LoadSeer: Exploiting Tensor Graph Convolutional Network for Power Load Forecasting With Spatio-Temporal Characteristics

Power load forecasting plays a crucial role in ensuring the stable operation of the power system and avoiding system collapse or resource waste caused by power shortages or surpluses. However, the complex spatio-temporal property of power load makes it difficult to predict, which poses a great chall...

Full description

Saved in:
Bibliographic Details
Main Authors: Jiahao Zhang, Bin Yu, Hanbin Lai, Lin Liu, Jinghui Zhou, Fengliang Lou, Yili Ni, Yan Peng, Ziheng Yu
Format: Article
Language:English
Published: IEEE 2024-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10786973/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849220100534566912
author Jiahao Zhang
Bin Yu
Hanbin Lai
Lin Liu
Jinghui Zhou
Fengliang Lou
Yili Ni
Yan Peng
Ziheng Yu
author_facet Jiahao Zhang
Bin Yu
Hanbin Lai
Lin Liu
Jinghui Zhou
Fengliang Lou
Yili Ni
Yan Peng
Ziheng Yu
author_sort Jiahao Zhang
collection DOAJ
description Power load forecasting plays a crucial role in ensuring the stable operation of the power system and avoiding system collapse or resource waste caused by power shortages or surpluses. However, the complex spatio-temporal property of power load makes it difficult to predict, which poses a great challenge to the power system. Existing spatio-temporal prediction methods can only handle one factor in each dimension of time and space. In reality, power load is influenced by various factors. Especially in terms of time dimension, crowd flow, weather, and historical load all have significant impacts on load forecasting. Inspired by tensor time series, considering the structure of spatial geographic location, we propose Tensor Graph Convolutional Network for Power Load Forecasting, LoadSeer. A distance adjacency matrix is designed to represent the geographical location relationship and land use nature. A spatio-temporal processing layer integrating graph convolution module (GCN) and T-Transformer is mapped out to extract the spatio-temporal features, which are then sent to the fully connected layer to provide the refined expression. The experimental results on three public datasets, PeMSD4, PeMSD7, and PeMSD8 show that our proposed method outperforms baseline models on all indicators. To further validate the effectiveness of our proposed approach, we apply LoadSeer to real load data during the Asian Games in a certain city, and the results also demonstrate the superiority of our method.
format Article
id doaj-art-b1cfc8ed93c541f7bcf7b0326577bb4c
institution Kabale University
issn 2169-3536
language English
publishDate 2024-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-b1cfc8ed93c541f7bcf7b0326577bb4c2024-12-20T00:01:04ZengIEEEIEEE Access2169-35362024-01-011219033719034610.1109/ACCESS.2024.351417410786973LoadSeer: Exploiting Tensor Graph Convolutional Network for Power Load Forecasting With Spatio-Temporal CharacteristicsJiahao Zhang0Bin Yu1Hanbin Lai2Lin Liu3Jinghui Zhou4Fengliang Lou5Yili Ni6Yan Peng7https://orcid.org/0009-0005-4331-8195Ziheng Yu8State Grid Zhejiang Hangzhou Xiaoshan District Power Supply Company Ltd., Hangzhou, ChinaState Grid Zhejiang Hangzhou Xiaoshan District Power Supply Company Ltd., Hangzhou, ChinaState Grid Zhejiang Hangzhou Xiaoshan District Power Supply Company Ltd., Hangzhou, ChinaState Grid Zhejiang Hangzhou Xiaoshan District Power Supply Company Ltd., Hangzhou, ChinaState Grid Zhejiang Hangzhou Xiaoshan District Power Supply Company Ltd., Hangzhou, ChinaState Grid Zhejiang Hangzhou Xiaoshan District Power Supply Company Ltd., Hangzhou, ChinaState Grid Zhejiang Hangzhou Xiaoshan District Power Supply Company Ltd., Hangzhou, ChinaDepartment of Computer Science and Technology, Shanghai University of Electric Power, Shanghai, ChinaDepartment of Computer Science and Technology, Shanghai University of Electric Power, Shanghai, ChinaPower load forecasting plays a crucial