Cheng et al. (2026) Dynamic water delay time estimation using dispatch data for improved inflow forecasting in cascade hydropower reservoirs

Identification

• Journal: Journal of Hydrology
• Year: 2026
• Date: 2026-02-19
• Authors: Xiong Cheng, Junwen Yang, Bin Luo, Miriam R. Aczel, Wenwu Li, Hao Zhong, Xianshan Li, Shumian Miao
• DOI: 10.1016/j.jhydrol.2026.135170

Research Groups

• Hubei Provincial Key Laboratory for Operation and Control of Cascaded Hydropower Stations and Renewable Energy, China Three Gorges University, Yichang 443002, China
• United Nations University Institute for Water, Environment and Health, 225 East Beaver Creek Road, Richmond Hill, L4B 3P4 Ontario, Canada
• College of Hydraulic and Environment Engineering, China Three Gorges University, Yichang 443002, China
• Sichuan Energy Internet Research Institute Tsinghua University, Chengdu, Sichuan 610042, China
• College of Electrical Engineering and New Energy, China Three Gorges University, Yichang 443002, China
• State Grid Sichuan Electric Power Company, Chengdu 610041, China

Short Summary

This study proposes an operation-data-driven probabilistic framework to improve inflow forecasting in cascade hydropower systems by dynamically estimating and quantifying the uncertainty of water delay time. Applied to the Lancang River cascade, the method significantly reduces forecasting errors compared to benchmark models, providing a practical, self-sufficient, and uncertainty-aware solution for dispatch centers.

Objective

• To develop an operation-data-driven probabilistic framework for dynamic water delay time estimation and uncertainty-aware inflow forecasting in cascade hydropower reservoirs, addressing the limitations of fixed delay times and lack of detailed hydraulic data.

Study Configuration

• Spatial Scale: Lancang River cascade hydropower system.
• Temporal Scale: Short-term operational forecasting, relevant for daily-regulated hydropower plants.

Methodology and Data

• Models used:
  • Proposed framework: Enhanced Dynamic Time Warping (DTW) for delay time extraction, Copula-Bayesian model for multivariate dependency and uncertainty quantification, and a probabilistic delay-matching scheme for forecasting.
  • Benchmark models for comparison: Back-Propagation Neural Network (BPNN), Convolutional Neural Network (CNN), Random Forest (RF), and Muskingum model.
• Data sources: Historical inflow/outflow records (dispatch data) from the cascade hydropower system.

Main Results

• The proposed method demonstrated superior robustness and accuracy compared to benchmark models (BPNN, CNN, RF, Muskingum).
• Under normal hydrological conditions, forecasting errors at two key stations were reduced by 63.61% and 68.69%.
• Under slight drought conditions, forecasting errors at the same two key stations were reduced by 24.48% and 27.07%.

Contributions

• Introduces an operation-data-driven probabilistic framework for dynamic water delay time estimation and uncertainty-aware inflow forecasting, shifting from deterministic to probabilistic predictions.
• Integrates an enhanced Dynamic Time Warping technique for high-resolution water delay time extraction directly from historical dispatch data, eliminating the need for external hydraulic monitoring.
• Develops a Copula-Bayesian model to capture multivariate dependencies and quantify water delay time uncertainty under varying conditions.
• Provides a practical, self-sufficient, and uncertainty-aware solution that enhances forecast reliability and supports adaptive scheduling under hydrological variability for hydropower dispatch centers.

Funding

• Not explicitly mentioned in the provided text.

Citation

@article{Cheng2026Dynamic,
  author = {Cheng, Xiong and Yang, Junwen and Luo, Bin and Aczel, Miriam R. and Li, Wenwu and Zhong, Hao and Li, Xianshan and Miao, Shumian},
  title = {Dynamic water delay time estimation using dispatch data for improved inflow forecasting in cascade hydropower reservoirs},
  journal = {Journal of Hydrology},
  year = {2026},
  doi = {10.1016/j.jhydrol.2026.135170},
  url = {https://doi.org/10.1016/j.jhydrol.2026.135170}
}