基于Transformer-集成学习的配电网短期负荷预测方法

doi:10.11930/j.issn.1004-9649.202511063

摘要/Abstract

摘要：

在新型电力系统背景下，配电网分布式能源渗透率不断提高，负荷特性日趋多元，传统短期负荷预测方法难以有效捕捉高维非线性时序特征。为此，提出一种基于Transformer-集成学习的短期负荷预测方法。首先，构建多维特征嵌入层，融合负荷时序、周期特征及环境变量；其次，采用多头自注意力机制建立跨时段动态关联，提取负荷的时空耦合特性；然后，设计分层随机化前馈网络，结合Dropout增强模型隐空间的多模态表征能力；最后，集成多个差异化Dropout模型，通过多次前向传播采样实现对预测不确定性的贝叶斯评估。实验结果表明，所提方法在预测精度与稳定性上均优于现有基准模型，可为配电网优化调度提供有效支持。

关键词: 短期负荷预测, Transformer, 集成学习, Dropout策略, 前向传播采样

Abstract:

Against the backdrop of the new power systems, the penetration rate of distributed energy resources in distribution networks is rising steadily, and the load characteristics are becoming increasingly diversified. Existing short-term load forecasting methods thus fail to effectively capture the high-dimensional nonlinear temporal characteristics of load data. To address this issue, this paper proposes a short-term load forecasting method for distribution networks based on Transformer and ensemble learning. First, a multi-dimensional feature embedding layer is constructed to fuse the temporal and periodic characteristics of loads as well as environmental variables. Second, a multi-head self-attention mechanism is adopted to establish dynamic cross-time interval correlations, thereby extracting the spatiotemporal coupling characteristics of loads accurately. Third, a hierarchical randomized feedforward network is designed, with the Dropout technique integrated to enhance the multimodal representation capability of the model’s latent space. Finally, multiple differentiated Dropout-based models are ensembled, and Bayesian evaluation of forecasting uncertainty is realized through sampling with multiple forward propagations. Experimental results demonstrate that the proposed method outperforms state-of-the-art benchmark models in both forecasting accuracy and stability, and can thus provide effective technical support for the optimal dispatching of distribution networks.

Key words: short-term load forecasting, Transformer, ensemble learning, Dropout, forward propagation sampling

张怀天, 贾东梨, 王帅, 何开元, 任昭颖, 刘佳静, 胡雪凯. 基于Transformer-集成学习的配电网短期负荷预测方法[J]. 中国电力, 2026, 59(5): 33-45.

ZHANG Huaitian, JIA Dongli, WANG Shuai, HE Kaiyuan, REN Zhaoying, LIU Jiajing, HU Xuekai. Short-term load forecasting method for distribution networks based on transformer and ensemble learning[J]. Electric Power, 2026, 59(5): 33-45.

