基于改进Informer的商业建筑短期用电负荷多步预测

doi:10.12141/j.issn.1000-565X.250024

华南理工大学学报(自然科学版) ›› 2026, Vol. 54 ›› Issue (1): 42-52.doi: 10.12141/j.issn.1000-565X.250024

• 能源、动力与电气工程 • 上一篇下一篇

基于改进Informer的商业建筑短期用电负荷多步预测

周璇¹^,³^,⁴(), 李可昕¹, 郭子轩²(), 俞祝良², 闫军威¹^,³^,⁴, 蔡盼盼¹

^1.华南理工大学机械与汽车工程学院，广东广州 510640
^2.华南理工大学自动化科学与工程学院，广东广州 510640
^3.广州现代产业技术研究院，广东广州 511458
^4.人工智能与数字经济广东省实验室（广州），广东广州 511442

收稿日期:2025-01-20 出版日期:2026-01-10 发布日期:2025-04-25
通信作者: 郭子轩 E-mail:zhouxuan@scut.edu.cn;zxguo020@foxmail.com
作者简介:周璇（1976—），女，博士，教授级高级工程师，主要从事建筑节能、数据挖掘等研究。E-mail： zhouxuan@scut.edu.cn
基金资助:
广东省自然科学基金项目(2022A1515011128)

Short-Term Power Load Multi-Step Forecasting for Commercial Building Based on Improved Informer

ZHOU Xuan¹^,³^,⁴(), LI Kexin¹, GUO Zixuan²(), YU Zhuliang², YAN Junwei¹^,³^,⁴, CAI Panpan¹

^1.School of Mechanical & Automotive Engineering，South China University of Technology，Guangzhou 510640，Guangdong，China
^2.School of Automation Science and Engineering，South China University of Technology，Guangzhou 510640，Guangdong，China
^3.Guangzhou Institute of Modern Industrial Technology，Guangzhou 511458，Guangdong，China
^4.Artificial Intelligence and Digital Economy Guangdong Province Laboratory（Guangzhou），Guangzhou 511442，Guangdong，China

Received:2025-01-20 Online:2026-01-10 Published:2025-04-25
Contact: GUO Zixuan E-mail:zhouxuan@scut.edu.cn;zxguo020@foxmail.com
Supported by:
the Natural Science Foundation of Guangdong Province(2022A1515011128)

摘要/Abstract

摘要：

商业建筑短期用电负荷多步预测是城市有序用电和虚拟电厂调度的关键环节。商业建筑用电负荷时间序列具有强随机性、非平稳、非线性等特点，针对传统的迭代式多步用电负荷预测方法存在误差累积效应影响预测精度的问题，提出一种基于频率增强通道注意力机制（FECAM）—麻雀优化算法（SSA）—Informer的短期用电负荷多步预测方法。该方法在Informer编码器输出时域特征的基础上，采用FECAM对各特征通道间的频率依赖性进行自适应建模，进一步提取多维输入序列的频域特征，生成式解码器利用融合的时、频域信息直接输出未来多步用电负荷序列。此外，由于改进Informer超参数设置缺乏理论依据，使用SSA寻优学习率、批处理大小、全连接维度和失活率的最佳组合。以广州某商业建筑全年用电负荷数据作为实际算例，结果表明，与其他深度学习模型相比，所提模型在不同预测步长（48、96、288、480、672步）下的预测精度显著提升，具有更优的短期用电负荷多步预测性能。

关键词: 商业建筑用电负荷预测, 频率增强通道注意力机制, Informer, 麻雀优化算法

Abstract:

Short-term power load multi-step forecasting for commercial buildings plays a pivotal role in urban orderly power consumption and virtual power plant scheduling. The power load time series in commercial buildings is characterized by strong stochasticity, non-stationarity, and nonlinearity, and traditional iterative multi-step power load forecasting strategy suffers from error accumulation effects that degrade prediction accuracy, a short-term power load multi-step forecasting method based on Frequency Enhanced Channel Attention Mechanism (FECAM)-Sparrow Search Algorithm (SSA)-Informer is proposed. Based on the time-domain features output by the Informer encoder, the method uses FECAM to adaptively model the frequency dependence between feature channels, and further extractings the frequency-domain features of multi-dimensional input sequences. The decoder then integrates both time-frequency domain information to directly generate future multi-step load sequences. Furthermore, due to the lack of theoretical basis for the improved Informer hyperparameter settings, the SSA is used to optimize model hyperparameters such as learning rate, batch size, fully connected dimensions, and dropout rate. Experimental validation using annual load data from a commercial building in Guangzhou demonstrates that, compared with other deep learning models, the proposed model significantly improved prediction accuracy across varying forecast horizons (steps of 48, 96, 288, 480 and 672), exhibiting superior performance in short-term power load multi-step forecasting.

