How wind farm owners get value from using WinJi’s Anomaly Detection module
Increased availability and energy yield / reduced short- and long-term costs with WinJi True Power solution
Avoiding unexpected failures and scheduling maintenance windows with least downtimes are key drivers to maximize the energy production of every wind park. Understanding data generated by different components and deriving failure predictions with high confidence are key to reach this goal. WinJi’s anomaly detection module allows its customers to detect data anomalies that do not conform to the normal plant behavior. Owners and operators gain actionable insights to optimize interventions and thus realize the true power of their assets.
Anomaly detection enables timely insights and optimal management of wind farms
Early detection of anomalies is key, it can determine whether a component needs a quick fix or a more extensive treatment to avoid future failures. It can save a lot of time, reduces costs, allows for higher energy yield and results in increased revenues. WinJi’s True Power solution enables wind farm owners to initiate further diagnosis, reduce turbine downtimes and extend the operational lifetime of their assets.
WinJi’s Anomaly Detection algorithm detects operational abnormalities for each component of a wind turbine. Therefore, digital models are created that predict the normal behaviour of the system of interest based on its operating conditions. In this way, deviations between the predicted and actual behaviour of the system are identified. The algorithm can thus continuously detect significant deviations from the normal behaviour of the systems. A deviation from this model can potentially indicate anomalies in the corresponding component. The flagging of anomalies is not only based on the deviation to the predicted behaviour but also on the frequency of significant deviations and a statistical test to assess whether the deviation is increasing over time.
WinJi’s Anomaly Detection has been successfully identifying operational issues in numerous wind farms up to date, proving its effectiveness and importance. As an example, anomalies were detected in the gearbox of a turbine in a Nordex wind farm. Later it could be verified that the maintenance work that was carried out, was effective as the detected temperature anomalies disappeared afterwards.
Another example is presented in the figure below. It shows the deviation of two turbines stator temperature to their predictions over time. While the deviation values of Turbine 1 are constant over the entire time period, Turbine 17 shows unusual operating behaviour of the respective component. Excessive stator temperatures first started to evolve in April 2022, indicating maintenance need.
For another farm, which showed anomalies in different components over several years, an analysis of the probability that a fault causing downtime will actually appear was carried out (see figure).
The results showed that when an anomaly for the gearbox or the converter is detected, it is almost certain (>97%) that within a month (k = 30), the wind turbine will experience a fault and stoppage. For transformer anomalies the probability is lower but still above 75%. Interestingly the gear bearing temperature anomalies do rarely lead to a stoppage and never exceed 20% probability in this case.
The presented framework allows for manufacturer-agnostic early fault detection and provides early notifications for wind farm managers and owners. This enables optimised operation and generation of additional value from the asset while also avoiding unplanned expenses and downtimes.
Conclusion
Customers of WinJi benefit from the Anomaly Detection and recommendations module in the True Power solution. Applying intelligent algorithms on vast datasets allows accurate prediction of future behaviour in up to 97% of cases. Owners appreciate the efficient management of large portfolios of wind farms and increase the value creation of each asset.
Contact us for more information at sales@win-ji.com or directly book a discovery call here.
