# Simultaneous identification of linear building dynamic model and disturbance using sparsity-promoting optimization

arXiv: https://arxiv.org/abs/1711.06386 · HTML: https://mlsanity.com/paper/1711.06386/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 17 Nov 2017.

## Built on by (most cited first, 1 of 1)

- [Physics-informed linear regression is competitive with two Machine Learning methods in residential building MPC](https://mlsanity.com/paper/2110.15911/) (2021)

## Abstract (arXiv)

We propose a method that simultaneously identifies a linear time-invariant model of a building's temperature dynamics and a transformed version of the unmeasured disturbance affecting the building. Our method uses l1-regularization to encourage the identified disturbance to be approximately sparse, which is motivated by the slowly-varying nature of occupancy that determines the disturbance. The proposed method involves solving a convex optimization problem that guarantees the identified black-box model possesses known properties of the plant, especially input-output stability and positive DC gains. These features enable one to use the method as part of a self-learning control system in which the model of the building is updated periodically without requiring human intervention. Results from the application of the method on data from a simulated and real building are provided.