As a person in charge of ensuring the wind turbines in your portfolio are generating the maximum amount of energy and that they are maintained properly to minimize unplanned outages, more than once you’ve probably asked yourself “How on earth can this be done for all of my turbines all the time?”.
Well, rest assured you’re not alone. Many organizations find themselves in a similar conundrum and this undesired situation stems from one or more of the reasons listed below:
Experienced folks know that each turbine behaves differently even if they are of the same make, model and age. Hence, in order to do a thorough job, each turbine (and its components) needs to be tracked and analyzed individually. But everyone’s busy these days and performance management, O&M and asset management personnel are no exception. Companies with a small portfolio of turbines have small teams, where each team member wears multiple hats. On the other hand companies with large teams where people can be dedicated to a particular responsibility tend to have a large portfolio of turbines. In either case, it is extremely difficult to make the time to care for each turbine individually.
Lack of the appropriate skill set combination
Ensuring that each turbine is performing at its best and that potential unplanned downtimes are prevented by a proactive approach is not an easy task. It requires significant expertise, not only in terms of engineering and of running effective O&M practices but also in terms of extracting, handling and analyzing data for generating insights. With so much data generated by these turbines (new generation turbines produce 70TB of data each year - that’s enough data to fill an iPhone 14 every other day) making the most out of this data requires skills that are hard to come by in the industry.
The need for consistency
Experience shows that any turbine can fail or have a performance drop at any time for a variety of reasons. Our research shows about 1 out of 4 turbines experience some sort of performance drop at some point during the course of the year. Therefore each turbine and its components must be tracked and analyzed pretty much all the time in a continuous manner to make sure that nothing is “missed”. Doing this manually is an inefficient use of resources, if not straight out impossible.
What can we do about it?
With so many obstacles in the way, it’s easy to give up and simply say “we’ll do the best we can”. But, considering how dire the needs of the planet are for more green energy, as an industry we have an obligation to maximize the energy generated by installed turbines. So, how can these challenges be overcome?
We believe the answer lies in A3I. “What is A3I?” you’re probably asking yourself. Well, it comes from the manner in which intelligence is developed and delivered to the end user. Namely, it refers to intelligence that is;
Despite its over- and often incorrect use, artificial intelligence and in this case its subfield of machine learning is critical for being able to learn the behavior of each individual turbine in a scalable manner. For example, thanks to artificial intelligence the normal behavior of main bearings for each turbine can be “learned” taking into account about 20 different parameters such as wind speed, rotor rpm, ambient temperature, pressure, etc. If the measured temperature of the main bearing falls outside of the “normal behavior” range then a warning can be flagged. Without artificial intelligence we would have to rely on generic “straight line” thresholds that do not take into consideration the uniqueness of each turbine, components and conditions.
When analyzing turbines for performance drops or reliability issues, there are a large number of parameters to consider, many time horizons to investigate and a myriad of analyses to run. For example, build up of ice on the blades impacting the aerodynamics of the blades could be the reason for a drop in the power curve for a few days. If the power curve is analyzed only once a month and not combined with temperature data for each data timestamp, then more than likely the performance drop will be not captured. Even if it is captured, the reasons will not be identifiable, making it impossible to take remedial action. Only with automation can this particular analysis be run on a daily basis across the entire portfolio.
In this day and age, how well the turbines are operated and maintained should not be bounded by the capabilities of the team. The ideal system should bring with it intellectual knowledge that the team may not have, in essence augmenting the capabilities of the team. For example, a team may not think that a loss in production could be due to the more than expected occurrences of rewinding of the yaw chains. Equipped with experience from other farms the ideal system should include in its capabilities to check and calculate the loss due to rewinding of yaw chains and alert should it fall outside of the expected ranges. Only with augmentation of the skills of individual teams can we democratize the access to knowledge and capability across the industry.
As Kavaken, we have embraced these three pillars and built it into the core of the platform. Our second to none artificial intelligence expertise is automated to be applied to all turbines continuously and the breadth of analyses is constantly expanding with new experience from different plants to augment the expertise of the user in a way they could have never imagined.