Knowledge is Power: SEL 351s- Significant systematic settings errors found on 150 feeder relays.

This article was previoulsy written and published by our electrical engineering director Thierry Julio Epassa PE        

EEE 003: SEL 351s- Significant systematic settings errors found on 150 feeder relays.

Power System Description

A substation feeder is rated 34.5 kV and 20 MVA with primary protection an SEL relay 351A. A fault occurred on that feeder at 15:07, and the relay generated the report shown in figure 2; I modified the digital chart from the original report, I added some digital elements to show all the settings errors.

Figure 1 below shows pre-fault currents and voltages data at the orange time cursor. The top chart shows balanced voltage before the incident (red dotted line); the bottom graphs show 2 phasor diagrams, voltage, and current.

1- IA mag = IB mag = IC mag = 325 A. 120 degrees apart, ABC rotation.

2- VA mag = VB mag = VC mag = 19.5 kV. 120 degrees apart, ABC rotation.

3- You should notice that voltages and currents' respective phases are 180 degrees apart.

4- Just for cool learning, this facility makes $110 per MWh. With 325 A and phase voltage 19.5 kV, 0.95 pf, MW output was 325 * 19.5 * 0.95* 3 = 18 MW. (notice there is no sqrt 3, tricky because we use phase voltages, not line voltage in the equations). In 24 h, this facility would have made that day $110 * 18 * 24 = $47,520. In case you want to justify the salary of a power/protection engineer that can help mitigate/prevent outages, there it goes.

Figure 1: Relay event showing pre-fault data (orange cursor) of the system

Significant Errors/Issues: In figure 2 below, all the critical elements (word bits) on the digital chart that are supposed to detect excessive high current are flickering during the fault:?

- 51A and 51B behave identically. Assert, deassert, assert, deassert, and then assert

- 51C and 51G assert, deassert, and assert

- 51P asserts, deasserts, asserts, deasserts, and then asserts

- 51Q asserts, deasserts, and then asserts

- 67P1 asserts, deasserts, asserts, deasserts, and then asserts

Almost all the word bits are gone at the magenta and the orange time cursors. We are in the middle of a 3 phase/arc flash fault, and the word bits that are supposed to protect us have disappeared:?

- 51A and 51B pick-up values in this relay are 500A. The magenta time cursor shows IA.mag = 10495.5 A, IB.mag = 10017.2A. The orange time cursor shows IA.mag= 3393.17A and IB.mag = 8857.29A. Why are 51A and 51B disappearing??

- 51C pick-up value is also 500A. The magenta and orange time cursor respectively shows IC.mag = 10332.8A and 8721.67A. Why does 51C assert at the magenta time but mysteriously disappear at the orange time cursor?

- 67P1 and 51P deassert at both timelines.

- The voltage values at the magenta time cursor are VA_mag =2.16 kV, VB_mag = 0.34 kV, VC_mag = 0.55 kV. Voltage dropped from 19 kV to almost zero.

Figure 2:?Relay event showing voltage and current data during the word bits dropping.

Why did the word bits drop?

The following conditions dictate when all the word bits shown in figure 1 shall assert:

1- Current magnitudes are above their respective pick-up values.

2- If directional is enabled, the current should be in forward or reverse (defined by the user). If directional is not enabled, disregard this condition.

During the fault, the current magnitudes were above their respective pick-up values.

Directional was enabled, DIR1 = F, E32=AUTO, all the torque control set to 1.

With those settings, the word bit 32PF will indicate to the relay that currents are forward direction and 32PR means reverse direction. Be aware that there are 1 million ways to set directional, and I love that flexibility about SEL.?

In figure 3 below, I removed all current and voltage charts; I inserted digital charts and phasor diagrams.

First timeline: orange cursor?

At the orange time cursor of figure 3, 51A, 51B, 51P, and 67P1 are deasserted. 32PF is also off. To fully understand how word bit 32PF works, read pages 4.15 through 4.42 of the 351A manual. All the 351 series work the same way concerning directionality.

Figure 3:?Relay event showing phasor diagram at orange time cursor

Few interesting facts:

- 32PF asserts during the fault initiation, deasserts, reasserts, deasserts, and reasserts. It does not seem to make sense. There was a sustained 3 phase/arc flash fault; Consequently, 32PF should have stayed on the whole time.?

- The phasor diagram shows Z1 in a forward direction. How can we have a 3 phase/arc flash fault, Z1 in a forward direction, but 32PF did not assert.?

- Relay manual for 351A (in SEL website) figure 4.26 dictates all the conditions for F32P ( when F32P assert, 32PF also assert) to assert. There are 4 conditions

?Condition 1: All internal phase to phase distance type elements forward on

Condition 2: Z1 in forward condition

Condition 3:?R32Q is 0

?Condition 4:??See figure 4.26 of the manual

The relay met Conditions 1 and 4 ( not discussed in this article), achieved condition 2 (the phasor in figure 3 shows it), and fulfilled condition 3 (the digital chart in figure 3 shows it).

Despite the relay fulfilling the 4 conditions, why did the relay not assert 32PF at that orange time cursor?

It seems to contradict the manual. No, the relay is not to blame; The designer did not program the settings adequately.

Second timeline: magenta cursor

At the magenta time cursor of figure 4 below, all the word bits are deasserted. However, all the 4 conditions discussed earlier were also met. At the magenta time cursor, the phasor diagram shows Z1 forward, R32Q is 0, and all internals phase to phase elements are active.

Notice the word bit 32PR was on during the pre-fault, confirming that the current was in the reverse direction in normal conditions. Moreover, as I noted earlier in figure 1, the voltage and current were 180 degrees apart, confirming that current is reverse during nominal conditions.?

During the fault initiation, 32PF is asserted, which tells us that the current is indeed in forward condition. Somehow, the relay got confused and dropped that 32PF a few times during the fault.

The relay was confused because the designer improperly programmed it.

Figure 4:?Relay event showing phasor diagram at the magenta time cursor

Summary

The current was indeed in the forward direction, as indicated by Z1 vector position. Still, somehow, at a few timelines during the fault, the relay refused to declare that the current was in the forward direction. All the elements in this event worked in the forward direction, and they were therefore dropped when 32PF was removed.

The root cause is not the relay malfunction, not the CT measurement errors, not a system error, not a glitch. The root cause is human error, engineering error: improper settings design.

I have seen 150 relays with that improper design, and I have documented a few significant incidents. I am sure there are more relays currently running with those dangerously wrong settings. There was no incident in this case because, fortunately, the relay initiated the trip before those elements were deasserted. Several years later, what could go wrong went wrong, and the relay did not trip as designed.

In April 2020, I met a lady in ETAP training, and we discussed her ETAP model and SEL 351s relays. She showed me their settings, and I immediately noticed the flaws discussed in this article. In November 2020, we talked on the phone; an incident had occurred, and the relay did not trip as designed.?

She later confided that she changed their procedures after our meeting and would have been significantly hurt had she not done so. She was pregnant at the time.

Their preliminary conclusion was: SEL Relay malfunction. Unfortunately, this often happens in power system significant failures: manufacturers are easy targets to blame. Would you be surprised if I tell you that 80% of the 1,000s significant incidents I reviewed were caused by human error: Improper power system studies and improper relay settings—more articles to come.

This event is highly complex and extremely rich in learnings. We can spend a day reviewing this event in detail.

---------------------------------------------------------------------------------------------------------

La Fin, The End, El Fin