The sealed prediction: how much of its own power does a neighbourhood that generates 140% actually use?

Image above: AI impression, not a construction drawing.

In Olst, in the east of the Netherlands, stand the Aardehuizen: an eco-neighbourhood of 23 homes and a shared house, with no gas connection. Their solar panels produce about 150,000 kWh a year. They use about 115,000. On paper they generate 130 to 140 percent of what they need.

Since 28 June 2025 they also have a 120 kW neighbourhood battery, built in the European research project SERENE with Saxion University of Applied Sciences and the University of Twente. The first full year with that battery is over.

How much of their own power did they really use that year? I do not know. I have not seen their data. So on 24 September I recorded what my model says, before anyone could look back.

My prediction for 1 July 2025 to 30 June 2026:

• Self-sufficiency 59%, with a range of 52 to 62%. More than four in every ten kWh the neighbourhood uses come from the grid.
• Self-consumption 47%. More than half of what they generate goes to the grid.
• May to August: close to 100% own power. December and January: around 25%.

In July the neighbourhood has three times the power it uses. In December a quarter. In summer the battery bridges every night. In January there is too little sun during the day to fill it. An annual balance averages those two months away. That is how a neighbourhood can generate 140% and be 59% self-sufficient at the same time. It is the gap I call the winter wall.

Only with what is public: annual generation, annual consumption, battery power, number of buildings, heat pumps with wood-fired backup. Plus the real weather of that year, hour by hour, from the Dutch weather service station at Heino, twelve kilometres from Olst.

The model runs 8,760 hours. In every hour the energy balance closes: what goes in comes out, is stored, or is lost. I calibrated two numbers: generation in an average weather year and annual consumption. Everything below that is prediction. The split across months, self-consumption, imports, exports, what the battery does.

One number I do not know, and it matters: how many kWh the battery can store. Its power is public, its capacity is not. At 120 kWh I get 52%, at 240 kWh 62%. That is the range. Panel orientation and the heat pump share move the result by less than one percentage point.

My model is dumb, on purpose. It uses solar power in the house first, then charges the battery, and sends the rest to the grid. No forecasting, no smart control.

The Aardehuizen do more. They have an energy management system built by Saxion, controlled boilers in four homes, and residents who shift cooking, washing and charging to sunny hours. Saxion reported in February that this behaviour really changed.

My bet: the measurement comes out above my range. That difference would then be the first measurement in a real neighbourhood of what control and behaviour are worth. In July I estimated control at a few percentage points at most, and had to retract my own favourite claim along the way. Now someone else can correct me. If the measurement comes out lower, the cause is probably winter: more heat pump load than I assume, or a battery that does not serve every building.

The full prediction, per month and in five variants, sits in one file with this fingerprint (SHA-256):

`499ed99a43b8752f82a8dbd5ebd126cfde01a05ca1611d5c594e10011f82b71c`

If a single number in that file changes after 24 September, the fingerprint changes. I will publish the file together with the comparison. Check it yourself then.

I am asking the Aardehuizen and the SERENE team whether they want to put their measurement next to it. I have not found anyone in the Netherlands who has done this for a neighbourhood like theirs: record the prediction first, then publish the measurement next to it. If I am wrong, you will read it here.

The weakest points, for anyone who wants to attack this:

• Battery capacity. Unknown, hence a range. It is the largest uncertainty in the prediction and the first question I will ask.
• Who the battery serves. I calculate as if all 24 buildings sit behind the same battery. If homes have their own meters and the battery only feeds part of the neighbourhood, the summer value drops and the model is structurally wrong.
• The demand profile. I use the Dutch standard household profile plus a heat layer. Residents of an eco-neighbourhood probably live differently from the average Dutch household. That is exactly what the measurement will show.
• Heat. Heat pumps with wood-fired backup, estimated at 1,200 kWh of electricity per building per year. That number drives the winter.
• Panel capacity. Derived from annual generation (142 kWp), not from an installation list. 2025 was a sunny year; generation in the measured period comes from the weather itself.

This piece builds on the winter wall. The cost of the winter wall is on the explorer.