A Review on the Utilization of Reinforcement Learning and Artificial Intelligence Techniques for Buildings Heating, Ventilation, and Air Conditioning Automation System: building industry, #0

A Review of the Utilization of reinforcement learning and Artificial Intelligence techniques for Buildings Heating, Ventilation, and Air Conditioning Automation System

Ahmed M. Paridie

Corresponding author’s email 

[email protected]

Keywords

Artificial intelligence, Building industry, Energy modeling, HVAC

Abstract

Artificial intelligence (AI) has been widely used in 20th century to find optimized solutions for real-time problems in different disciplines. Since buildings consume around 40% of direct energy consumption in United States based on United States Green Building Counsel (USGBC) reports, a review is done on the opportunities where AI, especially reinforcement learning, is utilized to reduce the energy consumption of the heating ventilation and air conditioning (HVAC) system used in building industry.

This discussion starts with a review on the commonly AI algorithms used in the control sequences of HVAC systems. Since most (not all) of AI algorithms need information about the environment being studied, an additional review is done on the methods used to collect simulated information that represent the HVAC environment of new buildings and the methods used to obtain data for existing buildings. Next, the architectures of recent AI algorithms are further discussed, and the methodologies used to interface the AI algorithm with a building HVAC system model are explained for different case studies. Finally, real-time applications where AI is used as an assistive algorithm to enhance energy savings are reviewed and the gaps that prevent AI from being widely used as a stand-alone control system for HVAC systems are discussed.

Review Highlights

Using artificial intelligence to enhance energy savings for heating, ventilation, and air conditioning systems.

The evolution of energy modeling software used to model real-time buildings to perform energy analysis and enhance energy savings.

Graphical abstract

Specifications table

Nomenclature

BAS  Building Automation System

AI   Artificial Intelligence

NN  Neural Networks

RL   Reinforcement Learning

HVAC  Heating, Ventilation, and Air Conditioning Automation

VAV  Variable air volume

RBC  Rule-based controllers

USGBC United States Green Building Counsel

GUI  Graphical user interface

1  Introduction

HVAC consumes 40 to 60% of the total energy use of a building, especially commercial buildings (Wetter et al., 2021). To reduce building energy consumption of HVAC, different customized improvements have been done by engineers like energy modeling and improved building automation system (BAS) control sequences (For example, ASHRAE 36). However, a building is considered a cyber-physical system where human, and weather physical systems interact with the different cyber building systems like heating ventilation and air conditioning (HVAC), lighting and plumbing systems. The loads that human and weather impose on a building can’t be fully predicted, it is challenging to find a generic control sequence that can achieve minimal energy consumption for different buildings that have different weather conditions and occupancy patterns.

With the rapid development of computational power in 20th century, artificial intelligence (AI) gained a recognizable reputation of numerous successes in different business and industrial fields (Yuan et al., 2021). For example, it is used in business analytics to predict stock prices and in automotive industry to improve the engine efficiency, implement autonomous driving vehicles and reduce simulations computational power (Paridie et al., 2022). In this paper, a review is done on