1.CSG Electric Power Research Institute,Guangzhou 510663,China
2.Guangdong Provincial Key Laboratory of Intelligent Measurement and Advanced Metering of Power Grid,Guangzhou 510700,China
3.College of Electrical and Information Engineering,Hunan University,Changsha 410082,China
Show less
文章历史+
Received
Published
2025-01-21
2025-10-25
Issue Date
2026-02-12
PDF (3257K)
摘要
针对现有短期电力负荷预测方法未充分考虑负荷在不同时间尺度和空间特征的多维度变化,无法有效捕捉负荷变化而导致预测准确度不高的问题,提出一种基于双重注意力融合长短期时间序列网络(dual-attention fusion of long-and short-term time series networks,DALSTNet)的短期电力负荷预测方法. 首先,构建双重注意力模块,分别对不同时间和空间特征进行加权,有效区分电力负荷数据中不同时空特征的重要性;其次,构建多尺度因果卷积神经网络(convolutional neural network,CNN)和两个堆叠的双向长短期记忆(bidirectional long and short-term memory,BiLSTM)模块,用于学习电力负荷数据中存在的短期、长期与超长期重复特征,实现对电力负荷数据多维多尺度特征的联合提取;最后,利用融合层融合上述模块提取到的多维多尺度时空特征,并通过全连接层输出获得最终短期电力负荷预测结果. 基于IEEE Dataport住宅综合能源系统负荷数据集算例分析结果表明,本文提出方法能有效提取负荷数据的多维多尺度时空特征,相比现有方法在短期电力负荷预测中的均方根误差、平均绝对误差、平均绝对百分比误差等指标均表现更好,实现了更高的预测准确度和更优的稳定性.
Abstract
Aiming at the problem that the existing short-term power load forecasting methods do not fully consider the multidimensional changes of loads in different time scales and spatial characteristics, and are unable to effectively capture the load changes, which leads to low forecasting accuracy, a short-term power load forecasting method based on the dual-attention fusion of long and short-term time series networks (DALSTNet) is proposed. First, a dual-attention module is constructed to weight different temporal and spatial features to effectively distinguish the importance of different spatial and temporal features in the power load data. Secondly, a multiscale causal convolutional neural network (CNN) and two stacked bidirectional long and short-term memory (BiLSTM) modules are constructed to learn short-term, long-term and ultra-long-term repetitive features in the power load data, and to realize the joint extraction of multi-dimensional and multi-scale features in the power load data. Finally, the fusion layer is utilized to fuse the extracted multidimensional and multiscale spatio-temporal features from the above modules, and the final short-term power load prediction results are obtained through the output of the fully connected layer. The analysis results based on the IEEE Dataport residential integrated energy system load dataset show that the proposed method can effectively extract the multidimensional and multiscale spatial and temporal features of the load data, and compared with the existing methods, it performs better than the existing methods in short-term power load forecasting in terms of root mean square error, mean absolute error, mean absolute percentage error, and other indexes, and achieves higher forecasting accuracy and better stability.
在进行短期电力负荷预测时,历史负荷数据所包含的时间特征及待预测负荷所处位置的空间特征都对模型的预测性能有重要影响. 在此基础上,本文以IEEE Dataport的住宅综合能源系统数据集[14]为例开展相关研究. 该数据集收集了2010年12月1日到2018年11月28日期间该住宅中多个客户各种负载概况汇总的负荷数据,以及该住宅楼所处位置的相关天气变量如气温、气压、相对湿度、太阳辐射和风速等,其采样间隔为1 h.
式中: Q 、 K 、 V 分别为查询矩阵、键矩阵和值矩阵; WQ、WK、WV 分别为对应的权重矩阵;A(∙)为缩放点积注意力函数;Softmax(∙)为归一化指数函数;dk 为K的维度;headi 为第i个头的自注意力输出矩阵;h为多头自注意力的头数;Concat(∙)表示拼接操作;MH为空间注意力模块最终输出的多头自注意力权重向量; WO 为转换矩阵.
ZHUJ Z, DONGH J, LIS L,et al .Review of data-driven load forecasting for integrated energy system[J].Proceedings of the CSEE,2021,41(23):7905-7924.(in Chinese)
[3]
QUK, SIG Q, SHANZ H,et al .Forwardformer:efficient transformer with multi-scale forward self-attention for day-ahead load forecasting[J].IEEE Transactions on Power Systems,2024,39(1):1421-1433.
[4]
TALAATM, FARAHATM A, MANSOURN,et al .Load forecasting based on grasshopper optimization and a multilayer feed-forward neural network using regressive approach[J].Energy,2020,196:117087.
LIUM, SHANGS .Research on power load forecasting method based on K-means++ clustering and RF-GRU combined model[J].Computer & Digital Engineering,2024,52(6):1662-1667.(in Chinese)
ZHANGY S, LIY L, YAOZ Y,et al .Short-term power load forecasting method based on VMD-GWO-SVR[J].Northeast Electric Power Technology,2024,45(7):27-31.(in Chinese)
LUB, HUOZ J, YUM .Multi load ultra short-term forecasting of integrated energy system based on LSTNet-Skip[J].Proceedings of the CSEE,2023,43(6): 2273-2282.(in Chinese)
LID, SUNG F, MIAOS W,et al. Short-term power load forecasting method based on multidimensional time series information fusion[J]. Proceedings of the CSEE,2023,43(Sup.1): 94-106.(in Chinese)
ZHANGK M, CAIS S, ZHANGT H,et al .Medium and long-term load forecasting method considering multi-dimensional time domain characteristics[J].Automation of Electric Power Systems,2023:47(20):104-114.(in Chinese)
KONGX Y, LIC, ZHENGF,et al .Short-term load forecasting method based on empirical mode decomposition and feature correlation analysis[J].Automation of Electric Power Systems,2019,43(5):46-52.(in Chinese)
DENGD Y, LIJ, ZHANGZ Y,et al .Short-term electric load forecasting based on EEMD-GRU-MLR[J].Power System Technology,2020,44(2): 593-602.(in Chinese)
[19]
LÜL L, WUZ Y, ZHANGJ H, et al .A VMD and LSTM based hybrid model of load forecasting for power grid security[J].IEEE Transactions on Industrial Informatics, 2022, 18(9):6474-6482.
