This paper presents the construction of a RC building model with a good identifiability and a good predictive ability. The identifiability refers to the convergence of a calibration algorithm towards parameters that correspond to the physical properties of the building. This property allows accurate parameterization of the models and robust calibration. The calibrated RC models can be used for predictive control and the non-calibrated RC models are appropriate for urban energy simulation platforms for the analysis of flexibility and retrofit scenarios. From the state of the art, a modified RC model based on the standard ISO 13790 is proposed and then studied on a simplified mono-zone case study. A Bayesian calibration method has highlighted a problem of correlation between the parameters. A selection of parameters and the re-parameterization of the model have been successfully proposed to reach a good identifiability of the model.

本文介绍了具有良好可识别性和良好预测能力的 RC 建筑模型的构建。可识别性是指校准算法对对应于建筑物物理特性的参数的收敛性。此属性允许模型的准确参数化和稳健的校准。校准后的 RC 模型可用于预测控制，未校准的 RC 模型适用于城市能源模拟平台，用于分析灵活性和改造方案。根据现有技术，提出了基于 ISO 13790 标准的改进型 RC 模型，然后在简化的单区案例研究中进行了研究。贝叶斯校准方法突出了参数之间的相关性问题。