Comparison of Electricity Load Prediction Errors Between Long Short-Term Memory Architecture and Artificial Neural Network on Smart Meter Consumer

No Thumbnail Available
Salleh N.S.M.
Suliman A.
J�rgensen B.N.
Journal Title
Journal ISSN
Volume Title
Springer Science and Business Media Deutschland GmbH
Research Projects
Organizational Units
Journal Issue
Machine learning can perform electricity load prediction on the demand side. This paper compared the electricity prediction errors between two machine learning algorithms: Artificial Neural Network (ANN) and Long Short-Term Memory (LSTM) architecture. LSTM can solve the regression problem in time-series. Due to that, this paper applied LSTM. The traditional machine learning approach, ANN, was used to compare the effectiveness of LSTM in performing the time-series prediction. A dataset that consisted of historical electricity consumption data with independent variables was used in this study. The mean squared error (MSE) and mean absolute error (MAE) evaluation metrics were used to evaluate the models. The model generated using LSTM showed the lowest error with MSE value of 0.1238 and MAE value of 0.0388. These results indicated that choosing a suitable machine learning algorithm for the time-series problem could improve the model generated from the training session. � 2021, Springer Nature Switzerland AG.
Brain; Errors; Forecasting; Learning algorithms; Mean square error; Memory architecture; Network architecture; Smart meters; Time series; Demand-side; Electricity load; Error values; Load predictions; Machine learning algorithms; Mean absolute error; Mean squared error; Prediction errors; Regression problem; Times series; Long short-term memory