A Polynomial Regression-Based Approach for Efficiency and Performance Trend Monitoring of Hydropower Turbines
Open Access
Abstract
Hydropower plays an increasingly important role in sustainable energy systems. While efficient plant operation is essential, hydropower efficiency is often assessed only during commissioning or over short measurement periods, leaving long-term trends largely unexplored. This limits the early detection of performance degradation and reduces opportunities for timely maintenance and optimization. To address this gap, this paper presents a framework for continuous efficiency trend monitoring based on polynomial regression, combined with measurement uncertainty analysis to distinguish genuine performance changes from noise. Particular attention is given to data preprocessing and aggregation. The resulting regression models used to estimate turbine efficiency from operational and discharge measurements achieved root mean square error (RMSE) values between 0.149% and 0.176%. An alternative model that predicts power output from head and guide vane opening yielded an RMSE of 0.28% in the most frequently used operating range. The resulting residuals were tracked over time and remained within the confidence levels defined by measurement uncertainty. Seasonal effects were also evident in the results, demonstrating the potential of the proposed approach for long-term monitoring of hydropower plant performance.Keywords
Funding Information
- European Union’s Horizon Europe research and innovation program (101147310)
This publication has 3 references indexed in Scilit:
- A hybrid methodology for assessing hydropower plants under flexible operations: Leveraging experimental data and machine learning techniquesApplied Energy, 2025
- Silt erosion and cavitation impact on hydraulic turbines performance: An in-depth analysis and preventative strategiesHeliyon, 2024
- Performance monitoring of kaplan turbine based hydropower plant under variable operating conditions using machine learning approachSustainable Computing: Informatics and Systems, 2024