The IEEE 1584 equations used for three phase arc flash hazard calculation studies make a distinction between solidly grounded and un-grounded / impedance grounded power systems.

A “solidly grounded� system is typically wye-grounded and an ungrounded system is often a delta system. Some systems use a grounding impedance such as a grounding resistor.

The type of grounding makes a difference in the results. The calculated incident energy will be slightly higher if the system is not solidly grounded.

This week’s question is about the type of grounding that is used:

Which type(s) of system grounding does your company/clients have?
(Select all that apply)

• Solidly Grounded
• Un-Grounded
• Resistance Grounded

I have all three. There are more examples of major differences between them:

1. Phase-to-ground faults are the most common type of fault. Although it is possible to detect and interrupt a phase-to-ground fault on a solidly grounded system, experience is that this is really hard to do without a significant arc occurring. Single phase-to-ground faults in ungrounded or high impedance grounded systems are effectively harmless. In a low impedance system, YMMV and some may have significant arc flash events. A 23 kV, 2000 A impedance grounded system is used here and the arc flash is significant, though obviously controlled. Numerically though these account for over 90% of ALL faults. The single best way to reduce the risk of arc flash by an order of magnitude is switching over to say a high resistance grounded system for between 400 and around 10 kV. From 10 kV to 35 kV it is still possible but due to resistor limitations best to go to a low impedance grounded system. Above 35 kV the majority of power systems are ungrounded.
2. Impedance and ungrounded systems increase the incident energy in a phase-to-phase condition by a slight amount, which is the price paid for a 10 fold decrease in likelihood, as mentioned.
3. Coordination for ground faults in an ungrounded system is nearly impossible to achieve. Most "fault locators" really just pulse a resistor in and out of the circuit (alternating between ungrounded and high resistance grounding).
4. The biggest advantage of solidly grounded systems is that at relatively low currents (say under 100 HP motors), ground fault tripping and coordination is almost "for free" because phase overcurrent protection usually also captures ground fault overcurrents as well.
5. The purported advantage of ungrounded systems (and even high impedance alarm-only systems) is that they can theoretically operate indefinitely in spite of a single fault in the system, although a second fault on another phase results in a very destructive double phase to ground fault. However, this is misleading in an ungrounded system. Ungrounded systems have significantly more phase-to-ground faults due to the fact that the system typically creates massive transients during arcing faults (600-800% of line-to-line voltage) from system capacitance acting like a spark gap voltage multiplier. These transients eventually overcome the weakest insulation in the system (typically motors) and result in significantly increased overall system failure rates and maintenance costs. I have never found a study on the subject but naval ship "below deck" motor failures (where ungrounded systems are required) are 4 times more likely than other applications.

Costs vary quite a bit. The cost of the resistor itself for a 5 kV system varies between $3K and $8K depending on the enclosure, and then down to under $1K for 480 V systems for high impedance grounds. Since resistor cost is based on absorbed power and P=I*I*R, doubling the resistance cuts the current in half and cuts the price roughly in half. The big issue is when it comes to breakers. Above 1000 V where relays are usually independent of the circuti breaker, typical multifunction relays have a separate CT input and the cost of the CT's are under $500, if a circuit breaker was in the design. Microprocessor controlled multifunction relays have reached a point where they are far more economical and reliable than any other relay alternative. A top of the line SEL 751A relay which is ideal for most industrial switchgear applications runs around $1K-$2.5K depending on options chosen, compared to around $1K for an equivalent single function GE or ABB relay. Although "smart fuses" such as an S&C "fault fiter" can theoretically accept an external trip input, I've never seen one available so at this time impedance grounding makes lower cost fused disconnect arrangements impractical. At 480 V, breakers with external trip inputs or ground fault tripping are considerably more rare than typical molded case thermal-magnetic breakers and the price is significantly higher, often adding as much as $1K to the cost. On the other hand electronic overload relays with ground fault tripping are almost equal in price to the cost of a eutectic overload relay with the thermal elements. Retrofitting existing ungrounded systems with the additional of either a zigzag (cheaper) or delta-wye auxiliary transformer (usually more expensive) adds another $3K to the cost. All prices are U.S. and based on recent experience.

