#21 By: PWE, April 28th, 2014 07:16
Things appear to be settling on the CTS area from all comments received thus far. There is very little building plant in Queens generally, just two air handling units in addition to the new over door heaters. There is some electrical testing being carried out in queens over the next couple of months which may give us some better clues as to which distribution boards the load in on.
#22 By: Richard Snape, May 23rd, 2014 11:42
I note there is fixed wire testing overnight during a 10 day period (25th June - 4th July). Can we keep tabs on which distro boards are isolated when to give us clues as to what's in the baseload (e.g. which distro boards were isolated when we see big drops).
What say you, @DMU_Energy, @carlholland, @lozawa00 and @ggstuart?
P.S. I found a copy of the as-built electrical drawings for Queen's. I note there's one particularly large distro board (isolator rated at 630A per phase) which could be a good place to start looking - the large capacity implies that a large load is anticipated there (or, at least, was when the building was built). It is DB 7-1-25 located in room 1.37. A quick skim of the prints suggests there's not an individual likely suspect connected to this board, but it seems to feed a lot of the ground floor including the machines and lighting.
#23 By: PWE, May 28th, 2014 05:15
Already on that one Richard! Well done for finding the drawings!
#24 By: Richard Snape, May 28th, 2014 06:04
@DMU_Energy I thought you would be
#25 By: PWE, June 26th, 2014 04:10
Well, the results from night 1 of the Fixed Wire testing in Queens are in. I'm not sure these are much help. I've compared last night with the previous three Wednesdays (and other days around the 25/06 in other graphs) and yesterdays profile was low all day, not just at night. What I can't see is a dip and then a recovery, which I was initially expecting. However, I am also thinking that the technicians in Queens are probably going to make sure all equipment likely to be affected by the testing is shut down on a controlled manner before they go home, and then they will come in early the next day to turn it back on again. This means that we will see the dip possibly around 5pm and then the baseload will remain low all night and come back up the next day. I was initially excited about the drop from 4pm onwards, thinking this was equipment being shut down for the testing, but Monday and Tuesday are very much the same. Check back tomorrow for the next instalment...
#26 By: Leticia Ozawa-Meida, June 26th, 2014 04:44
Thank you very much for posting the graphs and the explanation. It is really intriguing for a lot of us around here. Please keep posting them until all the fixed wiring testing is completed. One question, the graph says "night 1 testing" was done in the Substation, is it possible to know which parts of the building are shut down or was it the entire building? Sorry, if my question is silly. If it is done by parts, can we know which wings or floors are we talking about?
#27 By: Richard Snape, June 26th, 2014 05:00
I wouldn't be too disheartened, this is great. Personally, I wouldn't expect to see huge amounts on most switch offs, but then see a significant drop for one or two (maybe that's just what I'd hope for in trying to isolate the baseload). I guess if they're all like this, we have to think harder.....
#28 By: PWE, June 26th, 2014 05:28
@lozawa00 - Hopefully Graeme's sheet that he has asked the contract manager to complete as the tests are done will be able to provide us with an idea as to what areas are being tested each night, we can then compare this with the data. The big area of interest is the CTS area (where the studios, etc are). This is being tested on 1/7. However the whole building blackout is also being carried out on the same night. Hopefully we will be able to see the CTS being shut down before the blackout test.
#29 By: PWE, July 2nd, 2014 06:00
So, we have found 25% of the baseload last night. I'm waiting for confirmation of where the contractors were testing, but this is the first real move in the baseload that we can see since the testing started. My assumption is that this is associated with shutting down (part) of the CTS area (where the radio studios are based). No sign of the blackout though?!?!
I'll come back to you once I have confirmation.
#30 By: Graeme Stuart, July 2nd, 2014 04:58
The testing on 1st of July has had a huge impact on the baseline of queens building. It even looks like some of the impact has persisted until this morning. However, the baseline is still very high. Some loads have been removed and not come back whilst others have come back. We may see other loads come in during the day and the rest of the week. How does this relate to energy wastage and opportunities for efficiency gains?
It will be interesting to see if some savings persist until the weekend.
#31 By: Richard Snape, July 2nd, 2014 05:41
A careful correlation with what actually took place last night might yield some really interesting results. In particular - I'd ask whether the planned "whole building off" occurred last night and, if so, when and for how long? Because the baseline change looks pretty constant, but a long way from zero. If the whole building was supposed to be off - what's the meter feeding that is not on the switch??...
#32 By: PWE, July 3rd, 2014 04:08
Most things had come back on by 10am the next day (2nd). The whole building off did occur, but the test must have been for only a few minutes as it doesn't register on the data. The meter I'm using is on the primary LV panel, which is where the building was isolated from. Something also happened last nigh. I am hoping to go through the certificates issued by the contractor to know where they were and when (unfortunately they weren't given Graeme's sheet to complete!).
