Draganа KNEŽEVIĆ1*, Marija BLAGOJEVIĆ2, Aleksandar RANKOVIĆ2
1 Western Serbia Academy of Applied Studies, Užice 31000, Trg Svetog Save 34, Serbia
firstname.lastname@example.org (*Corresponding author)
2 University of Kragujevac, Faculty of Technical Sciences Čačak, 32102 Čačak, Svetog Save 65, Serbia
Abstract: Continuous population growth is causing an increasing electricity demand. In order to provide enough electricity, it should be possible to predict the prospective consumption. This is especially important nowadays, when energy-saving measures aimed at improving the energy efficiency of all energy sources, especially electrical ones, are gaining importance. Neural networks play an important role in predicting electricity consumption. This paper aims to provide the neural network architecture that will facilitate the prediction of the monthly consumption of different types of consumers with a minimum error. The proposed model is based on two uncommon types of layers, and its reliability is tested on a real dataset related to the electricity consumption of all consumers on the territory of the City of Užice in Serbia. To ensure that more precise results are obtained, this paper also sets forth another approach involving the dataset partitioning into meaningful units (subclusters) before applying the proposed model to them. Finally, the architecture of the Electricity Consumption Prediction System (ECPS) is presented, as an interactive GUI intended for the end user. The dataset employed for training the implemented models contains the consumption data collected over a period of three years, whereas the test set contains data from the fourth year, which corresponds to the actual conditions in which the application will be used.
Keywords: Neural network, Electricity consumption, Prediction, Custom neural network model, Electricity Consumption Prediction System (ECPS).
>>FULL TEXT: PDF
CITE THIS PAPER AS:
Draganа KNEŽEVIĆ, Marija BLAGOJEVIĆ, Aleksandar RANKOVIĆ, Еlectricity Consumption Prediction Model for Improving Energy Efficiency Based on Artificial Neural Networks, Studies in Informatics and Control, ISSN 1220-1766, vol. 32(1), pp. 69-79, 2023. https://doi.org/10.24846/v32i1y202307