Understanding how to know if you can trust automation and AI applications on a wind farm
Wind is complicated and experience in the field is critical to incorporate into the use of modern tech tools. For instance, automation, AI and machine learning (ML) alone will not deliver value if they are not used and applied properly with the right data quality and domain expertise.
For example, others set out with the goal of applying AI, which can lead to many false positives due to a lack of understanding of the physics and engineering principles of the systems. Many AI and ML-based approaches can be taken to detect a cow from imagery, but without insight into the physical characteristics of a cow, mistakes will be made.
To extend the example above to a wind farm, consider Turbine Performance Optimization (TPO) on GE turbines. If you were unfamiliar with TPO and how the solution works by changing the parameters, it would be difficult to understand the minimum pitch changes over time and which of the situations below were correct operating conditions..
How do you know WindESCo applies it correctly?
WindESCo starts with a well-defined problem and clear intent about what we’re trying to detect, and only then does WindESCo build the tools and machinery necessary to get there, applying AI when it is advantageous, but not for its own sake, or just because we are able to. The first step is starting with engineering first principles, layering artificial intelligence to scale and operationalizing our subject matter experience. When an issue is first released, every issue is reviewed by an expert before it is released to the customers.
Our team uses multiple data streams, domain expertise, and feedback from the field (or site level) to continuously improve our algorithms and refine our fix database.