导出引用管理器 EndNote|Ris|BibTeX

链接本文: https://www.electricpower.com.cn/CN/10.11930/j.issn.1004-9649.202511063

https://www.electricpower.com.cn/CN/Y2026/V59/I5/33

图/表 11

参考文献 35

1	盛万兴, 刘科研, 李昭, 等. 新型配电系统形态演化与安全高效运行方法综述[J]. 高电压技术, 2024, 50 (1): 1- 18.
	SHENG Wanxing, LIU Keyan, LI Zhao, et al. Review of basic theory and methods of morphological evolution and safe & efficient operation of new distribution system[J]. High Voltage Engineering, 2024, 50 (1): 1- 18.
2	王杰, 郑飞, 张鹏城, 等. 基于数据驱动的高比例新能源配电网规划模型[J]. 中国电力, 2025, 58 (3): 175- 182.
	WANG Jie, ZHENG Fei, ZHANG Pengcheng, et al. Model of high-proportion new energy distribution network planning based on data-driven approach[J]. Electric Power, 2025, 58 (3): 175- 182.
3	叶宇剑, 吴奕之, 胡健雄, 等. 城市电力-交通耦合系统的联合推演与协同优化: 研究综述、挑战与展望[J]. 中国电机工程学报, 2025, 45 (11): 4144- 4163.
	YE Yujian, WU Yizhi, HU Jianxiong, et al. Joint prediction and coordinated optimization of integrated urban power distribution and transportation systems: literature review, challenges and prospects[J]. Proceedings of the CSEE, 2025, 45 (11): 4144- 4163.
4	肖勇, 陆文升, 李云涛, 等. 城市配电网发展形态指标体系及其评估方法研究[J]. 电力系统保护与控制, 2021, 49 (1): 62- 71.
	XIAO Yong, LU Wensheng, LI Yuntao, et al. Research on index system and its evaluation methods of urban distribution network development form[J]. Power System Protection and Control, 2021, 49 (1): 62- 71.
5	李晨朝, 陈佳佳, 王敬华. 基于主从博弈和IGDT的含电动汽车需求响应光伏园区储能优化配置[J]. 电力建设, 2025, 46 (4): 126- 136.
	LI Chenzhao, CHEN Jiajia, WANG Jinghua. Optimal configuration of energy storage in photovoltaic park with electric vehicle demand response based on Stackelberg game and information gap decision theory[J]. Electric Power Construction, 2025, 46 (4): 126- 136.
6	陈景文, 黄羽倩, 刘耀先, 等. 基于复合因子构造的KAN-BiLSTM电力负荷预测方法[J]. 中国电力, 2025, 58 (12): 178- 189, 198.
	CHEN Jingwen, HUANG Yuqian, LIU Yaoxian, et al. A KAN-BiLSTM-based power load forecasting method utilizing composite factor construction[J]. Electric Power, 2025, 58 (12): 178- 189, 198.
7	于多, 曹燚, 王海荣, 等. 基于ICEEMDAN-PE和IDBO-Informer组合模型的短期负荷预测[J]. 中国电力, 2025, 58 (6): 19- 32.
	YU Duo, CAO Yi, WANG Hairong, et al. Short-term load forecasting based on a combined ICEEMDAN-PE and IDBO-informer model[J]. Electric Power, 2025, 58 (6): 19- 32.
8	杨佳泽, 王灿, 王增平. 新型电力系统背景下的智能负荷预测算法研究综述[J]. 华北电力大学学报(自然科学版), 2025, 52 (3): 54- 67.
	YANG Jiaze, WANG Can, WANG Zengping. Review on intelligent load forecasting algorithms for the new-type power system[J]. Journal of North China Electric Power University (Natural Science Edition), 2025, 52 (3): 54- 67.
9	梁宏涛, 刘红菊, 李静, 等. 基于机器学习的短期负荷预测算法综述[J]. 计算机系统应用, 2022, 31 (10): 25- 35.
	LIANG Hongtao, LIU Hongju, LI Jing, et al. Survey on short-term load forecasting algorithm based on machine learning[J]. Computer Systems & Applications, 2022, 31 (10): 25- 35.
10	李志军, 徐博, 张家安, 等. 基于TD3可变长度时间窗口最优加权的短期负荷预测策略[J]. 电力建设, 2024, 45 (6): 140- 148.
	LI Zhijun, XU Bo, ZHANG Jia'an, et al. Short-term load optimal weighted forecasting strategy based on TD3 variable length time window[J]. Electric Power Construction, 2024, 45 (6): 140- 148.
