基于BiLSTM网络与误差修正的超短期负荷预测

doi:10.3969/j.issn.2097-0706.2023.01.004

摘要/Abstract

摘要：

电力负荷预测对于电力系统电量供需平衡、经济运行具有重要意义，电力负荷具有随机性、波动性等不确定性特征且易受天气因素影响，负荷准确预测存在技术挑战。提出了一种基于双向长短期记忆（BiLSTM）神经网络与误差修正的超短期负荷预测模型，采用最大信息系数描述各影响因素与负荷的关系，并进一步对输入特征进行筛选；考虑负荷变量数值序列的时序性，利用BiLSTM网络建立负荷预测模型，针对预测结果误差，采用自适应噪声的完备集合经验模态分解（CEEMDAN）算法将误差结果序列分解为若干分量，每个误差分量分别再建立BiLSTM预测模型。以我国北方某地区配电网实际负荷数据为算例，采用不同神经网络模型进行对比测试，结果表明该模型具有更高的准确度。

关键词: 电力负荷, 超短期负荷预测, BiLSTM神经网络, CEEMDAN算法, 误差修正, 最大信息系数

Abstract:

Accurate power load forecasting is important to maintain the balance of power supply and demand and the economy and stability of power system. Since power load are random and volatile under the influence of meteorological factors，accurate forecasting on power load is a technical challenge. An ultra-short-term load forecasting model is built based on bi-directional long short-term memory （BiLSTM） neural network and error correction. Maximum information coefficient （MIC） is used to describe the nonlinear relationships between the various influencing factors and load data， so as to filter the input characteristics. Considering the time-series characteristics of the load sequence， the initial load forecasting model is established based on the BiLSTM network. To lessen the prediction error， complete ensemble empirical mode decomposition with adaptive noise（CEEMDAN）algorithm is used to decompose the error sequence into several error components， whose BiLSTM prediction models are built respectively to modify the initial predicted results. The simulation experiment on a power distribution network in Tianjin， a northern municipality in China， is carried out. The experimental results show that the prediction values made by the BiLSTM-based model is more accurate that made by the models based on other neural networks.

Key words: power load, ultra-short-term load forecasting, BiLSTM neural network, CEEMDAN, error correction, MIC

中图分类号:

TK01：TM715

高明, 郝妍. 基于BiLSTM网络与误差修正的超短期负荷预测[J]. 综合智慧能源, 2023, 45(1): 31-40.

GAO Ming, HAO Yan. Ultra-short-term load forecasting based on BiLSTM network and error correction[J]. Integrated Intelligent Energy, 2023, 45(1): 31-40.

