Thesis
Field-data based reliability modelling of wind turbine subsystems
- Creator
- Rights statement
- Awarding institution
- University of Strathclyde
- Date of award
- 2025
- Thesis identifier
- T17465
- Person Identifier (Local)
- 202063588
- Qualification Level
- Qualification Name
- Department, School or Faculty
- Abstract
- Operations and maintenance (O&M) costs account for up to one third of the levelized cost of energy of wind farms. Wind turbine component failures lead to significant repair costs and revenue losses, making the use of operational knowledge crucial for reducing associated costs and risks. However, uncertainty persists due to a lack of quantified reliability of wind turbines and their components, particularly for newer turbine generations. This gap directly impacts O&M decision-making, as simulations and calculations depend on accurate reliability inputs. The increasing size and further evolving technology of wind turbines further complicate projections, which are essential for future wind farm planning and competitive auction bidding. This thesis presents a series of analyses, which address the research question “How can operations and maintenance data be utilised more efficiently to further reduce costs and risks during wind farm planning as well as operation?” For this, a comprehensive review of existing reliability data, highlighting the shortcomings of previous studies, is provided and advantages and limitations of different reliability assessment methods are investigated. A detailed economic life cycle simulation and assessment framework is developed, integrating a cost-revenue model that accounts for CAPEX, OPEX, and revenue factors, as well as wake and blockage effects for offshore wind farms in the German North and Baltic Seas. A digitalisation workflow is introduced to transform unstructured, non-standardised maintenance reports into machine-readable data classifying components worked on during turbine visits. The feasibility of using text classifiers for preprocessing maintenance reports is evaluated, demonstrating their potential to reduce manual data processing efforts. Furthermore, the impact of classification methods on reliability key performance indicators is analysed. The thesis utilises a unique dataset of 1335 onshore and offshore turbines with rated capacities of up to 9 MW, which covers maintenance records from 2006 to 2024, offering a highly diverse and recent data resource compared to previous studies. A thorough analysis of failure rates, repair times, and maintenance resource requirements is conducted, providing O&M simulation input for 29 subsystems, covering major component replacements, further corrective maintenance as well as preventive maintenance interventions. Failure behaviour over time for the entire wind turbine system and key subsystems is analysed using Nelson-Aalen plots, while the influence of covariates is assessed with a non-homogeneous Poisson process (NHPP) model. A comprehensive analysis of component failures within the pitch and converter subsystems is conducted comparing electrical and hydraulic pitch systems as well as low-voltage and medium voltage power converters, respectively. Finally, the thesis compares the developed reliability modelling approaches against a previously published study and the impact of these models on O&M simulations is assessed, highlighting the limitations of average failure rates and the advantages of NHPP regression modelling. The results indicate that although onshore wind turbines experience lower failure rates per turbine and year, their failure rates per megawatt of rated turbine capacity per year are higher than those of offshore turbines. The pitch, control, and converter subsystems are identified as the most critical with respect to high failure rates. The analysis reveals distinct reliability patterns across wind turbine subsystems over wind turbine operating age. While some subsystems follow a classical bathtub curve, others transition directly from early failures to deterioration, highlighting the need for time-dependent, subsystem-specific reliability modelling rather than assuming uniform failure behaviour. The results of NHPP regression in combination with a covariate selection process confirm that multiple factors significantly influence wind turbine and subsystem reliability. Newer turbine commissioning years generally enhance reliability, reflecting technological and design advancements. However, higher rated turbine capacity negatively impacts reliability, aligning with previous findings that larger turbines experience higher failure intensities. These opposing trends underscore the advantages of NHPP modelling in separating and quantifying individual covariate effects. Additionally, subsystem design choices are found to be a key determinant of reliability. Keywords: wind turbines, operations and maintenance, reliability analysis, failure rate, corrective and preventive maintenance, maintenance reports, field data, failure data, reliability modelling, Nelson-Aalen plot, non-homogeneous Poisson process, digitalisation.
- Advisor / supervisor
- Fischer, Katharina
- Kolios, Athanasios
- Resource Type
- DOI
- Funder
Relations
Items
| Thumbnail | Title | Date Uploaded | Visibility | Actions |
|---|---|---|---|---|
|
|
PDF of thesis T17465 | 2025-09-30 | Public | Download |