All,
With all due respect to the many people who have been involved in the review and comment on this fire event over the past 1.5 years, there is more misinformation and misinterpretations about this particular fire than any we have had to date. It stems from a variety of information that was posted about this event including a good, but not complete, report from Pete Jackson, the inspecting electrical authority. BAKERSFIELD FIRE not Target Fire: The “Target Fire” I prefer we refer to as the “Bakersfield Fire” since it happened near Bakersfield. The fact that it happened on a Target store is unfortunate, but that retail establishment should not be referred to in subsequent postings, since it has a lot of PV and has been very supportive of the PV industry. SunPower, while allowing poor wire management for years, is also not to blame ultimately for this fire. THE PROBLEM: The problem with the Bakersfield Fire is endemic to all PV systems over about 30 kW in size that have been built in the last 5 years. Since 2005, large inverters have been employing ground fault protection equipment that lifts the grounded conductor in a fault. Prior to 2005, the Trace Technology 3-phase inverters sensed the grounded conductor to ground connection, shutting the inverter down if the current got above 10 amps. The 2008 NEC required that most PV systems have a GFP and that the fault path be interrupted, so subsequent designs started employing a ground fault fuse similar to small residential inverters. NOT GOOD: The presupposition of this ground fault protection (GFP) method is that it must be the first fault. If any fault exists prior to opening the ground fault fuse, the GFP system is ineffective and potentially worse than not doing anything at all. If a fault occurs on a string-level grounded conductor on a larger system (30 kW would do it), the ground fault current generated cannot exceed a few amps. This is due to the fact that there is very little voltage pushing the current. At best, we can probably run 3 amps through a fully bolted fault in the source circuit grounded conductor. This is problematic since the ground fault fuses on large PV inverters are set at 4 amps. PERFECT STORM: Therefore, since we can’t see the fault in the grounded conductor, we proceed with operation although the status of the system has gone from “safe” to “fire hazard”. If the next fault (and one fault always results in a second fault eventually) occurs in a large feeder circuit (100 or more amps), then the GFP fuse instantly opens removing the main ground connection on the array. However, this is exactly what you don’t want to do if you already have a fault in array. Now, instead of having a large equipment grounding conductor to carry the fault current, a 10AWG wire (or smaller) gets the entire return current as the short circuit now has to return at the point where the first undetected fault occurred. Shortly that conductor gets so hot that the insulation on the conductor catches on fire and then the fire is driven into whatever flammable material is next to the wire (PV module underside). THE BAKERSFIELD MISLEADING MESSAGES: Until the Bakersfield thermal event (fire), very few believed that this could happen. Many still believe this to be a “two-fault” event that we generally do not have to design for. I disagree. The likelihood that this will happen multiple times in the future is not only possible but very likely. Since the first fault is not visible to the GFP, it is a dangerous blind spot. The first fault could exist at installation if proper commissioning tests are not done (as pointed out by Matt). It is very easy to repeat experimentally. Unfortunately, the current messages coming from the Bakersfield Fire miss the point. The first common message is that we need to focus on expansion joints for raceways. This is a good message, and improper techniques by both installer and inspector probably resulted in the feeder fault that drove the fire, all the best expansion joints in the world will not solve the fire problem. Another common message is that disconnects would have solved the problem. While I helped develop the code requirement for the 2011 NEC requirement for disconnects in sight of combiner boxes, disconnects could not have stopped the fire. Disconnects would have only helped with cleanup and overhaul operations. THE REAL MESSAGE—CHANGE GFP METHODS: While expansion joints, disconnecting means, and proper wire management are all excellent messages from this fire, the real message to the codes and standards community is to change the way we design our GFP circuits. Good array design and installation practices can certainly reduce the possibility of faults, but it is impossible to eliminate these problems in a system that is intended to operate for 25 years, regardless of how much O&M is done. The only way to resolve this problem is with GFP devices that can detect faults in the source circuit grounded conductor, or better yet—don’t have a grounded conductor at all. Arc fault detectors are not the answer since most of these faults start as ground faults—just as at Bakersfield. The only way to get GFP fault detection below 1 amp for large PV systems is to UNGROUND or resistively ground the array circuit—just as they do it in Europe and Japan. Currently, European inverters can detect faults below 300 mA. That is an order of magnitude better than we can with our solidly grounded systems. It is also low enough to detect faults in just about all array types. RELATIVELY EASY FIX: Moving to ungrounded or resistively grounded arrays is relatively easy and the NEC covers the requirements in 690.35. We already have several transformerless designs from Power-One and SMA (new TL line in the U.S.), but there is nothing holding back larger inverters using transformers with ungrounded arrays. Existing arrays could be made much safer if they were to be retrofitted with a circuit at the inverter that would lift the grounded conductor each morning and test for ground faults on the grounded conductor prior to starting the inverter. While not quite as safe as “European-style” systems, it is a large improvement over where we are today. Don’t be surprised if things change about how we do GFPs soon. If I have anything to do with it, it is going to happen. The current hazardous situation, and bad information about the Bakersfield fire, cannot continue. Bill. Bill Brooks, PE Principal Brooks Engineering 873 Kells Circle Vacaville, CA 95688 707-332-0761 (office and mobile) 707-451-7739 (fax) b...@brooksolar.com (email) www.brooksolar.com (web)
_______________________________________________ List sponsored by Home Power magazine List Address: RE-wrenches@lists.re-wrenches.org Options & settings: http://lists.re-wrenches.org/options.cgi/re-wrenches-re-wrenches.org List-Archive: http://lists.re-wrenches.org/pipermail/re-wrenches-re-wrenches.org List rules & etiquette: www.re-wrenches.org/etiquette.htm Check out participant bios: www.members.re-wrenches.org
