Transformer Differential Protection

Author: David Costello, Schweitzer Engineering Laboratories, Inc., USA

In many ways, microprocessor-based transformer differential relays have simplified installations. They perform internal computations for phase-angle compensation and zero-sequence filtering. One can use these relays with lower burden Y-connected CTs regardless of transformer winding connections and can accommodate virtually any transformer application. Built-in metering displays secondary current magnitude and phase angle, operate and restraint quantities useful in commissioning.

Despite advanced technology, a multitude of potential mistakes and problems conspire to make the installation of differential relays challenging. There can be wiring errors such as rolled phase wires, reversed polarity, and incorrect CT taps. Phase-angle compensation and zero-sequence filtering settings errors continue to be widespread. System phase rotation, phase-to-bushing connections, CT secondary wiring, and mobile transformer arrangements can all complicate installations. Installers must also overcome challenges from power system events that include inrush, overexcitation, external faults with CT saturation, and operating conditions that bypass CTs. A number of proposed solutions attempt to reduce the complexity of transformer differential relay installation and commissioning. Primary injection tests attempt to find problems before transformer energization. Commissioning worksheets and process checklists help engineers and technicians step through in-service metering checks and troubleshooting methodically. Software tools graphically display currents, operate and restraint quantities, and harmonics. Relays also offer the promise of automated or relay -  assisted commissioning.

Commissioning tests are intended to find wiring and settings mistakes before protection schemes go into service. However, mistakes result from improper or incomplete tests, the lack of sufficient load current during in-service meter readings, misinterpretation of in-service readings, or the lack of complete testing altogether. Relay-assisted commissioning may not be able to uncover more than single-contingency installation errors. Unfortunately, multiple-contingency errors are not uncommon.

Problems developing over time include CT wiring short circuits and loose connections. Event report analysis provides the best way to identify these problems.

Proper application requires careful settings development and thorough commissioning. Technology has not yet replaced the invaluable contribution of a knowledgeable and experienced engineer and technician.

Recommended Event Report Analysis Procedure

During commissioning tests, you can trigger relay event reports to document the procedure. After a fault or field event, event reports become invaluable tools for finding root cause and solving problems. The following process is handy when analyzing event report data.

Step 1: To understand what is expected to happen for given conditions you must look at settings, installation drawings, reference texts, and instruction manuals

Step 2:  Collect all relevant information, including eyewitness testimony, information about the fault, sequence-of-events records, trip targets, and relay event data

Step 3:  Gather available analysis tools, such as instruction manuals, reference texts, and event analysis software

Step 4:  Compare the actual operation to expectations. If there are any differences, resolve them by determining root cause. Do not waste time analyzing unused elements or settings. Focus, instead, on trip logic and output contact programming. Do not forget to look at prefault information, and use such data to perform an “offline” commissioning test to prove that system installation is correct. Before and during the analysis process, save data intelligently, naming files in a coherent way

Step 5:  Document your findings, proposed solutions, and test results. When you have validated a correct operation, or developed a proven solution for an incorrect operation, you are done

Direct CT Polarity

As load increases during first-time energization of a new transformer, the transformer differential relay trips. The low current magnitudes and balanced conditions, shown in Figure 1, confirm that there was no system fault. The power transformer is 69 kV delta to 12 kV grounded-wye, with ABC system phase rotation and ABC phase-to-bushing connections. We would expect the high side to lead the low side by 30 degrees for this transformer, and the phase-angle compensation to be W1CTC = 0 or 12 and W2CTC = 1. Figure 2 shows the current phase angles at cycle two of the event data. The high-side does indeed lead the low side by 30 degrees, but we expect this only on the primary of the CTs.

Remember that we would expect the secondary current from the low side CT to be 180 degrees out of phase for a through-load condition. Therefore, the relay should observe low side currents leading high side currents by 150 degrees. We can use the operate currents from differential event reports to confirm that this is a problem involving all three phases. Upon inspection of the low-side CTs, we found that these CTs had been connected with reverse polarity.

Lessons Learned Use relay event report data to perform an “offline” commissioning test and identify CT polarity mistakes. Commissioning tests discussed previously should identify this problem before the transformer goes into service. One of the most common mistakes is to confuse expected phase angles on the primary versus the secondary side of a CT. Remember that the 30-degree high side lead or lag relates to primary currents. For through-load or fault current, one of the CT secondaries will see current that is reverse polarity (current leaving the zone). For a transformer in which the high side current leads the low-side current by 30 degrees, the secondary low side current will lead the high-side current by 150 degrees.

CT Saturation

A distribution power transformer is connected high-side delta and low-side grounded-wye. Note that Winding 1 of a relay generally connects to the high-voltage side of a power transformer. This is an arbitrary association, however. In this application, Winding 2 of the transformer connects to the high-voltage delta side, and Winding 1 connects to the low-voltage wye side. The transformer delta is connected polarity of H1-to-nonpolarity of H3. The system phase rotation is ABC.

