唐雅洁,阎洁,李玉浩,龚迪阳,杜倩昀,叶碧琦
TANG Yajie,YAN Jie,LI Yuhao,GONG Diyang,DU Qianyun,YE Biqi

• 1:国网浙江省电力有限公司电力科学研究院
• 2:新能源电力系统国家重点实验室(华北电力大学)
• 3:国网浙江省电力有限公司丽水供电公司

摘要(Abstract)：

对于风光水联合发电系统筛选出具有代表性的典型场景，是进行资源特性分析、调度优化研究的基础。传统聚类算法由于“维度效应”的影响，无法直接应用于高维数据聚类问题，而现有的“先降维，后聚类”技术路线无法保证降维后的低维特征适用于聚类任务，导致聚类结果不稳定。针对存在的问题，提出了一种基于深度嵌入聚类的风光水典型出力场景提取方法，在实现高维出力数据聚类的同时避免了降维后的低维特征不适用于聚类任务的问题。首先借助堆叠自编码器对数据的非线性表征能力，对高维风光水联合出力数据进行深层表征实现数据降维，然后结合K-means聚类方法对深层低维特征进行聚类，并在聚类过程中同时优化调整堆叠自编码器，得到适用于聚类空间的低维风光水联合出力特征，基于此实现对风光水联合出力场景的精准划分。最后以我国南方某区域风电、光伏、水电出力数据为研究对象对其进行深度嵌入聚类，并以PCA-K-means算法设置对比算例，验证了深度嵌入聚类在风光水典型联合出力场景选取上的有效性。
It is a prerequisite for resource characteristics analysis and scheduling optimization research to screen out the representative typical scenarios of wind-solar-hydropower integrated generation systems. Due to the influence of the “dimension effect”, traditional clustering algorithms cannot be directly applied to high-dimensional data clustering, and the existing technical route based on “dimension reduction before clustering” cannot guarantee that lowdimensional features after dimensionality reduction are suitable for clustering tasks, resulting in unstable clustering results. In view of the existing problems, this paper proposes a method for extracting typical output scenarios of windsolar-hydropower based on DEC(deep embedding clustering). The method can realize high-dimensional output data clustering and avoid that low-dimensional features after dimensionality reduction are not suitable for clustering tasks.First, with the help of the nonlinear representation ability of the stacked autoencoder, the high-dimensional windsolar-hydropower combined output data is deeply represented to achieve data dimensionality reduction. Then, the Kmeans clustering method is used to cluster the deep low-dimensional features, and the stacked encoder is optimized and adjusted at the same time in the clustering process to obtain the low-dimensional wind-solar-hydropower combined output feature suitable for the clustering space. Moreover, the precise division of wind-solar-hydropower combined output scenarios is realized. Finally, the DEC is performed on the wind-solar-hydropower output data of a region in south China. The PCA-K-means algorithm is used to set up a comparison example to verify the effectiveness of the DEC in selecting typical combined output scenarios of wind-solar-hydropower generation.

关键词(KeyWords)： 风光水一体化;典型场景;深度嵌入聚类;降维;特征提取
wind-solar-hydropower integration;typical scenarios;deep embedding clustering;dimensionality reduction;feature extraction

Abstract:

Keywords:

基金项目(Foundation): 国网浙江省电力有限公司科技项目（5211LS21N003）

作者(Author): 唐雅洁,阎洁,李玉浩,龚迪阳,杜倩昀,叶碧琦
TANG Yajie,YAN Jie,LI Yuhao,GONG Diyang,DU Qianyun,YE Biqi

