Abdi et al. (2025) Advancing Hydrological Prediction with Hybrid Quantum Neural Networks: A Comparative Study for Mile Mughan Dam

Identification

• Journal: Water
• Year: 2025
• Date: 2025-12-18
• Authors: Erfan Abdi, Mohammad Taghi Sattari, Saeed Samadianfard, Sajjad Ahmad
• DOI: 10.3390/w17243592

Research Groups

• Department of Water Engineering, Faculty of Agriculture, University of Tabriz, Tabriz, Iran
• Water Sciences and Hydroinformatics Research Center, Khazar University, Baku, Azerbaijan
• Department of Agricultural Engineering, Ankara University, Ankara, Türkiye
• Department of Environmental Engineering, Izmir Institute of Technology, Izmir, Türkiye
• Department of Civil and Environmental Engineering and Construction, University of Nevada Las Vegas, Las Vegas, NV, USA

Short Summary

This study compares a hybrid quantum neural network (HQNN) with two classical models (bidirectional CNN-LSTM and SVR) to predict monthly inflow to the Mile Mughan Dam. The HQNN demonstrated superior performance across all metrics in both multivariate and univariate scenarios, confirming its reliability for hydrological prediction.

Objective

• To conduct a comprehensive comparative analysis of a Hybrid Quantum Neural Network (HQNN), a Convolutional Neural Network–Bidirectional Long Short-Term Memory (CNN-BiLSTM) model, and Support Vector Regression (SVR) for predicting monthly inflow into the Mile Mughan Dam under univariate (historical inflow lags only) and multivariate (inflow lags and meteorological parameters) frameworks.

Study Configuration

• Spatial Scale: Mile Mughan Dam, a transboundary hydroelectric and irrigation complex located on the Aras River between Azerbaijan and Iran, supporting approximately 400,000 hectares of farmland.
• Temporal Scale: Monthly inflow prediction using a 14-year dataset spanning January 2010 to June 2023.

Methodology and Data

• Models used: Hybrid Quantum Neural Network (HQNN), Convolutional Neural Network–Bidirectional Long Short-Term Memory (CNN-BiLSTM), Support Vector Regression (SVR).
• Data sources: Monthly inflow records from the Water Department (Regional Water Company of East Azerbaijan Province, Iran). Complementary meteorological data (precipitation, mean temperature, mean humidity) retrieved from the Mathematica 13.3 software database.

Main Results

• The Hybrid Quantum Neural Network (HQNN) consistently outperformed classical models in both forecasting scenarios.
• Scenario 1 (Multivariate: mean temperature, mean humidity, precipitation, and three-month inflow lags):
  • HQNN achieved R² = 0.915, RMSE = 37.318 million cubic meters (MCM), NSE = 0.908, and MAPE = 8.343%.
  • CNN-BiLSTM achieved R² = 0.867, RMSE = 46.506 MCM, NSE = 0.858, and MAPE = 10.795%.
  • SVR achieved R² = 0.846, RMSE = 52.372 MCM, NSE = 0.821, and MAPE = 12.772%.
• Scenario 2 (Univariate: three-month inflow lags only):
  • HQNN maintained strong performance with R² = 0.855, RMSE = 48.561 MCM, NSE = 0.845, and MAPE = 9.979%.
• HQNN increased prediction accuracy by 19.76% in Scenario 1 and 13.47% in Scenario 2 compared to the CNN-BiLSTM model.
• HQNN demonstrated superior distributional fidelity, yielding the smallest Jensen–Shannon and Wasserstein distances to the actual inflow distribution, indicating better capture of overall inflow patterns.

Contributions

• First comprehensive comparative analysis of a Hybrid Quantum Neural Network (HQNN) against advanced classical deep learning (CNN-BiLSTM) and statistical (SVR) models for dam inflow prediction.
• Demonstrated the superior performance and reliability of HQNN in capturing complex, nonlinear temporal dynamics in hydrological systems, particularly in both multivariate (data-rich) and univariate (data-limited) contexts.
• Validated the potential of quantum-inspired machine learning to overcome classical computational limits and enhance hydrological forecasting accuracy, offering a robust framework for operational water inflow forecasting.
• Provided a transferable framework relevant to other dam-regulated systems globally, with significant implications for transboundary water resource management.

Funding

• This research received no external funding.

Citation

@article{Abdi2025Advancing,
  author = {Abdi, Erfan and Sattari, Mohammad Taghi and Samadianfard, Saeed and Ahmad, Sajjad},
  title = {Advancing Hydrological Prediction with Hybrid Quantum Neural Networks: A Comparative Study for Mile Mughan Dam},
  journal = {Water},
  year = {2025},
  doi = {10.3390/w17243592},
  url = {https://doi.org/10.3390/w17243592}
}