Wind power is now a significant cog in the UK’s electricity production machine. Wind turbines are a familiar sight from UK beaches, trains, and motorways, and have been key to decarbonising electricity production. The UK regularly reports record-breaking wind power outputs, with wind power producing a record-breaking 17.5GW of power on 13th February 2021.
Durham professor Dr Peter Matthews specialises in the monitoring and maintaining of turbines. I sit down with him to chat about his contributions, past and present, to keeping the lights on.
Matthews’ work involves analysing wind turbine sensor datasets to identify signs of part failures. Each turbine is fitted with sensors to measure everything from external wind speed, power output and orientation to internal generator temperatures and oil pressures. A typical turbine has between 300 and 600 sensors in total. Multiplied by the size of the wind farm (the offshore London Array has 175), that’s a lot of numbers to be crunched. So, Matthews and his colleagues use data mining – combining statistics and machine learning – to identify which sensors can best predict future failure.
But why do wind farm operators value such advanced warning of failure?
Matthews points out that offshore maintenance work requires planning as farms can be 10 to 14 hours out to sea. “You need to know you’re going to have clear weather by the time you get there to do the work”. If the weather window is too short then you may arrive only to find yourself in the middle of a storm.
This means repairs to the largest components like the generator and gearbox can often only be carried out in the summer. Winter failures leave turbines out of operation until the following year.
One early success story for Matthews was achieving a two-week prediction window for the failure of pitch motors for onshore turbines. These change the angle of individual turbine blades as wind conditions vary.
More recently, Matthews’ attention has turned to sudden failure of the power electronics converting the turbine output to the 50Hz frequency used in the National Grid. One cause identified is that the connection board between the converter and the sensor can fail, leading to the data drifting to indicate that the converter is cooler than it really is. Another current project looks at high-resolution wind data to investigate what wind conditions lead to convertor damage.
To end our conversation I asked Matthews if there were any misconceptions about wind power he wanted to debunk. He explained that when we drive past wind farms where a few turbines aren’t turning, often it’s because they are in self-maintenance mode rather – not because they’re broken. “People don’t realise quite how complex these pieces of kit are, and what they’re capable of in terms of looking after themselves.”
Illustration: Sophie Draper