Blade Soiling Detection

Blade soiling never trips an alarm. There is no fault code, no alert, and no threshold breached — the turbine reports itself as healthy while quietly giving up a fifth of its output at the wind speeds where it earns most of its money.

Because it degrades gradually and affects several machines at once, it also hides inside fleet averages. A monthly availability report will show 97% and tell you nothing is wrong.

Power curve scatter analysis against the OEM reference, machine by machine, binned by wind speed. What separates soiling from ordinary scatter is its shape: a consistent cluster of points sitting below the curve in the mid-wind band, on several turbines at once, persisting across weeks.

Sensor error moves one machine. Grid curtailment moves all of them together and shows in the event log. Soiling produces a cluster — and the cluster is the signature.

• Bin ten-minute data by wind speed and compare against the OEM curve
• Isolate systematic offset from ordinary scatter across rolling windows
• Cross-check against event logs to rule out curtailment and faults
• Quantify lost generation at the site's own tariff

A ranked list of affected turbines with the output gap quantified per machine, and a wash-cycle recommendation aligned to the site's seasonal dust pattern rather than a fixed calendar.

• 15–25% output gap identified at matched wind speeds
• ₹45–60L per turbine per year recoverable via blade wash cycles
• Detected entirely from existing SCADA — no site visit to make the first call
• Wash cadence set by measured soiling rate, not by calendar