#1 By: David Young, December 5th, 2014 05:02
Just been looking at the graph for 16 New Walk, looks like there is a fault with the data on Wednesday 03 December, as the Electricity predicted area is well into the minus. Any ideas why?
#2 By: Richard Snape, December 9th, 2014 06:37
@ggstuart would be far better placed than me to comment, but it looks like a failing in the consumption model to me. There's a less obvious version of what seems like the same problem here - in the middle of Wednesday 26th November. The fact that they're both Wednesdays around the middle of the day might be significant.
My understanding is that smartspaces calculates a model based on temperature for each timeslot in the week. If data for Wednesday mornings over the year is poor in some way (e.g. a big spread of consumptions for the same temperature, or consumption unexpectedly going down as temperature falls rather than vice-versa) that might lead to a bad model for that timeslot. Again, @ggstuart is the man to confirm or deny this theory
My other thought is this - we have hit more extreme temperatures for the first time this week - i.e. the temperature has gone lower than 5oC and sometimes even below zero. The model will have less data at these more rare temperatures, so we are relying on extrapolation of the model to areas where there are less data points. This will tend to highlight more problems with the data model that predicts expected consumption.
P.S. Water shows a similar, albeit reversed, problem for the same building on Friday 5th - where the expected consumption shoots off to an unreasonably high level
#3 By: Graeme Stuart, December 9th, 2014 07:40
I agree, extrapolation and/or erroneous models are the most likely cause. This could be because the temperature is more extreme lately than has occurred in the previous year. It could also be that in the previous year there are a few cases of low temperatures but they were also associated with unusual consumption patterns (e.g. over holiday periods). A final possibility is that the temperature data coming in are corrupted and we are extrapolating to e.g. -15 degrees which could also cause this kind of thing.
#4 By: Jas Odell, December 11th, 2014 06:28
Over the last two days, the electricity consumption and the base load has increased. There is a poblem with the district heating and it is manually controlled at the moment. The engineers are aware of the problem.