Development of dynamic simplified thermal models of active pipe-embedded building envelopes using genetic algorithm

• Active pipe-embedded building envelope is proposed for using low-grade energy sources.
• FDFD model is developed for frequency thermal characteristic prediction.
• Simplified RC model is developed for this structure.
• Model parameters can be determined in frequency domain by using genetic algorithm.
• Results show simplified RC model can predict frequency thermal characteristic well.

Active pipe-embedded building envelope is a new building envelope, which is an external wall or roof with pipes embedded in it. This structure has the advantages to utilize directly low-grade energy sources for reducing building cooling/heating load and improving indoor thermal comfort. This structure may also use some phase change materials (PCM) to be pasted as a thin layer for further enhancing these benefits. This paper presents a dynamic simplified thermal model of this structure with the thermal network structure of lumped thermal mass, and the parameter identification of the simplified model based on frequency characteristic analysis. These resistances and capacitances are identified in frequency domain by using generic algorithm (GA) by comparing the frequency characteristics of the simplified model with the theoretical frequency characteristics of this structure obtained with Frequency-Domain Finite Difference (FDFD) method. Firstly, the FDFD model of this structure is established, and the theoretical frequency characteristics under various disturbances are calculated for the reference of parameter identification. Then, an equivalent dynamic simplified thermal model with lump thermal network structure is developed, and its frequency characteristics are also deduced and calculated. Finally, GA estimator is used to identify these parameters of the simplified model for allowing the frequency responses of the simplified model to match the theoretical frequency responses by using FDFD method. Various case studies are presented to validate the accuracy of the simplified models and the effectiveness of the parameter identification for the model.