = 0.201 and 0.206). The BAT and GWO-optimized models generally provided more stable and reliable forecasts across different horizons, outperforming traditional methods in terms of accuracy. These results confirm that bio-inspired optimization enhances ELM-based hourly discharge forecasts, particularly for short- to mid-term predictions. This approach provides a reliable solution for real-time streamflow prediction, improving decision-making for sustainable water resource management and flood risk mitigation.</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

Accurate Multi-horizon Hourly Streamflow Forecasting Using ELM Optimized by GWO, BAT, DE, WOA, and GA Algorithms

  • 期刊论文

  • 发表日期：

    2025年12月8日

查看全文
  • Noureddine Daif,
  • Aziz Hebal,
  • Salah Difi,
  • Djillali Fettam,
  • Bilel Zerouali,
  • Salim Heddam

摘要

Hourly streamflow prediction is a critical component in various fields, including water resource management, energy production, network optimization, and flood risk mitigation. In flood management, accurate short-term discharge forecasts are essential for anticipating river surges and implementing preventive or mitigation strategies. This study proposes a forecasting approach for hourly streamflow at multiple time horizons (t, 1 h, 3 h, 6 h, 12 h, 24 h) using Extreme Learning Machine (ELM) models optimized by five bio-inspired algorithms: Grey Wolf Optimizer (GWO), Bat Algorithm (BAT), Differential Evolution (DE), Whale Optimization Algorithm (WOA), and Genetic Algorithm (GA). The models leverage temporal dependencies from past discharge values (t−1 to t−6) to improve predictive performance. Experimental results show that model performance varies significantly across time horizons. For immediate forecasts (t + 1h), the ELM-GWO1, ELM-BAT1, and ELM-DE1 models achieved the best accuracy, with R = 0.991, 0.990, and 0.988, respectively, and NSE values above 0.97. At longer horizons (t + 24h), predictive accuracy decreases, with models such as ELM-GA6 and ELM-WOA6 showing the lowest performance (R = 0.415 and 0.435, NSE = 0.201 and 0.206). The BAT and GWO-optimized models generally provided more stable and reliable forecasts across different horizons, outperforming traditional methods in terms of accuracy. These results confirm that bio-inspired optimization enhances ELM-based hourly discharge forecasts, particularly for short- to mid-term predictions. This approach provides a reliable solution for real-time streamflow prediction, improving decision-making for sustainable water resource management and flood risk mitigation.