Internet connected smart meters share your total electricity usage with your power company. What most people don't realize is that by tracking this simple variable over time, energy companies can actually deduce which devices you own and use.
It's called "Non-instrusive Load Monitoring" or "Power Disaggregation".
https://en.wikipedia.org/wiki/Nonintrus ... monitoring
For example, your fridge turns on at certain intervals, creating a unique power-usage fingerprint over time. All devices have unique fingerprints. This can be recognized in the total signal, as long as you measure that precisely enough.
The advantage of using these algorithms is that you don't need to wire up every device separately if you want to know how much power they are using, and when they are used. The downside should be obvious: this is a huge privacy risk, which is aggravated by the fact that few people know this is happening.
To get the advantages without the disadvantages, the data analysis should be done on the local network, and not in the cloud.
There are a few options here:
A. Adding Open Source NLM algorithms to Domoticz
I think the most logical place for this analysis would be in a domotica hub like Domoticz:
- A lot of people already use Domoticz and other platforms to also record the the total power usage of their home themselves, both from their smart meters or with clamps.
- Platforms like Domoticz already have a (partal) list of devices in the home.
There could be a lot of synergy here:
- The power usage of devices in the home could be automatically deduced and presented. That's a big feature.
- The user could be alerted to differences between expected and actual behavior of their devices. For example: that their fridge door may not completely closed, because it has used lots more power in the last hour.
- When a device fails (for example a pet feeder), the user could be informed.
- When devices still work but are slowly starting to fail, their power usage may start to subtly fluctuate. This could be detected. Dangerous situations could even be avoided: detecting an overloading device which could catch fire, like an electric mattress heater, and then turning it off before it catches fire.
- It could do all the other things commercial versions do: http://neur.io/iem/
Still, I couldn't find much about using these algorithms in Domoticz:
Even though there are open source algorithms available that could be used:
And even a database of power-signatures:
Domoticz could start with a very simple version. Domoticz powers on each device it controls the power to, one at a time. The base load of each of there can then be recorded, which offers a good starting point. The rest of the devices in the home are then turned on one by one manually, by the user (no need to be 'non-intrusive'). This should create a good starting point for further signature detection. At first, the most power hungry devices with easy signatures (heating, television) will be easiest to monitor, which would already be valuable. Over time this could be improved.
B. Using an external energy analyzer device
Specialized energy monitors like OpenEnergyMonitor would also be a nice place to do disaggregation work. By placing the algorithms outside of any specific smart hub, all smart hubs could connect
https://forum.mysensors.org/topic/1464/ ... -sensor/38
Oddly enough, I haven't found an open source one that does disaggregation either.
Plenty of commercial ones though:
A lot of money is being pumped into this field:
https://www.greentechmedia.com/articles ... ggregation
C. Uploading data to online analysis services.
A final way of keeping the data away from your power provider, while getting the benefits of Non-intrusive Load Monitoring, is to use a cloud-based analysis provider with your own 'dumb' measuring device. Of course this means the data is still escaping your network, just not to your electricity provider. You stay in control.
So I'm very curious:
- Has anyone combined energy disaggregation with Domoticz (or another hub?) in some way?
- Would you agree that Domoticz is a logical place for these algorithms? Or is OpenEnergyMonitor a better place for this?
Posts on this forum:
Outside of this forum:
https://forum.mysensors.org/topic/1464/ ... -sensor/26
1 post • Page 1 of 1
Users browsing this forum: No registered users and 2 guests