Paper Conference

Proceedings of Building Simulation 2017: 15th Conference of IBPSA

     

Sequential Monte Carlo for States and Parameters Estimation in Dynamic Thermal Models

Loïc RAILLON, Christian GHIAUS
loic.raillon1@insa-lyon.fr, christian.ghiaus@insa-lyon.fr Univ Lyon, CNRS, INSA-Lyon, Université Claude Bernard Lyon 1, CETHIL UMR5008, F-69621, Villeurbanne, France

DOI: https://doi.org/10.26868/25222708.2017.020
Abstract: Experimental identification of grey box model is the key of two social needs, the energy performance measurement and the energy management. Obtaining a reliable model may be time consuming and depends on the knowledge of building characteristics available. Furthermore, on-site measurements have to be collected before starting the identification process. From these facts, this paper investigates the capabilities of a sequential Monte Carlo method to learning models. An unoccupied house under real weather conditions has been used to test the proposed method. It has been shown that even if the chosen model structure is not the best, the sequentially learnt model provides satisfactory simulation results. Indeed, information on the building were used in the initialization which has prevented the algorithm to diverge from physical meaning. Afterwards, insights from the identified model may be used to improve the model with the complete data set.
Pages: 75 - 84
Paper:
BS2017_020