基于数据挖掘的电力系统不良数据检测与辨识算法研究

发布时间：2018-04-25 01:03

本文选题：数据挖掘 + 电力系统　；参考：《南京理工大学》2017年硕士论文

【摘要】：随着电力系统自动化水平的不断提升,电网运行对实时数据质量的要求越来越高,因此针对不良数据的检测与辨识就显得尤为必要。传统的不良数据检测方法基本都是估计后的算法,需要反复进行多次状态估计,计算量较大的同时还有可能造成"残差淹没"或"残差污染"现象,所以较为理想的状态就是在估计前就能把不良数据检测辨识出来,而本文研究的基于关联规则挖掘的不良数据检测与辨识方法可以在一定程度上解决这些问题。本文研究内容主要有以下几点:1)研究传统的不良数据检测与辨识方法,分析它们的局限性与不足之处,根据状态估计前不良数据检测的优越性提出关联规则挖掘算法在这一课题中应用的可行性;2)对关联规则中的经典算法—Apriori算法进行深入研究,分析其不足,并在传统Apriori算法基础上,提出一种改进的多维关联规则算法,并将两种算法进行对比分析,验证改进算法的有效性;3)引进时态关联规则的概念,重点研究其中的一类——周期性关联规则,阐述周期性关联规则的基本思想和算法实现等,并针对基本算法提出改进并将其实现,使能减少不必要的周期性支持度计数、降低时间成本;4)将改进后的关联规则算法应用到估计前不良数据的检测与辨识中:在不知道系统拓扑结构的前提下以历史数据样本为基础,通过关联规则挖掘获得每个时刻对应的量测预测量;之后通过人为设定不良数据的方法进行实验分析,利用时间属性与状态量的相关联系验证当前时刻状态值的正确性,从而验证了本文提出的方法的可行性和有效性。
[Abstract]:With the improvement of power system automation, the real-time data quality is becoming more and more important in power network operation, so it is necessary to detect and identify the bad data. The traditional bad data detection methods are basically estimated algorithms, which need repeated state estimation, which may cause "residual inundation" or "residual pollution". Therefore, the ideal state is to detect and identify the bad data before the estimation, and the method based on association rule mining can solve these problems to some extent. The main contents of this paper are as follows: 1) to study the traditional methods of bad data detection and identification, and to analyze their limitations and shortcomings. According to the superiority of bad data detection before state estimation, the feasibility of applying association rule mining algorithm in this subject is put forward. (2) the classical algorithm-Apriori algorithm in association rules is deeply studied, and its shortcomings are analyzed. Based on the traditional Apriori algorithm, an improved multidimensional association rule algorithm is proposed, and the two algorithms are compared and analyzed to verify the effectiveness of the improved algorithm and introduce the concept of temporal association rules. This paper focuses on a class of periodic association rules, expounds the basic idea and algorithm implementation of periodic association rules, and puts forward some improvements to the basic algorithms and implements them, which can reduce the unnecessary periodic support count. The improved association rule algorithm is applied to the detection and identification of bad data before estimation: based on historical data samples without knowing the topological structure of the system. By mining association rules, the corresponding measurements and predictions at each time are obtained, and then the correctness of the current state value is verified by the correlation between the time attribute and the state quantity, and the experimental analysis is carried out by the method of artificial setting bad data. The feasibility and effectiveness of the proposed method are verified.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM73;TP311.13

【相似文献】