Scholarlink

<p>Water resource management in dams requires accurate forecasting of the Allocated Water Volume (AWV) to optimize drinking water supply (DWS) and irrigation. However, forecasting AWV remains challenging due to the nonlinear dynamics of hydrological data and the limited availability of high-quality time series. This study develops a Convolutional Neural Network (CNN)-based model to predict AWV in the ZIT EMBA dam at different time horizons (T, T + 1, T + 3, T + 7, T + 15). The baseline CNN is enhanced using Grey Wolf Optimizer (GWO) and Particle Swarm Optimization (PSO) algorithms. Historical AWV values (T − 6 to T − 1) are used as inputs. The model evaluation uses key performance metrics (R, NSE, MAE, RMSE, KGE, PBIAS, and Combined Accuracy—CA). Results demonstrate that short-term forecasts (T, T + 1) provide superior accuracy. The best performance is obtained with the CNN-GWO1 model, which achieved R = 0.986, NSE = 0.971, MAE = 0.002 hm<sup>3</sup>, RMSE = 0.003 hm<sup>3</sup>, KGE = 0.983, PBIAS = 0.009, and CA = 0.009. CNN-PSO1 also outperformed the baseline CNN model (R = 0.970 vs 0.933; CA = 0.023 vs 0.047). In contrast, the CNN5 model at horizon T + 15 showed reduced accuracy (R = 0.517, KGE = 0.367, RMSE = 0.018, CA = 0.280). This optimized approach enhances AWV forecasting, offering a robust and efficient tool for improving water allocation strategies and ensuring long-term resource sustainability.</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

Short- and Medium-Term Forecasting of Allocated Water Volume in the ZIT EMBA Dam: Improving CNN with GWO and PSO Optimization

  • 期刊论文

  • 发表日期：

    2025年6月9日

查看全文
  • Bilal Lefoula,
  • Noureddine Daif

摘要

Water resource management in dams requires accurate forecasting of the Allocated Water Volume (AWV) to optimize drinking water supply (DWS) and irrigation. However, forecasting AWV remains challenging due to the nonlinear dynamics of hydrological data and the limited availability of high-quality time series. This study develops a Convolutional Neural Network (CNN)-based model to predict AWV in the ZIT EMBA dam at different time horizons (T, T + 1, T + 3, T + 7, T + 15). The baseline CNN is enhanced using Grey Wolf Optimizer (GWO) and Particle Swarm Optimization (PSO) algorithms. Historical AWV values (T − 6 to T − 1) are used as inputs. The model evaluation uses key performance metrics (R, NSE, MAE, RMSE, KGE, PBIAS, and Combined Accuracy—CA). Results demonstrate that short-term forecasts (T, T + 1) provide superior accuracy. The best performance is obtained with the CNN-GWO1 model, which achieved R = 0.986, NSE = 0.971, MAE = 0.002 hm3, RMSE = 0.003 hm3, KGE = 0.983, PBIAS = 0.009, and CA = 0.009. CNN-PSO1 also outperformed the baseline CNN model (R = 0.970 vs 0.933; CA = 0.023 vs 0.047). In contrast, the CNN5 model at horizon T + 15 showed reduced accuracy (R = 0.517, KGE = 0.367, RMSE = 0.018, CA = 0.280). This optimized approach enhances AWV forecasting, offering a robust and efficient tool for improving water allocation strategies and ensuring long-term resource sustainability.