| Title |
Particle Filter for Estimation of Uncertain Variables in Building Simulation Models |
| Authors |
Kim, Deuk Woo ; Park, Cheol Soo ; Kim, Yu Min |
| DOI |
https://doi.org/10.5659/JAIK_SC.2017.33.10.79 |
| Keywords |
Kalman filtering ; Non-linear filtering ; Particle filter ; Parameter estimation |
| Abstract |
Kalman filtering has been used to estimate state variables of a building simulation model such as room air or surface temperatures. However, the Kalman filter can be used only for a linear model and unfortunately most of the building models are strongly non-linear. For example, the convective heat transfer coefficient is strongly influenced by state variables such as room or surface temperatures. In this study, the authors introduce the particle filter (PF), one of the representative nonlinear filters. The particle filter can estimate state variables and uncertain model parameters simultaneously by using the state augmentation technique. For a virtual experiment, a single zone model was developed based on the heat balance method. A series of experiments were conducted to test PF’s estimation performance with regard to three uncertain model parameters (the heat capacity of the room air, indoor convective heat transfer coefficient, and the thermal resistance of a wall) and unknown internal heat gain. It is concluded that PF can successfully estimate the aforementioned four parameters. In addition, the more particles are used, the more accurate estimation can be achieved. |