Hi all,
I’ve just published EV Charging Predictor. It learns your car’s routine from your EV charger: when it leaves, when it comes home, and how many kWh it will need tomorrow. You can then use those predictions in your own charging Flows.
App Store: EV Charging Predictor | Homey
Source: GitHub - wez3/homey-ev-charging-predictor · GitHub
Why
Smart-charging apps (dynamic prices, solar, etc.) all ask the same two questions: when do you leave? and how much do you need? Most of us type in a fixed 07:00 and “charge to 80%”. This app fills in both answers by learning from what your charger already sees.
How it works
Pairing is one step: add a device and pick your charger.
- It watches when the cable is plugged in and out, and learns a departure and arrival pattern per weekday. Recent weeks count more, and workdays learn from other workdays.
- It reads the charger’s energy meter to predict tomorrow’s charge need in kWh, with a low–high band.
- Every prediction comes with a confidence %. A car that leaves at 08:00 ± 5 min every day scores high; an irregular schedule honestly scores low.
- Fully local. It only reads your charger.
Supported chargers
- Easee: gives predictions from day one, because the app seeds itself from 14 days of Homey Insights history. It also handles Easee reporting “unplugged” when a session merely completes.
- Any charger app with Homey’s standard `evcharger_charging_state` capability: works out of the box, and learns in 1–2 weeks.
- Zaptec, Ohme, Mennekes, go-e, Smappee, evcc, GARO, KEBA, OCPP 1.6 and Wallbox: supported through brand profiles. These are built from each app’s source code and not yet tested on real hardware, so feedback from owners is very welcome!
What you get in Flows
Device capabilities:
- predicted departure and arrival time
- minutes until departure and arrival
- confidence for each prediction
- car present
- predicted charge need (kWh and a band)
Triggers:
- A new prediction was calculated
- Departure / arrival predicted within X minutes
- The predicted daily charge need changed (tokens for the expected, low and high kWh)
Conditions:
- Departure / arrival confidence is above X %
- Departure / arrival is predicted within X minutes
- Predicted daily charge need is above X kWh
Example: smart charging
My own setup uses the predictor together with a smart charging app. It dynamically sets when the car needs to be ready, instead of a static schema.
The result: car gets more hours to charge in “cheap hours”.
Feedback welcome
This is the first public release, so I’d love to hear:
- Which charger you use, and whether it paired and detects plug in/out correctly. This matters especially for non-Easee chargers.
- How accurate the predictions are after a week or two.
- Flow cards or tokens you’re missing.
- Any bugs.
If your charger isn’t supported yet, support for a new brand is a small data-only change in `lib/ChargerProfiles.js`. Pull requests are welcome.
Thanks for trying it out!
