A Bayesian Data Analytics Approach to Buildings' Thermal Parameter Estimation

Abstract

Modeling buildings' heat dynamics is a complex process which depends on various factors including weather, building thermal capacity, insulation preservation, and residents' behavior. Gray-box models offer an explanation of those dynamics, as expressed in a few parameters specific to built environments that can provide compelling insights into the characteristics of building artifacts. In this paper, we present a systematic study of Bayesian approaches to modeling buildings' parameters, and hence their thermal characteristics. We build a Bayesian state-space model that can adapt and incorporate buildings' thermal equations and postulate a generalized solution that can easily adapt prior knowledge regarding the parameters. We then show that a faster approximate approach using Variational Inference for parameter estimation can posit similar parameters' quantification as that of a more time-consuming Markov Chain Monte Carlo (MCMC) approach. We perform extensive evaluations on two datasets to understand the generative process and attest that the Bayesian approach is more interpretable. We further study the effects of prior selection on the model parameters and transfer learning, where we learn parameters from one season and reuse them to fit the model in other seasons. We perform extensive evaluations on controlled and real data traces to enumerate buildings' parameters within a 95% credible interval.

Trending machine learning models in cyber‐physical building environment: A survey

“…Their experiments in offices and restaurant settings showed a significant reduction in energy usage. Pathak et al (2019) focused on learning the distribution of building envelope parameters such as thermal resistance per unit (R-value), heat transfer coefficient (U-value). The authors proposed a computational efficient and precise automatic differentiation variational interface (VI) based algorithm to find the distribution of building parameters.…”

Identifying Grey-box Thermal Models with Bayesian Neural Networks

“…Compared to that work, we form clusters based on metadata, determine the level of complexity of the RC model for each building, and use transfer learning to reduce the amount of data needed for system identification. Pathak et al [32] use Bayesian learning to estimates the RC parameters for one season and then utilize the learned parameters in another season. Our work is similar to that work but we explore the possibility of adapting and reusing an RC model in buildings that have similar characteristics to the building for which this model was initially built.…”

“…We initially restrict our study to winter due to the additional complexity that mixing data from different seasons presents [32]. For the winter season, we consider homes in our dataset that have at least 3 months worth of data between November 2018 to February 2019.…”

“…However, identifying even a low-order thermal model (e.g., 2R2C or 3R3C) requires several days of data. Should this data be available, the parameters of the RC model can be estimated by solving a constrained optimization problem [12,14], using a Kalman filter [23] or through supervised learning [32]. While sufficient training data is being gathered by a smart thermostat to identify the building's thermal model, the thermostat may take suboptimal control decisions which violate the thermal comfort requirements and increase the energy use.…”

Addressing partial observability in reinforcement learning for energy management

“…Wang et al [36] used LSTM in an RL actor-critic algorithm, whereas Zhang et al [39] used LSTM in a model-based RL algorithm to learn environmental dynamics. Sequence-to-sequence models [6][7][8] and Bayesian networks [16,28] were applied to make predictions in model predictive control. Soft actor-critic (SAC) [10] is chosen in this work due to its promising results in energy management [4,21,30,40].…”