Short-Term Load Forecasting with Missing Data using Dilated Recurrent Attention Networks
Poster presented at the Northern Lights Deep Learning Workshop (NLDL) 2020, 19.01.20 - 21.01.20, UiT The Arctic University of Norway, Tromsø Forecasting the dynamics of time-varying systems is essential to maintaining the sustainability of the systems. Recent studies have discovered that Recurrent N...
Main Authors: | , , , |
---|---|
Format: | Conference Object |
Language: | English |
Published: |
Septentrio Academic Publishing
2020
|
Subjects: | |
Online Access: | https://hdl.handle.net/10037/19949 |
id |
ftunivtroemsoe:oai:munin.uit.no:10037/19949 |
---|---|
record_format |
openpolar |
spelling |
ftunivtroemsoe:oai:munin.uit.no:10037/19949 2023-05-15T14:25:20+02:00 Short-Term Load Forecasting with Missing Data using Dilated Recurrent Attention Networks Choi, Changkyu Bianchi, Filippo Maria Kampffmeyer, Michael Jenssen, Robert 2020-02-06 https://hdl.handle.net/10037/19949 eng eng Septentrio Academic Publishing Choi, C., Bianchi, F.M., Kampffmeyer, M. & Jenssen, R. (2020). Short-Term Load Forecasting with Missing Data using Dilated Recurrent Attention Networks. Poster presentation at the Northern Lights Deep Learning Workshop (NLDL) 2020, 19.01.20 - 21.01.20, UiT The Arctic University of Norway, Tromsø FRIDAID 1854060 https://hdl.handle.net/10037/19949 openAccess Copyright 2020 The Author(s) VDP::Mathematics and natural science: 400::Physics: 430 VDP::Matematikk og Naturvitenskap: 400::Fysikk: 430 Conference object Konferansebidrag publishedVersion 2020 ftunivtroemsoe 2021-06-25T17:57:49Z Poster presented at the Northern Lights Deep Learning Workshop (NLDL) 2020, 19.01.20 - 21.01.20, UiT The Arctic University of Norway, Tromsø Forecasting the dynamics of time-varying systems is essential to maintaining the sustainability of the systems. Recent studies have discovered that Recurrent Neural Networks(RNN) applied in the forecasting tasks outperform conventional models that include AutoRegressive Integrated Moving Average(ARIMA). However, due to the structural limitation of vanilla RNN which holds unit-length internal connections, learning the representation of time series with missing data can be severely biased. The goal of this paper is to provide a robust RNN architecture against the bias from missing data. We propose Dilated Recurrent Attention Networks (DRAN). The proposed model has a stacked structure of multiple RNNs which layer of each having a different length of internal connections. This structure allows incorporating previous information at different time scales. DRAN updates its state by a weighted average of the layers. In order to focus more on the layer that carries reliable information against bias from missing data, it leverages attention mechanism which learns the distribution of attention weights among the layers. We report that our model outperforms conventional ones with respect to the forecast accuracy from two benchmark datasets, including a real-world electricity load dataset. Conference Object Arctic Tromsø Arctic University of Norway UiT The Arctic University of Norway University of Tromsø: Munin Open Research Archive Arctic Norway Tromsø |
institution |
Open Polar |
collection |
University of Tromsø: Munin Open Research Archive |
op_collection_id |
ftunivtroemsoe |
language |
English |
topic |
VDP::Mathematics and natural science: 400::Physics: 430 VDP::Matematikk og Naturvitenskap: 400::Fysikk: 430 |
spellingShingle |
VDP::Mathematics and natural science: 400::Physics: 430 VDP::Matematikk og Naturvitenskap: 400::Fysikk: 430 Choi, Changkyu Bianchi, Filippo Maria Kampffmeyer, Michael Jenssen, Robert Short-Term Load Forecasting with Missing Data using Dilated Recurrent Attention Networks |
topic_facet |
VDP::Mathematics and natural science: 400::Physics: 430 VDP::Matematikk og Naturvitenskap: 400::Fysikk: 430 |
description |
Poster presented at the Northern Lights Deep Learning Workshop (NLDL) 2020, 19.01.20 - 21.01.20, UiT The Arctic University of Norway, Tromsø Forecasting the dynamics of time-varying systems is essential to maintaining the sustainability of the systems. Recent studies have discovered that Recurrent Neural Networks(RNN) applied in the forecasting tasks outperform conventional models that include AutoRegressive Integrated Moving Average(ARIMA). However, due to the structural limitation of vanilla RNN which holds unit-length internal connections, learning the representation of time series with missing data can be severely biased. The goal of this paper is to provide a robust RNN architecture against the bias from missing data. We propose Dilated Recurrent Attention Networks (DRAN). The proposed model has a stacked structure of multiple RNNs which layer of each having a different length of internal connections. This structure allows incorporating previous information at different time scales. DRAN updates its state by a weighted average of the layers. In order to focus more on the layer that carries reliable information against bias from missing data, it leverages attention mechanism which learns the distribution of attention weights among the layers. We report that our model outperforms conventional ones with respect to the forecast accuracy from two benchmark datasets, including a real-world electricity load dataset. |
format |
Conference Object |
author |
Choi, Changkyu Bianchi, Filippo Maria Kampffmeyer, Michael Jenssen, Robert |
author_facet |
Choi, Changkyu Bianchi, Filippo Maria Kampffmeyer, Michael Jenssen, Robert |
author_sort |
Choi, Changkyu |
title |
Short-Term Load Forecasting with Missing Data using Dilated Recurrent Attention Networks |
title_short |
Short-Term Load Forecasting with Missing Data using Dilated Recurrent Attention Networks |
title_full |
Short-Term Load Forecasting with Missing Data using Dilated Recurrent Attention Networks |
title_fullStr |
Short-Term Load Forecasting with Missing Data using Dilated Recurrent Attention Networks |
title_full_unstemmed |
Short-Term Load Forecasting with Missing Data using Dilated Recurrent Attention Networks |
title_sort |
short-term load forecasting with missing data using dilated recurrent attention networks |
publisher |
Septentrio Academic Publishing |
publishDate |
2020 |
url |
https://hdl.handle.net/10037/19949 |
geographic |
Arctic Norway Tromsø |
geographic_facet |
Arctic Norway Tromsø |
genre |
Arctic Tromsø Arctic University of Norway UiT The Arctic University of Norway |
genre_facet |
Arctic Tromsø Arctic University of Norway UiT The Arctic University of Norway |
op_relation |
Choi, C., Bianchi, F.M., Kampffmeyer, M. & Jenssen, R. (2020). Short-Term Load Forecasting with Missing Data using Dilated Recurrent Attention Networks. Poster presentation at the Northern Lights Deep Learning Workshop (NLDL) 2020, 19.01.20 - 21.01.20, UiT The Arctic University of Norway, Tromsø FRIDAID 1854060 https://hdl.handle.net/10037/19949 |
op_rights |
openAccess Copyright 2020 The Author(s) |
_version_ |
1766297753178079232 |