Integrated optimization of load and heat exchanger models for improved building energy prediction

초록

Conventional sequential calibration of building energy models often yields physically inconsistent parameters and converges to local minima, hindering optimal control performance, for instance, when applying model predictive control (MPC). To address this issue, this study developed an integrated multiobjective optimization framework. The method simultaneously calibrates the load (resistance-capacitance) and heat exchanger models of a building, using their physical interdependencies as constraints via the following three coupled objective functions: indoor temperature, supply air temperature, and the rate of temperature variation. The integrated approach was validated using operational data and demonstrated superior stability and robustness. Its indoor temperature predictions remained stable across five test cases (RMSE 0.82-1.74 degrees C, R2 0.78-0.88), whereas the sequential method was highly sensitive to parameter bounds (RMSE 0.79-2.47 degrees C, R2 0.22-0.89). As regards the heating rate prediction of the air handling unit, the prediction errors were approximately one-tenth that of the conventional method. The proposed integrated framework provides a more reliable and efficient pathway for developing physically consistent MPC models and mitigates the need for laborious iterative tuning.

키워드