Scholarlink

<p>The Water Quality Index (WQI) is a widely utilized tool for assessing water quality issues, playing a crucial role in identifying pollution sources and informing water quality management strategies. However, conventional WQI models typically require input from dozens of water quality parameters, the measurement of which is both time-consuming and labor-intensive. To address this limitation, this study proposes a novel water quality assessment model, named MVIE-LSTM, designed to achieve rapid cost-effective WQI evaluation. First, a Missing Value Interpolation Expansion (MVIE) method is introduced to fill in missing water quality parameters as monthly averages. This approach reduces the need for frequent chemical property measurements by substituting actual values with interpolated data. Next, a Long Short-Term Memory (LSTM)-based model is built to process large-scale datasets, generating WQI predictions based on a comprehensive set of water quality parameters. Experimental results demonstrate that the MVIE-LSTM model maintains a high level of accuracy even when utilizing only six water quality parameters. Furthermore, the proposed model exhibits superior prediction performance (MAE = 3.5128, RMSE = 4.5019, and R<sup>2</sup> = 0.9024) compared to common models, including SVM, RNN and CNN.</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

MVIE-LSTM: a deep learning-based method for water quality assessment using monthly river data

  • 期刊论文

  • 发表日期：

    2025年2月26日

查看全文
  • Sha Xiong,
  • Junjie Cui,
  • Feifei Hou

摘要

The Water Quality Index (WQI) is a widely utilized tool for assessing water quality issues, playing a crucial role in identifying pollution sources and informing water quality management strategies. However, conventional WQI models typically require input from dozens of water quality parameters, the measurement of which is both time-consuming and labor-intensive. To address this limitation, this study proposes a novel water quality assessment model, named MVIE-LSTM, designed to achieve rapid cost-effective WQI evaluation. First, a Missing Value Interpolation Expansion (MVIE) method is introduced to fill in missing water quality parameters as monthly averages. This approach reduces the need for frequent chemical property measurements by substituting actual values with interpolated data. Next, a Long Short-Term Memory (LSTM)-based model is built to process large-scale datasets, generating WQI predictions based on a comprehensive set of water quality parameters. Experimental results demonstrate that the MVIE-LSTM model maintains a high level of accuracy even when utilizing only six water quality parameters. Furthermore, the proposed model exhibits superior prediction performance (MAE = 3.5128, RMSE = 4.5019, and R2 = 0.9024) compared to common models, including SVM, RNN and CNN.