基于MSCNN-BiGRU-Attention的短期电力负荷预测

doi:10.11930/j.issn.1004-9649.202406098

摘要/Abstract

摘要：

为解决电力负荷关键特征难以提取的问题，提出一种结合多尺度卷积神经网络-双向门控循环单元-注意力机制（multi-scale convolutional neural network-bi-directional gated recurrent unit-Attention，MSCNN-BiGRU-Attention）的组合模型进行短期电力负荷预测。首先，通过Spearman相关系数分析电力负荷数据集的相关性，筛选出相关性较高的特征，构建电力负荷数据集；其次，将数据输入到多尺度卷积神经网络（multi-scale convolutional neural network，MSCNN），对电力负荷数据进行多尺度的时序提取；然后，将提取后的时序特征输入到双向门控循环单元（bi-directional gated recurrent unit，BiGRU）神经网络进行时序预测，并通过注意力（Attention）机制对时序特征进行过滤和筛选；最后，通过全连接层整合输出预测值。以澳大利亚某地区3年的多维电力负荷数据作为数据集，并设置5种对照组模型。同时选用国内南方某地区2年的多维电力负荷数据作为模型验证数据集。结果表明，相较其他模型，MSCNN-BiGRU-Attention组合模型能够取得更好的预测效果，有效解决区域级电力负荷关键特征难以提取的问题。

关键词: 电力负荷预测, 多尺度卷积神经网络, 双向门控循环单元, 注意力机制, 深度学习, Spearman相关系数

Abstract:

To address the problem of difficult extraction of key features in power load, a multi-scale convolutional neural network-bi-directional gated recurrent unit-Attention (MSCNN-BiGRU-Attention) hybrid model is proposed for short-term power load forecasting. Firstly, the Spearman correlation coefficient was used to analyze the correlation of power load data set, and the features with high correlation were screened out to build the power load data set. Secondly, the data was input into the multi-scale convolutional neural network (MSCNN) to extract the multi-scale time sequence of power load data. Then, the extracted temporal features were input into the bidirectional gated recurrent unit (BiGRU) neural network for temporal prediction, and the temporal features were filtered and screened by attention mechanism. Finally, the outputs are integrated through a fully connected layer to generate the predicted values. With the 3 years of multidimensional power load data from a region in Australia as a data set, five control models were established. Meanwhile, we selected two years of multidimensional power load data from a region in southern China as the validation dataset for the models. The results show that MSCNN-BiGRU-Attention hybrid model can achieve better prediction effects than other models, thus effectively solving the problem of difficult extraction of the key features of regional power load.

Key words: power load forecasting, multi-scale convolutional neural network, bi-directional gated recurrent unit, Attention mechanism, deep learning, Spearman correlation coefficient

李科, 潘庭龙, 许德智. 基于MSCNN-BiGRU-Attention的短期电力负荷预测[J]. 中国电力, 2025, 58(6): 10-18.

LI Ke, PAN Tinglong, XU Dezhi. Short-Term Power Load Forecasting Based on MSCNN-BiGRU-Attention[J]. Electric Power, 2025, 58(6): 10-18.

导出引用管理器 EndNote|Ris|BibTeX

链接本文: https://www.electricpower.com.cn/CN/10.11930/j.issn.1004-9649.202406098

https://www.electricpower.com.cn/CN/Y2025/V58/I6/10

图/表 14

图 1 MSCNN结构

Fig.1 MSCNN structure

图 2 BiGRU结构

Fig.2 BiGRU structure

图 3 注意力机制流程

Fig.3 Flowchart of Attention mechanism

图 4 预测流程

Fig.4 Flowchart of prediction

表 1 模型超参数配置

Table 1 Model hyperparameter configuration

超参数		取值
过滤器数量1/卷积核尺寸1		32/2
过滤器数量2/卷积核尺寸2		32/16
过滤器数量3/卷积核尺寸3		64/64
迭代次数		300
批量大小		50
单元数		20
随机失活		0.2

图 5 Spearman相关性分析

Fig.5 Spearman correlation analysis

表 2 不同特征筛选下的预测误差

Table 2 Prediction errors under different feature screening

特征筛选	E_MAP/%	E_RMS/MW	E_MA/MW	R²
全部	0.904	96.342	79.073	0.9949
去掉DBT	0.731	79.873	63.790	0.9965
去掉WBT	0.745	82.916	65.266	0.9962
去掉DBT、WBT	0.635	71.877	55.285	0.9971

