Unlock Wind Turbine Potential with Machine Learning and Condition Monitoring

As offshore wind rapidly evolves, so does its complexity in terms of the sheer scale of wind farms and their distance from shore and sea depths. New technologies are keeping pace with the changes, and wind farm operators need to stay one step ahead to not only optimise wind revenue and extend asset integrity but also ensure that the focus remains on the health and safety of personnel.

Taking more control of their assets to shift the focus from ‘availability’ to ‘productivity’

Current business models tend to pass the responsibility of an asset’s wind generation capacity to OEMs and third parties. If wind developers were to shift their focus from project development, construction and energy trade and more towards raising the average annual capacity of wind turbine generators, which can be achieved through closer asset monitoring and control of maintenance activities, the possibilities for optimisation, and ultimately increased revenues, could be endless.

There is a major challenge in changing the industry mindset from ‘availability’ to ‘productivity’ by managing wind turbine generator services and minor breakdowns based on the wind resource and trade forecasting via the wind farm regional control rooms.

By establishing regional control centres closer to operational offshore wind farms, developers can more closely monitor O&M activities, allowing data capture and performance history to map trends and, in turn, enable the development of better maintenance strategies for optimal output.

Access to more detailed data will also allow a shift towards predictive and reliability-based maintenance strategies, leading to a reduction in operational expenditure, particularly at sites generating more than 500M. When we’re considering larger wind farms farther offshore with harsher weather conditions, the cost and risk of O&M activities rise, making the reliable uptime of wind turbine generators all the more critical.

If wind farm operators establish themselves as ‘complete asset managers’ instead of ‘remote customers’ and remove their complete reliance on OEM contracts, they will be able to directly influence and develop O&M strategies for large-scale wind farms to focus on maximum generation and increased reliability.

How operators can harness the power of machine learning and digitalisation to drive output

We live in a time when wind turbine generators are constantly being redeveloped and improved, but when we look specifically at O&M, there are still areas where their performance can be enhanced. Digitalisation and machine learning can be used to predict and control wind turbine generator downtime and provide an alternative approach to how wind farms are operated and maintained to drive output and ultimately optimise revenue.

Consider this. Wind turbine availability is a time-based ratio of the amount of time a wind turbine is ready to operate in a given time period divided by the total time in that period. A guarantee of ‘uptime’ is often agreed between an OEM and customer based on contractual availability, which uses a similar measure in which the turbine is not ready to operate. Compensation is paid to the customer if the contracted availability is not met. Typically contractual availability guarantees are 95% for offshore wind farms. However, what if the turbine is available 95% of the time but isn’t generating because of wind speeds outside the design operating range? This leaves a 5% window where 20% of the month’s revenue could be lost.

By using a mix of machine learning and condition monitoring to track wind turbine performance, it’s possible to minimise stoppages by scheduling wind turbine inspections and maintenance outside the design operating range.

First, you could calculate the theoretical maximum revenue based on the measured actual wind resource. Establish the wind speed in the operating range (>cut in, <cut out). Then establish a KPI for whether the wind turbine generator is “generating” or “not”. And finally, calculate the lost revenue for when the turbine is “not” generating.

Looking deeper into O&M activities, machine learning can also be used to create standardised reason codes to improve fault tracking and analysis. Examples could include identifying high-frequency and long-duration stops. Looking at performance, patterns, predictors, protagonists and promises.

All of this leads to increased revenue by one, minimising faults causing stops, and two minimising the response time to faults and planning maintenance and inspections at wind speeds below ‘cut in’.

Although rapidly developing, offshore wind as an energy source is still in its infancy. It was only 30 years ago when the first wind farm was commissioned and only 20 years ago when the first wind farm in the UK was installed, consisting of just two turbines.

In a developing industry, where so much attention is paid to scaling the many obstacles to establishing operational wind farms and ensuring their availability, attention now needs to shift to how these projects, which are a feat to modern engineering, generate the maximum amount of wind power during their lifecycle.

This article was originally published in Offshore Energies UK.