Open AccessAccurate electricity consumption prediction using enhanced long short‐term memory
Author(s) -
Chinnaraji Ragupathi,
Ragupathy Prakash
Publication year - 2022
Publication title -
iet communications
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.355
H-Index - 62
eISSN - 1751-8636
pISSN - 1751-8628
DOI - 10.1049/cmu2.12384
Subject(s) - mean squared error , benchmark (surveying) , electricity , computer science , approximation error , term (time) , exponential function , mean absolute percentage error , energy consumption , consumption (sociology) , deep learning , artificial intelligence , artificial neural network , statistics , algorithm , mathematics , engineering , mathematical analysis , physics , geodesy , quantum mechanics , electrical engineering , geography , social science , sociology
In the present era, the exponential growth in the human population and technological advancements has dramatically increased the power demand. As electricity is being used at the same time as it is produced at the power plant, effective forecasting of energy usage is crucial for maintaining a reliable power supply. In this work, a novel deep learning model named enhanced long short‐term memory (E‐LSTM) is proposed to accurately predict electricity consumption in advance as this deep learning model will accurately predict the electricity consumption by adjusting the number of hidden layers and optimizing the hyper‐parameters. Until attaining high accuracy, hyper‐parameters like learning rate (LR) and epochs along with hidden layers are adjusted to analyze the relative performance of the model. Also, rectified linear activation function alleviates the vanishing gradient problem of the model. Also, this work evaluated the seven LR policies using crucial policy (CP) benchmark on keras framework. Experiments using various evaluation criteria demonstrate that the developed E‐LSTM model is efficient. When comparing some of the other state‐of‐the‐art forecasting techniques, it has the lowest values of mean square error (MSE), mean absolute error (MAE), and root mean square error (RMSE) over the UC Irvine (UCI) residential building dataset.