Real-time thermal dynamic analysis of a house using RC models and joint state-parameter estimation

• Published in: Building and environment Vol. 188; p. 107184
• Main Authors: Li, Yong, Castiglione, Juan, Astroza, Rodrigo, Chen, Yuxiang
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ltd 15.01.2021; Elsevier BV
• Subjects: Boundary conditions; Building thermal dynamics; Capacitance; Economic models; Energy management; Kalman filters; Mathematical models; Model updating; Numerical models; Parameter estimation; Predictions; Process parameters; RC models; Real-time online prediction; Residential areas; State vectors; State-parameter estimation; Unscented kalman filter; Building thermal dynamics; Real-time online prediction; RC models; State-parameter estimation; Unscented kalman filter
• ISSN: 0360-1323; 1873-684X
• DOI: 10.1016/j.buildenv.2020.107184

To enable optimal building energy management in response to the ever-changing building and boundary conditions, it is critical to have numerical models that can provide accurate online prediction based on economically measurable inputs and feedback. The present study explores the capabilities of using the unscented Kalman filter (UKF) in combination with resistance-capacitance (RC) models for online estimation of the thermal dynamics of single detached houses. A joint state-parameter UKF estimation approach is applied to estimate unknown state and model parameters by using fictitious process equations to augment the state vector to include model parameters. The performance of this approach is evaluated by comparing the estimated state values to the monitored data. In addition, the prediction capability of the updated model is also investigated. The estimation procedure, mathematical operations, and result analysis are presented in detail. The remarkable model performance achieved shows that the UKF can efficiently improve RC models’ predictability and enable timely online model updating and response prediction. •A novel method for state-parameter estimation of RC models is presented.•Online estimation of temperatures and parameters is provided.•Probabilistic approach also provides estimation uncertainty.•Method is validated using data collected from an instrumented single-family detached house.•Accurate and robust model predictability is obtained based on the updated model.