| 引用本文: | 丁艺鼎,范宏翔,徐力刚,蒋名亮,吕海深,朱永华,程俊翔.可解释性长短期记忆模型用于预测湖泊总磷浓度变化.湖泊科学,2024,36(4):1046-1059. DOI:10.18307/2024.0415 |
| Ding Yiding,Fan Hongxiang,Xu Ligang,Jiang Mingliang,Lv Haishen,Zhu Yonghua,Chen Junxiang.The interpretable long-term and short-term memory model was used to predict the change of total phosphorus concentration in lakes. J. Lake Sci.2024,36(4):1046-1059. DOI:10.18307/2024.0415 |
|
| |
|
|
| 本文已被：浏览 1291次 下载 1586次 |
码上扫一扫！ |
|
|
| 可解释性长短期记忆模型用于预测湖泊总磷浓度变化 |
|
丁艺鼎1,2, 范宏翔2, 徐力刚2,3,4, 蒋名亮2, 吕海深1, 朱永华1, 程俊翔2
|
|
1.河海大学水文与水资源学院, 南京 210098;2.中国科学院南京地理与湖泊研究所, 南京 210018;3.中国科学院大学南京学院, 南京 211135;4.江西省鄱阳湖流域生态水利技术创新中心, 南昌 330029
|
|
| 摘要: |
| 对湖泊总磷的变化预测和来源识别对水资源调度和流域生态治理有着重要的意义,然而复杂的生化反应和水动力条件导致的非平稳性给湖泊总磷浓度的准确预测带来极大的困难。为克服这一挑战,本文引入了基于加权回归的季节趋势分解(seasonal and trend decomposition using Loess,STL)技术和夏普利加法(SHapley additive exPlanations,SHAP)结合长短期记忆网络(long short-term memory neural network,LSTM)和门控循环单元(gated recurrent unit,GRU)构建了一个可解释的预测框架,以增强对湖泊总磷浓度演变的预测并提高其可解释性。研究表明:(1)在骆马湖总磷浓度的预测中,该框架拥有较好的预报精度(R2=0.878),优于LSTM和卷积长短期记忆模型(convolutional neural networks and long short term memory network,CNN-LSTM)。当预测时间步长增加到8 h时,该框架有效提高了总磷浓度的预测精度,平均相对误差和均方根误差分别降低了47.1%和33.3%。从预测趋势来看,骆马湖在汛期的总磷平均浓度为0.158 mg/L,相较于非汛期的平均浓度,增加了202.1%。(2)运河来水是骆马湖总磷浓度最重要的影响因素,贡献权重为60.0%,并且不同断面(三湾、三场)的污染源受水动力、气象等因素的影响存在显著的时空差异。本文凸显了神经网络模型在预警水体污染方面的可实施性,并且为提高传统神经网络的学习能力和可解释性的开发与验证提供了重要方向。 |
| 关键词: 深度学习 LSTM SHAP 总磷 可解释性研究 骆马湖 |
| DOI：10.18307/2024.0415 |
| 分类号: |
| 基金项目:国家自然科学基金项目(42307106,U2240224,42071033)、江西省科技计划项目(20232BAB213053,20213AAG01012,20222BCD46002,20224BAB213035)、江西省水利厅科技项目(202325ZDKT08)和长春市科技发展计划项目(23SH03)联合资助。 |
|
| The interpretable long-term and short-term memory model was used to predict the change of total phosphorus concentration in lakes |
|
Ding Yiding1,2, Fan Hongxiang2, Xu Ligang2,3,4, Jiang Mingliang2, Lv Haishen1, Zhu Yonghua1, Chen Junxiang2
|
|
1.College of Hydrology and Water Resources, Hohai University, Nanjing 210098, P. R. China;2.Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences, Nanjing 210018, P. R. China;3.University of Chinese Academy of Sciences, Nanjing, Nanjing 211135, P. R. China;4.Jiangxi Poyang Lake Basin Ecological and Water Conservancy Technology Innovation Center, Jiangxi 330029, P. R. China
|
| Abstract: |
| The prediction and source identification of total phosphorus (TP) in lakes is critical for the management of water resource and watershed ecology. However, non-stationarity caused by inconstant hydrodynamic conditions and the complex biochemical reactions pose significant challenges in accurate forecast of lake TP concentrations. To address this challenge, this study introduced the Seasonal and Trend decomposition using Loess (STL) technique and SHapley additive exPlanations (SHAP), and combined them with Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) to develop an interpretable prediction framework. The framework was applied to enhance the prediction of lake TP concentrations and improving their interpretability. The study achieved the following results. (1) In the prediction of TP concentrations in Lake Luoma, this framework achieved a higher model fit with an R2 value of 0.878, outperforming LSTM and CNN-LSTM. By increasing the prediction time step to 8 hours, the framework achieved a better model fit with a decrease of MRE and RMSE by 47.1% and 33.3%, respectively. An analysis of the prediction trend for Lake Luoma revealed that the average TP concentration is 0.158 mg/L during the flood season, 202.1% higher than that during non-flood seasons. (2) Canal inflow was the most influential factor on TP concentrations, with a contribution of 60%. Different sections (Sanwan and Sanchang) had large spatiotemporal variations in phosphorus sources influenced by hydrodynamics and meteorological factors. This study highlighted the potential of neural network models in predicting water pollution, and offered valuable insights into enhancing the learning capabilities and interpretability of traditional neural networks. |
| Key words: Deep learning LSTM SHAP total phosphorus interpretability Lake Luoma |
|
|
|
|
