Eventos Anais de eventos
COBEM 2019
25th International Congress of Mechanical Engineering
MODELING THE THERMAL PERFORMANCE OF A WINDOW TYPE AIR-CONDITIONING SYSTEM WITH ARTIFICIAL NEURAL NETWORKS
Submission Author:
Diogo Lôndero Da Silva , SC
Co-Authors:
João Victor Fabri, Pablo Andretta Jaskowiak, Diogo Lôndero Da Silva
Presenter: João Victor Fabri
doi://10.26678/ABCM.COBEM2019.COB2019-0789
Abstract
Mathematical simulations of air-conditioning systems, based on physical principles, usually produce high precision results. Such simulations, however, tend to be time consuming and/or computationally expensive, turning out to be prohibitive in applications where a fast response is needed. A promising alternative to tackle this issue is the use of Artificial Neural Networks (ANNs), data driven, bio-inspired models that mimic human cognition to solve specific problems. A remarkable characteristic of ANNs is their ability to generalize, that is, learn based on a set of a priory labeled data and produce predictions for new (unseen) data quickly. In this work we employ ANNs to model the steady-state operation ofa window type air-conditioning system. More specifically, we evaluate different configurations of multilayer feedforward neural networks. Our results suggest that several configurations are capable of modeling the system with considerable accuracy (R2 up to 0.97) in relation to the experimental results, considering only non-invasive measurements of temper-ature, humidity, and air flow levels from the device. Given that such measurements are readily available at a low cost,these models are appealing for applications where low latency is desired, e.g., Internet of Things (IoT) platforms, Smart Buildings, and Control Systems.
Keywords
HVAC Systems, Air-Conditioning Systems, Artificial neural networks, machine learning
