基于图卷积网络和风速差分拟合的中长期风功率预测

doi:10.11930/j.issn.1004-9649.202303050

摘要/Abstract

摘要：

为充分利用数据特征间的先验关系，提高风电场中长期发电功率预测精度，提出一种基于图卷积神经网络（GCN）、风速差分拟合（DF）、粒子群优化算法（PSO）的中长期风功率预测模型。通过分析风力发电全过程，挖掘风功率影响因素及因素间的相互关联性，搭建GCN模型，分别拟合风速和功率利用效率，进一步结合基于DF的风速-功率计算模型计算风功率，模型的损失包含功率损失、风速损失和功率利用效率损失3个部分，采用粒子群优化算法为这3部分损失确定合适的权重。2个风电场的实际算例表明，该模型未来10天风功率预测的相对均方根误差分别为11.44%和13.09%，具有较高的预测精度。

关键词: 风力发电, 风功率预测, 图卷积神经网络, 风速差分拟合, 粒子群优化算法

Abstract:

In order to make full use of the prior relationships among data features and improve the prediction accuracy of medium and long term wind power at wind farms, a medium and long term wind power prediction model based on graph convolution neural network (GCN), wind velocity differential fitting (DF), and particle swarm optimization (PSO) is proposed. By analyzing the whole process of wind power generation, the influencing factors of wind power and the interrelation among them are explored, and the GCN model is built. The wind velocity and power utilization efficiency are fitted respectively. The wind power is calculated by combining with the wind velocity–power calculation model based on DF. The loss of the model includes three parts: power loss, wind velocity loss and power utilization efficiency loss. PSO algorithm is used to determine the appropriate weight for the three losses. The on-site examples of two wind farms show that the relative root mean square error of the wind power prediction model in the next 10 days is 11.44% and 13.09%, respectively, which has a high prediction accuracy.

Key words: wind power generation, wind power prediction, graph convolutional neural network, wind velocity differential fitting, particle swarm optimization algorithm

陈子含, 滕伟, 胥学峰, 丁显, 柳亦兵. 基于图卷积网络和风速差分拟合的中长期风功率预测[J]. 中国电力, 2023, 56(10): 96-105.

Zihan CHEN, Wei TENG, Xuefeng XU, Xian DING, Yibing LIU. Medium and Long Term Wind Power Prediction Based on Graph Convolutional Network and Wind Velocity Differential Fitting[J]. Electric Power, 2023, 56(10): 96-105.

导出引用管理器 EndNote|Ris|BibTeX

链接本文: https://www.electricpower.com.cn/CN/10.11930/j.issn.1004-9649.202303050

https://www.electricpower.com.cn/CN/Y2023/V56/I10/96

图/表 11

图 1 图网络示例

Fig.1 Graph network sample

图 2 风电机组发电功率与风速关系

Fig.2 Relationship between wind power and wind velocity

图 3 风速影响因素的图网络

Fig.3 Graph network of wind velocity influencing factors

图 4 GCN模型结构

Fig.4 GCN model structure

表 1 GCN图网络节点个数和特征关系

Table 1 Number of nodes and characteristic relation of GCN graph network

节点类型	节点个数	关联节点
时间向量-月份	12	气温、湿度、空气密度
时间向量-小时	24	气温、湿度、空气密度
气温	n	月份、小时、气压梯度
湿度	n	月份、小时、气压梯度
空气密度	n	月份、小时、气压梯度
气压梯度	n(n–1)/2	气温、湿度、空气密度、风向、风速
风向	8n	气压梯度、风速
风速	n	空气密度、气压梯度、风向

图 5 模型训练过程

Fig.5 Model training process

表 2 测试集部分气象数据范围

Table 2 Partial meteorological data range of test set

风电场	风速/(m·s^–1)	气温/℃	湿度/%	气压/Pa	天气情况
A	0.41~10.42	–6.53~7.41	24.67~94.52	1021.79~1034.56	晴、阴、雪、雨
B	0.48~10.70	–14.95~4.69	20.89~93.92	850.60~861.31	阴、雪、冰雹、晴、扬沙

图 6 风电场的风功率散点图

Fig.6 Wind power scatter plot of wind farms

图 7 GCN_DF_PSO模型的风速和风功率预测结果

Fig.7 Wind velocity and wind power prediction results of GCN_DF_PSO model

图 8 GCN_DF_PSO模型第10天的风速和风功率预测结果

Fig.8 Wind velocity and wind power predicted by GCN_DF_PSO model on day 10

表 3 各方法预测结果评价指标对比

Table 3 Comparison of evaluation index of prediction results of each method

风电场	方法	E_RMS/%	P_P/%
A	GCN_DF_PSO	11.44	95.04
	CNN_DF_PSO	14.38	90.11
	DNN_DF_PSO	12.14	93.66
	LSTM_DF_PSO	14.29	90.89
	GCN	13.81	90.28
	CNN	13.08	91.71
	DNN	15.50	87.05
	LSTM	19.90	81.36
B	GCN_DF_PSO	13.09	91.76
	CNN_DF_PSO	14.73	90.43
	DNN_DF_PSO	16.11	87.14
	LSTM_DF_PSO	16.23	88.24
	GCN	16.57	86.99
	CNN	15.15	88.09
	DNN	16.45	85.93
	LSTM	17.85	83.99

