How WINNIIO and VÖFAB (Växjö, Sweden) replaced traditional cloud-tethered BMS with an EdgeLake fabric — self-learning heating control, federated AI, and full data sovereignty.
WINNIIO and the Swedish city of Växjö (about 70,000 residents) deployed a decentralized smart-heating digital twin across VÖFAB's municipal real-estate portfolio, running on the EdgeLake fabric. The case study formalizes work the partners have been doing for years, now packaged as an LF Edge reference deployment.
The technical premise is straightforward. Real estate is roughly 40% of global energy consumption, more than 80% of the buildings that will be standing in 2050 already exist, and Sweden has committed to zero greenhouse emissions by 2045. Retrofitting existing buildings is the only path that matches the math, and traditional Building Management Systems lock that retrofit behind multi-year integration projects with proprietary vendor stacks.
EdgeLake replaces that vendor lock with a federated layer. Wireless mesh sensors and actuators feed a self-learning control loop that optimizes from the individual radiator up to the energy producer, all running on edge nodes inside each building. Data stays in place. The cloud is used for cross-portfolio analytics and backup, not as the primary data plane. Deployments measure in weeks, not multi-month integrations.
The next layer is federated learning. Each building's edge nodes train a local model on its own thermal and weather data, and an aggregator combines the local models into a portfolio-level model that gets redeployed for local inference. No raw building data crosses the network. WINNIIO is now in dialogue with Industry 5.0 and critical-infrastructure pilots applying the same pattern.