检索
E-mail
RSS

作者投稿 主编审稿 专家审稿 远程编辑

热带海洋学报

• • 上一篇    

深圳湾海水水质预测深度学习模型对比分析*


王凯敏1, 陈瑶2, 陈俊嘉1, 陈正茜2   

  1. 1. 深圳市海洋发展促进中心, 广东 深圳 518067;

    2. 深圳市朗诚科技股份有限公司, 广东 深圳 518029

  • 收稿日期:2025-08-15 修回日期:2025-10-11 接受日期:2025-10-20
  • 通讯作者: 陈瑶
  • 基金资助:

    深圳市科技计划资助(KCXFZ20211020164015024)


Comparative analysis of deep learning models for seawater quality forecasting in Shenzhen Bay*

WANG Kaimin  1, CHEN Yao 2, CHEN Junjia 1, CHEN Zhengxi  2   

  1. 1. Shenzhen Ocean Development Promotion Center, Shenzhen 518067, China;

    2. Shenzhen Lightsun Technology Co.,Ltd, Shenzhen 518029, China

  • Received:2025-08-15 Revised:2025-10-11 Accepted:2025-10-20
  • Supported by:
    Shenzhen Science and Technology Progra (KCXFZ20211020164015024)

PDF (PC)

赞

摘要： 由于深圳湾水体交换能力较弱，且周边河流输入大量陆源污染物，导致赤潮频发，水质波动显著，及时准确地预测深圳湾水质动态变化至关重要。本文系统比较了基于变分模态分解的深度学习组合模型以及基础深度学习模型在深圳湾水质预测中的表现。“分解-预测-重构”组合模型通过变分模态分解（Variational Mode Decomposition, VMD）将水质和营养盐数据分解为5个平稳本征模态函数，输入到时间卷积网络（Temporal Convolutional Network, TCN）、长短期记忆网络（Long Short-Term Memory, LSTM）模型当中预测各个本征函数，将预测结果叠加生成最终预测值。研究结果表明，（1）TCN-VMD和LSTM-VMD组合模型较基础TCN、LSTM模型预测效果显著提升，最高改进率达到91.69%；（2）TCN-VMD相比LSTM-VMD纳什系数的表现更优，最大改进率为95.56%，其中水温（0.998）、pH值（0.971）、盐度（0.988）、溶解氧（0.960）、叶绿素a（0.909）、氨氮（0.911）、磷酸盐（0.909）、硝酸盐（0.969）、亚硝酸盐（0.968），TCN-VMD均能准确捕捉真实值的波峰和波谷。

关键词: VMD, TCN, LSTM, 水质预测, 深圳湾

Abstract: Owing to the bay’s limited hydrodynamic exchange and the substantial influx of land-derived pollutants from adjacent rivers, Shenzhen Bay experiences frequent harmful algal blooms and pronounced water-quality fluctuations. Accurately and promptly forecasting these dynamics is therefore essential. This study systematically evaluates the predictive skill of deep-learning ensemble models that integrate Variational Mode Decomposition (VMD) against conventional, non-decomposed deep-learning models for water-quality forecasting in Shenzhen Bay. Within the proposed “decomposition-prediction-reconstruction” framework, the water-quality and nutrient time-series are first decomposed by VMD into five stationary intrinsic mode functions (IMFs). These IMFs are then input into Temporal Convolutional Network (TCN) and Long Short-Term Memory (LSTM) models to predict each individual component function. Finally, the predicted results are aggregated to generate the ultimate prediction.Key findings are as follows.(1). The hybrid TCN-VMD and LSTM-VMD models demonstrated significantly improved prediction performance compared to the baseline TCN and LSTM models, with the maximum improvement rate reaching 91.69%.(2). TCN-VMD performs better than the LSTM-VMD model in terms of the Nash-Sutcliffe efficiency coefficient, achieving a maximum improvement of 95.56%. The corresponding Nash-Sutcliffe efficiencies are 0.998 for water temperature, 0.971 for pH, 0.988 for salinity, 0.960 for dissolved oxygen, 0.909 for chlorophyll-a, 0.911 for ammonium, 0.909 for phosphate, 0.969 for nitrate, and 0.968 for nitrite; the TCN-VMD framework faithfully captures observed peaks and troughs.

Key words: VMD, TCN, LSTM, seawater quality forecasting, Shenzhen Bay