Notes from modeling operating fleets.

Technical writing from the team building the models: how a per-turbine expectation is constructed, why a manufacturer's power curve is the wrong baseline for an ageing machine, and what a predictive alert has to contain before an operations team will act on it.

Latest·Engineering·Jul 24, 2026

The average that hides the loss

A site reporting a healthy capacity factor can be carrying three turbines that are each four percent down. The arithmetic explains why nobody notices.

Read article

Notes from modeling operating fleets.

The writing here comes out of the work: what it takes to distinguish a genuine underperformance from a period of low wind resource, why a manufacturer's power curve is the wrong baseline for a machine in its eleventh year of operation, and what a predictive alert has to contain before an operations team will act on it.

We would rather publish one piece containing a measurement than four containing a thesis. For the vocabulary rather than the argument, the glossary defines the terms these pieces rely on.

The recurring subjects are the ones that determine whether a finding can be trusted: how an expectation is constructed, how a loss is attributed to a cause, and what separates a model that flags everything from one an operations team can plan around.