The change log ran to forty-one pages for the holiday build, and the line she wanted was on page thirty.
Wednesday, January 5, 07:20. Night shift was still in the chairs. The window clock on its putty read forty seconds ahead of the EMS banner, as it had all winter, and Trey was in the far seat with his headphones half on, writing a keep line for a feeder nobody had called. She had come in early to find one number, and Abel had told her where to look in two lines the morning before: Display change log. Holiday week. Not RTU.
DSP-1228 · Master alarm display · Intermediate-state alarm suppression raised 1 s to 2 s, system-wide. Reason: reduce chatter on coast feeders during wind event. Requested by: on-call. Approved: weekend queue.
One line. Signed by a display engineer whose name she did not know, at 01:14 on a Tuesday night between Christmas and New Year's, with the wind off the Gulf making every coastal recloser on the board flicker 00 for half a second at a time. She could see how it had happened. Somebody with a pane full of yellow and a wind event had made the yellow stop. She might have done the same at 01:14.
It was also why Tuesday's 1.8 seconds of 11 on PL BKR4 had not alarmed. On December 23 the pane had still been at one second, and the same 11 would have lit the row and sounded the tone, and Trey would have had something to call about instead of a photo captioned slow?
She did not ask for it back. The coast feeders had their own reasons for chatter and their own people to worry about it. She wrote to Abel.
DSP-1228 is the 2 s. Don't change it back for the coast. I want an exception list. Any breaker flagged for called events alarms on any 00 longer than 1 s and on any 11 at all, whatever the pane setting is. PL BKR4 first. MC
Then Saucedo, shorter, because she was about to ask him for something he might not have.
Tuesday's meter bins could not tell the breaker from the stair. On the one-minute EMS trend, the bin ending 07:12 had already come down to 31 megawatts before the breaker opened at 07:12:03.4, an RTU stamp shown in Central by the old display, because Hall 4's scheduler had been checkpointing jobs and dropping load in steps for eleven minutes by then. The bid card did not get paid on the trend. It got paid on the settlement meter, which binned on the five-minute grid, and it would credit the breaker with whatever that meter saw fall. Some of that fall was the machine room deciding to stop, not the breaker deciding to open.
Do you log per-hall checkpoint MW with SCHED CLOCK stamps? If yes, I want to subtract the stair from the bin and see what the breaker shed by itself. And I want that number on the card next to the 0.99. If no, write that down somewhere I can point to.
She copied both ticket numbers onto the sticky under Rubén's card, where the stickies had started to overlap like shingles. Under line 10, in Trey's square letters that still ran onto the paint, she pencilled small:
Pane setting is not ours. Exception list, DSP-1228.
Trey watched her do it and did not take his headphones off.
Thursday, January 6, 06:50. Two replies.
Abel's was four lines, which for Abel was a speech.
Can't put an exception in the pane timer. One timer, whole display. Can move points into their own alarm group with their own intermediate-state delay. Drafted DSP-1236: group DR-CALLED, 0 s on 11, 1 s on 00. PL BKR4 only member until the desk sends names. Needs your signature as requester and a display approver who is not on holiday. AQ
She opened the ticket. The form had a requester field with her name already in it and a grey button that said Sign. She pressed it before she had taken her coat off, and then she took her coat off and got coffee, and when she came back the ticket said Awaiting approver (Display) with a little clock icon beside it, which she had learned meant nothing about time.
Saucedo's reply was longer and less happy.
We do log it. Per hall, SCHED CLOCK. But one-minute resolution, and the row is written when the checkpoint job finishes, not when the hall load actually steps. So you can subtract it from a 5-min bin to within a minute. If the job wrote on time. Usually it does. I don't love "usually" on a card that gets paid. L.
She didn't love it either. She did it anyway, on the back of Tuesday's call sheet, in pen, because she wanted to know whether the number was worth asking for before she asked for it twice.
Bin ending 07:12, METER, 31.0 MW.
