Introduction - Overcurrent protection (50/51) is the golden standard of device protection. It's used everywhere, mirroring the behavior of molded case breakers and fuses. Sometimes, though, this type of protection can't quite get the job done. This is where differential protection comes into play.
Overcurrent protection curves suffer from a couple of problems.
First, 50/51 curves often need to be "coordinated" to avoid nuisance tripping off more of the system than is necessary. This coordination can add significant time delays upstream in the distribution network. These time delays can lead to unacceptable time durations for fault clearing, whether because it leads to arc flash hazards or infringes on equipment damage curves.
Second, Time overcurrent protection is inherently unable to tell where the fault is occurring. This means a sufficiently large current can send a signal to all the system relays that they need to trip (or at least start their countdown to tripping). Because of this, 51 protection relies on a time delay to only trip if the condition persists. This could lead to unnecessary damage to the system. Even thought the official damage curve may not be impacted, that doesn't mean that a fault piece of equipment won't still experience damage (since it's operating out of its normal conditions). The faster we could trip that off, the better.
Differential protection solves both of these problems by approaching the fundamental question of protection differently:
Instead of asking, "Is this current too large?", differential protection asks "Does the current in match the current out?"
Figure 1: Differential Protection in a Nutshell. Faults inside of a Protected Area are Detected by Noticing that the Input and Output Currents are NOT the Same
When everything is going normally, the current into the protected area (e.g. a transmission line, switchgear lineup, transformer) should match the current leaving that same area (after accounting for any turns ratio corrections in the case of a transformer). Even when there's a fault outside of the protected region, everything going in is equal to everything going out. However, when there's a fault in the monitored area, the current going in will NOT be the same as the current going out.
Figure 2: The Zone of Protection for a Bus Protected by Differential
This change in protection philosophy allows the user to trip on a fault condition much faster and with greater selectivity, only taking out the effective area with potentially an instantaneous trip setting.
Implementing Differential - Practical implementations of differential protection are a bit more complicated that simple overcurrent. Differential requires the use of multiple current transformers (CT), potentially with substantially different ratings. These CTs define a "zone of protection" where a fault can be detected. Bigger zones trip off larger areas, so they're safer but less selective.
In order to perform an assessment of whether "current in = current out", we need to normalize the values we obtain from each CT. At a high level, this just means accounting for the CT ratio and the transformer turns ratio (if differential is being used for a bus, this doesn't apply).
For example, a transformer may have a turns ratio of 10, so 100A on the primary corresponds to 1000A on the secondary. We would only trip if there was a difference in normalized versions of these values. During regular operation, we would expect to see a difference of 900A here!
87 Protection Pickup - So now we know how to assess the in/out currents and we know that an imbalance in these inputs equates to a trip, but what is the pickup? How much imbalance is acceptable?
First, we need to define exactly what we are measuring to pick up. Differential protection is about measuring the difference between inputs and outputs. Mathematically
Id = Iin - Iout / (Iin + Iout)
Where:
Id is the difference current, the value the relay operates on when high enough, measured in %
Iin is the input current to the protected area,"
Iout is the output current from the protected area,
At first glance, we may think that any difference in input to output current should cause a trip. However, this just isn't a realistic design. Unfortunately, current transformers and relays have limitations on accuracies. Standard C-class CT's are only accurate to +\-3% at rated current. Relays may have an error of ~1% as well. This means, even if there are no other contributing sources of error, a relay could see an error of:
3% * 2 + 1% * 2 = 8%
Setting below a difference current of 8% could lead to nuisance trips in this case. Where differential protection is used for transformer protection, there are additional sources of error to consider: tap changes, excitation, etc.
Types of Differential Protection Relays - A detailed discussion is beyond the scope of this article, but it's important to understand that differential protection relays generally come in two forms: High Impedance and Low Impedance. High impedance differential is generally used for bus protection while low impedance differential is normally used for transformer protection.
Do I Need Differential Protection? - The question of whether or not differential protection is necessary will likely be driven by a project's contract. The National Electrical Code does not require the use of differential protection. NFPA 70E only requires mitigating measures as required by a risk assessment plan, which leaves a ton of room for engineering judgment.
Contracts may require differential protection for buses, like switchgear or motor control centers, to support a lower arc flash hazard level (by tripping faster).
Even more typical is the use of differential protection around large transformers. Big transformers are high cost items that can be excessively damaged by prolonged internal faults. Time overcurrent may NEVER detect internal faults of low magnitude. Only when those faults develop into a serious hazard will the time overcurrent finally pick up.
If contractual requirements are not in place, then nothing explicitly requires 87 protection. A system can be designed for compliance with the NEC purely with time overcurrent. However, there may be loss of selectivity in some cases (tripping off more if the system than intended during a fault) and there may be excessive arc flash hazard levels (that prevent energized work or require very heavy-duty PPE).