南京邮电大学自动化学院 / 人工智能学院,江苏 南京 210023
王继波(2004年生),男;研究方向:机器学习;E-mail:b22051311@njupt.edu.cn
丁卉(1989年生),女;研究方向:人工智能;E-mail:hui-gmid@njupt.edu.cn
收稿:2025-09-02,
录用:2025-10-09,
网络首发:2025-10-27,
纸质出版:2026-01-25
移动端阅览
王继波,丁卉,刘明鑫等.多因素融合驱动的电动汽车充电负荷时空预测[J].中山大学学报(自然科学版)(中英文),2026,65(01):103-110.
WANG Jibo,DING Hui,LIU Mingxin,et al.Spatio-temporal prediction of electric vehicle charging load driven by multi factor integration[J].Acta Scientiarum Naturalium Universitatis Sunyatseni,2026,65(01):103-110.
王继波,丁卉,刘明鑫等.多因素融合驱动的电动汽车充电负荷时空预测[J].中山大学学报(自然科学版)(中英文),2026,65(01):103-110. DOI: 10.13471/j.cnki.acta.snus.ZR20250184.
WANG Jibo,DING Hui,LIU Mingxin,et al.Spatio-temporal prediction of electric vehicle charging load driven by multi factor integration[J].Acta Scientiarum Naturalium Universitatis Sunyatseni,2026,65(01):103-110. DOI: 10.13471/j.cnki.acta.snus.ZR20250184.
针对多因素影响下城市电动汽车充电负荷的预测问题,提出了时空图卷积与多通道注意力网络融合模型(STGC-SENet)。该模型构建了周期性时序片段提取模块以获取近期、日、周三尺度周期特征,通过时空图卷积实现时序依赖与空间拓扑关联的同步建模,引入SE通道注意力层对多因素特征通道进行重标定,动态强化关键特征。基于真实数据的实验表明:相比于长短时记忆网络(LSTM)、多视角时空图卷积网络(MSTGCN)、基于注意力的时空图卷积网络(ASTGCN)等基线模型,STGC-SENet的平均绝对误差分别下降3.36、0.56、1.10。在因素敏感性方面,相比于近期、周时序片段,日周期因子的加入对模型预测提升效果最为显著;在历史充电负荷输入下,充电桩占用数与实时电价融入能够获得最优的预测效果,整体平均绝对误差下降至5.08,而充电桩数目和气象因素的融入并未带来较好的预测效果提升。
To address the spatial-temporal prediction of urban electric vehicle charging load under multi-factor influences,a fusion model combining spatial-temporal graph convolutional and Squeeze-and-Excitation network(STGC-SENet) was proposed. A periodic temporal segment extraction block was constructed to acquire features across recent,daily and weekly scales. Temporal dependencies and spatial topological correlations were synchronously modeled through spatio-temporal graph convolution. A Squeeze-and-Excitation channel attention layer was introduced to recalibrate feature channels, dynamically enhancing key feature. Experiments based on real-world urban charging datasets demonstrate that: compared to baseline model(LTSM, MSTGCN, ASTGCN),STGC-SENet reduced the mean absolute error by 3.36, 0.56 and 1.10, respectively. Regarding factor sensitivity, incorporating the daily periodicity factor demonstrates the most significant prediction improvement compared to recent and weekly temporal segments. Under historical charging load inputing, integrating charging pile occupancy rate and real-time electricity price yields optimal prediction performance,with MAE decreasing to 5.08;conversely, incorporating the number of charging piles and meteorological factors did not yield significant prediction improvement.
黄学良 , 刘永东 , 沈斐 , 等 , 2024 . 电动汽车与电网互动:综述与展望 [J]. 电力系统自动化 , 48 ( 7 ): 3 - 23 .
黄一修 , 肖仕武 , 2025 . 考虑气象因素的高速公路服务区电动汽车充电负荷预测 [J]. 电力自动化设备 , 45 ( 1 ): 192 - 199 .
胡博 , 张鹏飞 , 黄恩泽 , 等 , 2022 . 基于图WaveNet的电动汽车充电负荷预测 [J]. 电力系统自动化 , 46 ( 16 ): 207 - 213 .
