[Paper Review] Machine Learning for Building Energy and Indoor Environment: A Perspective

Zhijian Liu, Di Wu+2 more|arXiv (Cornell University)|Dec 31, 2017

Solar Radiation and Photovoltaics55 references3 citations

TL;DR

This paper presents a perspective on applying machine learning, particularly artificial neural networks (ANN), to optimize building energy systems and predict indoor environmental quality—specifically culturable fungi concentration. The authors demonstrate improved prediction accuracy using ANN models, especially with hybrid methods like HTS, and advocate for broader adoption due to enhanced reliability and potential for energy conservation and indoor air quality improvement.

ABSTRACT

Machine learning is a promising technique for many practical applications. In this perspective, we illustrate the development and application for machine learning. It is indicated that the theories and applications of machine learning method in the field of energy conservation and indoor environment are not mature, due to the difficulty of the determination for model structure with better prediction. In order to significantly contribute to the problems, we utilize the ANN model to predict the indoor culturable fungi concentration, which achieves the better accuracy and convenience. The proposal of hybrid method is further expand the application fields of machine learning method. Further, ANN model based on HTS was successfully applied for the optimization of building energy system. We hope that this novel method could capture more attention from investigators via our introduction and perspective, due to its potential development with accuracy and reliability. However, its feasibility in other fields needs to be promoted further.

Motivation & Objective

• To address the challenge of predicting indoor environmental conditions, particularly culturable fungi concentration, with higher accuracy.
• To explore the feasibility and effectiveness of machine learning models in optimizing building energy systems.
• To promote the use of hybrid machine learning methods, such as HTS-based ANN, for improved reliability and performance in building applications.
• To highlight the untapped potential of machine learning in advancing energy conservation and indoor environmental quality.

Proposed method

• Application of artificial neural network (ANN) models to predict indoor culturable fungi concentration.
• Development and implementation of a hybrid machine learning method combining ANN with HTS (Hierarchical Tree Search) for enhanced optimization.
• Use of HTS to refine model structure and improve prediction accuracy in energy and environmental applications.
• Integration of machine learning techniques into building energy system optimization workflows.
• Evaluation of model performance through comparative analysis of prediction accuracy and computational convenience.
• Adoption of a perspective-based approach to discuss theoretical and practical challenges in applying machine learning to building energy and indoor environment systems.

Experimental results

Research questions

• RQ1 Can machine learning models, particularly ANN, effectively predict indoor culturable fungi concentration with high accuracy?
• RQ2 How does the hybrid HTS-ANN method improve model performance and reliability in building energy and indoor environment applications?
• RQ3 What are the key challenges in selecting optimal model structures for machine learning in this domain?
• RQ4 To what extent can machine learning contribute to energy conservation and improved indoor environmental quality in buildings?
• RQ5 How transferable are these machine learning approaches to other building system applications?

Key findings

• The ANN model achieved better prediction accuracy for indoor culturable fungi concentration compared to conventional methods.
• The hybrid HTS-based ANN approach demonstrated improved model structure selection and enhanced prediction reliability.
• The proposed method showed practical advantages in terms of computational convenience and performance in energy system optimization.
• Machine learning applications in building energy and indoor environment remain underdeveloped due to challenges in model structure determination.
• The study highlights the potential of machine learning to significantly contribute to energy efficiency and indoor air quality management in buildings.
• The authors emphasize the need for further research to validate the feasibility and scalability of these methods in broader building science applications.