The phase-to-bushing connections are A-H1, B-H2, C-H3. We expect the low-side phase currents to lead the high-side phase currents by 30 degrees. Prefault phasors confirm that this relationship is as we expect. The differential relay tripped for a distribution feeder phase-to-phase fault that was external to the differential zone of protection. The low-side fault was B-to-C phase. We would therefore expect IAW2 and IBW2 to be about equal on the high side (Winding 2) and in phase with ICW1 on the low side. The magnitude of ICW2 on the high-side should equal the sum of IAW2 and IBW2, and the ICW2 phase angle should be opposite those two currents and in phase with IBW1 on the low side.

In Figure 3, these observed relationships match expectations. Figure 3 shows four unexpected items. First, the unrestrained 87U element operated for an external fault. Second, second-harmonic blocking is asserted throughout the fault. Third, there is significant residual 3I0 current present on the low side (during a phase-to-phase fault) and on the high-side (a delta-connected winding). Fourth, the residual current magnitude decreases over time, even while the fault persists. In Figure 4 there is significant operate current in all differential elements, but this operate current decreases over time even while the phase fault current persists.

Filtered 60 Hz phase currents remaining constant throughout a fault, while residual and operate currents decrease, indicate that the CTs experienced saturation. Harmonics are present when a CT saturates, which explains the second-harmonic blocking element asserting during the fault. This blocking element assertion prevented the 87R restrained-differential element from operating in this event. Note that the 51N1T high-side ground element tripped for a subsequent fault. Also, when a CT saturates, the replicated secondary current is not a true ratio current, diminished in magnitude and shifted in angle.

Each CT phase behaves differently, resulting in calculated residual current (3I0 = IA + IB + IC). Over time, as the CT recovers, the false residual current subsides.

We suspect that saturating CTs are the root cause of the unexpected and diminishing zero-sequence and operating current. The only way to prove this theory, is to view the raw or unfiltered event data. Figure 8 shows the raw event data, and we can see that the CTs, especially on the high side, saturated significantly. The low side had C400, 600:5 multiratio CTs, tapped at 100:5. Remember that the accuracy class here is for the full winding of the CT, so this CT is derated to approximately C67 performance (if such a CT existed). The CTs either need to be improved, or the tap must increase to access full CT capabilities.

Lastly, we found the 87U element to be too sensitive. A normal setting for this element would be in the range of 8 to 10 per unit of tap. In this case, the element had been set to 3 per unit. We therefore increased the 87U setting.

Lessons Learned Decreasing residual or operate current over time, while filtered phase currents remain steady throughout a fault, is an indication of CT saturation. One can verify CT saturation only by looking at raw or unfiltered event data. The raw waveforms from Winding 2 in Figure 8 are a perfect example of CT saturation. Memorize these waveforms, so that you can easily recognize them in the future.

Second-harmonic blocking prevented the 87R restrained- differential element from operating in this case, but one cannot rely on such restraining of the differential for all events involving CT saturation. One should check CTs to ensure that they will perform as necessary during faults. Selecting lower taps on a multiratio CT derates CT performance.

Protection Transformer Inrush Restraint

A 50 MVA power transformer is connected 69 kV delta to 12 kV grounded-wye. Upon first energization of the transformer, with the low side open, the differential relay tripped. Figure 5 shows the digitally filtered 60 Hz phase currents on the high side during energization. The current magnitudes are low. We suspect a problem with inrush restraint settings.

To verify our suspicion, we must analyze the raw data. The unfiltered waveforms in Figure 6 are perfect examples of magnetizing inrush current. Investigation of the installed settings finds that independent second-harmonic blocking was enabled as the inrush restraint. The second-harmonic threshold was set to 15 percent, a common value. Independent blocking means that second-harmonic content in excess of 15 percent of the fundamental frequency on A-phase will only block the 87R1 differential element. Likewise, B-phase second-harmonic will block 87R2, and C-phase second-harmonic will block 87R3. This mode emulates a traditional scheme of three discrete, or independent, differential relays.

The waveforms in Figure 6 obviously contain second-harmonic, but we must analyze the unfiltered data further to determine how much second-harmonic exists on each phase. Event analysis software allows us to view this information quickly. In Figure 7, A-phase and B phase had second-harmonic content at or above 30 percent of the fundamental, while C-phase had second-harmonic content just below 15 percent.

Modern event analysis software generates IEEE COMTRADE files from event data. The COMTRADE file is a standard way to represent field data so that a test set can replay the field event exactly.

We replayed this event in a lab to prove that a variety of settings solutions would solve the problem. Lab results showed that either common harmonic blocking, harmonic restraint, or dc ratio blocking would solve the problem and prevent a trip during inrush. It was easy to use settings in the existing relay to enable any or all of these elements.

Lessons Learned A differential trip during energization, where current magnitudes are low and no other backup device operates, is likely the result of energizing inrush. One can best verify the existence of inrush current by viewing raw or unfiltered data. The raw waveforms in Figure 6 are a perfect example of inrush. Memorize these waveforms, so that you can easily recognize them in the future. Unfiltered data is necessary for performance of harmonic analysis. Different phases of the transformer will produce different levels of harmonics during inrush.

When a wide variety of settings solutions are possible, convert the field event data to IEEE COMTRADE files, and replay them into a relay in the lab with modified setting. Prove theories with real-world data before implementing them in the field.

