Optimal building community operation using data-driven mixed-integer convex model predictive control

We propose a new approach to coordinate the energy consumption of the rooftop unit (RTU) heating, cooling, ventilation, and air conditioning (HVAC) system of commercial buildings within an aggregation, e.g., building or units of a shopping center. Our approach uses a new specialized solver for neural network-constrained integer programs to optimize the energy consumption while ensuring that (i) temperature preferences are respected in each building and (ii) electric network constraints like voltage magnitude or maximum line current are satisfied in the vicinity of the aggregation. The latter is important when taking part to demand response programs as can induce a sudden drop in power consumption or if the aggregation includes electric vehicle charging stations or solar panels that can significantly alter its power consumption. The expected benefit to BrainBox AI is a readily implementable approach that can (i) optimally, efficiently, and safely be used on one of their main client types, commercial buildings, and (ii) be built on to integrate new, refined building specificities by the partner.

Faculty Supervisor:

Antoine Lesage-Landry

Student:

Partner:

BrainBox AI

Discipline:

Engineering

Sector:

Professional, scientific and technical services

University:

Polytechnique Montréal

Program: