基于多维聚类的配变负荷注意力短期预测方法

doi:10.16183/j.cnki.jsjtu.2021.263

摘要/Abstract

摘要：

在大规模配变负荷预测中,由于负荷特性差别以及受影响因素不同,若使用统一模型,准确率低且泛化能力差,若针对单台配变进行负荷预测建模,计算资源消耗过大.提出了一种基于多维聚类的配变负荷注意力长短期记忆网络(Attention Long Short-Term Memory,Attention-LSTM)短期预测方法.首先提取每个配变日负荷特征序列并利用非参数核方法进行概率拟合,形成配变负荷的典型日负荷序列;以欧式归整距离以及影响因素相似性作为相似度评判标准,使用改进的k均值聚类(k-means)双层聚类对日典型负荷序列进行负荷聚类分析;利用近邻传播(Affinity Propagation,AP)聚类提取影响因素相似时间序列,构建训练集,训练Attention-LSTM模型;针对不同的配变负荷类型以及不同的相似时间序列得到不同的Attention-LSTM模型.通过选取某市级配电网实测负荷数据以及气象等影响因素数据,验证了所提方法的有效性和实用性,准确率提升了2.75%且效率提升了616.8%.

关键词: 短期负荷预测, 日负荷序列, 负荷聚类, 相似时间序列, 长短期记忆网络

Abstract:

Due to the difference in load characteristics and influencing factors in large-scale distribution transformer load forecasting, if all the distribution transformers share a unified model, the prediction accuracy is low, and if the model is built for each distribution transformer, the computational resources will be excessively consumed. An Attention-LSTM short-term forecasting method of distribution load based on multi-dimensional clustering is proposed. The non-parametric kernel method is used to perform probability fitting on the daily load characteristics to form a typical daily load sequence. Improved two-level clustering is applied for load clustering, taking the Euclidean warping distance and influence factors as the similarity evaluation criteria. AP clustering is utilized for obtaining similar time-series, and training sets are formed to train the Attention-LSTM model. Different Attention-LSTM models are obtained by training for different distribution load types and time-series. The effectiveness and practicability of the method proposed are verified by the load data and meteorological data of a municipal distribution network. The accuracy rate is increased by 2.75% and the efficiency is increased by 616.8%.

Key words: short-term load forecasting, daily load sequence, load clustering, similar time-series, long short-term memory network

中图分类号:

TM715

钟光耀, 邰能灵, 黄文焘, 李然, 傅晓飞, 纪坤华. 基于多维聚类的配变负荷注意力短期预测方法[J]. 上海交通大学学报, 2021, 55(12): 1532-1543.

ZHONG Guangyao, TAI Nengling, HUANG Wentao, LI Ran, FU Xiaofei, JI Kunhua. Attention Short-Term Forecasting Method of Distribution Load Based on Multi-Dimensional Clustering[J]. Journal of Shanghai Jiao Tong University, 2021, 55(12): 1532-1543.

