The Bakersfield Fire: A Lesson in Ground-Fault Protection
Inside this Article
On Sunday afternoon, April 5, 2009, smoke was seen rising from the roof of a big box store, home to a 383 kW PV array, in Bakersfield, California. The store manager quickly investigated, finding one row of eight modules on fire and a smaller fire some 200 feet away. Fire extinguisher in hand, the manager soon realized this was a job for the fire department. A 911 call was placed at 4:15 pm and first responders were on-site 5 minutes later.
The subsequent investigator’s report, which is named after the retail store, is the most widely read incident report related to PV systems. The fact that this retail establishment, which has been very supportive of the PV industry, inadvertently lent its name to a two-alarm fire is both unfortunate and unwarranted. For this reason, I refer to this incident as the Bakersfield Fire. Similarly, the product manufacturer and installer, while not without fault, are also not ultimately to blame for this fire. Therefore, in the analysis that follows certain manufacturer and installer-specific details particular to the PV system in Bakersfield have intentionally been changed. The generic circuit diagrams used here represent the majority of PV systems deployed in North America.
It is important not to get lost in the details of this specific installation. Instead, I want to emphasize an underlying problem, one that is endemic to all grid-connected PV systems larger than 30 kW that have been built in the past 5 years. The “thermal event” that occurred on April 5, 2009, is clearly cause for alarm. More alarming, however, is the fact that it could happen again.
THE INVESTIGATOR'S REPORT
The investigator’s report on the Bakersfield Fire is quite good, even if it does not tell the whole story. It is available on numerous websites, most notably the National Fire Protection Agency website (see Resources). The author of the report is Pete Jackson, an electrical specialist for Kern County, California, and the chief electrical inspector for the City of Bakersfield. Both the Kern County and the Bakersfield Fire Departments responded to the fire.
I had the pleasure of meeting Mr. Jackson. He was particularly familiar with this installation, since he was the person who performed the project plan review. His report on the roof fire provides a reasonable outline of the events that transpired and the fire department’s response to those events. It includes two requirements and three recommendations intended to improve the safety of the Bakersfield PV installation and other similar installations.
In summary, the following corrective items were required:
1. Perform high-voltage insulation testing on all PV array conductors.
2. Use expansion joints in long conduit runs while ensuring that these are properly installed.
The actions recommended in the report include:
1. Check the ampacity of all conductors to see that they comply with the temperature requirements of Table 310.15(B)(2)(c) found in the 2008 National Electrical Code.
2. Install disconnect switches at or near the combiner boxes so that it is possible to deenergize power in the large feeders that run from combiner boxes to the inverter.
3. Reconfigure the combiner boxes and feeder conductors for a maximum of 100 amps per PV output circuit, so that the feeder fuses will be more likely to open under fault conditions.
While I wholeheartedly support the two corrective actions and the first two recommendations, they do not convey the whole story, nor will these measures prevent a repeat of the Bakersfield Fire. While the third recommended action would help reduce the fire hazard associated with a similar event, this would likely not have prevented the Bakersfield Fire because it does not address the fundamental problem.
GROUND-FAULT PROTECTION BLIND SPOT
The problem exposed by the Bakersfield Fire is that large inverters manufactured since 2005 employ ground-fault equipment that lifts the grounded conductor in the event of a ground fault. In practice, this is fine as long as it eliminates the only return path for the ground-fault currents. However, if a return path exists in the source-circuit conductors, a 30 kW array is capable of delivering approximately 100 amps of fault current, which is enough to burn a 12 AWG conductor.