Key words: power load forecasting for commercial buildings, frequency enhanced channel attention mechanism, Informer, sparrow search algorithm

中图分类号:

TU111.195

周璇, 李可昕, 郭子轩, 俞祝良, 闫军威, 蔡盼盼. 基于改进Informer的商业建筑短期用电负荷多步预测[J]. 华南理工大学学报(自然科学版), 2026, 54(1): 42-52.

ZHOU Xuan, LI Kexin, GUO Zixuan, YU Zhuliang, YAN Junwei, CAI Panpan. Short-Term Power Load Multi-Step Forecasting for Commercial Building Based on Improved Informer[J]. Journal of South China University of Technology(Natural Science Edition), 2026, 54(1): 42-52.

图/表 15

图1

图2

图3

图4

图5

表1

表2

表3

表4

图6

表5

图7

图8

表6

图9

参考文献 24

[1]	汤军，高洪超，余涛，等．电力市场环境下超大城市虚拟电厂的建设理念与创新实践［J］．电网技术，2025，49（1）：103-112.
	TANG Jun， GAO Hongchao， YU Tao，et al ．The construction concept and innovative practice of virtual power plants in megacities under electricity market environment ［J］．Power System Technology，2025，49（1）：103-112.
[2]	HONG T， FAN S ．Probabilistic electric load forecas-ting：a tutorial review［J］．International Journal of Forecasting，2016，32（3）：914-938.
[3]	孔祥玉，马玉莹，艾芊，等．新型电力系统多元用户的用电特征建模与用电负荷预测综述［J］．电力系统自动化，2023，47（13）：2-17.
	KONG Xiangyu， MA Yuying， AI Qian，et al ．Review on electricity consumption characteristic modeling and load forecasting for diverse users in new power system［J］．Automation of Electric Power Systems，2023，47（13）：2-17.
[4]	MATHUMITHA R， RATHIKA P， MANIMALA K ．Intelligent deep learning techniques for energy consumption forecasting in smart buildings：a review［J］．Artificial Intelligence Review，2024，57（2）：35-67.
[5]	陈纬楠，胡志坚，岳菁鹏，等．基于长短期记忆网络和LightGBM组合模型的短期负荷预测［J］．电力系统自动化，2021，45（4）：91-97.
	CHEN Weinan， HU Zhijian， YUE Jingpeng，et al ．Short-term load prediction based on a combined model of long short-term memory network and light gradient boos-ting machine［J］．Automation of Electric Power Systems，2021，45（4）：91-97.
[6]	DONG X， LUO Y， Yuan S，et al ．Building electricity load forecasting based on spatiotemporal correlation and electricity consumption behavior information［J］．Applied Energy，2025，377，124580/1-15.
[7]	WANG C， WANG Y， DING Z，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.
[8]	遆宝中，李庚银，武昭原，等．基于循环扩张机制的ConvGRU-Transformer短期电力负荷预测方法［J］．华北电力大学学报（自然科学版），2022，49（3）：34-43.
	TAI Baozhong， LI Gengyin， WU Zhaoyuan，et al ．A short-term load forecasting method based on recurrentand dilated mechanism of ConvGRU-Transformer［J］．Journal of North China Electric Power University（Natural Science Edition），2022，49（3）：34-43.
[9]	ZHOU H， ZHANG S， PENG J，et al ．Informer：beyond efficient Transformer for long sequence time-series forecasting［J］．Proceedings of the AAAI Conference on Artificial Intelligence，2021，35（12）：11106-11115.
[10]	LI F， WAN Z， KOCH T，et al ．Improving the accuracy of multi-step prediction of building energy consumption based on EEMD-PSO-Informer and long-time series［J］．Computers and Electrical Engineering，2023，110：108845/1-10.
[11]	陈辰，马恒瑞，陈来军，等．基于用户群体划分的多步短期负荷预测方法［J］．高电压技术，2023，49（10）：4213-4222.
	CHEN Chen， MA Hengrui， CHEN Laijun，et al ．Multi-step short-term load forecasting method based on user group division［J］．High Voltage Engineering，2023，49（10）：4213-4222.
[12]	DARWAZEH D， DUQUETTE J， GUNAY B，et al ．Review of peak load management strategies in commercial buildings［J］．Sustainable Cities and Society，2022，77：103493/1-19.
[13]	YANG K， SHI F ．Medium- and long-term load forecas-ting for power plants based on causal inference and Informer［J］．Applied Sciences，2023，13（13）：7696-7714.
[14]	WANG K， ZHANG J， LI X，et al ．Long-term power load forecasting using LSTM-Informer with ensemble learning［J］．Electronics，2023，12（10）：2175-2193.
[15]	XU H， PENG Q， WANG Y，et al ．Power-load forecasting model based on informer and its application［J］．Energies，2023，16（7）：3086-3099.
[16]	ZHANG Q， ZHOU S， XU B，et al ．TCAMS-Trans：efficient temporal-channel attention multi-scale transformer for net load forecasting［J］．Computers and Electrical Engineering，2024，118：109415/1-16.
[17]	林涵，郝正航，郭家鹏，等．基于TCA-CNN-LSTM的短期负荷预测研究［J］．电测与仪表，2023，60（8）：73-80.
	LIN Han， HAO Zhenghang， GUO Jiapeng，et al ．Research on short-term load forecasting based on TCA-CNN-LSTM［J］．Electrical Measurement & Instrumentation，2023，60（8）：73-80.
[18]	刘杰，周博文，田明，等．基于迁移学习和TCN-BIGRU的短期负荷预测［J］．北京航空航天大学学报，2024，doi：10.13700/j.bh.1001-5965.2024.0056 .
	LIU Jie， ZHOU Bowen， TIAN Ming，et al ．Short-term load forecasting based on transfer learning and TCN-BIGRU［J］．Journal of Beijing University of Aeronautics and Astronautics，2024，doi：10.13700/j.bh.1001-5965.2024.0056
[19]	赵婧宇，池越，周亚同．基于SSA-LSTM模型的短期电力负荷预测［J］．电工电能新技术，2022，41（6）：71-79.
	ZHAO Jingyu， CHI Yue， ZHOU Yatong ．Short-term load forecasting based on SSA-LSTM model［J］．Advanced Technology of Electrical Engineering and E-nergy，2022，41（6）：71-79.
[20]	DENG H， FENG Q ．A new approach for electricity demand forecasting based on improved informer［C］∥Proceeding of the 2023 5th International Conference on Electrical Engineering and Control Technologies （CEECT）．Chengdu：IEEE，2023：507-514.
[21]	JIANG M， ZENG P， LIU H，et al ．FECAM：frequency enhanced channel attention mechanism for time series forecasting［J］．Advanced Engineering Informatics，2023，58：102158/1-11.
[22]	LI D， LIU Q， FENG D，et al ．A medium- and long-term residential load forecasting method based on discrete cosine transform-FEDformer［J］．Energies，2024，17（15）：3676-3689.
[23]	HU J， SHEN L， SUN G ．Squeeze-and-excitation networks［C］∥Proceeding of the 2018 IEEE/CVF Confe-rence on Computer Vision and Pattern Recognition．Salt Lake City：IEEE，2018：7132-7141.
[24]	XUE J， SHEN B ．A novel swarm intelligence optimization approach：sparrow search algorithm［J］．Systems Science and Control Engineering，2020，8（1）：22-34.