She had written that much before she saw it. 07:12:00 was three seconds before the breaker moved. Nothing in that bin was the breaker. Every megawatt it showed falling was the stair, and it was a trend bin anyway, not the one the card got paid on. She drew a line through it and started again on the settlement bin.
Settlement bin 07:10 to 07:15, METER, 14.1 MW average. Hall at 38 before. Checkpoint log, Hall 4, 07:11 SCHED: 9.4 MW to 1.7 MW. Step 7.7 MW. Breaker open 07:12:03.4 (RTU, shown Central), about 30 MW.
A bin was an average. A step only counted for the part of the bin after it landed. If the stair landed at 07:11, it took four-fifths of 7.7 off the bin, 6.2. Saucedo's log could be a minute off either way, so anywhere from 07:10, all of it, 7.7, to 07:12, three-fifths, 4.6. The breaker had been closed for the first 123 seconds of the bin and open for 177, so it showed a little under three-fifths of its thirty: about 17.7.
Stair 4.6 to 7.7. Breaker what was left of 23.9: 16.2 to 19.3. Not one number. A range, and the range was the honest part.
She wrote roughly beside it, and then crossed that out too and wrote range, and circled it.
Then she went to the hook wall and added line 11 in pencil. Not green. Green was for lines she had watched work.
11. Settlement bin = breaker + stair. Subtract checkpoint MW weighted by where in the bin it landed (SCHED, ±1 min). Write the range, not one number.
She sent Saucedo one line back.
Put the subtracted range on the card, both ends, or don't put a number at all.
He read it at 07:03. He did not answer. She took that to mean he was off finding out who owned the label field on the card, which on the scheduler side was never the person who owned the number in it.
Friday, January 7, 06:55. End of a week of days, and she was on her way out when she stopped at her own desk and unmuted Trey's thread.
She had muted it on December 2, the week before the false open, after the ninth photo of a keep line at three in the morning, with a message that she could still quote: Don't send me keeps. Write them in the book. He had kept sending them anyway, to the thread, to no one. Thirty-one messages. Photos of the green book open on the night desk, his handwriting getting smaller over the weeks as he learned how much fit on a line. Not one question in all of it.
The last message was Tuesday, 09:43. A photo of the event list on the night console after the restore, PL BKR4, the 11 row visible in the middle of it, and under it one word.
slow?
He had seen the 1.8 seconds before she had. He had seen it and not called her, because the breaker had opened when asked and closed when asked and nothing had failed, and her rule was that he called when something failed. The rule had worked exactly the way she wrote it. He had put the question in a place he knew she would not read.
She scrolled back up through December and found herself in three of the photos, a sleeve at the edge of the frame, a coffee cup, the back of her head at the hook wall. Then she did not answer any of the thirty-one. She walked to the hook wall, photographed lines 7 through 11, the green ones and the pencil ones and the word target on the paint, and sent the photo to the thread with three lines under it.
Clock first (line 8). 11 on close is the pane, not the breaker. DSP-1228 / DSP-1236, sign-off pending. Settlement bin is breaker plus stair. Subtract checkpoint MW, weighted by where it landed. Write the range.
She read it, and then she typed a fourth line, slower.
If you see 11 for more than 2 s on a called point, call me. Day or night.
It was the first exception to don't send me keeps she had ever written down, and she had written it to the person she had written the rule for. She went back to the wall and put his initials and the date in the margin beside line 11, TH 1/7, so the next lead who stood there would know that Trey had this one.
DSP-1236 still said Awaiting approver (Display). The clock icon had not moved.
She left the thread unmuted and went home.
Monday, January 10, 05:20. The drafter release was in the overnight build. Saucedo's note had said the 10th in December and had said it again on Friday, and it had landed on the 10th, which made two promises in a row from the scheduler side that had beaten or met their dates. She still did not trust a release note she had not watched.
Abel was on the phone from the display room, the call on speaker on her desk, and she could hear the hum of the rack fans behind him.
"PL BKR4 to test position," she said.
"Test position." A pause. "It's yours."
"Hold 00 on the double point."