DOI: 10.19585/j.zjdl.202304005

参考文献(References)：

• [1]张雯.水电水利规划设计总院副院长赵增海：水风光一体化开发优化水电发展[N].中国能源报，2022-06-13(8).
• [2]刘永前，王函，韩爽，等.考虑风光出力波动性的实时互补性评价方法[J].电网技术，2020,44(9):3211-3220.LIU Yongqian,WANG Han,HAN Shuang,et al. Realtime complementarity evaluation method for real-time complementarity of wind and solar power considering their volatility[J]. Power System Technology,2020,44(9):3211-3220.
• [3]刘永前，林爱美，阎洁，等.基于深度学习的风光场群功率预测方法研究[J].分布式能源，2021,6(2):14-21.LIU Yongqian,LIN Aimei,YAN Jie,et al. Research on power forecasting method for wind farms and photovoltaic stations based on deep learning[J]. Distributed Energy,2021,6(2):14-21.
• [4] WANG H,ZHANG N,DU E,et al.A comprehensive review for wind,solar,and electrical load forecasting methods[J].Global Energy Interconnection,2022,5(1):9-30.
• [5]陈文进，朱峰，张童彦，等.基于AFSA-BP神经网络的光伏功率预测方法[J].浙江电力，2022,41(4):7-13.CHEN Wenjin,ZHU Feng,ZHANG Tongyan,et al. A photovoltaic power prediction method based on AFSA-BP neural network[J].Zhejiang Electric Power,2022,41(4):7-13.
• [6]刘晓，王洪伟.基于抽水蓄能的含高比例风电农业配电网优化调度[J].浙江电力，2022,41(1):48-54.LIU Xiao,WANG Hongwei.Optimal dispatch of agricultural distribution networks with a high proportion of wind power based on pumped storage[J]. Zhejiang Electric Power,2022,41(1):48-54.
• [7]赵亚威.多电源电力系统多目标优化调度与决策方法研究[D].北京：华北电力大学，2021.ZHAO Yawei. Research on multi-objective dispatch and decision method for power system with multi-type power stations[D].Beijing:North China Electric Power University,2021.
• [8]金新峰，廖胜利，刘战伟，等.考虑风电置信区间的水风火短期优化调度方法[J].水力发电学报，2020,39(9):33-42.JIN Xinfeng,LIAO Shengli,LIU Zhanwei,et al. Shortterm optimal operation method of water,wind and thermal power considering wind power confidence intervals[J].Journal of Hydroelectric Engineering,2020,39(9):33-42.
• [9]肖舒.基于实测运行数据的地区规模化风电出力场景划分方法[D].北京：华北电力大学，2020.XIAO Shu.Method for regional scaled wind power output scenario division based on measured operating data[D].Beijing:North China Electric Power University,2020.
• [10]刘汝琛.基于实测数据的地区风电出力的典型场景选取[D].北京：华北电力大学，2016.LIU Ruchen. Typical scenarios selection of wind power output based on measured data[D].Beijing:North China Electric Power University,2016.
• [11]李亮，唐巍，白牧可，等.考虑时序特性的多目标分布式电源选址定容规划[J].电力系统自动化，2013,37(3):58-63.LI Liang,TANG Wei,BAI Muke,et al. Multi-objective locating and sizing of distributed generators based on timesequence characteristics[J].Automation of Electric Power Systems,2013,37(3):58-63.
• [12]邱宜彬，欧阳誉波，李奇，等.考虑多风电场相关性的场景概率潮流计算及无功优化[J].电力系统保护与控制，2017,45(2):61-68.QIU Yibin,OUYANG Yubo,LI Qi,et al.Scenario probabilistic load flow calculation and reactive power optimization considering wind farms correlation[J].Power System Protection and Control,2017,45(2):61-68.
• [13]赵书强，要金铭，李志伟.基于改进K-means聚类和SBR算法的风电场景缩减方法研究[J].电网技术，2021,45(10):3947-3954.ZHAO Shuqiang,YAO Jinming,LI Zhiwei.Wind power scenario reduction based on improved K-means clustering and SBR algorithm[J].Power System Technology,2021,45(10):3947-3954.
• [14]王群，董文略，杨莉.基于Wasserstein距离和改进Kmedoids聚类的风电/光伏经典场景集生成算法[J].中国电机工程学报，2015,35(11):2654-2661.WANG Qun,DONG Wenlue,YANG Li.A wind power/photovoltaic typical scenario set generation algorithm based on Wasserstein distance metric and revised Kmedoids cluster[J]. Proceedings of the CSEE,2015,35(11):2654-2661.
• [15]姚剑峰，凌静，曲立楠，等.基于改进FCM聚类算法的清洁能源典型场景构建[J].电网与清洁能源，2019,35(4):76-82.YAO Jianfeng,LING Jing,QU Linan,et al.The construction method of typical scenario set for renewable energy based on improved FCM clustering algorithm[J]. Power System and Clean Energy,2019,35(4):76-82.
• [16]易锦桂，朱自伟，谢青.基于改进场景聚类算法的海上风电储能优化配置研究[J].中国电力，2022,55(12):2-10.YI Jingui,ZHU Ziwei,XIE Qing.Research on optimal configuration of offshore wind power energy storage based on improved scene clustering algorithm[J]. Electric Power,2022,55(12):2-10.
• [17]林俐，肖舒，费宏运，等.基于曲线形态特征的地区规模化风电出力场景划分[J].电网与清洁能源，2020,36(3):74-81.LIN Li,XIAO Shu,FEI Hongyun,et al.Regional scaled wind power output scene segmentation based on curve morphological features[J]. Power System and Clean Energy,2020,36(3):74-81.
