Scalable Probabilistic Load Forecasting via Compact Parameter Adaptation

This framework addresses scalability in customer-level power load forecasting by using a shared model with a small bank of low-dimensional adaptation components. This hybrid approach captures heterogeneous demand patterns without the high training and storage costs of independent models for every load.