Predicting of temperatures profiles along an Earth-Air Heat Exchanger (EAHE) using artificial neural networks

Abstract

In Brazil, buildings account for proximately 51% of electricity consumption. In commercial buildings, air conditioning systems are responsible for about 70% of this use. To collaborate to reduce this demand, this work presents an Earth-Air Heat Exchanger (EAHE) used for environment climatization. This passive system uses the soil as a heat exchanger, heating or cooling the air depending on the climatic conditions. The system, which includes 100 mm diameter Polyvinyl Chloride (PVC) ducts, and a fan for airflow control, was built at the Federal University of Technology of Paraná (UTFPR), Campus Ponta Grossa. A series of k-type thermocouples were inserted along the EAHE, the ground, and the environment. Artificial Neural Networks (ANNs) were used to obtain the temperature distribution along the exchanger to predict the performance of these heat exchangers subjected to different climatic conditions. Air temperature at the exchanger inlet, soil temperature, and airflow rate were used as input data for the model. Four different structures of Multilayer Perceptron (MLP) networks were used in this study, and all of them were capable of adequately predicting the temperatures of the thermocouples along the heat exchanger.