基于SVM的负载识别技术研究

发布时间：2018-07-11 15:58

本文选题：负载识别 + 支持向量机　；参考：《杭州电子科技大学》2017年硕士论文

【摘要】：负载识别技术能够识别电网中正在使用的负载类型,可以应用于电力公司对于电网的监管以及提高能源的利用效率,也可以应用于某些应用场合的电网违禁电器监管。对负载识别技术的研究有着重要意义。论文重点研究基于SVM的负载识别技术,主要研究工作有:1、提出基于支持向量机的负载识别方法总体框架,方法包括负载电流数据采集、负载识别特征量(用于负载识别的电流特征量)提取、特征量数据预处理以及负载识别多分类器四个部分,分为离线训练与在线测试两个流程。2、对采集的各种负载电流数据,通过对比负载电流的时域和频域特征,分析各种负载之间的时域和频域特征差异,提出若干用于负载识别的负载电流特征量。3、采用one-against-one组合多个SVM的方法设计负载识别多分类器,并针对不同的惩罚参数和核参数对SVM的分类性能影响较大的问题,运用遗传算法结合K折交叉验证寻找最优的SVM惩罚参数c和核参数g组合,以此训练出来的SVM二分类器构成用于负载识别的SVM多分类器。论文构建了负载识别多分类器训练样本集和测试集,然后对负载识别多分类器进行了训练和测试,训练和测试实验分为单负载以及混合负载两组。实验结果表明,本文提出的基于SVM的负载识别方法具有较好的负载识别效果。
[Abstract]:Load identification technology can be used to identify the load types used in the power network, can be applied to power companies to regulate the grid and improve the efficiency of energy use, but also can be used in some applications of the monitoring of prohibited electrical appliances. The research of load identification technology is of great significance. This paper focuses on the load recognition technology based on SVM. The main research work is: 1. The overall framework of load recognition method based on support vector machine (SVM) is proposed, which includes load current data acquisition. The load identification feature (current characteristic used for load identification) extraction, feature data preprocessing and load identification multi-classifier are divided into two parts: offline training and on-line testing. By comparing the time-domain and frequency-domain characteristics of the load current, the differences between the time-domain and frequency-domain characteristics of the load are analyzed, and some load current characteristics. 3, which are used to identify the load, are proposed. The method of combining one-against-one with multiple SVM is used to design the multi-classifier for the load identification. Aiming at the problem that different penalty parameters and kernel parameters have great influence on the classification performance of SVM, genetic algorithm combined with K-fold cross-validation is used to find the optimal combination of SVM penalty parameter c and kernel parameter g. The SVM two classifier which is trained by this method is a SVM multi classifier for load recognition. In this paper, the load identification multi-classifier training sample set and test set are constructed, and then the load identification multi-classifier is trained and tested. The training and testing experiments are divided into two groups: single load and mixed load. Experimental results show that the proposed load recognition method based on SVM has better load recognition effect.
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18;TM732

【参考文献】