图 6 不同特征筛选预测曲线

Fig.6 Prediction curves of different features

表 3 不同滑窗宽度下的预测误差

Table 3 Prediction errors under different sliding window widths

滑窗宽度	E_MAP/%	E_RMS/MW	E_MA/MW	R²
24	0.827	98.836	71.872	0.9946
48	0.707	82.788	62.791	0.9963
72	0.696	76.252	60.165	0.9968
96	0.635	71.877	55.285	0.9971
128	0.670	74.018	58.375	0.9969
192	0.875	94.865	76.318	0.9950
384	1.287	134.845	109.954	0.9899

图 7 不同滑窗宽度预测曲线

Fig.7 Prediction curves of different sliding window widths

表 4 不同模型的预测误差

Table 4 Prediction errors of different models

预测模型	E_MAP/%	E_RMS/MW	E_MA/MW	R²
MSCNN-BiGRU-Attention	0.635	71.877	55.285	0.9971
CNN-BiGRU-Attention	0.700	76.110	60.922	0.9968
CNN-BiGRU	0.904	103.141	79.401	0.9941
BiGRU	1.181	137.806	102.788	0.9895
GRU	1.306	140.546	108.669	0.9891
LSTM	1.365	146.344	114.527	0.9881

图 8 不同模型预测曲线

Fig.8 Prediction curves of different models

表 5 不同模型的预测误差（国内案例）

Table 5 Prediction errors of different models (Chinese case)

预测模型	E_MAP/%	E_RMS/MW	E_MA/MW	R²
MSCNN-BiGRU-Attention	1.038	82.224	67.581	0.9979
CNN-BiGRU-Attention	1.461	148.765	105.015	0.9931
CNN-BiGRU	1.955	173.029	131.370	0.9907
BiGRU	2.227	187.769	144.359	0.9891
GRU	2.502	207.248	162.0459	0.9867
LSTM	3.865	310.789	251.897	0.9701

图 9 不同模型预测曲线（国内案例）

Fig.9 Prediction curves of different models (Chinese case)

