Explainable machine learning model for multi-step forecasting of reservoir inflow with uncertainty quantification

Ming Fan; Siyan Liu; Dan Lu; Sudershan Gangrade; Shih Chieh Kao

doi:10.1016/j.envsoft.2023.105849

Explainable machine learning model for multi-step forecasting of reservoir inflow with uncertainty quantification

Ming Fan, Siyan Liu, Dan Lu, Sudershan Gangrade, Shih Chieh Kao

Research output: Contribution to journal › Article › peer-review

7 Scopus citations

Abstract

We propose an explainable machine learning (ML) model with uncertainty quantification (UQ) to improve multi-step reservoir inflow forecasting. Traditional ML methods have challenges in forecasting inflows multiple days ahead, and lack explainability and UQ. To address these limitations, we introduce an encoder–decoder long short-term memory (ED-LSTM) network for multi-step forecasting, employ the SHapley Additive exPlanation (SHAP) technique for understanding the influence of hydrometeorological factors on inflow prediction, and develop a novel UQ method for prediction trustworthiness. We apply these methods to forecast 7-day inflow in snow-dominant and rain-driven reservoirs. The results demonstrate the effectiveness of the ED-LSTM model, with high forecasting accuracy for short lead times. Our UQ method provides reliable uncertainty estimates, covering 90% of data with a 90% confidence level. The SHAP analysis reveals the importance of historical inflow and precipitation as influential factors. These findings and methods may support reservoir operators in optimizing water resources management decisions.

Original language	English
Article number	105849
Journal	Environmental Modelling and Software
Volume	170
DOIs	https://doi.org/10.1016/j.envsoft.2023.105849
State	Published - Dec 2023

Funding

We thank Lujun Zhang and Tiantian Yang for processing data in the Upper Colorado River Basin. This study is supported by the US Department of Energy (DOE) Water Power Technologies Office, USA. Additional supports are from the Artificial Intelligence Initiative as a part of the Laboratory Directed Research and Development Program of Oak Ridge National Laboratory (ORNL), USA. The research used resources of the Oak Ridge Leadership Computing Facility at ORNL, which is a DOE Office of Science User Facility. ORNL is managed by UT-Battelle for DOE, USA, under Contract DE-AC05-00OR22725. Accordingly, the US government retains and the publisher, by accepting the article for publication, acknowledges that the US government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this manuscript, or allow others to do so, for US Government purposes. We thank Lujun Zhang and Tiantian Yang for processing data in the Upper Colorado River Basin. This study is supported by the US Department of Energy (DOE) Water Power Technologies Office, USA . Additional supports are from the Artificial Intelligence Initiative as a part of the Laboratory Directed Research and Development Program of Oak Ridge National Laboratory (ORNL), USA . The research used resources of the Oak Ridge Leadership Computing Facility at ORNL, which is a DOE Office of Science User Facility. ORNL is managed by UT-Battelle for DOE, USA , under Contract DE-AC05-00OR22725 . Accordingly, the US government retains and the publisher, by accepting the article for publication, acknowledges that the US government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this manuscript, or allow others to do so, for US Government purposes. Copyright statement: This manuscript has been authored by UT-Battelle, LLC under Contract No. DE-AC05-00OR22725 with the US Department of Energy (DOE). The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan ( http://energy.gov/downloads/doe-public-access-plan ).

Keywords

Encoder–decoder LSTM
Explainable machine learning
Multi-step forecasting
Reservoir inflow
SHAP
Uncertainty quantification

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1016/j.envsoft.2023.105849

Cite this

@article{f1540de3d0374dc7ae5d67e1f7e35f44,

title = "Explainable machine learning model for multi-step forecasting of reservoir inflow with uncertainty quantification",

