A New Approach Based on Deep Learning in Electricity Consumption Forecasting with Cross Dataset Analysis
3rd International Congress of Electrical and Computer Engineering, ICECENG 2024, Bandirma, Turkey, 27 - 30 November 2024, pp.45-56, (Full Text)
- Publication Type: Conference Paper / Full Text
- Doi Number: 10.1007/978-3-031-88999-8_4
- City: Bandirma
- Country: Turkey
- Page Numbers: pp.45-56
- Keywords: Decision fusion, Deep learning, Electricity consumption forecasting, GRU, RNN
- Trakya University Affiliated: Yes
Abstract
Electricity is essential for many devices we use in daily life, and saving electrical energy is important for managing power grids used to provide electricity. Knowing the electricity consumption demand is an important issue to save electrical energy. Future estimations are made to predict electricity consumption demand. Deep learning methods may take over an important role in literature on this subject. In this study, electricity consumption estimation was performed on a different test dataset using electricity consumption data obtained from Türkiye as a training dataset. Three different deep learning models were employed, and the first two models that gave the best results were selected. As an individual model, RNN gave the best result, and its MSE value was 0.07263. As a second individual model, GRU gave the second-best result, and its MSE value was 0.07784. By combining the predictions of the two best individual models, a more successful prediction performance was achieved in various metrics. In summary, the lowest MSE value of 0.07168 was achieved with the proposed approach.