• [18]叶林，李镓辰，路朋，等.基于近邻传播聚类与MCMC算法的风电时序数据聚合方法[J].中国电机工程学报，2020,40(12):3744-3754.YE Lin,LI Jiachen,LU Peng,et al.Wind power time series aggregation approach based on affinity propagation clustering and MCMC algorithm[J]. Proceedings of the CSEE,2020,40(12):3744-3754.
• [19]林俐，费宏运，刘汝琛，等.基于分层聚类算法的地区风电出力典型场景选取方法[J].电力系统保护与控制，2018,46(7):1-6.LIN Li,FEI Hongyun,LIU Ruchen,et al.A regional wind power typical scenarios’ selection method based on hierarchical clustering algorithm[J]. Power System Protection and Control,2018,46(7):1-6.
• [20] WASSERMAN L,AZIZYAN M,SINGH A.Feature selection for high-dimensional clustering[EB/OL]. 2014:arXiv:1406.2240.https：//arxiv.org/abs/1406.2240.
• [21] JIN J S,WANG W J.Influential features PCA for high dimensional clustering[J].The Annals of Statistics,2016,44(6):2323-2359.
• [22] AGGARWAL C C.On the effects of dimensionality reduction on high dimensional similarity search[C]//Proceedings of the twentieth ACM SIGMOD-SIGACT-SIGART symposium on Principles of database systems.SIGMOD/PODS01:ACM SIGMOD International Conference on Management of Data.New York:ACM,2001:256-266.
• [23]唐海国，张志丹，康童，等.考虑场景聚类的配电网-天然气联合系统双层随机运行优化[J].现代电力，2021,38(6):681-694.TANG Haiguo,ZHANG Zhidan,KANG Tong,et al.Bilevel stochastic operation optimization of distributionnatural gas combined system considering scenario clustering[J].Modern Electric Power,2021,38(6):681-694.
• [24]朱云毓.城镇综合能源系统供用能方式分析与运行场景构建技术[D].南京：东南大学，2021.ZHU Yunyu.Analysis of energy supply and consumption methods and construction of operation scenarios for urban integrated energy systems[D].Nanjing:Southeast University,2021.
• [25]杨子豪.基于聚类分析的需求侧灵活性资源预测与应用研究[D].北京：华北电力大学，2021.YANG Zihao. Study on the demand-side flexibility resources prediction and application based on cluster analysis[D]. Beijing:North China Electric Power University,2021.
• [26]彭晨宇.基于轨迹特征的风电场功率预测及自适应控制方法研究[D].南京：东南大学，2020.PENG Chenyu. Research on trajectory characteristics based power forecasting and adaptive control method for wind farms[D].Nanjing:Southeast University,2020.
• [27]孙海蓉，曹瑶佳，张雨晴.基于WKFCM-SMOTE和随机森林的风电机组故障诊断[J].山东电力技术，2022,49(3):65-70.SUN Hairong,CAO Yaojia,ZHANG Yuqing. Wind turbine fault diagnosis based on WKFCM SMOTE and random forest[J].Shandong Electric Power,2022,49(3):65-70.
• [28]康佳乐，余浩，段瑶，等.风电场次同步振荡等值建模方法研究[J].发电技术，2022,43(6):880-891.KANG Jiale,YU Hao,DUAN Yao,et al.Equivalent modeling method of sub-synchronous oscillation in wind farm[J].Power Generation Technology,2022,43(6):880-891.
• [29]胡雪凯，尹瑞，时珉，等.基于改进粒子群算法的分布式光伏集群划分与无功优化策略[J].电力电容器与无功补偿，2021,42(4):14-21.HU Xuekai,YIN Rui,SHI Min,et al.Distributed photovoltaic cluster division and reactive power optimization strategy based on improved particle swarm optimization[J]. Power Capacitor&Reactive Power Compensation,2021,42(4):14-21.
• [30]祁乐，唐健，江平，等.考虑机组分类的海上风电短期功率预测-校正模型[J].山东电力技术，2021,48(5):16-22.QI Le,TANG Jian,JIANG Ping,et al. Short-term offshore wind power prediction-correction model considering classification of wind farm units[J]. Shandong Electric Power,2021,48(5):16-22.
• [31]赵莎莎，朱雅魁，王悦.基于大数据分析的综合能源系统负荷特性聚类分析[J].电测与仪表，2023,60(2):10-15.ZHAO Shasha,ZHU Yakui,WANG Yue.Cluster analysis of load characteristics of comprehensive energy system based on big data analysis[J].Electrical Measurement&Instrumentation,2023,60(2):10-15.
• [32]严明辉，谢雄，李维劼，等.基于高斯核密度估计的典型负荷曲线形态聚类算法[J].电测与仪表，2023,60(2):37-44.YAN Minghui,XIE Xiong,LI Weijie,et al.Morphological clustering algorithm of typical load curve based on Gaussian kernel density estimation[J]. Electrical Measurement&Instrumentation,2023,60(2):37-44.
• [33] XIE J Y,GIRSHICK R,FARHADI A. Unsupervised deep embedding for clustering analysis[C]//Proceedings of the 33rd International Conference on International Conference on Machine Learning-Volume 48.New York,NY,USA.New York:ACM,2016:478-487.
• [34] YANG B,FU X,SIDIROPOULOS N D,et al.Towards k-means-friendly spaces:Simultaneous deep learning and clustering[C]//international conference on machine learning.PMLR,2017:3861-3870.
• [35]杨晶显，刘俊勇，韩晓言，等.基于深度嵌入聚类的水光荷不确定性源场景生成方法[J].中国电机工程学报，2020,40(22):7296-7306.YANG Jingxian,LIU Junyong,HAN Xiaoyan,et al. An uncertain hydro/PV/load typical scenarios generation method based on deep embedding for clustering[J]. Proceedings of the CSEE,2020,40(22):7296-7306.