Hourly Demand Forecasting for an Electric Vehicle Charging Network
An operator of 1,500 fast-charging points buys energy on the day-ahead market: it must commit today how many MWh it will need tomorrow, hour by hour. If it falls short it pays for energy at the imbalance price (up to 3 times higher); if it buys too much, it resells at a loss. Demand has clear patterns: hour of the day, day of the week, long weekends and summer.
Where the money comes from
What to do now
- 1Measure the current monthly cost of imbalances.
- 2Hourly forecasting model with a confidence band.
- 3Asymmetric purchasing policy based on the cost of falling short.
For your committee
One page with the decision, the impact, the risks, the steps and the KPI to track.
The print dialog opens: choose "Save as PDF".
Watch out: more volume is not always more profit
Selling more charging at peak hours, when energy is most expensive, can have a negative margin. Use the forecast also to move demand to cheap hours with dynamic pricing.
KPI for the committee
Energy cost per kWh sold (including imbalances).
Detail for your data team
Proposed technique: Time Series (Prophet / ARIMA)
Why: Demand depends fundamentally on time: growth trend, daily and weekly cycles and calendar events. A time series model breaks down those effects, projects into the future and provides a confidence band, which is what the trader needs to decide how much to buy.
Cost of getting it wrong: Underestimating forces buying at the imbalance price (very expensive); overestimating forces reselling at a loss. The confidence band makes it possible to choose the level of risk.
Data needed: Hourly consumption per station over the last 3 years.; Working calendar, public holidays, long weekends and holiday exodus periods.; Weather forecast (cold reduces range and increases charging).; Openings of new stations.