Scholarlink

<p>Water discharge (WD) is a crucial element of river hydrodynamics. The assessment of WD in dam design and river engineering studies is crucial. Proper WD from a barrage or dam must be regulated to prevent downstream floods during rainy season. WD encompasses numerous <b>complex</b> nonlinear processes and factors that traditional approaches are <b>not able</b> to address such complexities. To address these chllanges, hybrid multi-objective optimisation Genetic algorithm(GA) based artificial neural network(ANN) (MOO-GA-ANN) with automated parameter tuning model is proposed to optimizes the variance-bias trade-off for WD estimation in Mahanadi river (MR), India. These approaches optimises the two conflicting responses bias and variance with all artificial neural network(ANN) model parameters optimisation simulateneously to enhance robustness and generalisation capability of the model. The MOO-GA-ANN is developed using hydro-climatic data (Tempearature(T), Rainfall(R), water level(WL), and Suspended sediment yield(SSY)) as input parameters to estimate WD at the most downstream gauge location(Tikarapara) in the MR. Multiple linear regression (MLR), standalone ANN and single-objective GA-based ANN (GA-ANN) models are used for comparison from the MOO-GA-ANN on the basis of the statistical error metrics performances evaluation. Results indicates that the hybridised MOO-GA-ANN model has produced the most accurate and efficient results compared to other models for estimating WD in MR. The proposed model can be potentially used to estimate WD at both ungauged or gauged sites in the absence of measured WD because of its high performance and simplicity of use.</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

Multi-Objective Variance-Bias Tradeoff Optimization of ANN by Hybridizing with Multi-Objective GA for Water Discharge Estimation

  • 期刊论文

  • 发表日期：

    2026年3月28日

查看全文
  • Arvind Yadav,
  • Ashwini Kumar Pradhan,
  • Manish Kumar,
  • Priyaranjan Sharma,
  • Sunil Kr. Maakar,
  • Shivangi Singh,
  • Umamaheswararao Batta,
  • Devendra Joshi

摘要

Water discharge (WD) is a crucial element of river hydrodynamics. The assessment of WD in dam design and river engineering studies is crucial. Proper WD from a barrage or dam must be regulated to prevent downstream floods during rainy season. WD encompasses numerous complex nonlinear processes and factors that traditional approaches are not able to address such complexities. To address these chllanges, hybrid multi-objective optimisation Genetic algorithm(GA) based artificial neural network(ANN) (MOO-GA-ANN) with automated parameter tuning model is proposed to optimizes the variance-bias trade-off for WD estimation in Mahanadi river (MR), India. These approaches optimises the two conflicting responses bias and variance with all artificial neural network(ANN) model parameters optimisation simulateneously to enhance robustness and generalisation capability of the model. The MOO-GA-ANN is developed using hydro-climatic data (Tempearature(T), Rainfall(R), water level(WL), and Suspended sediment yield(SSY)) as input parameters to estimate WD at the most downstream gauge location(Tikarapara) in the MR. Multiple linear regression (MLR), standalone ANN and single-objective GA-based ANN (GA-ANN) models are used for comparison from the MOO-GA-ANN on the basis of the statistical error metrics performances evaluation. Results indicates that the hybridised MOO-GA-ANN model has produced the most accurate and efficient results compared to other models for estimating WD in MR. The proposed model can be potentially used to estimate WD at both ungauged or gauged sites in the absence of measured WD because of its high performance and simplicity of use.