ABM4EnergyTransition: Making Renovation Decisions Visible – with Agent-Based Simulation
Fluidly rendering a simulation that projects individual buildings over years in the browser – and doing so in a privacy-compliant way: that was the core GIS challenge in the research project ABM4EnergyTransition (FFG “AI for Green” call, 2022–2024). As MovingLayers, we were responsible for the web GIS demonstrator – and had to solve three spatial problems where standard mapping falls short.
The engine: Repast Simphony
The simulation engine was Repast Simphony – the “Recursive Porous Agent Simulation Toolkit”, an open-source, Java-based framework for agent-based modelling that traces back to work at Argonne National Laboratory. Decisive for a spatial model: Repast natively supports geographic projections and GIS data and can be integrated into existing infrastructure via an open API. Both properties were the precondition for connecting behavioural simulation and mapping application in the first place.
Performance: pre-computing years instead of live calculation
Repast produces separate results for each year – each “tick”. These are data volumes a web frontend cannot compute in real time. We therefore pre-computed scenarios, stored the results as GeoJSON per time step directly in PostGIS, and aggregated them within Repast so the frontend can display them without reloading.
Integration: connecting simulation and map
Initially, no interface existed between the simulation engine and the mapping application. Via the Repast API, we connected the simulation to the geodata infrastructure and embedded it in an end-to-end pipeline of PostGIS, content management and frontend – versioned and rolled out through CI/CD.
Privacy in space: two resolutions, one data basis
The same results must serve two audiences differently: spatial planners see them down to building level, while the public receives only a representation aggregated to a 250 m raster – coarse enough to rule out inferences about individual households, fine enough for the spatial message.
Data basis and result
The data basis was a representative survey around the pilot city of Kapfenberg: 2,841 valid responses (9% response rate) fed the model. The result is openly accessible at app.abm4energytransition.eu – and remains reusable as geodata in any GIS, for example in QGIS with our optional extension Kermit GeoAI.
📋 Project fact sheet — ABM4EnergyTransition
Funding: FFG, “AI for Green” (cooperative project) · Duration: 2022–2024
Consortium: University of Salzburg / Geoinformatics (lead), MovingLayers, AEE INTEC, Energie Agentur Steiermark, City of Kapfenberg, et al.
MovingLayers’ role: Web GIS demonstrator, IT infrastructure, data integration
Orchestrate spatial analyses yourself
The same approach – from spatial question to a robust result – powers Kermit GeoAI, our optional QGIS extension for orchestrating geodata analyses.