参考文献 21

1	曹钰, 房磊. “双碳”背景下热电机组-储热联合运行消纳弃风策略[J]. 中国电力, 2022, 55 (10): 142- 149, 160.
	CAO Yu, FANG Lei. Combined operation strategy of CHP unit and heat accumulator for eliminate abandoned wind under "double carbon" background[J]. Electric Power, 2022, 55 (10): 142- 149, 160.
2	吴晓刚, 阎洁, 葛畅, 等. 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.
	WU Xiaogang, YAN Jie, GE Chang, et al. Ultra-short-term power forecasting method for wind-solar-hydro integration based on improved GRU-CNN[J]. Electric Power, 2023, 56 (9): 178- 186, 205.
3	赵永宁, 李卓, 叶林, 等. 基于时空相关性的风电功率超短期自适应预测方法[J]. 电力系统保护与控制, 2023, 51 (6): 94- 105.
	ZHAO Yongning, LI Zhuo, YE Lin, et al. A very short-term adaptive wind power forecasting method based on spatio-temporal correlation[J]. Power System Protection and Control, 2023, 51 (6): 94- 105.
4	赵倩, 黄景涛. 基于EMD-SA-SVR的超短期风电功率预测研究[J]. 电力系统保护与控制, 2020, 48 (4): 89- 96. DOI
	ZHAO Qian, HUANG Jingtao. On ultra-short-term wind power prediction based on EMD-SA-SVR[J]. Power System Protection and Control, 2020, 48 (4): 89- 96. DOI
5	苗长新, 王霞, 李昊, 等. 基于数值天气预报风速误差修正的风电功率日前预测[J]. 电网技术, 2022, 46 (9): 3455- 3464.
	MIAO Changxin, WANG Xia, LI Hao, et al. Day-ahead prediction of wind power based on NWP wind speed error correction[J]. Power System Technology, 2022, 46 (9): 3455- 3464.
6	陆秋贤, 马刚, 涂孟夫. 基于IEWT-FE-BO-LSTM模型的超短期风功率预测[J]. 水电能源科学, 2023, 41 (1): 217- 220.
	LU Qiuxian, MA Gang, TU Mengfu. Ultra-short-term wind power prediction based on IEWT-FE-BO-LSTM model[J]. Water Resources and Power, 2023, 41 (1): 217- 220.
7	姜明洋, 徐丽, 张开军, 等. 基于季节指数调整的循环神经网络风速时间序列预测[J]. 太阳能学报, 2022, 43 (2): 444- 450.
	JIANG Mingyang, XU Li, ZHANG Kaijun, et al. Recurrent neural network prediction of wind speed time series based on seasonal exponential adjustment[J]. Acta Energiae Solaris Sinica, 2022, 43 (2): 444- 450.
8	李金龙, 王义民, 畅建霞, 等. 基于并行机器学习的风功率超短期预测[J]. 水力发电学报, 2023, 42 (1): 40- 51. DOI
	LI Jinlong, WANG Yimin, CHANG Jianxia, et al. Ultra-short-term wind power prediction based on parallel machine learning[J]. Journal of Hydroelectric Engineering, 2023, 42 (1): 40- 51. DOI
9	YUAN C, TANG Y M, MEI R, et al. A PSO-LSTM model of offshore wind power forecast considering the variation of wind speed in second-level time scale[J]. Mathematical Problems in Engineering, 2021, 2021, 1- 9.
10	ZHA W T, LIU J, LI Y L, et al. Ultra-short-term power forecast method for the wind farm based on feature selection and temporal convolution network[J]. ISA Transactions, 2022, 129, 405- 414. DOI
11	陈海鹏, 李赫, 阚天洋, 等. 考虑风电时序特性的深度小波-时序卷积网络超短期风功率预测[J]. 电网技术, 2023, 47 (4): 1653- 1665.
	CHEN Haipeng, LI He, KAN Tianyang, et al. DWT-DTCNA ultra-short-term wind power prediction considering wind power timing characteristics[J]. Power System Technology, 2023, 47 (4): 1653- 1665.
12	武新章, 梁祥宇, 朱虹谕, 等. 基于CEEMDAN-GRA-PCC-ATCN的短期风电功率预测[J]. 山东大学学报(工学版), 2022, 52 (6): 146- 156.
	WU Xinzhang, LIANG Xiangyu, ZHU Hongyu, et al. Short-term wind power forecasting based on CEEMDAN-GRA-PCC-ATCN[J]. Journal of Shandong University (Engineering Science), 2022, 52 (6): 146- 156.
13	马丽. 基于机器学习的中长期风功率预测方法研究[D]. 北京: 华北电力大学, 2020
	MA Li. Research on medium and long-term wind power prediction method based on machine learning[D]. Beijing: North China Electric Power University, 2020.
14	何健伟, 曹渝昆. LSTM-RF的中长期风电功率组合预测方法[J]. 上海电力大学学报, 2020, 36 (4): 341- 350.
	HE Jianwei, CAO Yukun. Wind power mid-long term load forecasting based on LSTM-RF combination forecasting method[J]. Journal of Shanghai University of Electric Power, 2020, 36 (4): 341- 350.
15	李铭. 基于风电场气象数据的风功率预测[D]. 包头: 内蒙古科技大学, 2021.
	LI Ming. Wind power prediction based on meteorological data of wind farm[D]. Baotou: Inner Mongolia University of Science & Technology, 2021.
16	朱满庭. 基于深度学习的风电功率预测研究[D]. 北京: 华北电力大学, 2021.
	ZHU Manting. Research on wind power prediction based on deep learning[D]. Beijing: North China Electric Power University, 2021.
17	HAN S, QIAO Y H, YAN J, et al. Mid-to-long term wind and photovoltaic power generation prediction based on copula function and long short term memory network[J]. Applied Energy, 2019, 239, 181- 191. DOI
18	AHMED S I, RANGANATHAN P, SALEHFAR H. Forecasting of mid- and long-term wind power using machine learning and regression models[C]//2021 IEEE Kansas Power and Energy Conference (KPEC). Manhattan, KS, USA. IEEE, 2021: 1–6.
19	WU Y X, WU Q B, ZHU J Q. Data-driven wind speed forecasting using deep feature extraction and LSTM[J]. IET Renewable Power Generation, 2019, 13 (12): 2062- 2069. DOI
20	WANG X X, LIU Y R, HOU J C, et al. Medium- and long-term wind-power forecasts, considering regional similarities[J]. Atmosphere, 2023, 14 (3): 430. DOI
21	孙驷洲, 付敬奇, 朱峰. CGAQPSO优化LSSVM短期风电预测[J]. 电子测量与仪器学报, 2016, 30 (11): 1718- 1725. DOI
	SUN Sizhou, FU Jingqi, ZHU Feng. Short-term wind power prediction based on LSSVM optimized by chaos Gauss attractor quantum-behaved particle swarm optimization[J]. Journal of Electronic Measurement and Instrumentation, 2016, 30 (11): 1718- 1725. DOI

