Online Optimization for Controlling HVAC Systems
Abstract
Heating, ventilation, and air-conditioning (HVAC) systems account for a significant portion of buildings' energy consumption, motivating the design of energy- and cost-efficient control strategies. However, conventional strategies employ simple, reactive control, without incorporating a model. Therefore, in this paper we adopt an economic model predictive control framework and focus on designing online algorithms to solve the control input design problems with a limited computational budget. In particular, we propose two tailored online algorithms based on projected gradient descent, with a custom projection routine, and the projected primal-dual method. We discuss the pros and cons of both algorithm designs, focusing in particular on computational efficiency. We further compare them through numerical experiments simulating the control of an HVAC system over the course of a day. The results show that computationally efficient online solvers for HVAC systems control are indeed possible. Moreover, the performance closely matches that of a baseline algorithm with unlimited computational budget, in terms of reference temperature tracking and comfort temperature range enforcement.
Related
- Controlling the energy jump of multistable structures using shape optimization
- Online Optimization with Unknown Time-Varying Parameters Using Noisy Gradient Measurements
- Online Feedback Optimization for Constrained Stochastic Problems with Decision-Dependent Distributions
- Dec-BFTRL: Squre-Root Regret for Decentralized Online Upper-Linearizable Optimization under Separation Access with Application to Continuous Submodular Maximization
- Fractional-Memory Distributed Online Optimization with Long-Term Constraints
BibTeX
@misc{wu2026online,
title = {{Online Optimization for Controlling HVAC Systems}},
author = {Tong Wu and Dominik Baumann and Angela Fontan and Nicola Bastianello},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.07777},
eprint = {2611.07777},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.07777},
}
RIS
TY - UNPB
TI - Online Optimization for Controlling HVAC Systems
AU - Tong Wu
AU - Dominik Baumann
AU - Angela Fontan
AU - Nicola Bastianello
PY - 2026
DA - 2026/11/05
PB - arXiv
DO - 10.48550/arxiv.2611.07777
UR - https://arxiv.org/abs/2611.07777
AB - Heating, ventilation, and air-conditioning (HVAC) systems account for a significant portion of buildings' energy consumption, motivating the design of energy- and cost-efficient control strategies. However, conventional strategies employ simple, reactive control, without incorporating a model. Therefore, in this paper we adopt an economic model predictive control framework and focus on designing online algorithms to solve the control input design problems with a limited computational budget. In particular, we propose two tailored online algorithms based on projected gradient descent, with a custom projection routine, and the projected primal-dual method. We discuss the pros and cons of both algorithm designs, focusing in particular on computational efficiency. We further compare them through numerical experiments simulating the control of an HVAC system over the course of a day. The results show that computationally efficient online solvers for HVAC systems control are indeed possible. Moreover, the performance closely matches that of a baseline algorithm with unlimited computational budget, in terms of reference temperature tracking and comfort temperature range enforcement.
ER -