Reliability Modelling of Wind Turbines Using Spatial Analysis

  • Number of scholarships 1
  • Value £14,296 (pa for 3 years)
  • Opens 16 February 2016
  • Deadline 5 August 2016
  • Help with Tuition fees, Living costs
  • Duration 36 months


Candidates should have:


  • A good Honours degree (minimum 2:1) and/or a Master’s degree in a quantitively focused subject area (e.g. economics, management science, statistics, mathematics, engineering.)
  • They may also have appropriate experience or other skills which are relevant to this project. Relevant skills could include experience developing reliability models, spatial analysis models, experience in the energy sector, etc.
  • As well as a CV and relevant qualification transcripts, applications must be accompanied by a cover letter indicating how their skills and experience would contribute to this project.
  • Two references are required, of which at least one should be from an academic.

Project Details

Project Details

There has been a significant growth in onshore and offshore wind developments in the past decade in the EU. By 2030, it is expected that 400GW of EU energy will be generated using wind resource – establishing it as the main power technology. At this time, a large amount of focus is on designing, manufacturing and installing large scale wind farms, both onshore and off. Over the next ten years, it is expected that around 2000 wind turbines will be operational in the UK alone. As the number of operational turbines increases and the market matures, attention naturally moves to focusing on operations, maintenance and servicing issues. Much attention has focused on developing infrastructure to monitor the performance and health of the assets. Following this, models have been developed to analyse the data generated by the sensors to inform maintenance related decisions and the equipment needed to perform maintenance. There are already numerous papers in this area, however, one dimension that has been under explored in this literature is the  degree of  interdependence between neighbouring turbines in producing the output from SCADA and telemetry systems, e.g.  spillovers between turbines based on distance between each turbine, or shared environmental factors not controlled for by wind speed/direction, etc.

While power production models at the planning stage explicitly and extensively capture interdependencies between turbines, during operations, decision making fails to consider spatial impact when modelling for a host of decisions, e.g. curtailment strategies, fault detection and maintenance modelling. Anecdotal evidence provided by utilities and academics indicate that the location of a turbine can have an impact on both the volume and type of alarms that are observed, the likelihood of those alarms being false positives, different maintenance patterns, etc. One way to capture this lack of independence is through modelling these outages using a model that is robust to this omitted factor, using spatial econometric models. This would provide a more reliable basis to predict these disruptions. Using spatial econometric modelling techniques also enables a direct test of the assumption of independence that underlies much of the existing work in this area, where no account is taken of interactions between turbines generating spillovers, for instance in terms of the probability of faults or maintenance issues. Simply put, given both shared environment, close spatial proximity, and the 'wake'  effect of wind turbines (which generates eddies of turbulence) affecting the other turbines, there are good reasons to think that the probability of a given turbine generating a fault or error code, and requiring maintenance, is not independent of its location within the wind farm.  In light of this, this PhD will design and implement a statistical model to explain and predict the type and frequency of error codes and maintenance issues for on—shore wind turbines using data from a large wind farm, one of the largest in Europe.