Live data from Hacker News

Smart thermostats inadvertently strain electric power grids

news.cornell.edu

31–40 of 226 posts

Re: Smart thermostats inadvertently strain electric power grids

#31

Sounds like they need to program them with something like exponential backoff and retry with jitter. Do the smart thermostats have any information on grid demand? IF not, then they could just be programmed to apply some random time shift +/- a few minutes (or however long) to spread out the demand and lower the spikes.

Came here to propose jitter as well: https://en.m.wikipedia.org/wiki/Jitter

Re: Smart thermostats inadvertently strain electric power grids

#32
Seems like less of a problem for smart thermostats (which will likely adjust to an odd-number start time based on your actual patterns and the heating speed of your house) versus the clock-based thermostats where everyone just had set it to 6am because that was the default and it was a pain in the ass to change it.

Re: Smart thermostats inadvertently strain electric power grids

#33
Isn't this the same if you have a 'dumb' thermostat set on a schedule, which is presumably what most people have anyways? Even if you have a 'really dumb' thermostat, presumably people are waking up at similar times and turning on their heaters...

I sometimes wonder if the smartest thermostat isn't actually a 'really dumb' thermostat on a timer. As in, you want it to be warm? You have to get up and set the heater to 21C/70F, this turns off after 4 hours (or when you lower it). This way if someone hit snooze, their 'smart' thermostat isn't turning on the heater at 6AM. If they usually are home at 7PM but have dinner at a neighbors house, it doesn't think they are commuting home and start turning on the AC...

Re: Smart thermostats inadvertently strain electric power grids

#34

Ecobee has eco+ which has "Community Energy Savings" [0]. It was just way too intrusive. It seemed to be a reactive system vs proactive. Technology Connections goes into this more [1], but I believe they should be providing a system that proactively cools when there is less demand _prior_ to an expected increase in demand. Effectively using your house as a "battery". When I had it enabled, I was _always_ too hot (in…

Nest's does this; it cools below your set temperature in advance of the "Rush Hour" for a few hours. https://support.google.com/googlenest/answer/9244031

In my experience, it was not nearly sufficient. If your AC is appropriately sized (or undersized) for your house, you can't just drop several degrees in 2 hours during the hottest part of the day. Going from 78 to 72 to 84 in set temp looked more like 78 to 77 to 83 in practice. Granted this was in AZ in an older house, so extremely high outdoor temps combined with marginal insulation didn't help.

If they pre-cooled the night before it might be a different story.

Re: Smart thermostats inadvertently strain electric power grids

#35

Smart thermostats could really be a lot smarter, in particular if you had a little bit of local battery storage, but even without that it's possible to imagine a thermostat that could be set to lower electric bills based on real-time monitoring of electrical prices. > "With few exceptions, when the load peaks, the price also peaks. So, if you can move some of your electricity consumption away from the peak times, rea…

Every house is already a battery, with significant thermal mass. Unfortunately, this means that the is not real-time but forward-looking: The thermostat needs to predict when load and price will peak a few hours early, and turn the AC on before that happens. If you cool the house down lower than necessary in the morning, it will warm up through a high-load afternoon, but stay comfortable, spreading the load and reducing usage during that price peak.

Of course, if everyone does this on every morning when the national weather service predicts high temperatures, you've just moved the demand forward a few hours and not actually spread it out.

Re: Smart thermostats inadvertently strain electric power grids

#36
As one of the authors of both the Zigbee Smart Energy Profile v1 [0] and the pre-IEEE seed that became P2030.5 Smart Energy Profile 2.0 [1], we did anticipate this and included a randomized modifier to any demand response or price signals coming from the grid to provide the “jitter” discussed in other comments.

The problem here is that these thermostats have NTP-synched clocks and come with default schedules out of the box and failed to consider the importance of staggering over a 5-min time period. It’s understandable from a consumer acceptance standpoint (“I set it to start at 6a, why did it turn on at 6.03a?!”), but that could be solved with some messaging.

Better still would be for future thermostats to be required to connect to the Internet and to the grid’s price signals or utility’s servers in order to receive rebates. This would see them base behavior on pricing signals, which would include a device-specific random offset to the price taking effect.

Resideo (formerly Honeywell Home) acquired some very cool tech that would solve this issue, too. Developed by Whisker Labs (formerly WeatherBug) It combines a weather feed with a custom thermodynamic model of each home to adjust the start time of each setpoint. This provides both energy savings and more comfort. For example, if it’s an abnormally cold Feb day, it would start the heater earlier to reach the desired setpoint rather than a normal thermostat starting at a precise time and hitting the setpoint later in the morning.[2]

Final fun fact: ~40% of smart thermostats never get connected to the Internet. Many people are price insensitive enough to buy a communicating thermostat and just slap it on the wall without bothering to program the Wi-Fi settings. There are sneaky ways some manufacturers and ISPs are trying to overcome this, though it’s a borderline dark pattern. [source: I used to work for a large ISP’s connected home ecosystem team]

[0] https://zigbeealliance.org/wp-content/uploads/2019/11/docs-0...

[1]https://standards.ieee.org/ieee/2030.5/5897/

[2] https://www.resideo.com/us/en/corporate/newsroom/all-article...

Re: Smart thermostats inadvertently strain electric power grids

#37
post #13
post #3

How is this different from a regular timer thermostat which turns on exactly at the set time?

Syncing the time over the internet would make them a lot more consistent. Our non-smart thermostat has trouble keeping accurate time and drifts a minute per month or so. That would spread the load out more because “6am” is a ten minute window.

This doesn't make any sense, though, as nobody turns on a heater for 10 minutes. Even if there is drift of 10-15 mins, there would still be a peak usage at 6:15 when all of the 6AM 'starters' are sucking power from the grid.

Re: Smart thermostats inadvertently strain electric power grids

#39

Ecobee has eco+ which has "Community Energy Savings" [0]. It was just way too intrusive. It seemed to be a reactive system vs proactive. Technology Connections goes into this more [1], but I believe they should be providing a system that proactively cools when there is less demand _prior_ to an expected increase in demand. Effectively using your house as a "battery". When I had it enabled, I was _always_ too hot (in…

No post body was provided.

Re: Smart thermostats inadvertently strain electric power grids

#40

Earlier quoted context omitted.

Nest's does this; it cools below your set temperature in advance of the "Rush Hour" for a few hours. https://support.google.com/googlenest/answer/9244031

In my experience, it was not nearly sufficient. If your AC is appropriately sized (or undersized) for your house, you can't just drop several degrees in 2 hours during the hottest part of the day. Going from 78 to 72 to 84 in set temp looked more like 78 to 77 to 83 in practice. Granted this was in AZ in an older house, so extremely high outdoor temps combined with marginal insulation didn't help. If they pre-cooled…

I think a poorly insulated old house in Arizona is probably one of the worst-case scenarios here. Maybe there should be a configurable or learning-based pre-cool length, though.
Post reply on HN