Decentralized identification of building models

As energy efficient model-based controllers gain acceptance in the building domain, the problem of identifying accurate building models for control becomes even more important. Current literature provides analysis on the identifiability of building models. However, in practice, building models can be too large and too complex to properly identify parameter estimates. Towards that end, we present a strategy to decompose large building models into smaller building zone models so that these zone models can be individually identified. We find that the identifiability of zone models imply the identifiability of the building model, and we outline a decentralized approach to identify large building models. Finally, we demonstrate this approach with a simulated building example.