Hey @dickensbill - thanks for posting this!
This gives a great illustration of one of the challenges we run into with different EV chargers: that very slow (>1 hr) taper of the charger that you noted. Some other EV chargers exhibit similarly slow ramp up times. While it’s quite easy for us to all see it here in the image, it’s quite difficult to detect in real-time.
For a real-time model to be generated, it needs to identify with strong confidence when the device turns on and off. When we think about furnaces and coffee machines, they have fairly distinct “on” and “off” transients which occur across just a few seconds. Most of our model architecture is built around looking for these short transients, so we’re doing a lot of core infrastructure work to support substantially larger detection windows that can recognize these types of “on” and “off” signals that occur across this big chunk of time.
We’re also looking at ways we might be able to report usage statistics for devices like these historically in the meantime, so even if we don’t immediately have a model that can provide real-time device detection for your i3, you could still have data that reports “my EV used X kWh and cost $Y to charge this month”.
I hope that’s helpful! Thanks again for posting this.