Fourier-based periodicity analysis and lightweight RC calibration for heating load modeling under limited sensing

초록

Calibrating resistance-capacitance (RC) models with limited sensing is challenging because parameter identifiability depends strongly on whether the training data contain sufficient and representative thermal excitation. This study introduces a frequency-domain prescreening approach for multifamily housing, where the indoor temperature is unavailable and heating energy is the main signal. Two Fourier-based indices were defined to characterize diurnal heating: periodicity strength (PS), the relative power at the 24-h component, and periodicity balance (PB), the temporal stability derived from sliding-window variability. Households were clustered in the PS-PB space into regular, low-excitation, and irregular patterns, and a lightweight 1R1C model was calibrated using only weather and heating energy. The results on apartment data show that strong and stable periodicity is associated with more reproducible RC parameters and more stable responses in held-out periods, whereas lowexcitation and irregular patterns exhibit higher uncertainty. The framework provides a practical mechanism for screening data and selecting calibration windows for scalable RC-based heating-load modeling under limited sensing.

키워드