图/表 17

图1

图2

图3

图4

图5

表1

图6

图7

图8

图9

表2

图10

图11

图12

图13

表3

图14

参考文献 20

[1]	舒印彪, 赵勇, 赵良, 等. “双碳”目标下我国能源电力低碳转型路径[J/OL]. 中国电机工程学报:1-9[2022-10-12]. https://doi.org/10.13334/j.0258-8013.pcsee.221407. doi: https://doi.org/10.13334/j.0258-8013.pcsee.221407
	SHU Yinbiao, ZHAO Yong, ZHAO Liang, et al. Study on low carbon energy transition path toward carbon peakand carbon neutrality[J/OL]. Proceedings of the CSEE:1-9[2022-10-12]. https://doi.org/10.13334/j.0258-8013.pcsee.221407. doi: https://doi.org/10.13334/j.0258-8013.pcsee.221407
[2]	张照贝, 顾春华, 温蜜. 基于XGBoost和QRLSTM的超短期负荷预测方法[J]. 计算机仿真, 2022, 39(1):90-95,110.
	ZHANG Zhaobei, GU Chunhua, WEN Mi, et al. Ultra-short-term load forecasting method based on XGBoost and QRLSTM[J]. Computer Simulation, 2022, 39(1):90-95,110.
[3]	陈培垠, 方彦军. 基于卡尔曼滤波预测节假日逐点增长率的电力系统短期负荷预测[J]. 武汉大学学报(工学版), 2020, 53(2):139-144.
	CHEN Peiyin, FANG Yanjun. Short-term load forecasting of power system for holiday point-by-point growth rate based on Kalman filtering[J]. Engineering Journal of Wuhan University, 2020, 53(2):139-144.
[4]	邓带雨, 李坚, 张真源, 等. 基于EEMD-GRU-MLR的短期电力负荷预测[J]. 电网技术, 2020, 44(2):593-602.
	DENG Daiyu, LI Jian, ZHANG Zhenyuan, et al. Short-term electric load forecasting based on EEMD-GRU-MLR[J]. Power System Technology, 2020, 44(2):593-602.
[5]	方娜, 李俊晓, 陈浩, 等. 基于变分模态分解的卷积神经网络-双向门控循环单元-多元线性回归多频组合短期电力负荷预测[J]. 现代电力, 2022, 39(4):441-448.
	FANG Na, LI Junxiao, CHEN Hao, et al. Multi-frequency combination short-term power load forecasting with convolutional neural networks-bidirectional gated recurrent unit-multiple linear regression based on variational mode decomposition[J]. Modern Electric Power, 2022, 39(4):441-448.
[6]	甘景福, 晏坤, 马明晗, 等. 基于改进聚类算法的人工神经网络短期负荷预测研究[J]. 电工电能新技术, 2022, 41(9):40-46.
	GAN Jingfu, YAN Kun, MA Minghan, et al. Research on short-term load forecasting based on modified clustering and artificial neural networks[J]. Advanced Technology of Electrical Engineering and Energy, 2022, 41(9):40-46.
[7]	姚程文, 杨苹, 刘泽健. 基于CNN-GRU混合神经网络的负荷预测方法[J]. 电网技术, 2020, 44(9):3416-3424.
	YAO Chengwen, YANG Ping, LIU Zejian. Load forecasting method based on CNN-GRU hybrid neural network[J]. Power System Technology, 2020, 44(9):3416-3424.
[8]	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 Energy, 2022, 41(6):71-79.
[9]	王继东, 杜冲. 基于Attention-BiLSTM神经网络和气象数据修正的短期负荷预测模型[J]. 电力自动化设备, 2022, 42(4):172-177,224.
	WANG Jidong, DU Chong. Short-term load prediction model based on Attention-BiLSTM neural network and meteorological data correction[J]. Electric Power Automation Equipment, 2022, 42(4):172-177,224.
[10]	李彬, 胡纯瑾, 王婧. 基于EEMD-BiLSTM的可调节负荷预测方法[J]. 综合智慧能源, 2022, 44(9):33-39. doi: 10.3969/j.issn.2097-0706.2022.09.005
	LI Bin, HU Chunjin, WANG Jing. Prediction method for adjustable load based on EEMD-BiLSTM[J]. Integrated Intelligent Energy, 2022, 44(9):33-39. doi: 10.3969/j.issn.2097-0706.2022.09.005
[11]	李玉志, 刘晓亮, 邢方方, 等. 基于Bi-LSTM和特征关联性分析的日尖峰负荷预测[J]. 电网技术, 2021, 45(7):2719-2730.
	LI Yuzhi, LIU Xiaoliang, XING Fangfang, et al. Daily peak load prediction based on correlation analysis and bi-directional long short-term memory network[J]. Power System Technology, 2021, 45(7):2719-2730.