参考文献 25

1	徐鹏, 张飞龙, 杜景勃. 深度学习方法在能源负荷预测的应用[J]. 煤气与热力, 2023, 43 (4): 1- 14. DOI
	XU Peng, ZHANG Feilong, DU Jingbo. Application of deep learning methods in energy load forecasting[J]. Gas & Heat, 2023, 43 (4): 1- 14. DOI
2	朱明, 夏宇栋, 常凯, 等. 基于粒子群优化算法的空调负荷灰箱模型辨识[J]. 电力科学与技术学报, 2023, 38 (4): 214- 221.
	ZHU Ming, XIA Yudong, CHANG Kai, et al. Identification of grey box model for air conditioning load based on particle swarm optimization algorithm[J]. Journal of Electric Power Science and Technology, 2023, 38 (4): 214- 221.
3	贾巍, 雷才嘉, 方兵华, 等. 供电分区场景下基于数据驱动的负荷密度综合评估及预测方法[J]. 中国电力, 2023, 56 (6): 71- 81.
	JIA Wei, LEI Caijie, FANG Huabing, et al. A comprehensive evaluation and prediction method for load density based on big data under power supply partition scenarios[J]. Electric Power, 2023, 56 (6): 71- 81.
4	刘杰, 从兰美, 夏远洋, 等. 基于DBO-VMD和IWOA-BILSTM神经网络组合模型的短期电力负荷预测[J]. 电力系统保护与控制, 2024, 52 (8): 123- 133.
	LIU Jie, CONG Meilan, XIA Yuanyang, et al. Short-term power load prediction based on DBO-VMD and an IWOA-BILSTM neural network combination model[J]. Power System Protection and Control, 2024, 52 (8): 123- 133.
5	黄冬梅, 陈欢, 王宁, 等. 基于自适应图注意力网络的短期用户负荷预测[J]. 电力系统保护与控制, 2023, 51 (20): 140- 149.
	HUANG Dongmei, CHEN Huan, WANG Ning, et al. Short-term user load prediction based on an adaptive graph attention network[J]. Power System Protection and Control, 2023, 51 (20): 140- 149.
6	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. DOI
	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. DOI
7	姜东良, 李天昊, 刘文浩. 基于相似日和SAE-DBiLSTM模型的短期电力负荷预测[J]. 电气工程学报, 2022, 17 (4): 240- 249.
	JIANG Dongliang, LI Tianhao, LIU Wenhao. Short-term power load forecasting using similar day and SAE-DBiLSTM model[J]. Journal of Electrical Engineering, 2022, 17 (4): 240- 249.
8	李科, 李龙胜, 肖钰皓. 基于时间序列分析的负荷高峰时段用电预测与削峰研究[J]. 光学与光电技术, 2022, 20 (4): 153- 159. DOI
	LI Ke, LI Longsheng, XIAO Yuhao. Research on power consumption forecasting and peak shaving during load period based on time series analysis[J]. Optics & Optoelectronics Technology, 2022, 20 (4): 153- 159. DOI
9	MADHUKUMAR M, SEBASTIAN A, LIANG X D, et al. Regression model-based short-term load forecasting for university campus load[J]. IEEE Access, 2022, 10, 8891- 8905. DOI
10	王艳松, 申晓阳, 李强, 等. 基PCA-GRD-LWR模型的海上油田中长期最大电力负荷预测[J]. 中国石油大学学报(自然科学版), 2023, 47 (2): 129- 135.
	WANG Yansong, SHEN Xiaoyang, LI Qiang, et al. Forecasting of medium and long-term maximum power load for offshore oilfields based on PCA-GRD-LWR model[J]. Journal of China University of Petroleum (Edition of Natural Science), 2023, 47 (2): 129- 135.
11	曾囿钧, 肖先勇, 徐方维. 基于小波变换与BiGRU-NN模型的短期负荷预测方法[J]. 电测与仪表, 2023, 60 (6): 103- 109.
	ZENG Menjun, XIAO Xianyong, XU Fangwei. Short-term load forecasting method based on wavelet transform and BiGRU-NN model[J]. Electrical Measurement & Instrumentation, 2023, 60 (6): 103- 109.
12	魏步晗, 鲍刚, 李振华. 基于支持向量回归预测模型考虑天气因素和分时电价因素的短期电力负荷预测[J]. 电网与清洁能源, 2023, 39 (11): 9- 19. DOI
	WEI Buhan, BAO Gang, LI Zhenhua. Short-term electricity load forecasting based on support vector regression forecasting model considering weather factors and time-of-use tariff factors[J]. Power System and Clean Energy, 2023, 39 (11): 9- 19. DOI
13	蔡颖凯, 张冶, 曹世龙, 等. 基于决策树算法的短期电力负荷大数据预测模型[J]. 制造业自动化, 2022, 44 (06): 152- 155, 182. DOI
	CAI Yingkai, ZHANG Ye, CAO Shilong, et al. Short-term power load big data forecasting model based on decision tree algorithm[J]. Manufacturing Automation, 2022, 44 (06): 152- 155, 182. DOI
14	SHEN Q, MO L, LIU G J, et al. Short-term load forecasting based on multi-scale ensemble deep learning neural network[J]. IEEE Access, 2023, 11, 111963- 111975. DOI
15	钱育树, 孔钰婷, 黄聪. 电力负荷预测研究综述[J]. 四川电力技术, 2023, 46 (4): 37- 43, 58.
	QIAN Yushu, KONG Yuting, HUANG Cong. Review of power load forecasting[J]. Sichuan Electric Power Technology, 2023, 46 (4): 37- 43, 58.
16	LI Z L, LI Y Z, LIU Y, et al. Deep learning based densely connected network for load forecasting[J]. IEEE Transactions on Power Systems, 2021, 36 (4): 2829- 2840. DOI
17	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. DOI
18	曾囿钧, 肖先勇, 徐方维, 等. 基于CNN-BiGRU-NN模型的短期负荷预测方法[J]. 中国电力, 2021, 54 (9): 17- 23.
	ZENG Youjun, XIAO Xianyong, XU Fangwei, et al. A short-term load forecasting method based on CNN-BiGRU-NN model[J]. Electric Power, 2021, 54 (9): 17- 23.
19	张宇帆, 艾芊, 林琳, 等. 基于深度长短时记忆网络的区域级超短期负荷预测方法[J]. 电网技术, 2019, 43 (6): 1884- 1892.
	ZHANG Yufan, AI Qian, LIN Lin, et al. A very short-term load forecasting method based on deep LSTM RNN at zone level[J]. Power System Technology, 2019, 43 (6): 1884- 1892.
20	关蕾. 基于改进DBN的用电负荷预测建模与仿真[J]. 微型电脑应用, 2022, 38 (2): 42- 45, 48. DOI
	GUAN Lei. Modeling and simulation of power load forecasting based on improved DBN[J]. Microcomputer Applications, 2022, 38 (2): 42- 45, 48. DOI
21	赵厚翔, 沈晓东, 吕林, 等. 基于GAN的负荷数据修复及其在EV短期负荷预测中的应用[J]. 电力系统自动化, 2021, 45 (16): 143- 151. DOI
	ZHAO Houxiang, SHEN Xiaodong, LYU Lin, et al. Load data restoration based on generative adversarial network and its application in short-term load forecasting of electric vehicle[J]. Automation of Electric Power Systems, 2021, 45 (16): 143- 151. DOI
22	任爽, 杨凯, 商继财, 等. 基于CNN-BiGRU-Attention的短期电力负荷预测[J]. 电气工程学报, 2024, 19 (1): 344- 350.
	REN Shuang, YANG Kai, SHANG Jicai, et al. Short-term power load forecasting based on CNN-BiGRU-Attention[J]. Journal of Electrical Engineering, 2024, 19 (1): 344- 350.
23	安文杰, 陈长征, 田淼, 等. 基于MSCNNSA-BiGRU的变工况风电机组滚动轴承故障诊断研究[J]. 机电工程, 2022, 39 (8): 1096- 1103. DOI
	AN Wenjie, CHEN Changzheng, TIAN Miao, et al. Rolling bearing fault diagnosis of wind turbine under variable working conditions based on MSCNSA-BIGRU[J]. Journal of Mechanical & Electrical Engineering, 2022, 39 (8): 1096- 1103. DOI
24	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.
25	吴军英, 路欣, 刘宏, 等. 基于Spearman-GCN-GRU模型的超短期多区域电力负荷预测[J]. 中国电力, 2024, 57 (6): 131- 140.
	WU Junying, LU Xin, LIU Hong, et al. Ultra-short-term multi-region power load forecasting based on Spearman-GCN-GRU Model[J]. Electric Power, 2024, 57 (6): 131- 140.