A second potential disadvantage next to cost is training. Impedance grounded systems are more complicated than solidly grounded systems and require electricians to understand grounding in general. Impedance grounding systems turn what is essentially a passive safety system into a much more active system that requires a deeper understanding in order to operate it effectively. As impedance grounding systems are not the de facto "standard" in many work places, they are automatically viewed with suspicion and distrust as some sort of "new technology" that is perceived as less reliable than the de facto standard. Impedance grounding is only allowed by the NEC over 240 V as an alternative and only required for portable equipment operating over 1 kV. The almost lack of any rules whatsoever for impedance grounded systems in the NEC further confuses new installations.

Although the initial installation costs are significant for impedance grounding, the payback is that maintenance costs are significantly reduced. Since ground fault currents are typically less than operating currents, secondary damage that is typically seen in a solidly grounded system is nonexistant. Troubleshooting times are also substantially reduced because of coordination during ground faults. Finally, the improvement in personnel safety cannot be denied.

Impedance grounding is required in the U.S. and Canada for all mines (except surface metal/nonmetal in the U.S.). It is very popular among chemical plants in the U.S. and is required in a large number of countries. In spite of the reduced long term costs, the initial up front cost, though low, tends to continue to limit continued growth.

My personal preferences are these:
Under 250 V: Solidly grounded.
250 V and above: High resistance grounding on a delta-wye transformer. 15 A resistor under 1 kV, 25 A to 10 kV, and 400 A (low resistance) above that.
Relaying: Lowest level warns at 40% of resistance, and trips are in 0.5 seconds at 60% of current. Second tier distribution breakers in 1-2 seconds, then 2-4 seconds, and continue in 2 second increments through each additional level of breakers (if there are any). I prefer to use "residual Earth ground" aka "core balanced CT's" for tripping, but mounting a 120 V PT across the resistor itself which is set to trip at 60% (72 Volts) definite tripping in 4-6 seconds. This 59G trip function performs double duty. It is backup protection for the current-based relays. Second it provides a monitoring function so that in the event that the resistor fails open, it will provide a backup trip function.
Ground checking relays: Don't use unless required by regulation. These rely on the idea that the impedance of the wired ground is significantly less than an alternate fault or Earth ground. This is generally true at distances of under 500 feet but quickly becomes decidedly invalid at longer distances since the conductivity of Earth grounds increases linearly with distance while it decreases over wired grounds linearly with distance. In addiiton the assumption is that "electricity follows the path of least resistance" when this is an over simplification of Ohm's law which is that electricity follows all paths proportional to their conductivities.
Alarming vs. tripping: If plant procedures treat an alarm as the same as a fire alarm...that is, find the fault as quickly as practical and shutdown in an orderly fashion in order to fix the issue, with continuous monitoring to a security/control room location, then alarm-only can be a viable option. In practical experience the sense of urgency is not there. The alarm gets ignored and eventually a double-ground fault then forms resulting in significant and sometimes prolonged damage, and a possible arc flash.

A 15A resistor let-through current for systems under 1 kV seems fairly high. Many of the LV resistance grounding packages I'm aware of have taps available from 0.9A up to 10A or less, and many ship with the resistor connected to the 0.9A tap. I have had a few calls from folks who get elevated neutral to ground potentials during a ground fault. Each of them was aware that their resistor was connected to the 0.9A tap and they suspected that their capacitive charging current was higher than 0.9A. In each case the owner needed convincing that they should take an outage to change a resistor tap and that this would fix the issue, so they called me for a second opinion. Inevitably I never heard back to confirm which resistor tap they switched to in order to solve the problem. Generally the capacitive charging current was blamed on 6 pulse drives rather than cable capacitance.