[12]	任建吉, 位慧慧, 邹卓霖, 等. 基于CNN-BiLSTM-Attention的超短期电力负荷预测[J]. 电力系统保护与控制, 2022, 50(8):108-116.
	REN Jianji, WEI Huihui, ZOU Zhuolin, et al. Ultra-short-term power load forecasting based on CNN-BiLSTM-Attention[J]. Power System Protection and Control, 2022, 50(8):108-116.
[13]	孙辉, 杨帆, 高正男, 等. 考虑特征重要性值波动的MI-BILSTM短期负荷预测[J]. 电力系统自动化, 2022, 46(8):95-103.
	SUN Hui, YANG Fan, GAO Zhengnan, et al. Short-term load forecasting based on mutual information and bi-directional long short-term memory network considering fluctuation in importance values of features[J]. Automation of Electric Power Systems, 2022, 46(8):95-103.
[14]	JAVED U, IJAZ K, JAWAD M, et al. A novel short receptive field based dilated causal convolutional network integrated with bidirectional LSTM for short-term load forecasting[J]. Expert Systems with Applications, 2022, 205:117689. doi: 10.1016/j.eswa.2022.117689
[15]	ZHANG C, HUA L, JI C, et al. An evolutionary robust solar radiation prediction model based on WT-CEEMDAN and IASO-optimized outlier robust extreme learning machine[J]. Applied Energy, 2022, 322(9):119518. doi: 10.1016/j.apenergy.2022.119518
[16]	WANG J, LUO Y, TANG L, et al. A new weighted CEEMDAN-based prediction model: An experimental investigation of decomposition and non-decomposition approaches[J]. Knowledge-Based Systems, 2018, 160(11):188-199. doi: 10.1016/j.knosys.2018.06.033
[17]	肖小刚, 莫莉, 张祥, 等. 基于CEEMDAN+RF+AdaBoost的短期负荷预测[J]. 水电能源科学, 2020, 38(4):181-184,175.
	XIAO Xiaogang, MO Li, ZHANG Xiang, et al. Short-term load forecasting based on CEEMDAN+RF+AdaBoost[J]. Water Resources and Power, 2020, 38(4):181-184,175.
[18]	乔石, 王磊, 张鹏超, 等. 基于模态分解及注意力机制长短时间网络的短期负荷预测[J/OL]. 电网技术:1-13[2022-10-12]. https://doi.org/10.13335/j.1000-3673.pst.2022.0368. doi: https://doi.org/10.13335/j.1000-3673.pst.2022.0368
	QIAO Shi, WANG Lei, ZHANG Pengchao, et al. Short-term load forecasting by long and short-term temporal networks with attention based on modal decomposition[J/OL]. Power System Technology:1-13[2022-10-12]. https://doi.org/10.13335/j.1000-3673.pst.2022.0368. doi: https://doi.org/10.13335/j.1000-3673.pst.2022.0368
[19]	HU H, XIA X, LUO Y, et al. Development and application of an evolutionary deep learning framework of LSTM based on improved grasshopper optimization algorithm for short-term load forecasting[J]. Journal of Building Engineering, 2022, 57:104975. doi: 10.1016/j.jobe.2022.104975
[20]	葛磊蛟, 刘航旭, 赵康, 等. 面向商业和居民混合的配电网短期负荷预测HGWOACOA-LSTMN方法[J]. 天津大学学报(自然科学与工程技术版), 2021, 54(12):1269-1279.
	GE Leijiao, LIU Hangxu, ZHAO Kang, et al. An HGWOACOA-LSTMN method for short-term load forecasting of distribution network for commercial and residential users[J]. Journal of Tianjin University(Science and Technology), 2021, 54(12):1269-1279.

影响因素	MIC值
预测日的天气类型	0.21
预测日的最高气温	0.51
预测日的最低气温	0.53
预测日的平均气温	0.48
预测日的日期类型	0.22

模型	δ_MAE/MW	δ_RMSE/MW	δ_MAPE/%
BP	6.72	7.95	6.67
BiLSTM	2.52	3.27	2.41
BiLSTM-BiLSTM	2.49	3.19	2.39
BiLSTM-EMD-BiLSTM	2.37	3.31	2.37
CNN-BiLSTM-Attention	2.09	2.82	2.22
本文方法	1.23	1.49	1.27

模型	δ_MAE/MW	δ_RMSE/MW	δ_MAPE/%
BP	4.50	6.58	2.36
BiLSTM	3.92	6.19	2.05
BiLSTM-BiLSTM	9.16	12.41	4.79
BiLSTM-EMD-BiLSTM	3.79	5.28	1.98
CNN-BiLSTM-Attention	3.11	4.90	1.63
本文方法	2.86	4.83	1.53