By Jay Prigmore II, Ph.D., P.E. Last reviewed · Prepared and reviewed by a licensed Professional Engineer
Investigation of wind turbine electrical failures and nacelle fires — up-tower inspection, nacelle-down examination, and reconstruction from SCADA and protection records.
By Jay Prigmore II, Ph.D., P.E. Last reviewed · Prepared and reviewed by a licensed Professional Engineer
Wind turbine failures are among the harder forensic problems in power engineering, for reasons that have little to do with the electrical engineering. The evidence sits eighty to a hundred and forty metres in the air. If there has been a fire, much of it has fallen to the ground or burned in place. Access requires climb certification and coordination with the operator, and the decision to lower a nacelle is expensive enough that it is often made before anyone has decided what the investigation actually needs.
What survives, reliably, is data — SCADA, converter and controller logs, protective relay event records, and the substation and collector system record. Those are frequently sufficient to establish the failure sequence even when the physical evidence is badly compromised, which makes preserving them the single highest-value early action.
Inspection scope should follow from the evidence question, not the other way around. Depending on the matter, that has meant:
Where multiple parties are involved — owner, OEM, operator, insurers — inspection protocols and evidence handling should be agreed before anyone goes up. Disputes about how evidence was collected are avoidable and consume more time than the inspection itself.
A nacelle burns and the debris lands over a wide radius. Traditional fire scene methodology assumes a scene that can be walked; here the origin was 90 metres up and much of it fell in pieces.
The compensating evidence is that a modern turbine records itself. SCADA holds power output, rotor and generator speed, pitch and yaw position, converter status, gearbox and generator bearing temperatures, and the alarm and fault queue — usually at ten-minute averages, sometimes at one-second resolution for a window around a trip. The alarm sequence in the minutes before the fire will normally point to a subsystem: a converter fault, a generator overtemperature, a brake or yaw event, or an electrical fault on the medium-voltage side.
That narrows the physical search. Converter, transformer, generator terminals, brake assembly, and the hydraulic system each leave different residues and different damage patterns, and debris recovered from the ground can be sorted by subsystem and examined for arcing versus fire attack in the same way a switchgear lineup would be. The turbine’s own record and the physical evidence then have to agree, and where they do not, the disagreement is itself the finding.
Where the collector system or tower-base switchgear is in scope, the IEEE 1584 arc flash calculator at PowerEngCalc supports the incident energy reconstruction, and the DC arc flash calculator covers converter DC-link faults, which the AC methodology does not address. For turbine and collector protection questions the analysis links directly to the rotating machinery and transformer practice areas.
Questions that come up before a retention decision, answered by the engineer who would do the work.
The recurring electrical causes are converter and power electronics failure, generator winding and terminal faults, capacitor bank failure, connection overheating at high-current terminations, and lightning damage that was not cleared by the protection. Non-electrical contributors — brake overheating, hydraulic fluid release onto hot surfaces, and gearbox failure — often combine with an electrical ignition source, which is why a defensible investigation examines both. Establishing which subsystem the fire began in is normally the whole dispute, and the SCADA alarm sequence is usually where it starts to resolve.
By treating the turbine's own data as the primary scene record and the debris field as the secondary one. SCADA, converter logs, pitch system logs, and the alarm queue reconstruct the machine's state in the minutes before the event. Debris is mapped by recovery position and sorted by subsystem, then examined for arcing damage versus fire attack exactly as a ground-level scene would be. Up-tower inspection is used where the structure is safe; where it is not, nacelle-down inspection after controlled removal is preferable to no examination, provided a written protocol governs the removal.
It is asserted far more often than it is established. Lightning detection network data will confirm whether a strike occurred near the site in the relevant window, and the turbine's own lightning current sensors, where fitted, may record a strike to the blade receptors. What decides the question is whether the damage path is consistent with the lightning protection concept under IEC 61400-24: a properly zoned and bonded system routes the current from the blade receptor through the down-conductor and out through the foundation earth. Damage inconsistent with that path points to a bonding, grounding, or surge protection deficiency rather than to the strike itself — and that shifts responsibility.
Not by itself, but fleet history frequently reframes the case. Where the same component fails across multiple units at similar operating hours, the argument shifts from operation and maintenance toward design, manufacturing batch, or the adequacy of the manufacturer's qualification testing. Obtaining that history is often a discovery question rather than an engineering one, but the engineering can specify precisely what to ask for: failure records by serial number, component revision history, service bulletins, and the qualification test reports for the component in question.
Access is the binding constraint. Up-tower inspection requires the turbine to be locked out, weather within limits, and inspectors qualified for tower climbing and rescue — which typically means scheduling weeks out rather than days, and joint inspections harder still. Nacelle-down inspection requires a crane and a written removal protocol so no party can argue the evidence was altered during recovery. Because scheduling is slow, the data preservation steps must not wait for it: SCADA, converter logs, and relay records should be secured immediately, independent of when anyone gets to look at the machine.
SCADA and relay records usually survive when the nacelle does not. They are also the first thing to age out of retention.