Search papers, labs, and topics across Lattice.
This study introduces a personalized federated sparse adaptation framework for time-series foundation models (TSFMs) tailored to building energy forecasting, addressing the challenges of private and non-IID meter data. By employing a heterogeneous temporal mixture-of-experts (MoE) adapter, the method allows for specialized adaptation that balances building-specific behaviors with cross-building knowledge transfer. Results demonstrate that personalized adaptation consistently outperforms both global and local training approaches across 50 buildings and three different TSFM backbones, highlighting the importance of client and backbone awareness in federated learning settings.
Personalization in federated learning can significantly enhance energy forecasting accuracy by leveraging specialized expert knowledge tailored to individual building behaviors.
Federated adaptation of time-series foundation models (TSFMs) is attractive for building energy forecasting because meter data are private, distributed, and highly non-IID. However, a single parameter-sharing strategy is unlikely to serve all pretrained TSFMs or building clients: fully shared adapters can suppress building-specific temporal behavior, while fully local adaptation discards cross-building transfer. We propose a personalized federated sparse adaptation framework with a heterogeneous temporal mixture-of-experts (MoE) adapter placed after the pretrained TSFM representation. A sequence-level router maps each 168-hour context window to a top-$k$ subset of experts specialized for periodicity, long-range interactions, local variation, trend-residual structure, and multi-resolution behavior. We compare global FL, local training, and personalized FL variants with globally shared or client-private expert banks. Across 50 buildings and three TSFM backbones, personalization consistently outperforms Global FL-MoE and Local MoE, while the best sparse-adaptation strategy varies by backbone and metric. Routing behavior further reveals client-level expert specialization, expert concentration, and near-uniform routing across backbones, showing that federated TSFM adaptation should be both client-aware and backbone-aware.
