Electrical Protection Failure - $4 million Vestas Wind Turbine V120 on Fire - SEL, Vestas, Ormazabal, ETAP blamed for the incident; Wait a minute!

A debacle in protection and power studies:

• A Vestas wind turbine V120 was damaged, an electrical fire started at Ormazabal switchgear (repair was not practical).
• $4 million equipment loss.
• $2.2 MW, $100 MWh PPA, 40% net capacity factor, $20 MWh Production Tax Credits (PTC). It leads to $1 million/year in production and PTC losses. $20 million production and PTC losses over the lifetime of the projected facility.

Preliminary investigations assigned possible blame to:

• SEL relays- relay misoperate, SEL CT failures, metering issue?
• Vestas- Short circuit data were not adequate, turbine integrity inadequate failure?
• Ormazabal Switchgear- Equipment integrity inadequate?
• ETAP Software- WTG model not accurate enough, inverter short circuit contribution modeling inadequate?

Not so fast. A headache is usually a symptom of a sickness, not the root cause. What is the root cause?

Figure 1 below is the feeder E relay event that captured the fault.

Figure 1: SEL relay event showing lack of electrical protection during the fire incident

• Orange time cursor shows current peak values during fault: IB = -7595 A, IC = 7561 A.
• ?Magenta time cursor shows feeder was energized pre-fault: IA = 113, IB = -195
• IA stayed very low during fault. It appears that we have a BC fault
• Voltages confirm the BC fault. VB and VC peak values significantly dropped, while VA peak value remained unchanged
• 52A remained asserted ( breaker never opened). The trip never asserted.

At the end of the event, the peak of the curves starts to reduce. There is a catastrophic reason.

Why did the SEL relay never assert a trip? Moreover, where was the backup SEL relay? Why didn't the backup relay assert a trip?

I came across this long ongoing investigation via their facility electrical engineer. I proposed simulating the fault in their existing (ETAP model) model and comparing simulation results vs. real life. Numbers don't lie! The facility engineer responded:

“I have never heard of that; how is it even feasible? You don’t have enough data to simulate similar fault events in the software; What about utility data? Loading conditions before the fault? Current contributions from other turbines? Moreover, We already reviewed several times our simulation model with a competent firm and found no major issues, etc.”

I responded: "I have all the data needed! It would take less than 1 week to simulate the exact fault on your existing simulation model. Moreover, if you give me 2 additional weeks and your drawings, I can show you my modified version of your ETAP model". They did not believe me and reluctantly agreed.

Step 1: Retrieve all pre-fault current conditions and talk with site operations.

Notice in Figures 2 and 3 below, I highlighted in grey square some essential information; all the info are pre-fault conditions.

• Main Transformer T1-351: Current is 420A; Feeder E relay 52-FE: 138 A.
• Look at the time in both relay events; incredibly accurate to the nanoseconds. Transformer T1 time is 3:12:05:697853041 am. Feeder E time is 3:12:05:697994443 a.m. If you are not familiar with SEL, the format for that time is Hour:Minutes:Seconds: Milliseconds.
• Figure 4 later in this article shows the one-line diagram.

Figure 2: Relay event at Transformer T1 during the fault

Figure 3: Relay event at Feeder E during the fault

Step 2: Configure simulation software to match loading conditions pre-fault.

Run power flow in the ETAP simulation to show that loading conditions match pre-fault data in the field.

Figure 4 below shows transformer and Feeder E ampacities. 420 A and 138 A match the same current ampacity shown in Figures 2 and 3

We now have a simulation model that matched identically actual conditions in the field, nanoseconds before the fault occurred. The wind farm was operating at 20% of nominal conditions before the fault occurred.

Figure 4: ETAP model showing loading matching pre-fault conditions

Step 3: Fault the turbine in the software to simulate the real-life incident.

Place a BC fault at the Ormazabal Switchgear inside V120 and observe the fault currents in Feeder E and the Transformer main breaker. The fault occurred 24930 ft (4.7 miles / 7.6 km) from the substation. The technical detailed fault analysis will be part of another article earlier next year.

Figure 5 on the right is their current existing model. with a BC fault simulating the incident that occurred, fault current main transformer = 8,000 A. Feeder E fault current = 8,700 A

Figure 5: ETAP model showing flawed fault current values

Figure 6 on the right is my modification. With a BC fault simulating the incident, fault current main transformer = 6,200A. Feeder E fault current = 6,700 A. I want to stress that most data entries were the same as their previous model.

Figure 6: Modified ETAP model showing matching current values

Let's compare the values above with the real-life values during the incident at both the transformer and feeder E.

Look at figure 7 and 8 below. I removed the sinewaves' instantaneous values and inserted magnitudes values of currents and voltages. Orange time cursors show current values during the electrical fault.

Figures 7 and 8 below state:

1. Transformer 1 current seen by the relay during the fault was 6,100A. Voltage was VB = 13 kV and VC = 10 kV. As indicated earlier in Figure 5, their simulation value indicates 8,000 A.
2. Feeder E relay seen by the relay during the fault was 6,400A. Their simulation value in Figure 5 indicates 8,700 A.
3. I showed them my software modification. 6100 A vs. 6200 A at the transformer; 6,500A vs. 6,700 A at Feeder E. We all agree it is a Match! There was a long silence in the room.

Figure 7: Relay at the transformer showing current and voltages values during the fault

Figure 8: Relay at Feeder E showing current and voltages values during the fault

- SEL metering readings were on point. Both SEL relays (primary and backup) did not misoperate.

- ETAP software integrity is not an issue. the software works ideally and is accurate

- Vestas data were correct

- Ormazabal Switchgear and Vestas turbine had no issue; they were victims of an inadequate protection scheme. primary and backup relays were not programmed adequately because power studies and simulations were inaccurate

It is the 21st facility where I witnessed the same issues. Every power system model/simulation I tested with actual incidents has been significantly incorrect. The inaccuracies have led to significant incidents. 15,000 incidents, 600 relays reviewed or reprogrammed, 20,000 arc flash labels redone, 3 fatalities, 2 cogen generator complete rewound, 10 known wind turbine catastrophic failures, wind farms yearly production losses, etc.

• Do you believe Elon Musk and his team would approve a Tesla car simulation SOLELY after verifying that the data input in their simulation software is accurate? NO
• Do you believe Tim Cook and his team would approve a new iPhone simulation SOLELY after checking the data accuracy in their simulation software? NO
• Do you believe Boeing releases a new flight software SOLELY after checking the integrity of the data in their simulation software? NO

Why are we one of the rare disciplines on this planet that believe that if data from electrical drawings are correctly input into software (ETAP, SKM, PSSE, Easy power, Aspen, etc.), our power system simulation must be accurate!

Let me be as loud as I could be: it is highly unfeasible to produce accurate power studies on high and medium complexity power systems SOLELY based on electrical drawings and data input in simulation software! Our current best practices are significantly flawed.

For the next 20 years, I intend to write 1,040 articles (weekly articles) on this Linkedin newsletter, and countless of those articles will show you how flawed practices are in performing power studies and verifying their accuracy.

Power studies errors are the primary reason that failures and anomalies occur in power systems; unfortunately, we have no idea because we don't even realize how significantly flawed our simulation models are. We can't even accurately prove the accuracy or inaccuracy because we SOLELY rely on theory.

"In theory, theory and practice are the same. In practice, they are not."

Albert Einstein.

"Garbage In, Garbage Out" in power systems simulation does not mean "Not Garbage In, Not Garbage Out."

Thierry Epassa