Deep Learning for Household Load Forecasting – A Novel Pooling Deep RNN

Heng Shi, Minghao Xu, Ran Li

Research output: Contribution to journalArticle

82 Citations (Scopus)
1113 Downloads (Pure)

Abstract

The key challenge for household load forecasting lies in the high volatility and uncertainty of load profiles. Traditional methods tend to avoid such uncertainty by load aggregation (to offset uncertainties), customer classification (to cluster uncertainties) and spectral analysis (to filter out uncertainties). This paper, for the first time, aims to directly learn the uncertainty by applying a new breed of machine learning algorithms – deep learning. However simply adding layers in neural networks will cap the forecasting performance due to the occurrence of overfitting. A novel pooling-based deep recurrent neural network (PDRNN) is proposed in this paper which batches a group of customers’ load profiles into a pool of inputs. Essentially the model could address the over-fitting issue by increasing data diversity and volume. This work reports the first attempts to develop a bespoke deep learning application for household load forecasting and achieved preliminary success. The developed method is implemented on Tensorflow deep learning platform and tested on 920 smart metered customers from Ireland. Compared with the state-of-art techniques in household load forecasting, the proposed method outperforms ARIMA by 19.5%, SVR by 13.1% and classical deep RNN by 6.5% in terms of RMSE.
Original languageEnglish
Pages (from-to)5271-5280
Number of pages10
JournalIEEE Transactions on Smart Grids
Volume9
Issue number5
Early online date22 Mar 2017
DOIs
Publication statusPublished - 30 Sep 2018

Fingerprint

Uncertainty analysis
Recurrent neural networks
Spectrum analysis
Learning algorithms
Learning systems
Agglomeration
Uncertainty
Deep learning
Neural networks

Cite this

Deep Learning for Household Load Forecasting – A Novel Pooling Deep RNN. / Shi, Heng; Xu, Minghao; Li, Ran.

In: IEEE Transactions on Smart Grids, Vol. 9, No. 5, 30.09.2018, p. 5271-5280.

Research output: Contribution to journalArticle

@article{3f7339a99ec541f88cedd359f16d420e,
title = "Deep Learning for Household Load Forecasting – A Novel Pooling Deep RNN",
abstract = "The key challenge for household load forecasting lies in the high volatility and uncertainty of load profiles. Traditional methods tend to avoid such uncertainty by load aggregation (to offset uncertainties), customer classification (to cluster uncertainties) and spectral analysis (to filter out uncertainties). This paper, for the first time, aims to directly learn the uncertainty by applying a new breed of machine learning algorithms – deep learning. However simply adding layers in neural networks will cap the forecasting performance due to the occurrence of overfitting. A novel pooling-based deep recurrent neural network (PDRNN) is proposed in this paper which batches a group of customers’ load profiles into a pool of inputs. Essentially the model could address the over-fitting issue by increasing data diversity and volume. This work reports the first attempts to develop a bespoke deep learning application for household load forecasting and achieved preliminary success. The developed method is implemented on Tensorflow deep learning platform and tested on 920 smart metered customers from Ireland. Compared with the state-of-art techniques in household load forecasting, the proposed method outperforms ARIMA by 19.5{\%}, SVR by 13.1{\%} and classical deep RNN by 6.5{\%} in terms of RMSE.",
author = "Heng Shi and Minghao Xu and Ran Li",
year = "2018",
month = "9",
day = "30",
doi = "10.1109/TSG.2017.2686012",
language = "English",
volume = "9",
pages = "5271--5280",
journal = "IEEE Transactions on Smart Grids",
issn = "1949-3053",
publisher = "IEEE",
number = "5",

}

TY - JOUR

T1 - Deep Learning for Household Load Forecasting – A Novel Pooling Deep RNN

AU - Shi, Heng

AU - Xu, Minghao

AU - Li, Ran

PY - 2018/9/30

Y1 - 2018/9/30

N2 - The key challenge for household load forecasting lies in the high volatility and uncertainty of load profiles. Traditional methods tend to avoid such uncertainty by load aggregation (to offset uncertainties), customer classification (to cluster uncertainties) and spectral analysis (to filter out uncertainties). This paper, for the first time, aims to directly learn the uncertainty by applying a new breed of machine learning algorithms – deep learning. However simply adding layers in neural networks will cap the forecasting performance due to the occurrence of overfitting. A novel pooling-based deep recurrent neural network (PDRNN) is proposed in this paper which batches a group of customers’ load profiles into a pool of inputs. Essentially the model could address the over-fitting issue by increasing data diversity and volume. This work reports the first attempts to develop a bespoke deep learning application for household load forecasting and achieved preliminary success. The developed method is implemented on Tensorflow deep learning platform and tested on 920 smart metered customers from Ireland. Compared with the state-of-art techniques in household load forecasting, the proposed method outperforms ARIMA by 19.5%, SVR by 13.1% and classical deep RNN by 6.5% in terms of RMSE.

AB - The key challenge for household load forecasting lies in the high volatility and uncertainty of load profiles. Traditional methods tend to avoid such uncertainty by load aggregation (to offset uncertainties), customer classification (to cluster uncertainties) and spectral analysis (to filter out uncertainties). This paper, for the first time, aims to directly learn the uncertainty by applying a new breed of machine learning algorithms – deep learning. However simply adding layers in neural networks will cap the forecasting performance due to the occurrence of overfitting. A novel pooling-based deep recurrent neural network (PDRNN) is proposed in this paper which batches a group of customers’ load profiles into a pool of inputs. Essentially the model could address the over-fitting issue by increasing data diversity and volume. This work reports the first attempts to develop a bespoke deep learning application for household load forecasting and achieved preliminary success. The developed method is implemented on Tensorflow deep learning platform and tested on 920 smart metered customers from Ireland. Compared with the state-of-art techniques in household load forecasting, the proposed method outperforms ARIMA by 19.5%, SVR by 13.1% and classical deep RNN by 6.5% in terms of RMSE.

UR - https://doi.org/10.1109/TSG.2017.2686012

U2 - 10.1109/TSG.2017.2686012

DO - 10.1109/TSG.2017.2686012

M3 - Article

VL - 9

SP - 5271

EP - 5280

JO - IEEE Transactions on Smart Grids

JF - IEEE Transactions on Smart Grids

SN - 1949-3053

IS - 5

ER -