Akademik Veri Yönetim Sistemi
8th International Researchers, Statisticians and Young Statisticians Congress, Adana, Türkiye, 28 - 30 Kasım 2024
Electricity consumption, which constitutes a large part of the energy need, is of vital importance for the sustainability of the modern world. In this context, being able to make accurate and reliable electricity consumption estimates plays a critical role in ensuring supply-demand balance, providing electricity at affordable costs, and determining energy policies.
With the developments in the field of deep learning, the use of advanced algorithms in time series forecasting has attracted great interest. However, it has been observed that a single deep learning model may be insufficient in learning complex structures. In order to address this situation, a hybrid deep learning model combining Gated Recurrent Unit (GRU) and Bidirectional Long Short-Term Memory (BiLSTM) models was designed in our study. While GRU offers the ability to effectively capture temporal dependencies, the BiLSTM layer provides a more comprehensive analysis opportunity by considering future data as well as past data. This hybrid structure aims to increase forecast accuracy by better modeling complex relationships in time series data.
In this study, hourly electricity consumption data collected in Turkey from 2022 to 2024 was considered. The performance of the developed hybrid deep learning algorithm was evaluated using R² and RMSE metrics to compare it with traditional deep learning models. The findings show that the hybrid deep learning algorithm is more successful than the deep learning models used alone.