Identification of Network Dynamics and Disturbance for a Multizone Building

We propose a method that simultaneously identifies a sparse transfer matrix and a disturbance signal for a multizone building's temperature dynamics from the measurements of inputs and outputs. The proposed method is based on solving a convex optimization problem whose cost function involves an ℓ
₁
-penalty to promote a sparse solution. The method ensures that the transfer matrix is sparse, so that only dominant interactions among zones are retained in the model. The disturbance, which is mostly occupant-induced, is assumed to be a piecewise-constant signal, which aids in identification, since the derivative of a piecewise-constant signal is a sparse signal. We test our method on data from a virtual building (a simulation model) and a real building. Results from the virtual building show that the proposed method can accurately identify a sparse network model and a transformed disturbance. Results from the real building data-that does not have a ground truth-show that the method produces sensible results.