11	CHEN S, LIN R H, ZENG W. Short-term load forecasting method based on ARIMA and LSTM[C]//2022 IEEE 22nd International Conference on Communication Technology (ICCT). Nanjing, China. IEEE, 2023: 1913–1917.
12	AL AMIN M A, HOQUE M A. Comparison of ARIMA and SVM for short-term load forecasting[C]//2019 9th Annual Information Technology, Electromechanical Engineering and Microelectronics Conference (IEMECON). Jaipur, India. IEEE, 2019: 1–6.
13	田书欣, 韩雪. 基于正交小波变换的LSTM-ARIMA海上风速组合预测模型[J]. 智慧电力, 2023, 51 (7): 39- 43, 50.
	TIAN Shuxin, HAN Xue. LSTM-ARIMA offshore wind speed combined prediction model based on orthogonal wavelet transform[J]. Smart Power, 2023, 51 (7): 39- 43, 50.
14	TAN Z Q, ZHANG J, HE Y, et al. Short-term load forecasting based on integration of SVR and stacking[J]. IEEE Access, 2020, 8, 227719- 227728.
15	于润泽, 窦震海, 张志一, 等. 基于二次分解重构与多任务学习的综合能源系统多元负荷短期预测[J]. 电力建设, 2024, 45 (12): 149- 161.
	YU Runze, DOU Zhenhai, ZHANG Zhiyi, et al. Multi-energy load forecasting of integrated energy system based on secondary decomposition-reconstruction and multi-task learning[J]. Electric Power Construction, 2024, 45 (12): 149- 161.
16	邵必林, 纪丹阳. 基于VMD-SE的电力负荷分量的多特征短期预测[J]. 中国电力, 2024, 57 (4): 162- 170.
	SHAO Bilin, JI Danyang. Multi-feature short-term prediction of power load components based on VMD-SE[J]. Electric Power, 2024, 57 (4): 162- 170.
17	ZHANG M F, YU Z T, XU Z H. Short-term load forecasting using recurrent neural networks with input attention mechanism and hidden connection mechanism[J]. IEEE Access, 2020, 8, 186514- 186529.
18	TANG X L, DAI Y Y, LIU Q, et al. Application of bidirectional recurrent neural network combined with deep belief network in short-term load forecasting[J]. IEEE Access, 2019, 7, 160660- 160670.
19	陈媛, 段文献, 何怡刚, 等. 带降噪自编码器和门控递归混合神经网络的电池健康状态估算[J]. 电工技术学报, 2024, 39 (24): 7933- 7949.
	CHEN Yuan, DUAN Wenxian, HE Yigang, et al. State of health estimation of lithium ion battery based on denoising autoencoder-gated recurrent unit[J]. Transactions of China Electrotechnical Society, 2024, 39 (24): 7933- 7949.
20	AURANGZEB K, ALHUSSEIN M, JAVAID K, et al. A pyramid-CNN based deep learning model for power load forecasting of similar-profile energy customers based on clustering[J]. IEEE Access, 2021, 9, 14992- 15003.
21	RUBASINGHE O, ZHANG X N, CHAU T 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.
22	葛众, 隆交凤, 李健, 等. 结合CNN与软共享机制的综合能源系统多元负荷预测[J]. 电力建设, 2024, 45 (12): 162- 173.
	GE Zhong, LONG Jiaofeng, LI Jian, et al. Multivariate load forecasting of integrated energy systems based on convolutional neural network and soft sharing mechanism[J]. Electric Power Construction, 2024, 45 (12): 162- 173.
23	KONG W C, DONG Z Y, JIA Y 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.
24	杨国华, 祁鑫, 贾睿, 等. 基于CEEMD-SE的CNN & LSTM-GRU短期风电功率预测[J]. 中国电力, 2024, 57 (2): 55- 61.
	YANG Guohua, QI Xin, JIA Rui, et al. Short-term wind power forecast based on CNN & LSTM-GRU model integrated with CEEMD-SE algorithm[J]. Electric Power, 2024, 57 (2): 55- 61.