I recall an IEEE paper a few years ago that presented a method to easily field measure the capacitive charging current in high resistance grounding applications. I've never had an occasion to try their method, and it may be too esoteric for most folks who prefer to just change the tap on the ground resistor and move on.

First, I inherited a lot of this. It mostly follows coal mining standards.

Charging currents around here tend to run around 1-3 A for "small" systems here, although it is higher with lots of shielded cabling at medium voltage and less at lower voltages. We have a lot of 1500-2500 kVA transformers and lots of motor loads, all of which contribute charging current. The rule of thumb is usually to have the resistor at least 3 times the charging current. If it gets any smaller than that, then the system tends to respond more and more like an ungrounded system. There are some simple heuristics for estimating this if it is not measureable. That puts the minimum size of the resistor at 9 A if it is necessary to pick a constant size for all applications. Since I have dozens of them, standardization is desirable. In some small systems at low voltage it might be possible to run with a 0.9 A resistor but this would be the minority and I'd be concerned about sensitvity (trip at 0.5 A??).

Furthermore, the CT design can be troublesome as well. There are some practitioners running with very small output currents (say 25:0.025) and very sensitive ground fault relays. Using standard relays necessitates aiming for either 1 A or 5 A, with 5 A being the industry standard. Most microprocessor relays balk at the idea of setting their trip points below around 2.5 A (tap = 0.5), so unless the system uses something other than a "standard" relay (Basler, ABB, GE, SEL, etc.) or CT, that limits the trip setting to around 2.5 A as a bare minimum, so if using 60% of the resistor size as a trip setting, this sets the resistor size at 4.2 A...rounded up to 5 A. Granted those low burden, very sensitive relays with the 25:0.025 ratios are really nice if you don't mind less commonality amount available CT's.

The other argument for small size is the physical size and heat dissipation of the resistor. Since wattage defines resistor size for the most part if voltage is fixed and P=I*I*R, doubling the current cuts the resistance in half but doubles the size of the resistor and heat dissipated. Still, my 15 A resistors are about 12" x 12" x 4" in size...hardly any space at all in typical switchgear.

On the upper end current carrying capacity of the conductors can become an issue. At 15 A wire size is not really a major factor given the relatively thin insulation on 600 V insulation but at 25 A or larger, it can certainly become a consideration. At medium voltage the minimum conductor sizes are already pretty large.

Overall I agree that what you would normally receive from one of the major vendors (Post Glover, Igard, etc.) would be around 2-8 A. And I agree that there's a certain stigma among recent converts to go with fairly high currents when it is not necessary. However I would definitely disagree with running with very low currents such as under 5 A for the reasons I've suggested already, and 10 A is probably a more comfortable minimum that would serve nearly any situation well for a system rated under 1000 V.

I currently work in an industrial plant with a solidly grounded system, but have previously also worked in a facility with a high resistance grounded system (as well as another facility that had a solidly grounded system, and I was involved in the design phase for a plant that was going to be built with a resistance grounded system.)

Paul, thanks for the background. I should clarify that I don't advocate the use of very low resistor taps either. The petrochemical plants in my area installed primarily Fran Fox style systems years ago when they converted from ungrounded to HRG systems. They generally settled on 5A let-through current, although these systems use a voltage relay across the resistor so CT sensitivity was not an issue.

In the last few years I've worked on a number of 480V MCCs that have motor overload relays capable of detecting ground faults on HRG systems. The particular relays I was working with used 1A CT secondaries and a wired residual connection was used for ground fault detection. With a 30/1 or 50/1 CT the relay had enough sensitivity to use the residual to detect the 5A ground fault, but the 100/1, 200/1, and 400/1 CTs used in larger starters required an additional zero sequence CT for GF detection. There were a number of folks that were skeptical of spending this much money on a fancy 480V motor overload, but the troubleshooting time to find most ground faults has dropped from hours to a few minutes.