[20]
RUBASINGHEO, ZHANGX N, CHAUT K,et al .A novel sequence to sequence data modelling based CNN-LSTM algorithm for three years ahead monthly peak load forecasting[J].IEEE Transactions on Power Systems,2024,39(1): 1932-1947.
YAOC W, YANGP, LIUZ J .Load forecasting method based on CNN-GRU hybrid neural network[J].Power System Technology,2020, 44(9): 3416-3424.(in Chinese)
[23]
SAMANT, MOHAMMADJ, MOEINM A. 8 years of hourly heat and electricity demand for a residential building[EB/OL]. (2021-06-12)[2025-01-21].
[24]
ZHUK D, LIY P, MAOW B,et al .LSTM enhanced by dual-attention-based encoder-decoder for daily peak load forecasting[J].Electric Power Systems Research,2022,208:107860.
JIANGY C, YANGX Q, HEF,et al .Super-short-time wind power forecasting based on EEMD-IGSA-LSSVM[J].Journal of Hunan University (Natural Sciences),2016,43(10):70-78.(in Chinese)
LIY, JIAY J, LIL,et al .Short term power load forecasting based on a stochastic forest algorithm[J].Power System Protection and Control, 2020, 48(21): 117-124.(in Chinese)
SUNC, LÜQ, ZHUS T,et al .Ultra-short-term power load forecasting based on two-layer XGBoost algorithm considering the influence of multiple features[J].High Voltage Engineering,2021,47(8):2885-2898.(in Chinese)
LIUJ T, SUNY Q, GUOZ T, et al. Exploration of the application of Attention-GAT-LSTM algorithm model in new power systems[J]. Southern Power System Technology, 2025, 19(6): 95-104.(in Chinese)
WANGW Q, HANA Y, YUL T,et al .Short-term load forecasting model based on autoencoder and PSOA-CNN[J].Journal of Shandong University (Natural Science),2019,54(7):50-56.(in Chinese)
[35]
KONGW C, DONGZ Y, JIAY W,et al .Short-term residential load forecasting based on LSTM recurrent neural network[J].IEEE Transactions on Smart Grid, 2019, 10(1):841-851.
LIUR H, SHIJ F, SUNG P, et al. Short-term net load probability forecasting based on MV-WC and gated recurrent unit[J]. Southern Power System Technology, 2025, 19(6): 152-161.(in Chinese)
WANGZ P, ZHAOB, JIW J,et al .Short-term load forecasting method based on GRU-NN model[J].Automation of Electric Power Systems, 2019, 43(5): 53-58.(in Chinese)
WANGC C, WANGG, ZHAOQ Y, et al. A short-term load forecasting method for large scale electric vehicle charging stations considering characteristics of charging utilization rate and time-of-use electricity price[J]. Southern Power System Technology, 2024, 18(5): 75-84.(in Chinese)
ZHANGS Q, LIJ, JIANGA Q,et al .A novel two-stage model based on FPA-VMD and BiLSTM neural network for short-term power load forecasting[J].Power System Technology, 2022, 46(8): 3269-3279.(in Chinese)
LUJ X, ZHANGQ P, YANGZ H,et al .Short-term load forecasting method based on CNN-LSTM hybrid neural network model[J].Automation of Electric Power Systems,2019,43(8):131-137.(in Chinese)
[46]
WANA P, CHANGQ, AL-BUKHAITIK,et al .Short-term power load forecasting for combined heat and power using CNN-LSTM enhanced by attention mechanism[J].Energy,2023,282:128274.
[47]
DAIY M, YANGX Y, LENGM M .Optimized Seq2Seq model based on multiple methods for short-term power load forecasting[J].Applied Soft Computing,2023,142:110335.
RENJ J, WEIH H, ZOUZ L,et al .Ultra-short-term power load forecasting based on CNN-BiLSTM-Attention[J].Power System Protection and Control,2022,50(8): 108-116.(in Chinese)
OUYANGF L, WANGJ, ZHOUH X. Short-term power load forecasting method based on improved hierarchical transfer learning and multi-scale CNN-BiLSTM-Attention[J]. Power System Protection and Control,2023,51(2):132-140.(in Chinese)
LIW Y, YANGG C, WENM,et al .A short-term power load forecasting method based on spatiotemporal graph attention[J].Journal of Hunan University (Natural Sciences), 2024, 51(2):57-67.(in Chinese)
基金资助
智能电网国家科技重大专项(2030)(2024ZD0802100)
Smart Grid-National Science and Technology Major Project under Grant (2030)(2024ZD0802100)
南方电网科学研究院科技项目(SEPRI-K23B028)
Science and Technology Project of China Southern Power Grid Research Institute(SEPRI-K23B028)