25	茹瑶, 赵永宁, 叶林, 等. 超短期LSTM风电功率预测模型的混合专家模块化代理解释方法[J]. 电力建设, 2024, 45 (11): 114- 124.
	RU Yao, ZHAO Yongning, YE Lin, et al. Modular surrogate interpretation method based on decision tree mixture of experts for ultra-short-term LSTM wind power forecasting model[J]. Electric Power Construction, 2024, 45 (11): 114- 124.
26	黄文琦, 梁凌宇, 王鑫, 等. 基于变量选择与Transformer模型的中长期电力负荷预测方法[J]. 浙江大学学报(理学版), 2024, 51 (4): 483- 491, 500.
	HUANG Wenqi, LIANG Lingyu, WANG Xin, et al. Mid-long term power load forecasting based variable selection and transformer model[J]. Journal of Zhejiang University (Science Edition), 2024, 51 (4): 483- 491, 500.
27	孟衡, 张涛, 王金, 等. 基于多尺度时空图卷积网络与Transformer融合的多节点短期电力负荷预测方法[J]. 电网技术, 2024, 48 (10): 4297- 4305.
	MENG Heng, ZHANG Tao, WANG Jin, et al. Multi-node short-term power load forecasting method based on multi-scale spatiotemporal graph convolution network and transformer[J]. Power System Technology, 2024, 48 (10): 4297- 4305.
28	范士雄, 李东琦, 郭剑波, 等. 基于时变滤波经验模态分解-重构和独立自注意力机制的iTransformer超短期负荷预测方法[J]. 电网技术, 2025, 49 (6): 2436- 2445.
	FAN Shixiong, LI Dongqi, GUO Jianbo, et al. Ultra-short-term load forecasting method based on time-varying filter empirical mode decomposition-reconstruction and iTransformer with stand-alone self-attention mechanism[J]. Power System Technology, 2025, 49 (6): 2436- 2445.
29	梁睿, 金沫含, 朱慧君, 等. 基于多场景聚类和可解释时间融合Transformer网络的乡村地区净负荷区间预测[J]. 电网技术, 2025, 49 (12): 5019- 5027, I0016–I0018.
	LIANG Rui, JIN Mohan, ZHU Huijun, et al. Net load interval forecasting in rural areas based on multi-scenario clustering and explainable temporal fusion transformers network[J]. Power System Technology, 2025, 49 (12): 5019- 5027, I0016–I0018.
30	WANG C, WANG Y, DING Z T, et al. A transformer-based method of multienergy load forecasting in integrated energy system[J]. IEEE Transactions on Smart Grid, 2022, 13 (4): 2703- 2714.
31	ZHAO H S, WU Y C, MA L B, et al. Spatial and temporal attention-enabled transformer network for multivariate short-term residential load forecasting[J]. IEEE Transactions on Instrumentation and Measurement, 2023, 72, 2524611.
32	俞胜, 孙可, 蔡华, 等. 结合极端梯度提升决策树与改进Informer的短期电力负荷预测方法[J]. 中国电力, 2025, 58 (10): 195- 205.
	YU Sheng, SUN Ke, CAI Hua, et al. A short-term power load forecasting method combining extreme gradient boosting decision tree with an improved informer[J]. Electric Power, 2025, 58 (10): 195- 205.
33	孟浩, 徐飞, 符帅, 等. 考虑温控型负荷特性影响的集群用户超短期负荷预测方法[J]. 中国电力, 2025, 58 (12): 63- 72, 85.
	MENG Hao, XU Fei, FU Shuai, et al. Ultra-short-term load forecasting method for aggregated users considering the impact of temperature-controlled load characteristics[J]. Electric Power, 2025, 58 (12): 63- 72, 85.
34	Srivastava N, Hinton G, Krizhevsky A, et al. Dropout: a simple way to prevent neural networks from overfitting[J]. The journal of machine learning research, 2014, 15 (1): 1929- 1958.
35	Cheng B, Chen Y. Open Datasets for grid modeling and visualization: an Alberta power network case[J]. arXiv preprint arXiv, 2504, 07870, 2025.