图/表 14

图1

表1

图2

图3

表2

图4

图5

图6

表3

表4

图7

图8

表5

图9

参考文献 27

[1]	谭显东, 刘俊, 徐志成, 等. “双碳”目标下“十四五”电力供需形势[J]. 中国电力, 2021, 54(5):1-6.
	TAN Xiandong, LIU Jun, XU Zhicheng, et al. Power supply and demand balance during the 14th five-year plan period under the goal of carbon emission peak and carbon neutrality[J]. Electric Power, 2021, 54(5):1-6.
[2]	XIA C H, ZHANG M, CAO J. A hybrid application of soft computing methods with wavelet SVM and neural network to electric power load forecasting[J]. Journal of Electrical Systems and Information Technology, 2018, 5(3):681-696. doi: 10.1016/j.jesit.2017.05.008 URL
[3]	LI Y Y, HAN D, YAN Z. Long-term system load forecasting based on data-driven linear clustering method[J]. Journal of Modern Power Systems and Clean Energy, 2018, 6(2):306-316. doi: 10.1007/s40565-017-0288-x URL
[4]	康重庆. 电力系统负荷预测[M]. 北京: 中国电力出版社, 2017.
	KANG Chongqing. Power system load forecast[M]. Beijing: China Electric Power Press, 2017.
[5]	崔和瑞, 彭旭. 基于ARIMAX模型的夏季短期电力负荷预测[J]. 电力系统保护与控制, 2015, 43(4):108-114.
	CUI Herui, PENG Xu. Summer short-term load forecasting based on ARIMAX model[J]. Power System Protection and Control, 2015, 43(4):108-114.
[6]	谢开, 汪峰, 于尔铿, 等. 应用Kalman滤波方法的超短期负荷预报[J]. 中国电机工程学报, 1996, 16(4):245-249.
	XIE Kai, WANG Feng, YU Erken, et al. Very short-term load forecasting by Kalman filter algorithm[J]. Chinese Society for Electrical Engineering, 1996, 16(4):245-249.
[7]	路轶, 王民昆. 基于短期负荷预测的超短期负荷预测曲线外推法[J]. 电力系统自动化, 2006, 30(16):102-104.
	LU Yi, WANG Minkun. An ultra-short term load forecasting method based on short-term load forecasting[J]. Automation of Electric Power Systems, 2006, 30(16):102-104.
[8]	马文晓, 白晓民, 沐连顺. 基于人工神经网络和模糊推理的短期负荷预测方法[J]. 电网技术, 2003, 27(5):29-32.
	MA Wenxiao, BAI Xiaomin, MU Lianshun. Short term load forecasting using artificial neuron network and fuzzy inference[J]. Power System Technology, 2003, 27(5):29-32.
[9]	耿艳, 韩学山, 韩力. 基于最小二乘支持向量机的短期负荷预测[J]. 电网技术, 2008, 32(18):72-76.
	GENG Yan, HAN Xueshan, HAN Li. Short-term load forecasting based on least squares support vector machines[J]. Power System Technology, 2008, 32(18):72-76.
[10]	徐军华, 刘天琪. 基于小波分解和人工神经网络的短期负荷预测[J]. 电网技术, 2004, 28(8):30-33.
	XU Junhua, LIU Tianqi. An approach to short-term load forecasting based on wavelet transform and aritficial neural network[J]. Power System Technology, 2004, 28(8):30-33.
[11]	邹红波, 伏春林, 喻圣. 基于Akima-LMD和GRNN的短期负荷预测[J]. 电工电能新技术, 2018, 37(1):51-56.
	ZOU Hongbo, FU Chunlin, YU Sheng. Short-term load forecasting based on Akima-LMD and GRNN[J]. Advanced Technology of Electrical Engineering and Energy, 2018, 37(1):51-56.
[12]	ALMALAQ A, EDWARDS G. A review of deep learning methods applied on load forecasting[C]// 2017 16th IEEE International Conference on Machine Learning and Applications (ICMLA). Piscataway, NJ, USA: IEEE, 2017: 511-516.
[13]	CINAR Y G, MIRISAEE H, GOSWAMI P, et al. Period-aware content attention RNNs for time series forecasting with missing values[J]. Neurocomputing, 2018, 312:177-186. doi: 10.1016/j.neucom.2018.05.090 URL
[14]	杨龙, 吴红斌, 丁明, 等. 新能源电网中考虑特征选择的Bi-LSTM网络短期负荷预测[J]. 电力系统自动化, 2021, 45(3):166-173.
	YANG Long, WU Hongbin, DING Ming, et al. Short-term load forecasting in renewable energy grid based on Bi-directional long short-term memory network considering feature selection[J]. Automation of Electric Power Systems, 2021, 45(3):166-173.
[15]	陈振宇, 刘金波, 李晨, 等. 基于LSTM与XGBoost组合模型的超短期电力负荷预测[J]. 电网技术, 2020, 44(2):614-620.
	CHEN Zhenyu, LIU Jinbo, LI Chen, et al. Ultra short-term power load forecasting based on combined LSTM-XGBoost model[J]. Power System Technology, 2020, 44(2):614-620.
[16]	陆继翔, 张琪培, 杨志宏, 等. 基于CNN-LSTM混合神经网络模型的短期负荷预测方法[J]. 电力系统自动化, 2019, 43(8):131-137.
	LU Jixiang, ZHANG Qipei, YANG Zhihong, 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.
[17]	ALMALAQ A, EDWARDS G. A review of deep learning methods applied on load forecasting[C]// 2017 16th IEEE International Conference on Machine Learning and Applications (ICMLA). Piscataway, NJ, USA: IEEE, 2017: 511-516.
[18]	朱文俊, 王毅, 罗敏, 等. 面向海量用户用电特性感知的分布式聚类算法[J]. 电力系统自动化, 2016, 40(12):21-27.
	ZHU Wenjun, WANG Yi, LUO Min, et al. Distributed clustering algorithm for awareness of electricity consumption characteristics of massive consumers[J]. Automation of Electric Power Systems, 2016, 40(12):21-27.
[19]	刘洋, 许立雄. 适用于海量负荷数据分类的高性能反向传播神经网络算法[J]. 电力系统自动化, 2018, 42(21):96-103.
	LIU Yang, XU Lixiong. High-performance back propagation neural network algorithm for classification of mass load data[J]. Automation of Electric Power Systems, 2018, 42(21):96-103.
[20]	VIEGAS J L, VIEIRA S M, SOUSA J M C, et al. Electricity demand profile prediction based on household characteristics[C]// 2015 12th International Conference on the European Energy Market (EEM). Piscataway, NJ, USA: IEEE, 2015: 1-5.
[21]	AL-OTAIBI R, JIN N L, WILCOX T, et al. Feature construction and calibration for clustering daily load curves from smart-meter data[J]. IEEE Transactions on Industrial Informatics, 2016, 12(2):645-654. doi: 10.1109/TII.9424 URL
[22]	杨德昌, 赵肖余, 何绍文, 等. 面向海量用户用电数据的集成负荷预测[J]. 电网技术, 2018, 42(9):2923-2929.
	YANG Dechang, ZHAO Xiaoyu, HE Shaowen, et al. Aggregated load forecasting based on massive household smart meter data[J]. Power System Technology, 2018, 42(9):2923-2929.
[23]	MOORE B. Principal component analysis in linear systems: Controllability, observability, and model reduction[J]. IEEE Transactions on Automatic Control, 1981, 26(1):17-32. doi: 10.1109/TAC.1981.1102568 URL
[24]	SAINATH T N, KINGSBURY B, RAMABHADRAN B. Auto-encoder bottleneck features using deep belief networks[C]// 2012 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). Piscataway, NJ, USA: IEEE, 2012: 4153-4156.
[25]	张宁, 刘天键. 考虑影响因素的短期负荷预测核函数ELM方法[J]. 武汉大学学报(工学版), 2018, 51(8):703-707.
	ZHANG Ning, LIU Tianjian. Kernel function ELM method for short-term load forecasting considering influencing factors[J]. Engineering Journal of Wuhan University, 2018, 51(8):703-707.
[26]	陈鸿琳, 李欣然, 冷华, 等. 运用PSO和GRNN的短期负荷二维组合预测[J]. 电力系统及其自动化学报, 2018, 30(2):85-89.
	CHEN Honglin, LI Xinran, LENG Hua, et al. Bidirectional combined short-term load forecasting by using PSO and GRNN[J]. Proceedings of the CSU-EPSA, 2018, 30(2):85-89.
[27]	贾慧敏, 何光宇, 方朝雄, 等. 用于负荷预测的层次聚类和双向夹逼结合的多层次聚类法[J]. 电网技术, 2007, 31(23):33-36.
	JIA Huimin, HE Guangyu, FANG Chaoxiong, et al. Multi-level clustering method for hierarchical clustering and bidirectional capping combination for load forecasting[J]. Power System Technology, 2007, 31(23):33-36.

