D. Topalović and D. Gabrijelčič,  published by IEEE in 2024 47th MIPRO ICT and Electronics Convention (MIPRO), Opatija, Croatia, 20-24 May 2024, pp. 1931-1936, doi: 10.1109/MIPRO60963.2024.10569381.

Abstract:
Integrating renewable energy sources into power grids introduces challenges due to the decentralization and variability of power generation. Demand-side flexibility (DSF) is one solution for optimizing power consumption. Buildings in particular offer significant DSF potential due to their large thermal mass and controllable HVAC (Heating, ventilation, and air conditioning) systems. Maximizing DSF benefits requires accurate energy consumption and heat demand prediction. Therefore, the development of robust thermal models for consumer/prosumer households that adhere to international energy standards is needed. Thermal models are based on Ordinary Differential Equations (ODE) and explain the thermal behavior in view of the household’s physical parameters, e.g. floor area or thermal capacity. Since measuring these parameters is often impractical, this paper introduces a novel approach for household’s parameters identification. Our methodology involves adapting the model’s ODE for air temperature observations and enhancing parameter estimation through a comprehensive synthetic dataset. We then classify households into parameter ranges based on collected data, facilitating Neural ODEs training to fit measured temperatures to the ODE for parameter inference. The major contribution of our work is in providing a scalable solution that eliminates the need for individual parameter measurements, enhancing the feasibility of implementing DSF strategies in a broader context.