One Digital Twin per Asset: Catching What a Monthly Average Hides

Monthly averages hide the devices that are actually losing energy. Asset Performance puts one digital twin on each asset — expected power learned from that device's own history against the variable that drives it — then tells you every day whether it is behaving, and if not, why.

What the model produces, per device, per interval

• Expected power - what this device should be producing given irradiance, hour and season, learned from its own history, not a datasheet.
• Normal - inside its learned band. No action.
• Underperforming - below its own lower band: soiling, shading, a fault or a bad sensor.
• Underperforming vs neighbours - below the band and behind similar devices, so a real fault is not blamed on weather.
• Curtailment flagged - power abnormally flat during daylight: grid limiting, not equipment.

One model, any asset with a driving variable and a rating

The same logic runs on wind turbines, solar inverters, strings, trackers, turbine components and weather stations. What does not match is classified as curtailment, underperformance or noise — never a flat threshold. Typical model error is under 2% of rated power, validated per engagement.

The output is a work list

Crews stop chasing losses they cannot fix when curtailment is told apart from a fault. Each device is judged against its neighbours as well as its history, so a cloudy afternoon does not raise forty alarms. The list is ranked by lost energy — the difference between a dashboard and a maintenance decision.

Detection and forecast talk to each other

Catching a fault is only half the value. A confirmed cause can feed a live forecast so the next market gate sees the gap closed automatically; a genuinely unexplained case alerts the person who owns that asset instead.

Figures on this page are typical of the deployed model, validated per engagement.

Ready to talk?

Want a digital twin on each of your assets? Let's talk - schedule a meeting with our team.