Use · Vehicle-to-home & vehicle-to-grid

HomeGrid V2G

A parked EV holds 3–5 days of household electricity. This tool forecasts the home's hourly demand, learns the driver's habits to predict when the car will leave and how much energy the next trip needs, then plans every hour of charging and discharging against dynamic prices. It also answers the honest question: when does bidirectional charging actually pay?

Simulated household year · forecasts evaluated on 75 unseen days

Scenario

Plug-in evening

Battery capacity Bidirectional charger Rooftop PV Battery wear cost Charge at departure ≥ Bidirectional mode

Tonight's plan

House: demand forecast vs. what actually happened

€/kWh on the right axis

Over a year

When does V2G pay?

extra €/year over smart charging

Why it matters

If millions of EVs charge "as soon as they're plugged in", they all land on the evening peak, when the grid is already stressed. Smart and bidirectional charging can turn that load into flexibility. But owners only take part when the planner respects their next trip, and when the money is honest once losses and battery wear are counted.

Method

Load forecast: ridge regression on hour × weekend profile, heating/cooling degree-hours from a noisy weather forecast, and lagged demand.
Driver model: k-nearest-neighbour days (weekday, season). It plans for an early departure (10th percentile) and a high energy need (85th percentile).
Optimiser: exact dynamic programming over a 1 kWh SoC grid. Each plan is then scored on the actual load and departure.

Limits & next steps

Single synthetic household and price model. Next: real smart-meter data and market prices.
Wear is a flat €/kWh; a cycle-depth-aware model would be more accurate.
Next: aggregate many homes into a virtual power plant and bid flexibility.