Nearing limits : warning

Not sure if this came up. Assume a home with a standard 100A split phase mains. Or heck: 200A.

Request #1: as more and more stuff becoming power hungry, etc. a home may tend to utilize almost its full capacity. If one factors in the in rush due to mutilple motors, etc the main breaker may start to trip. So, would it not be great if this product post historical analysis start to determine if a particular home is nearing its stated capacity and sends out warnings? This way a homeowner can plan for an upgrade beforehand. If he is already at utility max capacity (usually 300A) then he can sit back and re look his usage?

Request #2: Almost all residential breakers are specifically built to sustain its max rated load for xx time. In other words, 15A sustained current for (say) 5 minutes. If sense can detect a big current draw and that too its longer than xx minutes it has a potential of overheating a circuit breaker. Case in point: with the deluge of EV charging stations an electrician was telling me cases of breakers overheating has quadrupled. Another case where Sense can do preemptive warnings!!


Interesting - I had talked with Sense a while back about the value of a “virtual circuit breaker” to catch these kinds of power issues. The challenge is that Sense can only really accurately observe what’s happening in the mains and perhaps a couple of instrumented circuits (DCM). Still useful.

In these 2 cases the entire technology should be already there.

For example: at any given time Sense already knows the total current draw. Now multiply that second by second for a week. [i do Machine Learning / Deep Learning for a living]. At this point thru unsupervised learning standard algorithms we have patterns which can be used to determine if reaching limits of main lug tolerance. This can then be sent as a warning to the homeowner.

For the 2nd case the approach again is similar to above but atomic instances are captured over time and patterns established. I don’t know what is being used but if Tensor Flow then its very straightforward. Again alerts can be sent out.

For the 2nd case its important to know deep technical details on capabilities of circuit breakers in terms of sustained currents, testing, etc. as 60% of homes are today using the same company products like QO or Homeline breakers, again this should be very doable.

I am NOT suggesting doing anything automated like tripping the circuit, etc. That has its own issues and complexities including code (NEC) etc.

Just alerting the end user.

The company’s sales team should love this idea BTW. As it gives a notification to end user and then gear up to either replace the panel or buy probablu different breakers. Like for example higher rated AIC?

Here’s a little bit more to ponder - a typical “trip curve” for a circuit breaker, which is a function of current usage and sustain time. This is just an example but gives a view of different sustain time scenarios under which a 100A main breaker would trip.

As for your second scenario, the Sense problem is that it has no idea which circuits (and breakers) a detected device is on nor does it know which devices are together on the same breaker. EVs and HVAC units might be special cases it could handle because they typically require dedicated circuits (often 240V).

It’s not as sophisticated as a breaker trip curve, but you can set “goals” in the Sense app that will push a notification to your phone if you exceed a certain power value. I have one set to trigger at 90A on my 100A service. I’ve never hit it, even with an EV charger, a 12kW steam generator, and electric kitchen appliances.

1 Like

These goals are fine. However, that’s an isolated use case where due to a particular atomic instance your load center may hit the 90% mark and you get an alert. The whole concept behind Sense is pattern and usage recognition and using which create identifiable and actionable patterns utilizing the power of machine learning. Thus my suggestion. If Sense can tabulate such instances it’ll be great. The technology is already there. Someone needs to write the code. For example - in an area where I work:

Approximately +1 million bank accounts segregated into clients. A client can have 10 to +10,000 accounts globally. Using machine learning I constantly track these accounts w.r.t. over-draft, interest accured, etc. Then via patterns I can make recommendations to the client’s treasury department on how best to structure these accounts into tiers for example which can and will greatly reduce occurrences of overdraft and get interest.

What I wrote above is +10 times more complex than what I proposed.

@kevin1 Thank you. This helps a lot.
Taking my 1st. case : Now that the curves are known, Sense can start accumulating every time the home hits that curve. For example: at 100A @ 1,000 sec. its envisaged the main breaker will trip. 80% of that will be 80A @ 800 sec. Store : Value #1 with time. Iteratively keep adding the values. Set a time period. If there are 6 stored values in that array of (say) 10 max size with time = 1 month, generate an alert. If within that time, the array gets exhausted - generate an immediate alert.

Taking my 2nd case: You’re 100% correct Sense will never know which breaker as it works off the main lug. That’s fair and square. However, it does know if (for example) there was a sudden sustained current which is +50% of stated load and that too for XX minutes. For a 100A, let’s say Sense detected 50A continuous usage across 1 pole(?), both poles and it’s for upwards of 15 min without any drops. Now, most residential circuit breakers as I’m being told are tested with sustained current only for few minutes at best. Even though here the individual circuit breakers - let’s say 1: 15A, 2: 20A and 1: 30A are involved its still within its individual limits so they don’t trip. But, however they are indeed at maximum and that too for so many minutes. There’s a strong chance of over heat here. Again: its NOT something which trips or breaks a circuit but a warning message to end user.

I think there’s a real opportunity here, but I think of it a little different. This would also be a big investment for Sense.

A lot of people are maxing out their panels as they purchase EV’s. Now we know most people charge their EV’s at night, so panel capacity is theoretically not much of an issue. Except this isn’t recognized in building codes, as there will always be that one person who charges an EV while running everything in the house.

What if Sense had an integration with one of the EV charger manufacturers that would only allow the EV to charge when the panel was working well below capacity? This could help people avoid some very expensive panel upgrades. Of course, I assume it would also require some extensive lobbying to factor this into the NEC.

1 Like

Seems like a custom notification field could be generated. Such as XX wattage exceeded over XX time. Obviously if would have to be for the whole service box and not a single circuit such as a clothes dryer, if the dry was on and the AC kicked on or a EV charger and you had it set at 8500 watts for 120 seconds its going to send you a warning but in reality your below 36 amps in a 200 amp box.

Maybe this would be better just added automatically to labs.

There used to be 120% rule in the NEC which pretty much means if you have 200 amp main than all your breakers cant exceed 240amps. However seems like every jackleg working on electric these days has never though about that. They wouldn’t hesitate to pull every 15amp breaker out and put 20 amp ones on 14/2 and run a few 100 amp sub panels out of 200amp panel that was already fully loaded. Then these EV chargers are 40-80amps that people are using for 10 hours at a time.

People also forget when adding solar this matters… the rating of the on the “box” is actually the rating of the busbar. So if you have a 60amp interconnect inverter and you are pumping 60 amps in to that bus and your in a 200amp box with a 200 main., you might end up feeding the bus 200 amps and 60 from the solar and those breakers think every thing is fine … and some point your feeding 2 EV chargers. a dryer and 2 AC units so each of the individual breakers also are fine… but your busbar has 260 amps on it which is 30% over its rating.

1 Like

@baivab.mitra, just noticed your reference to TensorFlow. Not sure which ML platform Sense leverages for ML, but the main detection learning type today is unsupervised. You might enjoy reading though what I have learned (and think I learned so far) and help extend the knowledge…

Wow!! I have to go thru this, this weekend. I wish we had better access to Sense Developers.

100% correct. Yep: nice for labs!!