2026年 05期

摘要(Abstract)：

针对电网线损率受气象、售电量等多因素耦合影响呈现非线性、多尺度和时变特性，难以实现准确预测的问题，提出一种基于自相关序列分解网络(Autoformer)、 Kolmogorov-Arnold网络(KAN)的时频融合的双支路电网线损率预测模型。时域支路采用多尺度下采样策略结合Autoformer捕捉不同时间尺度特征，通过KAN提升模型非线性拟合能力；频域支路利用经验小波变换分解线损序列，对中低频分量采用Autoformer-KAN组合网络建模，对高频分量采用卷积神经网络(CNN)、双向长短期记忆网络(BiLSTM)、 KAN提取特征；最终通过注意力机制融合时频两支路输出，预测电网线损率。依托真实台区数据开展多模型对比实验，选取LSTM、 BiLSTM、 Autoformer等作为基线模型，验证所提模型的预测性能。结果表明，所提模型通过时域多尺度挖掘与频域差异化建模相结合，能够更好捕捉电网线损率长期趋势、周期波动，同时通过KAN可学习激活函数提升多因素耦合非线性关系建模效果，电网线损率的预测精度优于对比模型，在3个典型台区的电网线损率预测结果平均绝对误差均值为0.049 9,均方根误差均值为0.130 2,平均绝对百分比误差均值为1.26%,证明该模型具备优异的泛化能力和鲁棒性。

关键词(KeyWords)：电网线损率;时频融合预测;自相关序列分解网络;卷积神经网络;长短期记忆网络;Kolmogorov-Arnold网络

基金项目(Foundation):国家自然科学基金项目(12104262);; 山东省科技型中小企业创新能力提升工程项目(2023TSGC0601);; 山东省自然科学基金项目(ZR2025QC1633)

作者(Author):兰鹏,刘赐奎,孙丰刚,马超,张华霞

DOI:10.13349/j.cnki.jdxbn.20260710.001

参考文献(References)：

[1] Fang Xi,Misral S,Xue Guoliang,et al.Smart grid—the new and improved power grid:a survey[J].IEEE Communications Surveys & Tutorials,2011,14(4):944.

[2] 生西奎，付强，于洋，等.基于深度学习GRU网络的配电网理论线损计算方法[J].电测与仪表，2021,58(3):54.

[3] Georgilakis P S,Hatziargyriou N D.A review of power distribution planning in the modern power systems era:Models,methods and future research[J].Electric Power Systems Research,2015,121:89.

[4] Chen Xi,Song Chunhe,Wang Tianran.Spatiotemporal analysis of line loss rate:a case study in China[J].Energy Reports,2021,7:7048.

[5] Tulensalo J,Seppnen J,Ilin A.An LSTM model for power grid loss prediction[J].Electric Power Systems Research,2020,189:106823.

[6] 张昆.基于改进BP神经网络的低压配网线损率自动预测方法[J].自动化应用，2025,66(6):186.

[7] 张一彦，陆嘉铭，贺静，等.基于改进随机森林算法的线损率预测研究[J].电力大数据，2025,28(3):19.

[8] Jiang Lin,Li Chen,Qiu Wei,et al.Research on short-term line loss rate prediction method of distribution network based on RF-CNN-LSTM[J].Frontiers in Smart Grids,2025,4:1612770.

[9] 李亚，刘丽平，李柏青，等.基于改进K-means聚类和BP神经网络的台区线损率计算方法[J].中国电机工程学报，2016,36(17):4543.

[10] 周王峰，李勇，郭钇秀，等.基于DAE-LSTM神经网络的配电网日线损率预测[J].电力系统保护与控制，2021,49 (17):48.

[11] 赵雅婷，林顺富，姜恩宇，等.基于mRMR-R-GCN的配电网低压台区线损率预测模型[J].上海电力大学学报，2025,41(2):120.

[12] Zhang Yishu,Li Yilun,Wang Xuanliang,et al.Research on hybrid variable weight prediction model of line loss in transformer district Based on CEEMDAN-WPT[C]//2020 Chinese Automation Congress (CAC),November 6-8,2020,Shanghai,China.New York:IEEE,2020:2179.

[13] Zhang Zhanlong,Yang Yu,Zhao Hui,et al.Prediction method of line loss rate in low-voltage distribution network based on multi-dimensional information matrix and dimensional attention mechanism-long-and short-term time-series network[J].IET Generation,Transmission & Distribution,2022,16(20):4187.

[14] Fan Shiming,Wang Hua,Zhang Fan.CAWformer:a cross variable attention with discrete wavelet denoising for multivariate time series forecasting[J].Knowledge-Based Systems,2025,324:113846.

[15] Klaar A C R,Stefenon S F,Seman L O,et al.Optimized EWT-Seq2Seq-LSTM with attention mechanism to insulators fault prediction[J].Sensors,2023,23(6):3202.

[16] Liao Kaili,Zhou Wuneng.An EWT-EnSEMLSTM-LSSA model for metro passengers volume prediction[J].IEEE Access,2023,11:92188.

[17] Zhou Tian,Ma Ziqing,Wen Qingsong,et al.FEDformer:frequency enhanced decomposed transformer for long-term series forecasting[J]//Proceedings of the 39th International Conference on Machine Learning,PMLR,2022.162:27268.

[18] Zeng Ailing,Chen Muxi,Zhang Lei,et al.Are transformers effective for time series forecasting?[C]//The 37th AAAI Conference on Artificial Intelligence,February 7-14,2023,Washington,USA.Washington:AAAI,2023,37(9):11121.

[19] Ouyang Jing,Zuo Zongxu,Wang Qin,et al.Seasonal distribution analysis and short-term PV power prediction method based on decomposition optimization Deep-Autoformer[J].Renewable Energy,2025,246:122903.

[20] Zhao Tianlong,Fang Lexin,Ma Xiang,et al.TFformer:a time-frequency domain bidirectional sequence-level attention-based transformer for interpretable long-term sequence forecasting[J].Pattern Recognition,2025,158:110994.

[21] Huang Songtao,Zhao Zhen,Li Can et al.TimeKAN:KAN-based frequency decomposition learning architecture for long-term time series forecasting[PP/OL].arXiv(2025-02-10)[2025-10-20].https://doi.org/10.48550/arXiv.2502.06910.

[22] Chen Yao,Li Shaorong,Zhao Na,et al.BiLSTM-KAN:a time series-based traffic flow forecasting model[C]//Proceedings of the 2024 13th International Conference on Computing and Pattern Recognition,October 25-27,2024,New York,USA.New York:ACM,2024:314.

[23] Leng Zhiyuan,Chen Lu,Yi Bin,et al.Short-term wind speed forecasting based on a novel KAN informer model and improved dual decomposition[J].Energy,2025,322:135551.