He held it. She watched the drafter pane, which for nine days had been rendering PL BKR4 as whatever it had last believed. The row went amber at a second and a fraction.
PL BKR4 · DP 00 >1 s · UNCONFIRMED
"How long?" she said.
"One point four."
"Now 11."
He forced both contacts. The row went red, and the words were exactly the ones she had typed on December 27.
PL BKR4 · BAD POINT, CALL FIELD
On the bid card, which she had open in the other pane, Hall 4 stopped carrying 0.99 and showed a grey chip: Point manual. The card had given up the fleet average and was waiting for a person. That was the thing she had asked for without quite asking for it.
"Good," she said. "Leave it there. Hold on."
The SOE column had a new tag next to every stamp. She had asked for that in December too, item (d): every time names its clock. The tag was there. It said CPT.
She looked at the digits beside it. Abel had forced the 00 at 05:22 by the banner, and the SOE line read 11:22:41.306 CPT.
That was the stamp straight off the RTU, which took its time from the substation GPS in UTC. Abel had read her the setting on December 24, standing at the relay panel with Dorantes, and she had written it on line 8 in green. The minutes matched the banner. The hour did not. She did not stop to count hours; the minutes were enough to tell her this was the raw stamp. The old display had converted it to Central before it showed her anything. The new one had stopped converting and started labeling. A UTC time, called Central. Every SOE line in every morning report from today would look hours off to whoever read it tired at eleven at night without knowing why.
Line 8 said Name the clock. The release had named one. It had named the wrong one, with confidence, in the exact place she had asked for a name.
She wrote it up for Saucedo in four lines.
Abel forced 00 1.4 s and 11 on PL BKR4, test position, 05:22. Row showed UNCONFIRMED and BAD POINT, CALL FIELD. Card went to manual. Accepted. SOE tag reads CPT. RTU stamp is GPS UTC. Rejected. Every SOE line will read about five hours off. Take the tag from the source (RTU, GPS, UTC). Not from the server's locale.
She marked the release in the tracker: accepted on 00 and 11, rejected on SOE tag. Then she went to the wall and pencilled under line 8:
Tag says CPT. Stamp is UTC. Until fixed, add 6.
She looked at it and crossed out the 6 and wrote 5 above it, because daylight time had ended in November, and the RTU did not care about daylight time either. She went back to her desk and the first call of the morning came in, and she did not think about the line again for an hour.
06:40. Coffee, and the pencil line under 8 again, because she walked past it on the way back from the machine.
Until fixed, add 6 5.
She read it twice, and it was wrong both ways.
She had the 05:22 test line still up on the SOE pane. 11:22:41 on the line, 05:22 on the banner when Abel forced it. Six. In January Central ran on standard time, UTC minus six. Five was last October's number, the one she had been subtracting since March without thinking about it. And you did not add. A stamp that read 11:12 UTC had happened at 05:12 at this desk, not at 16:12. She had written a rule about naming the clock, and then she had done the arithmetic from memory at five in the morning, which was the exact thing the rule was there to stop.
Nobody else had read it yet. Trey would be in at 18:00. The wall was hers for eleven more hours.
She erased the whole pencil line, hard enough that the paper went fuzzy, and wrote it out longhand.
Tag says CPT, stamp is RTU GPS UTC. Until fixed: subtract 6 (CST, Nov to Mar), subtract 5 (CDT, Mar to Nov). Check the date before you subtract.
Then she opened her 05:50 note to Saucedo, the one that said about five hours off, and replied under it.
Six, not five, this time of year. Which is the reason the tag has to come from the source.
She did not apologize in the thread. Saucedo had his own clocks to be wrong about. She went back to the wall and made a small tick next to line 8, and beside the tick, 1/10, so that whoever read it next would know the line had been corrected on the tenth, not written wrong and left.
The window clock said 06:47:40. The EMS banner said 06:47. DSP-1236 still said Awaiting approver (Display), and somewhere in Port Lavaca a slow b-contact was waiting for March. She sat down and started the morning report by hand, and when she came to the first SOE line she stopped, checked the date on the banner, and subtracted six.