abstract = "We propose an explainable machine learning (ML) model with uncertainty quantification (UQ) to improve multi-step reservoir inflow forecasting. Traditional ML methods have challenges in forecasting inflows multiple days ahead, and lack explainability and UQ. To address these limitations, we introduce an encoder–decoder long short-term memory (ED-LSTM) network for multi-step forecasting, employ the SHapley Additive exPlanation (SHAP) technique for understanding the influence of hydrometeorological factors on inflow prediction, and develop a novel UQ method for prediction trustworthiness. We apply these methods to forecast 7-day inflow in snow-dominant and rain-driven reservoirs. The results demonstrate the effectiveness of the ED-LSTM model, with high forecasting accuracy for short lead times. Our UQ method provides reliable uncertainty estimates, covering 90% of data with a 90% confidence level. The SHAP analysis reveals the importance of historical inflow and precipitation as influential factors. These findings and methods may support reservoir operators in optimizing water resources management decisions.",

keywords = "Encoder–decoder LSTM, Explainable machine learning, Multi-step forecasting, Reservoir inflow, SHAP, Uncertainty quantification",

author = "Ming Fan and Siyan Liu and Dan Lu and Sudershan Gangrade and Kao, {Shih Chieh}",

note = "Publisher Copyright: {\textcopyright} 2023",

year = "2023",

month = dec,

doi = "10.1016/j.envsoft.2023.105849",

language = "English",

volume = "170",

journal = "Environmental Modelling and Software",

issn = "1364-8152",

publisher = "Elsevier",

}

TY - JOUR

T1 - Explainable machine learning model for multi-step forecasting of reservoir inflow with uncertainty quantification

AU - Fan, Ming

AU - Liu, Siyan

AU - Lu, Dan

AU - Gangrade, Sudershan

AU - Kao, Shih Chieh

PY - 2023/12

Y1 - 2023/12

N2 - We propose an explainable machine learning (ML) model with uncertainty quantification (UQ) to improve multi-step reservoir inflow forecasting. Traditional ML methods have challenges in forecasting inflows multiple days ahead, and lack explainability and UQ. To address these limitations, we introduce an encoder–decoder long short-term memory (ED-LSTM) network for multi-step forecasting, employ the SHapley Additive exPlanation (SHAP) technique for understanding the influence of hydrometeorological factors on inflow prediction, and develop a novel UQ method for prediction trustworthiness. We apply these methods to forecast 7-day inflow in snow-dominant and rain-driven reservoirs. The results demonstrate the effectiveness of the ED-LSTM model, with high forecasting accuracy for short lead times. Our UQ method provides reliable uncertainty estimates, covering 90% of data with a 90% confidence level. The SHAP analysis reveals the importance of historical inflow and precipitation as influential factors. These findings and methods may support reservoir operators in optimizing water resources management decisions.

AB - We propose an explainable machine learning (ML) model with uncertainty quantification (UQ) to improve multi-step reservoir inflow forecasting. Traditional ML methods have challenges in forecasting inflows multiple days ahead, and lack explainability and UQ. To address these limitations, we introduce an encoder–decoder long short-term memory (ED-LSTM) network for multi-step forecasting, employ the SHapley Additive exPlanation (SHAP) technique for understanding the influence of hydrometeorological factors on inflow prediction, and develop a novel UQ method for prediction trustworthiness. We apply these methods to forecast 7-day inflow in snow-dominant and rain-driven reservoirs. The results demonstrate the effectiveness of the ED-LSTM model, with high forecasting accuracy for short lead times. Our UQ method provides reliable uncertainty estimates, covering 90% of data with a 90% confidence level. The SHAP analysis reveals the importance of historical inflow and precipitation as influential factors. These findings and methods may support reservoir operators in optimizing water resources management decisions.

KW - Encoder–decoder LSTM

KW - Explainable machine learning

KW - Multi-step forecasting

KW - Reservoir inflow

KW - SHAP

KW - Uncertainty quantification

UR - http://www.scopus.com/inward/record.url?scp=85174040694&partnerID=8YFLogxK

U2 - 10.1016/j.envsoft.2023.105849

DO - 10.1016/j.envsoft.2023.105849

M3 - Article

AN - SCOPUS:85174040694

SN - 1364-8152

VL - 170

JO - Environmental Modelling and Software

JF - Environmental Modelling and Software

M1 - 105849

ER -

Explainable machine learning model for multi-step forecasting of reservoir inflow with uncertainty quantification

Abstract

Funding

Keywords

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this