Guardado en:
Detalles Bibliográficos
Autores principales: Hu, Pengfei, Fan, Ming, Han, Xiaoxue, Lu, Chang, Zhang, Wei, Kang, Hyun, Ning, Yue, Lu, Dan
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2511.07649
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866912700980789248
author Hu, Pengfei
Fan, Ming
Han, Xiaoxue
Lu, Chang
Zhang, Wei
Kang, Hyun
Ning, Yue
Lu, Dan
author_facet Hu, Pengfei
Fan, Ming
Han, Xiaoxue
Lu, Chang
Zhang, Wei
Kang, Hyun
Ning, Yue
Lu, Dan
contents Reservoir inflow prediction is crucial for water resource management, yet existing approaches mainly focus on single-reservoir models that ignore spatial dependencies among interconnected reservoirs. We introduce AdaTrip as an adaptive, time-varying graph learning framework for multi-reservoir inflow forecasting. AdaTrip constructs dynamic graphs where reservoirs are nodes with directed edges reflecting hydrological connections, employing attention mechanisms to automatically identify crucial spatial and temporal dependencies. Evaluation on thirty reservoirs in the Upper Colorado River Basin demonstrates superiority over existing baselines, with improved performance for reservoirs with limited records through parameter sharing. Additionally, AdaTrip provides interpretable attention maps at edge and time-step levels, offering insights into hydrological controls to support operational decision-making. Our code is available at https://github.com/humphreyhuu/AdaTrip.
format Preprint
id arxiv_https___arxiv_org_abs_2511_07649
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Adaptive Graph Learning with Transformer for Multi-Reservoir Inflow Prediction
Hu, Pengfei
Fan, Ming
Han, Xiaoxue
Lu, Chang
Zhang, Wei
Kang, Hyun
Ning, Yue
Lu, Dan
Machine Learning
Artificial Intelligence
Reservoir inflow prediction is crucial for water resource management, yet existing approaches mainly focus on single-reservoir models that ignore spatial dependencies among interconnected reservoirs. We introduce AdaTrip as an adaptive, time-varying graph learning framework for multi-reservoir inflow forecasting. AdaTrip constructs dynamic graphs where reservoirs are nodes with directed edges reflecting hydrological connections, employing attention mechanisms to automatically identify crucial spatial and temporal dependencies. Evaluation on thirty reservoirs in the Upper Colorado River Basin demonstrates superiority over existing baselines, with improved performance for reservoirs with limited records through parameter sharing. Additionally, AdaTrip provides interpretable attention maps at edge and time-step levels, offering insights into hydrological controls to support operational decision-making. Our code is available at https://github.com/humphreyhuu/AdaTrip.
title Adaptive Graph Learning with Transformer for Multi-Reservoir Inflow Prediction
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2511.07649