This paper introduces a transformer-based generative network for rapid parameter estimation of Modelica building models using simulation data from a Functional Mock-up Unit (FMU). Utilizing the \texttt{MixedAirCO2} model from the Modelica Buildings library, we simulate a single-zone mixed-air volume with detailed thermal and \cotwo dynamics. By varying eight physical parameters and randomizing occupancy profiles across 100 simulated systems, we generate a comprehensive dataset. The transformer encoder, informed by room temperature and \cotwo concentration outputs, predicts the underlying physical parameters with high accuracy and without re-tuning (hence, ``zero-shot''). This approach eliminates the need for iterative optimization or can be used to warm-start such optimization-based approaches, enabling real-time control, monitoring, and fault detection in FMU-based workflows.

This session is chaired by

Numerical simulation of a thermofluid vapor compression cycle (VCC) model in Modelica, for example, can exhibit a variation in the total fluid (refrigerant) mass. This paper provides a dynamic analysis of a commonly used VCC model, identifies and analyzes the root cause of this variation, and proposes a number of remedies. The cause lies within the dynamic equations that result from application of the principle of mass conservation. In many common formulations, these equations express the conservation of mass as one or more differential equations that equate the time derivative of mass to zero. The resulting set of n ordinary differential equations (and a number of auxiliary algebraic equations) include the time derivative of a mass constraint function, but not the actual mass constraint function itself. As a result, this modeling formulation has the following properties: (1) equilibrium solutions of the system are neither isolated, nor exponentially stable; (2) a linearization about any equilibrium solution has at least one eigenvalue equal to zero, making an equilibrium solution stable, but not exponentially stable; (3) for a VCC model formulated using two fluid states per control volume, a one-dimensional equilibrium manifold exists containing all of the equilibrium solutions, and is parameterized by the total fluid mass; (4) an (n-1) dimensional, stable, invariant manifold exists transverse to the equilibrium manifold, defined by the mass constraint function, and on which analytic solutions to the model evolve and the total fluid mass remains constant; and (5) numerical solutions may drift off of this manifold, resulting in an observed drift of fluid mass. These properties have consequences for simulation, control design, numerical model reduction, and state estimation. A number of methods to stabilize the mass constraint are proposed and a number of examples that illustrate the behavior, analysis and remedies are provided.