Contributed by Won Shin, Vice President of Products, ONYX Insight
Every megawatt-hour counts. To the wind industry, to the energy transition, and perhaps most of all, to the sustainability of the planet. Therefore, every megawatt-hour lost is a problem. Downtime, maintenance, deratings, and other issues all result in lost energy and the industry has been trying to reduce these losses for many years. Curbing the amount of asset energy loss is critical to the effectiveness of wind to the energy transition – yet so far there hasn’t been a comprehensive solution.
Why is so much energy being lost, and what can be done to minimize it? It is often multiple small issues that contribute to an accumulation of lost energy, typically hidden within the data coming off a turbine. Automated data analysis, grounded in understanding the common causes most culpable for lost energy, could be the solution. By analyzing the raft of data produced by turbines and combining that with root cause analysis, it has become possible to predict when these common lost energy events might occur and notify operators before it starts costing them time and money.
Read more: What a year for wind
Lost energy: the big problem yet to be solved – and how advanced analytics can reveal hidden lost energy causes
Performance analytics were introduced initially to help identify the root causes of lost energy. Operators typically receive a monthly performance report which shows availability, total power production, downtime, and alarm analysis. But this is high-level, retrospective insight which only gives a broad picture of what is happening at a wind site.
With an abundance of time and a team of experts, operators can use performance analytics to identify the root cause of lost energy, and then take preventative action against it. But this resource-intensive process is seldom one that operators can afford to employ. It means that while site engineers focus on the critical issues, the smaller issues mount up, causing lost energy. Knowing the key issues to look out for is an essential first step.
The top 10 energy loss issues
With years of engineering skill, and a monitoring portfolio of over 7,000 wind turbines, Onyx Insight believes that 80% of lost energy is caused by just 10 common issues. These include:
Hydraulic system issues
Cooling system issues
Bad control parameters
Leading edge erosion to blades
These issues are encountered by operators encounter every day – and they’re costly. In some cases, just one issue can result in the loss of tens of thousands of dollars in revenue, as a case study example further on in this article reveals. There are many other issues that cause lost energy, but these are smaller contributors
Automation by combining performance and reliability models
Performance analytics will show how many megawatt-hours are being lost, but not why. That is where combining it with the component-specific reliability ML models – which is the model that shows the health condition of the specific component – allow the identification of the root cause of faults. The combination of performance analytics with a component-specific reliability model draws out actionable insights.
The next stage is to automate this. Analytics need to be actionable and automated in order to be sustainable. Smart software can use predictive analytics of the data collected by those two models, to automatically detect when faults are likely to occur, and alert owners and operators to this using real-time notifications.
This process becomes possible through understanding the top 10 causes of lost energy.
This innovative and revolutionary approach enables predictive maintenance to be carried out on common challenges, freeing up resources to tackle the less frequent, more critical O&M issues, while considerably reducing in-house O&M costs, time spent on understanding the data, and of course lost energy.
The approach in practice
This integration of two models – which means that software is effectively monitoring turbines and flagging escalations in challenges before they become lost energy events – has worked successfully in a number of instances, saving global renewable operators significant time and money.
One example involved 280MWh of lost energy caused by a hydraulic oil temperature error, which was responsible for more than 80% of the lost energy in just one turbine. This one fault meant that the owner of the turbine lost out on around $10,000 of revenue in four months.
The lost energy model, as referred to above, identified that this temperature error was responsible for much of the turbine’s lost energy, but this didn’t indicate what action should be taken to fix the problem. For this, a cooler reliability model was required, which honed in on the hydraulic oil SCADA temperature – where there were fluctuations in temperature – revealing that there was an intermittent blockage in the cooling system. The resulting recommendation was that the cooling fins be inspected and cleaned.
The next step was to build a model that would automatically predict this issue. That was done through training the cooler health ML model using the hydraulic oil temperature data and related SCADA tags. The modeling process also utilized engineering knowledge such as cooler control logic.
This automation is vital to predicting issues. In this case, the software provides early warning, and the operator can plan the maintenance in advance. Hence, a reactive problem becomes a predictable problem. As a result, maintenance can be planned in advance and lost energy can be saved.
When every megawatt-hour counts, predicting when energy might be lost and acting to prevent it, is hugely valuable for not only the return on investment of wind projects, but for the wind industry’s role in the energy transition and, of course, for the greater good the planet’s future.
About the author:
Won Shin, Vice President of Products, ONYX InSight
Won Shin head’s up the Machine Learning and analytic products for ONYX Insight, a leading predictive maintenance solution provider for the Wind industry. Won has 15 years of experience working in the automotive and renewable energy industry focusing on data science, machine learning, optimization.