#33 By: PWE, July 3rd, 2014 04:10
Looking at the graph on the previous post, it is possible that some things may not have come back on. Time will tell!
#34 By: Richard Snape, July 3rd, 2014 06:56
I guess so - if the whole building was off for e.g. 5 mins, I suppose you'd expect to see ~8% drop in the meter reading for that hour or ~16% if it's half hourly. Hard to isolate when the baseline is dropped ~25% anyway due to other tests and it could well be less than 5 mins, I guess, depending what the test was. I don't want to sound obsessive about the whole building off, I'm just looking for an definite, identifiable, event to convince myself that all the baseload really is within the Queen's building (not, e.g., outside lighting or something...). If the meter is right next to the isolator, that should account for that, I guess.
It is interesting that the consumption remains below green on smartspaces after the switch off. I think that implies that consumption is in the bottom 10% of consumption seen for these weather conditions - indicating the potential wastage by unnecessary things being left on. Also interesting to note that the ramp up to peak consumption started much later - 0900, rather than normally 0600, or so. I wonder if there are some automatically awakened systems that didn't come back on...
#35 By: Carl, July 9th, 2014 10:26
A week on from the drop in the base load. The yellow column indicates missing data. Could this have been the moment when the whole building shut down occurred? (loss of power to the AMR equipment). The minimum that the baseload drops to has gradually risen over the week, back to about 80kW from 70kW earlier in the week. The maximum height of the day time peaks appears to have dropped a little.
#36 By: Richard Snape, August 1st, 2014 11:15
I missed this at the time. It would seem likely that the whole building off might have turned the AMR off accounting for the missing data. As you say, it looks like any lasting reduction from the switch off has disappeared now. What's interesting to me is that only the switching off of the one big board made noticable different to the baseload. So there are still some 50-55 kW unidentifiable - evenly spread through other boards, perhaps.
#37 By: Graeme Stuart, August 4th, 2014 07:45
Exactly, unless we find out that some boards were turned off for literally a few minutes, which seems unlikely given the one on the 1st was off for several hours.
#38 By: Richard Snape, August 4th, 2014 12:04
I remain confused. It's hard to believe that every other board is basically serving a load that is indistinguishable from noise. But it looks like that's the upshot of it all.
#39 By: Graeme Stuart, August 5th, 2014 12:17
I have an untested pet theory that the consumption is partly due to network switches which need to be powered 24hrs even when they are not 'in use'. I can imagine the noisy boxes at either end of IESD are at least 1kW each (they require active cooling) and there could be many of them distributed around the building. Add to this the lighting loads which could also be distributed and we might be looking at a few kW of baseload per board.
So, one hypothesis we might attack is that the remaining load is evenly distributed across all main boards.
If there's 25 boards and 50 - 60kW of missing load, then they average at around 2.0 - 2.5kW per board. Could this amount of power be 'hiding' within the noise?
2.5kW for e.g. six minutes is 0.25kWh, i.e. a very small drop in the chart. Well within the level of the noise. So its looking plausible. Even if they were off for an hour, the movement is only 2.5kW which will likely be split over three half hour periods.
I also note that on the evening of the 26th/27th there was a sudden drop of about 5kW. So we may have a slightly smaller dispersed load than we think.
#40 By: Richard Snape, August 6th, 2014 05:26
Yeah - to pursue that pet theory a little further - the boxes each contain 3 or 4 48 port switches. A modern HP one (like the new boxes) reckons to consume 100W max. The active cooling units might take 15-25W or so. Can't find huge amounts on power in switches, but this white paper by cisco is the nearest I could find very quickly. It suggests (my quick summary)
- that switches vary considerably in power performance (factor of 1.5 in ostensibly same spec from different manufacturers)
- 10GB 48 port switches consume 140 - 210W at 100% uplink
- Power needed for cooling is significantly more than I quote above (~200W). More thought / investigation needed.
- Nothing about loading as a function of number of connected ports / traffic etc.
There is a page from D-Link Green Networking that looks at a much smaller switch and suggests that about 2/3 of power is consumed irrespective of workload (connected ports). 1
All in all - I think your 1kW per box estimate may be a little high, but not by orders of magnitude. It's probably reasonable for old kit - the Cisco suggests that a 3 switch HP rack running at 100% uplink and required cooling would be ~860W. Caluculation from manufacturer specs would suggest more like 400W. I think Cisco metric running full traffic all the time is a bit unrealistic, so I'd plump for 500W or so average.
So, 4 or 5 boxes per board on average would fill up the 2/2.5kW per board. This seems very plausible to me.
I am starting to be converted to your pet theory...
1 Interesting too that simply powering the switch from higher EU voltages vs US 120V consumes significantly more power.
← previous page
next page →