10th International Congress on Information and Communication Technology

Authors: David Attipoe, Donatien Koulla Moulla, Sree Ganesh Thottempudi, Lateef Adesola Akinyemi, Jelil Olatunbosun Agbo-Ajala, Olufisayo Sunday Ekundayo, Ernest Mnkandla, Alain Abran
Abstract: Accurately forecasting energy consumption in smart homes is essential for optimizing energy use and supporting the integration of renewable resources. However, current research on energy consumption forecasting in smart homes face several challenges, including the size and quality of the dataset collected, selection of suitable models, patterns of energy usage, and lack of scalability of the models. This study evaluates the scalability and suitability of three advanced machine learning models—Long Short-Term Memory (LSTM), Gated Recurrent Units (GRU), and Artificial Neural Networks (ANN)—for short-term energy consumption forecasting across datasets of varying sizes. Specifically, it utilised four generated datasets representing 20, 50, 100, and 200 smart homes, each covering 365 days of energy consumption data. The analysis focuses on how each model performs as the dataset size increases, considering criteria such as root mean squared error (RMSE), mean absolute error (MAE), coefficient of determination (R2), training time, and inference speed. We examine the models’ ability to generalize from smaller datasets to larger ones and their suitability in capturing diverse consumption patterns across different household datasets. Additionally, we assess the resource and time efficiency of each model. We found in this study that ANN models offer a robust and accurate approach for energy consumption prediction in smart homes and indeed smart cities, providing insights into their suitability for real-time energy management.