Adapting Machine Learning Models for Indoor Temperature Prediction from Kuwait to the French Riviera Climate

¹ Department of Mechanical Engineering, College of Engineering, Australian University, West Mishref, Safat 13015, Kuwait
² Université Côte d’Azur, Polytech’Lab, France
³ Department of Mechanical Engineering, Najafabad Branch, Islamic Azad University, Najafabad, Iran
⁴ Civil and Environmental Engineering Department, California State University, Fullerton, California, USA

^* Corresponding author: a.sedaghat@au.edu.kw

Abstract

This study investigates how machine learning models that were first created to forecast indoor temperatures in Kuwaiti portable cabins may be modified to replicate indoor temperature conditions in Nice, France. Two thermally identical portable cabins were built in Kuwait. Indoor weather conditions and energy consumption were measured, and the data was stored using Internet of Things (IoT). A transient system simulation (TRNSYS) model and several machine learning (ML) models were developed and validated against experimental data over the full years of 2023 and 2024. A total of nineteen regression (ML) models are examined. The Matern 5/2 Gaussian process regression model has done somewhat better at modeling interior temperature; all models are shown to function well. In this work, the predicted indoor temperatures are presented and discussed for Kuwait and Nice, highlighting the similarity of temperature profiles across these two distinct climate regions. The research aims to assess the effectiveness and accuracy of these models across different climatic conditions, contributing to the development of energy-efficient solutions for smart buildings.