Incorrect Phase-Angle Compensation Settings

A 50 MVA power transformer is 69 kV delta to 12 kV grounded-wye, with ABC system phase rotation and ABC phase-to-bushing connections. We would expect the high side to lead the low side by 30 degrees for this transformer. We would also expect the phase-angle compensation to be WHSCTC = 0 or 12 and WLSCTC = 1 (or 30 degrees rotation in the counterclockwise direction to align with our reference high-side winding).

After being in service for some time, the differential relay trips during the external phase-to-phase fault as shown in Figure 9. Using the prefault data to perform an “offline” commissioning test, we find that phasor magnitudes, angle relationships, and rotation are as expected for through load (see Fig.ure11). The 12 kV A-phase current, IA12, leads the 69 kV A-phase current, IA69, at the relay by 150 degrees. This corresponds to the 69 kV primary leading the 12 kV primary currents by 30 degrees.

An inspection of the relay settings uncovers an interesting mistake. Winding 1 of the relay most commonly connects to the high-voltage side of a power transformer. The connection of Winding 1 to the high-voltage side is an arbitrary association by the design engineer. The high-voltage side of the transformer connects to Winding 2 of the relay. The settings for CT ratio and winding voltages reflect that Winding 1 is connected to the low-voltage side, and that Winding 2 is connected to the high-voltage side. However, the settings for phase-angle compensation reflect that the Winding 1 (in this case low-side) will lead the Winding 2 (in this case high-side) currents by 30 degrees, which is incorrect.

CTR1 = 3000:5 WYE

CTR2 = 600:5 WYE

WDG1 Voltage = 12.00 kV

WDG2 Voltage = 69.00 kV

WDG1 Phase -Angle Compensation = 12

WDG2 Phase -Angle Compensation = 1

A correct setting for the Winding 2 phase-angle compensation would be 0 or 12, and for the Winding 1 phase-angle compensation would be 1. We can assume that these settings were transposed in the settings process. Reversing them corrects the problem.

Lessons Learned - As many of these events demonstrate, it is critical to perform the previously discussed commissioning tests to identify this problem before putting the transformer and scheme into service. Verify that actual phase angles match expectations, and agree with phase-angle compensation settings selected.  Insure that the operate current is less than 10% of the restraint current. Verify CT wiring, and ensure that transformer winding to relay connections are understood.

CT Saturation Resulting From DC Offset

A generator is online, but is not synchronized to the power system. The breaker on the high side of the generator step-transformer is open. The 270 MVA step-up transformer is connected with grounded-wye to the 345 kV side, and with delta to the 18 kV side. System phase rotation is ABC, and phase-to-bushing connections are A-H1, B-H2, and C-H3. The delta is made with polarity-of-X1 connected to nonpolarity-of-X3. This type of connection forms a yDAC or yD1 transformer. The relay is set “y11D,” with the Winding 2 delta side as the reference, and settings are correct.

A flashover occurs on the 345 kV side, causing a 345 kV C phase-to-ground fault. As we see in Figure 10a, the fault appears correctly as an A-to-C phase fault on the low side. We know from the event data that the fault was external to the differential zone, but the differential relay operated. The phase-angle relationships for the fault agree with what we would expect, given our transformer and settings. The event data show an unexpected 3I0 ground current on a delta winding and decaying 3I0 ground current magnitude (and operate current), despite the filtered faulted phase currents remaining constant. We therefore suspect CT saturation.

We need a raw event report to prove our theory. Figure 10b shows the unfiltered data. This saturated waveform looks very different from our earlier CT saturation example in Figure 8. Rather than the “sawtooth” waveform of that figure, we see here heavy dc offset with a very noticeable long time-constant delay. This condition can arise during external faults close in to generator step-up transformers. We can also see from the filtered data in figure10a the point at which there is a sharp spike of 3I0 current on the low side. This spike corresponds to the point where the dc offset of the two low-side currents has subsided and is nearly, but not completely, balanced. 2.5 cycles later, when the differential element trips, it is evident that the high-side CT still has a good deal of offset. This dissimilar CT performance causes the differential misoperation. We must adjust the minimum operate setting O87P and the slope settings to make this application more secure.

Lessons Learned Heavy dc offset is another contributor to CT saturation. We must have raw event data to see this offset. Decreasing residual or operate current over time, while filtered phase currents remain steady throughout a fault, is an indication of CT saturation. The raw waveforms from Winding 2 in Figure 10b are a perfect example of a long time-constant delay, dc offset. Memorize these waveforms, to recognize them in the future. Check CTs to ensure acceptable performance during faults. Increasing minimum operate and slope settings increases security (while decreasing sensitivity).


David Costello graduated from Texas A&M University in 1991 with a BSEE. He worked as a system protection engineer at Central Power and Light and Central and Southwest Services in Texas and Oklahoma. He has served on the System Protection Task Force for ERCOT. In 1996, David joined Schweitzer Engineering Laboratories working as a field application engineer and regional service manager. At present he is a senior application engineer. He is a senior member of IEEE, and a member of the planning committee for the Conference for Protective Relay Engineers at Texas A&M University. .