JIITA, vol.6 no.4, p.631-638, 2022, DOI: 10.22664/ISITA.2021.6.4.631
Sang C. Suh 1,*), Md. Abdur Rahman 2,*)
1) Dept. of Computer Science, Texas A&M University-Commerce, Texas, USA
2) Dept. of Mathematics, Jahangirnagar University, Dhaka, Bangladesh
Abstract: Energy consumption prediction is becoming popular research topic as many countries want to know the requirements of power consumption so that they
can generate sufficient power to provide uninterrupted electricity. The aim of our
work is to develop an effective predictive model for a building located in Clamart, France and then do performance analysis with the existing deep learning models.
For this purpose, we develop a long-short-term memory (LSTM) neural network model which uses acute parameters as well as monitored data with different time resolutions to determine the levels of accuracy for prediction. The model used a dataset to train and test for getting the best accuracy. The results showed that the developed LSTM model is more applicable to predict energy consumption using 1-min resolution dataset rather than other time resolutions. The analysis of the results using the dataset of 1-minute resolution showed that the proposed model outperformed other existing predictive deep learning models.
Keywords: Energy consumption, Energy prediction, Time series, Time series predicting strategies, Deep learning, Recurrent neural network, Long short-term memory