预测模型	数据预处理	特征
Ensemble Transformer	正弦余弦位置编码+嵌入编码	自注意力机制+差异化 Dropout正则化集成
STformer^[21]	负荷序列趋势- 波动分解	稀疏注意力机制+ Dropout正则化
XGBoost+ Informer^[22]	XGBoost关键特征选择	概率稀疏自注意力机制
标准Transformer	正弦余弦位置编码+嵌入编码	标准自注意力机制
LSTM	/	经典LSTM结构

预测模型	数据预处理	特征
Ensemble Transformer	正弦余弦位置编码+嵌入编码	自注意力机制+差异化 Dropout正则化集成
STformer^[21]	负荷序列趋势- 波动分解	稀疏注意力机制+ Dropout正则化
XGBoost+ Informer^[22]	XGBoost关键特征选择	概率稀疏自注意力机制
标准Transformer	正弦余弦位置编码+嵌入编码	标准自注意力机制
LSTM	/	经典LSTM结构

预测模型	验证集			高温日（6月15日—17日）			工作日（7月15日—19日）			周末/节假日（6月29日—30日）
预测模型	MAPE/ %	MAE/ MW	RMSE/ MW	MAPE/ %	MAE/ MW	RMSE/ MW	MAPE/ %	MAE/ MW	RMSE/ MW	MAPE/ %	MAE/ MW	RMSE/ MW
Ensemble Transformer	1.48	191.14	247.45	1.26	162.73	194.05	1.13	172.28	211.22	1.50	192.79	215.52
STformer	1.60	206.48	263.42	1.35	175.03	234.21	1.10	172.42	216.88	1.69	216.84	275.79
XGBoost+Informer	1.72	221.97	284.70	1.95	253.77	308.67	1.18	183.31	229.20	1.66	216.58	283.30
Transformer	1.87	243.76	298.32	1.67	217.69	276.02	1.75	265.34	328.65	1.94	245.97	302.94
LSTM	2.13	274.60	366.39	1.93	248.82	306.59	1.84	279.78	336.18	2.05	262.20	324.33

预测模型	验证集			高温日（6月15日—17日）			工作日（7月15日—19日）			周末/节假日（6月29日—30日）
预测模型	MAPE/ %	MAE/ MW	RMSE/ MW	MAPE/ %	MAE/ MW	RMSE/ MW	MAPE/ %	MAE/ MW	RMSE/ MW	MAPE/ %	MAE/ MW	RMSE/ MW
Ensemble Transformer	1.48	191.14	247.45	1.26	162.73	194.05	1.13	172.28	211.22	1.50	192.79	215.52
STformer	1.60	206.48	263.42	1.35	175.03	234.21	1.10	172.42	216.88	1.69	216.84	275.79
XGBoost+Informer	1.72	221.97	284.70	1.95	253.77	308.67	1.18	183.31	229.20	1.66	216.58	283.30
Transformer	1.87	243.76	298.32	1.67	217.69	276.02	1.75	265.34	328.65	1.94	245.97	302.94
LSTM	2.13	274.60	366.39	1.93	248.82	306.59	1.84	279.78	336.18	2.05	262.20	324.33

负荷范围	预测模型	高温日（6月15日—17日）			工作日（7月15日—19日）			周末/节假日（6月29日—30日）
负荷范围	预测模型	MAPE/%	MAE/MW	RMSE/MW	MAPE/%	MAE/MW	RMSE/MW	MAPE/%	MAE/MW	RMSE/MW
峰荷	Ensemble Transformer	1.25	167.92	205.21	1.27	211.66	257.93	1.94	261.35	276.08
	STformer	1.29	172.12	247.30	1.21	202.42	248.52	2.21	295.89	382.25
	XGBoost+Informer	1.89	253.82	315.74	1.38	232.16	267.08	2.87	385.11	437.36
	Transformer	1.67	226.16	281.75	1.39	231.86	291.16	1.97	266.52	321.05
	LSTM	1.80	239.74	308.34	1.70	285.68	336.12	1.99	267.98	286.60
谷荷	Ensemble Transformer	1.38	169.41	188.50	1.21	163.38	198.08	1.47	173.19	191.56
谷荷	STformer	1.49	183.48	231.71	0.94	129.14	161.79	1.73	204.81	219.36