[1]	沈鑫, 王钢, 赵毅涛, 骆钊, 李钊, 杨晓华. 融合SENet注意力机制和GA-CNN的非侵入式负荷识别方法[J]. 中国电力, 2025, 58(5): 33-42.
[2]	姜通海, 王峰, 刘子琪, 单帅杰. 基于改进生成对抗网络的风光气象资源联合场景生成方法[J]. 中国电力, 2025, 58(3): 183-192.
[3]	朱沐雨, 马宏忠, 郭鹏宇, 宣文婧. 典型调峰/调频工况下储能电池组荷电状态估计[J]. 中国电力, 2024, 57(6): 18-26.
[4]	吴军英, 路欣, 刘宏, 张彬, 柴守亮, 刘蕴春, 王佳楠. 基于Spearman-GCN-GRU模型的超短期多区域电力负荷预测[J]. 中国电力, 2024, 57(6): 131-140.
[5]	高岩, 吴汉斌, 张纪欣, 张华铭, 张沛. 基于组合深度学习的光伏功率日前概率预测模型[J]. 中国电力, 2024, 57(4): 100-110.
[6]	吴晓刚, 阎洁, 葛畅, 唐雅洁, 倪筹帷, 季青锋. 基于改进GRU-CNN的风光水一体化超短期功率预测方法[J]. 中国电力, 2023, 56(9): 178-186,205.
[7]	刘东, 张越, 皮俊波, 单连飞, 刘赫, 宋鹏程, 姜涛. 面向电网调控信息智能检索的知识图谱构建及应用[J]. 中国电力, 2023, 56(7): 78-84.
[8]	张鑫, 叶俊杰, 崔瑶, 黄鑫, 仲林林. 基于语义信息距离解耦的变电运维多类别缺陷图像检测[J]. 中国电力, 2023, 56(6): 209-218.
[9]	陆友文, 崔昊, 陈佳宁, 彭祥佳, 冯双, 刘栋. 基于RA-CNN和同步相量的风电场次/超同步振荡参数智能辨识方法[J]. 中国电力, 2023, 56(4): 46-55,67.
[10]	苏向敬, 宇海波, 符杨, 田书欣, 李海瑜, 耿福海. 基于DALSTM和联合分位数损失的海上风电功率概率预测[J]. 中国电力, 2023, 56(11): 10-19.
[11]	周震震, 宋云海, 何宇浩, 王黎伟, 黄和燕, 何珏, 朱志航, 闫云凤. 基于分组查询注意力的可扩展电力人员行为分类方法[J]. 中国电力, 2023, 56(11): 77-85.
[12]	李冰, 白云山, 赵宽, 郭聪彬, 翟永杰. 基于HSCA-YOLOv7的风电机组叶片表面缺陷检测算法[J]. 中国电力, 2023, 56(10): 43-52.
[13]	陈铁, 张治藩, 李咸善, 陈一夫, 李鸿鑫. 基于混合模态分解和LSTM-CNN的变压器油中溶解气体浓度预测[J]. 中国电力, 2023, 56(1): 132-141.
[14]	李文武, 石强, 李丹, 胡群勇, 唐芸, 梅锦超. 基于VMD和PSO-SVR的短期电力负荷多阶段优化预测[J]. 中国电力, 2022, 55(8): 171-177.
[15]	樊江川, 于昊正, 刘慧婷, 杨丽君, 安佳坤. 基于多分支门控残差卷积神经网络的短期电力负荷预测[J]. 中国电力, 2022, 55(11): 155-162,174.