A novel methodology based on artificial intelligence to achieve the formost Buildings' heating system. (15th June 2023)

• Record Type: Journal Article
• Authors: Mokhtarian Mobarakeh, Jalil; Sayyaadi, Hoseyn
• Is Part Of: Energy conversion and management. Volume 286(2023)
• Publication Date: 2023-06-15
• Subjects: Artificial Intelligence -- Artificial Neural Networks -- Energy Optimization -- HVAC systems -- Multi-layer Perceptron; Direct energy conversion -- Periodicals; Energy storage -- Periodicals; Energy transfer -- Periodicals; Énergie -- Conversion directe -- Périodiques; Direct energy conversion; Periodicals; 621.3105
• DOI: 10.1016/j.enconman.2023.116958
• Languages: English
• ISSNs: 0196-8904

Abstract

Highlights: An artificial intelligence system was designed to achieve buildings' heating system. A PSO minimizes life cycle costs and CO2 emissions. SVM was proposed to predict the separator lines between heating systems. A deep learning-based predictive model was presented using MLP networks. The performance of the model satisfied the ASHRAE standard. Abstract: This study investigates the possibility of creating an artificial intelligence system in the form of a predictive model that can recommend the foremost building heating system and their optimum performances for a building based on desired life cycle costs by the project manager and the permitted CO2 emission rate. Accordingly, the developed intelligent system would be able to recommend the most appropriate heating system when the building specifications are given to the system. Therefore, experts' knowledge can be substituted to recommend the foremost heating systems for a building. In this framework, a database is produced using a Multi-objective Particle Swarm Optimization (MOPSO) for the buildings with various numbers of floors, built areas, and the number of residents. Pareto front data for each building is fitted with a curve, and the coefficients of the curves are given to a Multi-Layer Perceptron (MLP) to create a predictive model for optimum performances. Moreover, all optimization data are assigned to a Support Vector Machine (SVM) to determine the classification lines between different heating systems with the lowest risk. The coefficients of separator lines are given to another multi-layer perceptron to create the predictive model for available HVAC systems. The models' performance was improved by evaluating different structures for the Artificial Neural Networks (ANN) and implementing single-objective models instead of multi-objective ones. The proposed models' performance is acceptable considering the requirements of ASHRAE guideline 14.