[1]	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.
[2]	董福贵, 夏美娟, 李婉莹. 基于PSO-GWO的省级能耗强度预测与碳减排潜力估算[J]. 中国电力, 2023, 56(9): 226-234.
[3]	王霄鹤, 杨海超, 宋爽, 杨玉新, 殷贵, 王克. 海上风电场经66 kV海缆送出系统过电压机理[J]. 中国电力, 2023, 56(11): 29-37, 48.
[4]	周诗嘉, 杨光源, 彭光强, 武霁阳, 辛清明. 基于多相风力发电系统的容错控制策略研究[J]. 中国电力, 2022, 55(7): 134-141.
[5]	刘櫂芮, 贾祺, 严干贵, 翟文超, 孙勇, 李宝聚. 基于惯量响应的双馈风电机组动态协调机理研究[J]. 中国电力, 2022, 55(7): 142-151.
[6]	黄书民, 蒋林高, 李志川, 杨光绪, 宋福根. 基于PSO寻优与DBN神经网络的电晕损耗预测[J]. 中国电力, 2022, 55(6): 95-102,214.
[7]	李存斌, 董佳. 中国风力发电绩效的区域差异及空间计量分析[J]. 中国电力, 2022, 55(3): 167-176.
[8]	沙骏, 徐雨森, 刘冲冲, 冯定东, 胥峥, 臧海祥. 基于变分模态分解和分位数卷积-循环神经网络的短期风功率预测[J]. 中国电力, 2022, 55(12): 61-68.
[9]	方日升, 林耀东, 江伟, 黄霆, 徐振华, 吴丹岳, 陈志, 苏毅, 林济铿. 调速侧电力系统稳定器优化设计方法[J]. 中国电力, 2021, 54(9): 74-82.
[10]	王奔, 牛洪海, 徐卫峰, 陈俊. 2 MW永磁直驱风机主控系统的设计与应用[J]. 中国电力, 2021, 54(9): 208-220.
[11]	游广增, 李华瑞, 李常刚, 刘超, 钱迎春, 李玲芳. 计及风电高频保护的送端电网多直流协同频率控制[J]. 中国电力, 2021, 54(5): 83-90,110.
[12]	陈岩, 靳伟, 王文宾, 李会彬, 韩胜峰, 王一鸣, 钟嘉庆. 基于电动汽车分群的“风-网-车”联合消纳调度策略[J]. 中国电力, 2021, 54(4): 107-118.
[13]	付红军, 陈惠粉, 赵华, 王凯丰, 鲁宗相, 乔颖. 高渗透率下风电的调频技术研究综述[J]. 中国电力, 2021, 54(1): 104-115.
[14]	于国强, 崔晓波, 史毅越, 汤可怡, 张天海. 基于改进群优化算法的深度调峰机组一次调频建模[J]. 中国电力, 2020, 53(6): 147-152.
[15]	都嘉慧, 王长江, 李凌, 杨晨光. 计及换流器损耗的风电经柔直并网系统的随机最优潮流模型[J]. 中国电力, 2020, 53(5): 77-88.