Artificial intelligence-assisted performance prediction of Francis turbine
ENGINEERING COMPUTATIONS, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Basım Tarihi: 2026
- Doi Numarası: 10.1108/ec-12-2025-1491
- Dergi Adı: ENGINEERING COMPUTATIONS
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, ABI/INFORM, Aerospace Database, Compendex, INSPEC, zbMATH, Academic Search Ultimate (EBSCO), Engineering Source (EBSCO), Materials Science & Engineering Collection (ProQuest), Pharma Collection (ProQuest), Technology Collection (ProQuest)
- Gazi Üniversitesi Adresli: Evet
Özet
Purpose-The purpose of this study is to assess the technical and economic feasibility of integrating a Francis-type micro-hydropower turbine into a pressurized drinking water pipeline using data-driven performance prediction. Daily flow rate time series are forecast using deep learning models to capture the dynamic operating conditions of the system. The predicted flow data are then used to construct flow-duration curves and head-discharge relationships for turbine selection and annual energy estimation. The study aims to demonstrate the applicability of machine learning-based forecasting for reliable micro-hydropower feasibility assessments. Design/methodology/approach-Daily flow rate data obtained from a pressurized drinking water transmission pipeline were analyzed as time-series. Flow-duration curves and head-discharge relationships were derived using classical hydraulic methods. To predict future operating conditions, deep learning-based long short-term memory (LSTM) and gated recurrent unit (GRU) models were developed and optimized through hyperparameter tuning. The forecasted flow series were then used to evaluate turbine operating conditions, estimate annual energy production, and assess the technical and economic feasibility of integrating a Francis-type in-pipe micro-hydropower turbine. Findings-The optimized LSTM model achieved a prediction accuracy exceeding R 2 5 0.90, demonstrating strong capability in capturing the short-term memory behavior of the pipeline flow system. Based on the predicted flow series, the optimal design discharge and head were identified as approximately 550 L/s and 70 m, respectively. The feasibility analysis indicates that installing a Francis-type in-pipe turbine with an installed capacity of about 290 kW is technically viable. The system is expected to generate approximately 2.01 GWh of electrical energy annually, yielding favorable economic returns. Research limitations/implications-The study is based on historical flow data from a single pressurized drinking water pipeline, which may limit the generalizability of the results to other network configurations. Future operating conditions were estimated using data-driven models rather than real measured values. Nevertheless, the proposed methodology provides a transferable framework for preliminary feasibility assessments in similar water infrastructure systems. Future research may incorporate longer datasets, multiple pipeline cases, transient hydraulic effects and alternative deep learning architectures to improve prediction robustness and broader applicability. Practical implications-The proposed approach offers a practical decision-support tool for water utilities and engineers to evaluate the feasibility of micro-hydropower installations in pressurized pipeline systems. By combining machine learning-based flow forecasting with classical hydraulic analysis, turbine selection and energy production can be assessed without extensive field testing. The methodology enables efficient identification of suitable operating conditions, supports energy recovery from excess pressure and facilitates cost-effective integration of Francis-type turbines into existing water infrastructure, contributing to sustainable energy generation and reduced operational losses. Social implications-The implementation of micro-hydropower systems in existing water infrastructure supports the transition toward cleaner and more sustainable energy production without additional environmental or land-use impacts. Recovering energy from excess pressure in drinking water pipelines can reduce carbon emissions, lower operational costs for water utilities and contribute indirectly to more affordable public services. By promoting energy efficiency and renewable energy integration within urban infrastructure, the proposed approach supports broader societal goals related to sustainability, climate resilience and responsible resource management. Originality/value-This study presents a novel integration of deep learning-based time-series forecasting with classical hydraulic feasibility analysis for micro-hydropower applications in pressurized water pipelines. Unlike existing studies that rely on static or short-term data, the proposed approach uses long-term daily flow predictions to support turbine selection and energy estimation. The application of LSTM and GRU models to a Francis-type in-pipe turbine feasibility problem provides a data-driven framework that enhances reliability, reduces uncertainty and extends the methodological toolkit for sustainable energy recovery in water infrastructure systems.