A novel method for Indoor Air Quality Control of Smart Homes using a Machine learning model. (November 2022)

• Record Type: Journal Article
• Authors: Majdi, Ali; Alrubaie, Ali Jawad; Al-Wardy, Alia Haider; Baili, Jamel; Panchal, Hitesh
• Is Part Of: Advances in engineering software. Volume 173(2022)
• Publication Date: 2022-11
• Subjects: Indoor Air Quality -- Artificial intelligence -- Neural network -- Radial base function -- Forecasting; Computer-aided engineering -- Periodicals; Engineering -- Computer programs -- Periodicals; Engineering -- Software -- Periodicals; Periodicals; 620.0028553
• DOI: 10.1016/j.advengsoft.2022.103253
• Languages: English
• ISSNs: 0965-9978

Abstract

Highlights: This article presents a model for predicting indoor air quality, considering the importance of smart buildings and new solutions for their smart management through machine learning. Around 3% error obtained due to the machine learning model in this work. Input data and qualities are used as network output for training, including 1104 samples from 138 days from the beginning of May to mid. Abstract: Smart buildings are one of the areas of interest for researchers today and are of special importance in large buildings. Air quality control in buildings is one of the crucial issues in this field that is directly related to the health and efficiency of people inside the building. This article presents a model for predicting indoor air quality, considering the importance of smart buildings and the need to provide new solutions for their smart management. Kian Center 2 commercial office center in Mashhad, Iran, with fifteen air conditioners used in current research article. Data are collected from this project's control and monitoring system on different days and hours, and air quality is performed using a neural network of the radial base function. The neural network of the radial base function has three inputs: temperature, air humidity, and carbon dioxide. The network output includes volatile organic compounds in the air. The inputs are collected from the return air sensors in the air conditioners and the amount of volatile organic compounds from the sensor located in the peripheral area of ​​the food court floor. Input data and qualities are used as network output for training, including 1104 samples from 138 days from the beginning of May to mid-August 2021 and test data including 24 samples in three days, 15, 16, and 17 August (8 h every day). The grid is tested with different radii of the Gaussian function, and the results are reported. The proposed model could learn the pattern of temperature, humidity, and carbon dioxide data on air quality and generate predictions with a 3% error.