200 trips in my XC60: here's what I wanted to know, and what the Aximote API let me answer
I had three questions after three months with my XC60 T8 PHEV.
First: is motorway driving actually cheaper than city driving?
That's where I started, and on electric power alone the answer surprised me: 21.5 kWh/100km on motorways vs 22.8 kWh/100km in town.
The steady highway run, the one that feels thirsty in a petrol car, is the more efficient one when the battery is involved.
But that's only half the answer, because a PHEV has two energy sources. The real number changes dramatically once petrol enters the equation. Which is where the second question took me.
Second: what's this actually costing me (both fuel types)?
A PHEV has two energy sources, and they're not equal.
On pure electric at home rates (€0.34/kWh, actual Dutch contract), a trip runs about €0.07–0.08/km.
But the moment the petrol engine kicks in, the same distance costs ~€0.50/km.
That's roughly 6× more.
The API gives you trip-level consumption for both electricity and petrol.
So, I wired it to real prices: my actual €/kWh from home charging, my actual €/L from pump receipts, split per trip, divided by distance.
No averages, no estimates.
The €/km number doesn't tell the whole story until you see the fuel mix behind it & that's where the car tells you what you're really spending.
Third: can I predict the next trip before I start?
Built a forecasting model on the trip history. This was the most fun one.
It guesses my consumption before I leave the driveway.
Is it always right? No.
Is it right enough to be useful? Yes
and it will keep getting better. with each trip made.
This is just the surface of the rabbithole, LOL. Then i had more questions.
Along the way my curiosity pulled me further:
- What road types am I driving on?
- Can the car learn my route families without me labelling anything?
- What's the efficiency difference on a road-by-road level?
A weekend project, then another, then another.
The GPX export gives speed and bearing at every trackpoint. Which turns out to be enough to start measuring how smooth a drive really is. That's a later build or maybe the dev-gods can make my life easy by adding G-force data to the GPX export someday. wink Hint hint.
I should also attribute the 2 specific git repos that help me build this:
google-research/timesfm
google-research/tabfm
They really helped fill the gaps. I am not a developer, just someone with a hermes agent, time and curiosity.
If I can wire this together, anyone with curiosity and an API key can.
All of this runs on the Aximote API. I pay for the Pro licence, and honestly, it's worth every cent.
The endpoints carry more detail than the official app shows now(it could show more, you never know), and the team has been responsive whenever I've hit a question. Just good data and the right questions.
So here is me peeping out of my rabbithole, asking:
What questions did your car make you ask?
Anyone else pulling Aximote data into their own stack, and what did you find?