Prediction of power consumption is an integral part of the operation and planning of the electricity supply company. In terms of power supply and demand, For the stable supply of electricity, the reserve power must be prepared. However, it is necessary to predict electricity demand because electricity is difficult to store. In this paper, we propose a CNN-LSTM hybrid network that can extract spatio-temporal information to effectively predict the house power consumption. Experiments have shown that CNN-LSTM hybrid networks, which linearly combine convolutional neural network (CNN), long short-term memory (LSTM) and deep neural network (DNN), can extract irregular features of electric power consumption. The CNN layer is used to reduce the spectrum of spatial information, the LSTM layer is suitable for modeling temporal information, the DNN layer generates a predicted time series. The CNN-LSTM hybrid approach almost completely predicts power consumption. Finally, the CNN-LSTM hybrid method achieves higher root mean square error (RMSE) than traditional predictive methods for the individual household power consumption data sets provided by the UCI repository.
|Title of host publication||Intelligent Data Engineering and Automated Learning – IDEAL 2018 - 19th International Conference, Proceedings|
|Editors||Hujun Yin, Paulo Novais, David Camacho, Antonio J. Tallón-Ballesteros|
|Number of pages||10|
|Publication status||Published - 2018|
|Event||19th International Conference on Intelligent Data Engineering and Automated Learning, IDEAL 2018 - Madrid, Spain|
Duration: 2018 Nov 21 → 2018 Nov 23
|Name||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Other||19th International Conference on Intelligent Data Engineering and Automated Learning, IDEAL 2018|
|Period||18/11/21 → 18/11/23|
Bibliographical noteFunding Information:
This research was supported by Korea Electric Power Corporation. (Grant number: R18XA05).
This research was supported by Korea Electric Power Corporation. (Grant
© 2018, Springer Nature Switzerland AG.
All Science Journal Classification (ASJC) codes
- Theoretical Computer Science
- Computer Science(all)