【作者】 张重阳；

【导师】 柳伟；

【作者基本信息】 南京理工大学 ， 电力系统及其自动化， 2021， 硕士

【摘要】 随着分布式能源、储能以及多元负荷的接入,配电网的建设日趋成熟,运行方式灵活多变。同时伴随着电力体制的不断改革,用户对供电可靠性需求不断提升,如何打通用户可靠性管理需求的最后一米,量化终端用户的供电需求,进而满足不同区域、不同类型用户的差异化供电可靠性需求,提高配电网的供电可靠性水平,是当前的亟需解决的问题。因此本文结合终端用户用电信息数据融合和深度挖掘方法分析用户的用电行为,量化分析终端用户的供电需求,采用大数据聚类与灰色关联度方法实现终端用户的停电属性划分,并通过融合多源数据实现基于属性划分的区域停电信息智能研判。具体工作如下:(1)研究了结合孤立森林理论与三次样条插值的终端用户用电数据融合技术,实现用户用电数据中异常值的定位与插值,形成了优良样本数据;然后针对高维数据特征,通过研究基于信息熵分段聚合近似实现样本数据的特征提取,进而采用谱聚类算法实现用户典型负荷曲线的提取,为下文用户的供电需求量化提供数据基础。(2)通过对不同类型终端用户的供电需求进行分析,以及终端用户对供电可靠性、电能质量、以及经济损失需求的量化,构建终端用户供电需求模型,研究基于大数据聚类和灰色关联度分析的方法,实现终端用户的停电属性和供电需求等级划分。为下文停电信息研判提供用户属性信息。(3)通过结合多源系统信息,确定停电区域与停电性质。针对事故停电,研究基于改进量子遗传算法求解开关量特征与希尔伯特-黄变换(Hilbert-Huang Transform,HHT)求解电气量特征。通过研究停电区域内不同属性用户对停电特征的影响,确定改进加权D-S证据理论(Dempster-Shafer evidence theory)的权重赋值,进而实现融合多源数据的停电信息综合决策,并结合停电区域用户供电需求等级与停电属性,为停电抢修提供信息支撑,提升终端用户的供电可靠性水平。更多还原

【Abstract】 With the access of distributed energy,energy storage and multiple loads,the construction of distribution network is becoming mature and its operation mode is flexible and changeable.With the continuous reform of electric power system at the same time,the user to improve power supply reliability requirements,how to get through the last mile of user reliability management needs,quantify the power supply requirements of end users,to satisfy different region,the differentiation of different types of user power supply reliability requirements,improve the level of the power supply reliability of distribution network,is the current needs to solve the problem.So this article combined with the end user electricity information data fusion and the power of the depth of mining methods to analyze the user behavior,the quantitative analysis of end user’s demand,big data clustering and grey correlation method was adopted to realize the end user attributes of power division,and through the fusion of multi-source data based on attribute partition area outage information to intelligence.The specific work is as follows:(1)The end-user power consumption data fusion technology combining the isolated forest theory and cubic spline interpolation is studied to achieve the location and interpolation of outliers in the user power consumption data and form excellent sample data;Then,in view of the features of high-dimensional data,feature extraction of sample data is realized by studying the segmentation aggregation approximation based on information entropy,and then the typical load curve of users is extracted by spectral clustering algorithm,which provides the data basis for the quantification of power supply demand of users in the following.(2)Based on the demand of different types of end user is analyzed,as well as to the end user to the power supply reliability,power quality,and economic loss quantitative requirements,building end-user demand model,the research based on large data clustering and grey correlation analysis method,realize the end user attributes blackouts and power supply demand hierarchy.Provide user attribute information for the following outage information analysis.(3)By combining multi-source system information,determine the blackout area and the nature of the blackout.Study forced outage,quantum genetic algorithm based on improved switch features and Hilbert huang transform to solve the electric parameters,by studying the different attributes blackout area user influence on power characteristics,determine the improved weighted weights assignment of Dempster-Shafer evidence theory,thus realize the fusion of multi-source data comprehensive decision-making power outage information,Combined with power supply demand level and power failure attribute of users in power failure area,information support for power failure emergency repair is provided.更多还原