Neighbors Selling Solar: AI Saves More When Batteries Are In Play

Imagine a street where every house has solar panels, and a few have big batteries. Instead of selling their extra power back to the grid at low rates, neighbors could sell it to each other at a fair price. But how should that price be set? A team of researchers wanted to see whether a simple rule-based approach beats an AI that learns as it goes.

First, they compared several straightforward pricing schemes—like splitting the bill evenly, charging the average of grid buy/sell prices, or setting prices based on supply and demand. To their surprise, these simple rules performed just as well as the AI when homes had only solar panels. That challenges the popular idea that advanced machine learning is always the smartest choice.

However, when batteries were added to the mix, the AI began to shine. The best AI-driven pricing raised total community savings from about €734 to €978 per year—roughly a 33% improvement. Why? Because the AI could learn to charge batteries when power was cheap and sell when prices were high, acting like a clever energy trader for the neighborhood.

The catch is that these savings weren't shared equally. Some households benefited far more than others, depending on their energy use and battery size. This fairness issue could matter for community projects, where everyone needs to feel they're getting a good deal. The study, still awaiting peer review, suggests that simple, transparent pricing is a solid starting point—and AI can add real value, but only in combination with storage, and not automatically for every home.