李波 , 王宁 , 吕叶林 , 等 , 2024 . 考虑环境因素的电动汽车充电站实时负荷预测模型 [J]. 同济大学学报 , 52 ( 6 ): 962 - 969 .
李宏伟 , 宋玉峰 , 李帅兵 , 等 , 2025 . 基于ArcGIS路网结构与交通拥挤度分析的电动汽车充电负荷预测方法 [J]. 电网技术 , 49 ( 5 ): 1920 - 1930 .
刘赏 , 陈浩 , 陈小玉 , 等 , 2023 . 面向交通流预测的双分支时空图卷积神经网络 [J]. 信息与控制 , 52 ( 3 ): 391 - 404+416 .
马恒瑞 , 袁傲添 , 王波 , 等 , 2025 . 基于深度学习的负荷预测研究综述与展望 [J]. 高电压技术 , 51 ( 3 ): 1233 - 1250 .
秦建华 , 潘崇超 , 张璇 , 等 , 2023 . 基于充电行为分析的电动汽车充电负荷预测 [J]. 电测与仪表 , 60 ( 4 ): 19 - 26 .
王长春 , 王果 , 赵倩宇 , 等 , 2024 . 考虑分时电价和充电利用率特征的大型电动汽车充电站负荷短期预测方法 [J]. 南方电网技术 , 18 ( 5 ): 75 - 84 .
杨建喜 , 郁超顺 , 李韧 , 等 , 2021 . 基于多周期组件时空神经网络的路网通行速度预测 [J]. 交通运输系统工程与信息 , 21 ( 3 ): 112 - 119+139 .
张夏韦 , 梁军 , 王要强 , 等 , 2023 . 电动汽车充电负荷时空分布预测研究综述 [J]. 电力建设 , 44 ( 12 ): 161 - 173 .
GUO S , LIN Y , FENG N , et al , 2019 . Attention based spatial-temporal graph convolutional networks for traffic flow forecasting [C]// AAAI Conference on Artificial Intelligence . Keystone, USA : 922 - 929 .
HAO G J , YANG J , XU Q , et al , 2024 . Prediction of urban electric vehicle charging load based on Monte Carlo method [C]// The 5th International Conference on Advanced Electrical and Energy Systems . Lanzhou,China : 473 - 477 .
HE K , ZHANG X , REN , S , et al , 2016 . Deep residual learning or image recognition [C]// IEEE Conference on Computer Vision and Pattern Recognition . Las Vegas, USA : 770 - 778 .
HU J , SHEN L , ALBANIE S , et al , 2019 . Squeeze-and-Excitation networks [EB/OL].( 2019-05-16 )[ 2025-05-20 ]. https://arXiv.org/abs/1709.01507 https://arXiv.org/abs/1709.01507 .
HUANG R Z , HUANG C Y , LIU Y B , et al , 2021 . LSGCN: Long short-term traffic prediction with graph convolutional networks [C]// The 29th International Joint Conference on Artificial Intelligence . Yokohama, Japan : 2355 - 2361 .
KIM H J , KIM M K , 2024 . Spatial-temporal graph convolutional-based recurrent network for electric vehicle charging stations demand forecasting in energy market [J]. IEEE Trans Smart Grid , 15 ( 4 ): 3979 - 3993 .
SU Z P , YU Z W , MA L , 2023 . Short-term load prediction of electric vehicle charging station based on Long-short-term memory neural network [C]// International Conference on Computer Engineering and Intelligent Control . Guangzhou, China : 595 - 599 .
WANG Y C , LI X C , YUAN R M , et al , 2024 . Electric vehicle charging load prediction model based on improved LSTM neural network [C]// International Conference on Energy Systems and Electrical Power . Wuhan, China : 1505 - 1509 .
YAN S J , XIONG Y J , LIN D H , et al , 2018 . Spatial temporal graph convolutional networks for skeleton-based action recognition [C]// AAAI Conference on Artificial Intelligence . New Orleans, USA : 7444 - 7452 .
0
浏览量
265
下载量
0
CSCD
关联资源
相关文章
相关作者
相关机构
京公网安备11010802024621