role in ensuring the stable operation of the power system and avoiding system collapse or resource waste caused by power shortages or surpluses. However, the complex spatio-temporal property of power load makes it difficult to predict, which poses a great challenge to the power system. Existing spatio-temporal prediction methods can only handle one factor in each dimension of time and space. In reality, power load is influenced by various factors. Especially in terms of time dimension, crowd flow, weather, and historical load all have significant impacts on load forecasting. Inspired by tensor time series, considering the structure of spatial geographic location, we propose Tensor Graph Convolutional Network for Power Load Forecasting, LoadSeer. A distance adjacency matrix is designed to represent the geographical location relationship and land use nature. A spatio-temporal processing layer integrating graph convolution module (GCN) and T-Transformer is mapped out to extract the spatio-temporal features, which are then sent to the fully connected layer to provide the refined expression. The experimental results on three public datasets, PeMSD4, PeMSD7, and PeMSD8 show that our proposed method outperforms baseline models on all indicators. To further validate the effectiveness of our proposed approach, we apply LoadSeer to real load data during the Asian Games in a certain city, and the results also demonstrate the superiority of our method.https://ieeexplore.ieee.org/document/10786973/Tensor time seriesgraph convolutional networkspatio-temporal sequenceelectricity load forecasting
spellingShingle Jiahao Zhang
Bin Yu
Hanbin Lai
Lin Liu
Jinghui Zhou
Fengliang Lou
Yili Ni
Yan Peng
Ziheng Yu
LoadSeer: Exploiting Tensor Graph Convolutional Network for Power Load Forecasting With Spatio-Temporal Characteristics
IEEE Access
Tensor time series
graph convolutional network
spatio-temporal sequence
electricity load forecasting
title LoadSeer: Exploiting Tensor Graph Convolutional Network for Power Load Forecasting With Spatio-Temporal Characteristics
title_full LoadSeer: Exploiting Tensor Graph Convolutional Network for Power Load Forecasting With Spatio-Temporal Characteristics
title_fullStr LoadSeer: Exploiting Tensor Graph Convolutional Network for Power Load Forecasting With Spatio-Temporal Characteristics
title_full_unstemmed LoadSeer: Exploiting Tensor Graph Convolutional Network for Power Load Forecasting With Spatio-Temporal Characteristics
title_short LoadSeer: Exploiting Tensor Graph Convolutional Network for Power Load Forecasting With Spatio-Temporal Characteristics
title_sort loadseer exploiting tensor graph convolutional network for power load forecasting with spatio temporal characteristics
topic Tensor time series
graph convolutional network
spatio-temporal sequence
electricity load forecasting
url https://ieeexplore.ieee.org/document/10786973/
work_keys_str_mv AT jiahaozhang loadseerexploitingtensorgraphconvolutionalnetworkforpowerloadforecastingwithspatiotemporalcharacteristics
AT binyu loadseerexploitingtensorgraphconvolutionalnetworkforpowerloadforecastingwithspatiotemporalcharacteristics
AT hanbinlai loadseerexploitingtensorgraphconvolutionalnetworkforpowerloadforecastingwithspatiotemporalcharacteristics
AT linliu loadseerexploitingtensorgraphconvolutionalnetworkforpowerloadforecastingwithspatiotemporalcharacteristics
AT jinghuizhou loadseerexploitingtensorgraphconvolutionalnetworkforpowerloadforecastingwithspatiotemporalcharacteristics
AT fenglianglou loadseerexploitingtensorgraphconvolutionalnetworkforpowerloadforecastingwithspatiotemporalcharacteristics
AT yilini loadseerexploitingtensorgraphconvolutionalnetworkforpowerloadforecastingwithspatiotemporalcharacteristics
AT yanpeng loadseerexploitingtensorgraphconvolutionalnetworkforpowerloadforecastingwithspatiotemporalcharacteristics
AT zihengyu loadseerexploitingtensorgraphconvolutionalnetworkforpowerloadforecastingwithspatiotemporalcharacteristics