时段		指标	定义	变量
全天	00:00—24:00	负荷率	P_av/P_max	y₁
		日峰谷差率	(P_max-P_min)/P_max	y₂
峰期	08:00—11:00	峰期负载率	P_{av, peak}/P_av	y₃
	18:00—21:00
平期	06:00—08:00	平期负载率	P_{av, sh}/P_av	y₄
	11:00—18:00
	21:00—22:00
谷期	22:00—24:00	谷期负载率	P_{av, val}/P_av	y₅
	00:00—06:00

天气	标记值
晴	0
多云	0.2
阴	0.4
小雨	0.6
中雨	0.8
大雨	1.0
工作日	0
周末	0.5
节假日	1

k	DBI
2	0.91
3	1.37
4	1.12
5	0.75
6	1.05
7	1.41
8	1.10
9	1.43
10	1.54
11	0.82
12	0.97

类别	信息熵
第1类	0.216
第2类	0.483
第3类	0.149
第4类	0.134
第5类	0.201

组别	MAPE/%						RMSE/KW						训练时间/s	泛化时间/s
组别	第1类	第2类	第3类	第4类	第5类	第6类	第1类	第2类	第3类	第4类	第5类	第6类	训练时间/s	泛化时间/s
实验组	1.39	1.56	5.27	1.31	1.50	1.21	0.043	0.047	0.159	0.040	0.047	0.037	93.7	0.076
对比组1	5.21	4.71	7.32	5.33	5.69	4.23	0.157	0.138	0.226	0.161	0.179	0.132	655.6	0.080
对比组2	3.35	3.47	6.12	4.15	4.22	3.09	0.104	0.106	0.189	0.128	0.129	0.093	671.7	0.079
对比组3	4.61	4.15	7.08	5.11	4.39	3.91	0.141	0.126	0.221	0.151	0.130	0.116	6.53	0.072