Development of numerical and data models for the support of digital twins in offshore wind engineering

Rights statement
Awarding institution
  • University of Strathclyde
Date of award
  • 2022
Thesis identifier
  • T16199
Person Identifier (Local)
  • 201879973
Qualification Level
Qualification Name
Department, School or Faculty
  • As offshore wind farms grow there is a continued demand for reduced costs. Maintenance costs and downtime can be reduced through greater information on the asset in relation to its operational loads and structural resistance to damage and so there is an increasing interest in digital twin technologies. Through digital twins, an operational asset can be replicated computationally, thus providing more information. Modelling these aspects requires a wide variety of models in different fields. To advance the feasibility of digital twin technology this thesis aims to develop the multi-disciplinary set of modelling domains which help form the basis of future digital twins. Throughout this work, results have been validated against operational data recorded from sensors on offshore structures. This has provided value and confidence to the results as it shows how well the mix of state-of-the art models compare to real world engineering systems. This research presents a portfolio of five research areas which have been published in a mix of peer-reviewed journal articles and conference papers. These areas are: 1) A computational fluid dynamics (CFD) model of an offshore wind farm conducted using a modified solver in the opensource software. This work implements actuator disk turbine models and uses Reynolds averaged Naiver Stokes approaches to represent the turbulence. This investigates the impact of modelling choices and demonstrates the impact of varied model parameters. The results are compared to operational site data and the modelling errors are quantified. There is good agreement between the models and site data. 2) An expansion on traditional CFD approaches through incorporating machine learning (ML). These ML models are used to approximate the results of the CFD and thereby allow for further analysis which retains the fidelity of CFD at comparatively negligible computational cost. The results are compared to operational site data and the errors at each step are quantified for validation. 3) A time-series forecasting of weather variables based on past measured data. A novel approach for forecasting time-series is developed and compared to two existing methods: Markov-Chains and Gradient Boosting. While this new method is more complex and requires more time to train, it has the desirable feature of incorporating seasonality at multiple timescales and thus providing a more representative time-series. 4) An investigation of the change in modal parameters in an offshore wind jacket structure from damages or from changing operational conditions. In this work the detailed design model of the structure from Ramboll is used. This section relates the measurable modal parameters to the operational condition through a modelling approach. 5) A study conducted using accelerometer data from an Offshore Substation located in a wind farm site. Operational data from 12 accelerometers is used to investigate the efficacy of several potential sensor layouts and therefore to quantify the consequence of placement decisions. The results of these developments are an overall improvement in the modelling approaches necessary towards the realisation of digital twins as well as useful development in each of the component areas. Both areas related to wind loading as well as structural dynamics have been related to operational data. The validation of this link between the measured and the modelled domains facilitates operators and those in maintenance in gaining more information and greater insights into the conditions of their assets.
Advisor / supervisor
  • Kolios, Athanasios
  • Brennan, Feargal
Resource Type
  • Error on title page. Date of award is 2022.