广义负荷建模中纵横聚类策略研究

发布时间：2019-06-05 01:03

【摘要】：考虑风电接入后广义负荷的不确定性,提出纵横聚类策略。针对聚类客观性与复杂场景分析需求,引入高质量的仿射传播(affinity propagation,AP)聚类算法。利用包含纵向聚类与横向聚类的纵横聚类策略对全年实测有功响应空间客观聚类,实现大尺度与小尺度数据在统一时间框架下的聚类分析:通过三步聚类实现全年大时间尺度下的纵向聚类,结果体现了细化季节特性;通过将纵向类内全部数据联排统一聚类,实现较小时间尺度下的精细横向聚类,结果体现日时段特性。引入带概率信息的广义负荷建模方法建模,并检验聚类策略的有效性。仿真结果表明,所提策略实现的聚类客观、合理,便于精确建模与现场应用,可为后续仿真分析与调度控制提供辅助参考。
[Abstract]:Considering the uncertainty of generalized load after wind power access, a vertical and horizontal clustering strategy is proposed. In order to meet the needs of clustering objectivity and complex scene analysis, a high quality affine propagation (affinity propagation,AP) clustering algorithm is introduced. The vertical and horizontal clustering strategy, which includes longitudinal clustering and transverse clustering, is used to objectively cluster the measured active power response space in the whole year. To realize the clustering analysis of large-scale and small-scale data in a unified time frame: the longitudinal clustering of large-scale and small-scale data in the whole year is realized by three-step clustering, and the results reflect the refined seasonal characteristics. Through the unified clustering of all the data in the longitudinal class, the fine lateral clustering at a small time scale is realized, and the results show the characteristics of the daily period. The generalized load modeling method with probability information is introduced, and the effectiveness of clustering strategy is tested. The simulation results show that the clustering of the proposed strategy is objective, reasonable, convenient for accurate modeling and field application, and can provide an auxiliary reference for subsequent simulation analysis and scheduling control.
【作者单位】：电网智能化调度与控制教育部重点实验室(山东大学);
【基金】：国家自然科学基金资助项目(51177091) 国家863高技术基金项目(2011AA05A101)~~
【分类号】：TM614

【参考文献】