参数	数值
编码器输入维度	4
解码器输入维度	4
模型输出维度	1
嵌入维度	512
多头自注意头数	8
编码器层数	2
解码器层数	1
堆叠编码器层数	［3，2，1］
稀疏注意力因子	5
通道数	512

超参数	数值上、下限
初始学习率	［0.000 1，0.001 0］
批处理大小	［8，64］
失活率	［0.05，0.20］
全连接神经层维度	［256，2 048］

参数	数值
种群数量	10
种群维度	4
最大迭代数	20
探索者比例	0.2
侦查者比例	0.1
安全阈值	0.8

预测步长/步	最优超参数组合
48	［0.000 1，8，694，0.06］
96	［0.000 2，13，560，0.05］
288	［0.000 2，41，1 210，0.05］
480	［0.000 2，53，1 139，0.08］
672	［0.000 2，30，1 990，0.06］

预测步长/步	FECAM-SSA-Informer			iTransformer			Autoformer
预测步长/步	E_MAPE/%	E_RMSE/kW	R²	E_MAPE/%	E_RMSE/kW	R²	E_MAPE/%	E_RMSE/kW	R²
48	8.210	90.739	0.984	10.277	104.430	0.979	21.546	162.673	0.957
96	8.838	109.187	0.978	11.553	119.452	0.973	31.248	203.375	0.921
288	10.029	149.822	0.960	13.749	150.662	0.957	33.513	234.330	0.901
480	12.185	165.058	0.951	14.775	167.864	0.946	40.899	267.466	0.851
672	13.198	171.359	0.943	15.724	183.126	0.936	40.733	277.684	0.869
预测步长/步	Reformer			Transformer			LSTM
预测步长/步	E_MAPE/%	E_RMSE/kW	R²	E_MAPE/%	E_RMSE/kW	R²	E_MAPE/%	E_RMSE/kW	R²
48	12.832	139.174	0.966	11.104	100.543	0.982	20.260	167.273	0.947
96	15.411	176.986	0.942	10.559	111.540	0.978	20.973	176.194	0.941
288	19.914	210.039	0.929	27.251	165.106	0.952	21.370	193.533	0.928
480	20.078	264.822	0.901	27.834	193.597	0.933	22.811	205.571	0.919
672	23.968	287.744	0.887	25.409	208.702	0.916	24.472	216.837	0.910

基于改进Informer的商业建筑短期用电负荷多步预测

Short-Term Power Load Multi-Step Forecasting for Commercial Building Based on Improved Informer

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 15

参考文献 24

相关文章 0

编辑推荐

Metrics

本文评价