| 廖胜利,王在能,程春田,吴慧军.耦合气象信息和可解释深度学习方法的日径流预报[J].水利学报,2025,56(7):831-843 |
| 耦合气象信息和可解释深度学习方法的日径流预报 |
| Coupling meteorological information and interpretable deep learning methods for daily runoff forecasting |
| 投稿时间：2024-11-14 修订日期：2025-05-09 |
| DOI：10.13243/j.cnki.slxb.20240737 |
| 中文关键词: 径流预报 SHAP CNN-BiLSTM ERA5 输入特征 可解释性 |
| 英文关键词: runoff forecasting SHAP CNN-BiLSTM ERA5 input features interpretability |
| 基金项目:国家自然科学基金项目(52379004,51979023) |
|
| 摘要点击次数: 104 |
| 全文下载次数: 70 |
| 中文摘要: |
| 高精度的径流预报信息是水利防洪和水电调度的重要依据。然而，输入特征的选择、模型结构的确定及预报结果的可解释性等因素严重限制了数据驱动模型在径流预报中的应用。本文以实测径流、降雨和包含丰富气象信息的ERA5再分析数据等多源数据为输入，将沙普利可加性解释（SHAP）方法和卷积神经网络-双向长短期记忆网络（CNN-BiLSTM）模型相结合，建立了一种通过“事后”特征选择机制优化模型输入的可解释径流预报框架。首先，多源数据为框架提供了丰富的输入信息，CNN和BiLSTM分别捕捉数据中的空间相关性和时序信息，提高模型预报精度。其次，模型训练后使用SHAP方法计算各输入特征的贡献度，通过比较不同输入条件下模型的预报表现，确定最佳输入特征及模型结构。最后，通过量化和可视化输入特征贡献度，提供了对模型预报机制和结果的解释。将所提方法应用于天生桥一级电站的区间入库流量预报并与通过偏自相关系数（PACF）、互相关系数（CCF）和随机森林（RF）等传统方法选择输入特征的模型进行比较。结果表明，与传统方法相比，SHAP方法不仅优化了模型特征空间选择而且显著改善了数据驱动模型在径流预报中的输入特征选择困难、无效以及预报结果难以解释的局限性。 |
| 英文摘要: |
| Accurate runoff forecast information is an important basis for flood control and hydropower scheduling.However,factors such as the selection of input features,determination of model structure,and interpretability of forecasting results severely limit the application of data-driven models in runoff forecasting.This study utilizes multisource data,including observed runoff,rainfall,and ERA5 reanalysis data enriched with comprehensive meteorological information,as inputs.The framework combines the SHapley Additive exPlanations （SHAP） method with a Convolutional Neural Network-Bidirectional Long Short-Term Memory （CNN-BiLSTM） model to establish an interpretable runoff forecasting framework optimized through a post-hoc feature selection mechanism.Firstly,the multisource data enriches the framework with diverse input information,with CNN and BiLSTM respectively capturing spatial correlations and temporal patterns in the data to enhance forecasting accuracy.Secondly,following model training,SHAP calculates the contribution of each input feature.By comparing model performance under different input conditions,optimal input features and model structures are determined.Finally,by quantifying and visualizing the contribution of input features,the framework provides insights into the forecasting mechanism and results.The proposed method is applied to interval inflow forecasting at the Tian Sheng Qiao Hydropower Station and compared with models using input features selected by traditional methods such as Partial Autocorrelation Function （PACF）,CrossCorrelation Function （CCF）,and Random Forest （RF）.The results demonstrate that the SHAP-based approach not only optimizes feature selection compared to traditional methods,but also significantly improves the challenges faced by data-driven models in runoff forecasting,including difficulties in input feature selection,inefficiencies,and interpreting forecasting results. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|