Gray-Box Modeling of an Energy Recovery Ventilator Using Streamed Data

<p dir="ltr">This paper presents the development of a complex emulator to support the energy prediction and controls optimization for an energy recovery ventilator. The emulator comprises gray-box models of several components, namely the enthalpy recovery wheel, heating and cooling coils, and supply and exhaust fans. For each constituent model, a physics-based model was created and tuned with streamed data from a test building. The emulator was tuned using data from the primary energy recovery ventilator serving as the primary ventilation system for a multi-use academic building and tested using the two parallel units of the same type to test its validity and the generalizability of the tuned parameters. This paper presents the full process for creating this model, including the building automation system data streaming method, pre-processing algorithms implemented to permit online learning, physics-based model development, and parameter optimization to tune unknown coefficients using the streamed data. Results demonstrate that the streaming and tuning method was effective to rapidly tune the emulator and that minimal effort was required to re-tune for previously-unseen pieces of equipment. Further, because of the modular nature of the emulator, its constituent models have significant potential to support gray-box modeling of fan coil units and air handling units, thus providing significant value for creating digital twins of air-side HVAC systems.</p>