I've just made arrangements to get a Cala heat-pump water heater thermal battery installed: https://www.calasystems.com/ We also looked at: https://www.reservoirhome.com/ This looks good too and I would be happy with either. Cala is made in the US and is already supported by Home Assistant Free/Libre/Open-source home automation software: https://www.home-assistant.io/ I refer to these as "thermal batteries" because they allow you to use energy at the time of your choosing, and use it later. This will be the first time I've had the option to choose how much electricity I'm using, and when. (Other than setting my electric car's charging time and level.) Both Reservoir and Cala come with adaptive control algorithms that learn your usage patterns. This is potentially *very useful* for a heat-pump water heater, because heat pumps heat water slowly but efficiently. Predictive/adaptive planning can minimize relying on backup electric resistance heating. A grand(iose) idea: getting Embedded Common Lisp (ECL) running on a Raspbery Pi running Home Assistant to predict need for hot water, car charging, and home heating. I'm imagining fitting single-hidden-layer quantile neural nets (to predict the median and, say, 10% and 90% quantiles, useful for planning purposes) using SPSA: https://www.jhuapl.edu/spsa/ I imagine a modest processor could chew on the fitting for a while, using little memory, and come up with something pretty good. It would be awesome if a small processor running Home Assistant could do this on its own. Totally scratching an itch rather than fulfilling a real need, except that (I think) most existing algorithms are proprietary. E.g., Kraken: https://www.kraken.tech/residential-flexibility #algorithms #AI #homeassistant #thermalbatteries