Scholarlink

<p>The suspended sediment load transported by rivers can be estimated using various methodologies, including those based on artificial intelligence. In this study, we employed the Long Short-Term Memory (LSTM) model to estimate the suspended sediment concentration in the Mississippi River, United States of America. The input variables for the LSTM model included river discharge, water depth, suspended sediment load, and flow velocity. To enhance the model's performance, the input data and initial parameters were optimized using the Red Fox Optimization (RFO) algorithm, resulting in a super-optimized LSTM model (SLSTM-RFO) developed through a two-phase optimization process. Additionally, sediment load estimations were conducted using alternative models, specifically the Artificial Neural Network (ANN) and Generalized Regression Neural Network (GRNN) models. The performance of these models was assessed using five performance indicators, the correlation coefficient (R<sup>2</sup>), root mean square error (RMSE), mean absolute error (MAE), Nash and Sutcliffe efficiency (NS), and RMSE observations standard deviation ratio (RSR), demonstrating that the SLSTM-RFO model significantly outperformed the other models. Specifically, the SLSTM-RFO yielded improved estimation results, achieving reductions in error (RMSE) of 73.30%, 81.50%, and 82.56% compared to the LSTM, ANN, and GRNN models, respectively.</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

Estimation of suspended sediment load utilizing a super-optimized deep learning approach informed by the red fox optimization algorithm

  • 期刊论文

  • 发表日期：

    2025年2月24日

查看全文
  • Mohammad Mahdi Malekpour,
  • Mohammad Mehdi Ahmadi,
  • Marcello Gugliotta,
  • Mahmoud Mohammad Rezapour Tabari,
  • Kourosh Qaderi

摘要

The suspended sediment load transported by rivers can be estimated using various methodologies, including those based on artificial intelligence. In this study, we employed the Long Short-Term Memory (LSTM) model to estimate the suspended sediment concentration in the Mississippi River, United States of America. The input variables for the LSTM model included river discharge, water depth, suspended sediment load, and flow velocity. To enhance the model's performance, the input data and initial parameters were optimized using the Red Fox Optimization (RFO) algorithm, resulting in a super-optimized LSTM model (SLSTM-RFO) developed through a two-phase optimization process. Additionally, sediment load estimations were conducted using alternative models, specifically the Artificial Neural Network (ANN) and Generalized Regression Neural Network (GRNN) models. The performance of these models was assessed using five performance indicators, the correlation coefficient (R2), root mean square error (RMSE), mean absolute error (MAE), Nash and Sutcliffe efficiency (NS), and RMSE observations standard deviation ratio (RSR), demonstrating that the SLSTM-RFO model significantly outperformed the other models. Specifically, the SLSTM-RFO yielded improved estimation results, achieving reductions in error (RMSE) of 73.30%, 81.50%, and 82.56% compared to the LSTM, ANN, and GRNN models, respectively.