Predicting the Household Power Consumption Using CNN-LSTM Hybrid Networks

Tae Young Kim, Sung-Bae Cho

Research output: Chapter in Book/Report/Conference proceedingConference contribution

4 Citations (Scopus)

Abstract

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.

Original languageEnglish
Title of host publicationIntelligent Data Engineering and Automated Learning – IDEAL 2018 - 19th International Conference, Proceedings
EditorsHujun Yin, Paulo Novais, David Camacho, Antonio J. Tallón-Ballesteros
PublisherSpringer Verlag
Pages481-490
Number of pages10
ISBN (Print)9783030034924
DOIs
Publication statusPublished - 2018 Jan 1
Event19th International Conference on Intelligent Data Engineering and Automated Learning, IDEAL 2018 - Madrid, Spain
Duration: 2018 Nov 212018 Nov 23

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11314 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other19th International Conference on Intelligent Data Engineering and Automated Learning, IDEAL 2018
CountrySpain
CityMadrid
Period18/11/2118/11/23

Fingerprint

Memory Term
Power Consumption
Electric power utilization
Neural Networks
Neural networks
Electricity
Network layers
Predict
Mean square error
Long short-term memory
Time series
Spatial Information
Hybrid Approach
Hybrid Method
Repository
Irregular
Planning
Linearly
Roots
Necessary

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Kim, T. Y., & Cho, S-B. (2018). Predicting the Household Power Consumption Using CNN-LSTM Hybrid Networks. In H. Yin, P. Novais, D. Camacho, & A. J. Tallón-Ballesteros (Eds.), Intelligent Data Engineering and Automated Learning – IDEAL 2018 - 19th International Conference, Proceedings (pp. 481-490). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11314 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-03493-1_50
Kim, Tae Young ; Cho, Sung-Bae. / Predicting the Household Power Consumption Using CNN-LSTM Hybrid Networks. Intelligent Data Engineering and Automated Learning – IDEAL 2018 - 19th International Conference, Proceedings. editor / Hujun Yin ; Paulo Novais ; David Camacho ; Antonio J. Tallón-Ballesteros. Springer Verlag, 2018. pp. 481-490 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{2c34be11da4e408dba8f145b4b110f35,
title = "Predicting the Household Power Consumption Using CNN-LSTM Hybrid Networks",
abstract = "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.",
author = "Kim, {Tae Young} and Sung-Bae Cho",
year = "2018",
month = "1",
day = "1",
doi = "10.1007/978-3-030-03493-1_50",
language = "English",
isbn = "9783030034924",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "481--490",
editor = "Hujun Yin and Paulo Novais and David Camacho and Tall{\'o}n-Ballesteros, {Antonio J.}",
booktitle = "Intelligent Data Engineering and Automated Learning – IDEAL 2018 - 19th International Conference, Proceedings",
address = "Germany",

}

Kim, TY & Cho, S-B 2018, Predicting the Household Power Consumption Using CNN-LSTM Hybrid Networks. in H Yin, P Novais, D Camacho & AJ Tallón-Ballesteros (eds), Intelligent Data Engineering and Automated Learning – IDEAL 2018 - 19th International Conference, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11314 LNCS, Springer Verlag, pp. 481-490, 19th International Conference on Intelligent Data Engineering and Automated Learning, IDEAL 2018, Madrid, Spain, 18/11/21. https://doi.org/10.1007/978-3-030-03493-1_50

Predicting the Household Power Consumption Using CNN-LSTM Hybrid Networks. / Kim, Tae Young; Cho, Sung-Bae.

Intelligent Data Engineering and Automated Learning – IDEAL 2018 - 19th International Conference, Proceedings. ed. / Hujun Yin; Paulo Novais; David Camacho; Antonio J. Tallón-Ballesteros. Springer Verlag, 2018. p. 481-490 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11314 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Predicting the Household Power Consumption Using CNN-LSTM Hybrid Networks

AU - Kim, Tae Young

AU - Cho, Sung-Bae

PY - 2018/1/1

Y1 - 2018/1/1

N2 - 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.

AB - 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.

UR - http://www.scopus.com/inward/record.url?scp=85057083690&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85057083690&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-03493-1_50

DO - 10.1007/978-3-030-03493-1_50

M3 - Conference contribution

SN - 9783030034924

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 481

EP - 490

BT - Intelligent Data Engineering and Automated Learning – IDEAL 2018 - 19th International Conference, Proceedings

A2 - Yin, Hujun

A2 - Novais, Paulo

A2 - Camacho, David

A2 - Tallón-Ballesteros, Antonio J.

PB - Springer Verlag

ER -

Kim TY, Cho S-B. Predicting the Household Power Consumption Using CNN-LSTM Hybrid Networks. In Yin H, Novais P, Camacho D, Tallón-Ballesteros AJ, editors, Intelligent Data Engineering and Automated Learning – IDEAL 2018 - 19th International Conference, Proceedings. Springer Verlag. 2018. p. 481-490. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-030-03493-1_50