Heat networks under uncertainty: GeoAI generative design

Stop designing heat networks around a single “best guess” future. In real neighbourhoods, who connects – and where – changes everything, and that uncertainty can stall investment.

In this session you’ll see how we used GeoAI and probabilistic simulation to turn uncertain heat network uptake into investment-grade evidence. Using a co-development case study with Vattenfall, we enhanced Ramboll’s District One automated modelling tools so they generate tens of thousands of plausible network designs, not just one. Each run samples connections using tenure-aware priors (owner-occupied, private rented, social), automatically routes the network across the road graph, sizes pipes to CP1 (2020) practice, and produces cost and risk distributions (P10/P50/P90 style) that funders and delivery teams can actually govern against.

You’ll learn how to interpret route probability overlays to identify the most robust spines you can commit to early, while keeping low-probability laterals as options. We’ll also show how to communicate results through interactive dashboards and GIS probability layers – helping stakeholders compare trade-offs, plan civils, and reduce redesigns as signups evolve.

If you work on district energy, utilities or infrastructure programmes and want a repeatable method to design under uncertainty, this session will give you a practical blueprint.