基于PSO优化SVM分类器的配电网电压波动分析
2020-12-23陈宁钊
陈宁钊



摘 要:为了进一步改善直流电能质量,需要提出一种基于PSO优化SVM分类器的配电网电压波动分析方法。在分析直流电压波动的基础上,设计了基于PSO优化的SVM分类器。通过仿真分析得到采在含有5各类簇的情况下,轮廓系数与1非常接近,可见采用特征集对各类直流电能质量扰动情况进行准确分析。测试发现,在30个测试样本内实现正确分类的是28,只出现了1 组电压波动以及1组电压纹波发生分类错误的情况,分类模型获得了高达92.1%的准确率。通过算法比较得到采用BP神经网络算法进行处理时形成了更低的分类准确率,相对于SVM需要更多的训练样本数。采用文中算法能够使分类过程达到更高的准确率,能够显著改善PSO优化效果,而在训练时间方面则比多分类SVM更弱。
关键词:配电网;电压波动;SVM分类器;准确率
中图分类号:TP 393
文献标志码:A
文章编号:1007-757X(2020)11-0174-03
Abstract:In order to further improve the DC energy quality, it is necessary to propose a PSO-based SVM classifier optimization method for distribution network voltage fluctuation analysis. Based on the analysis of DC voltage fluctuation, an SVM classifier based on PSO optimization is designed. The simulation analysis shows that in the case that there are 5 kinds of clusters, the contour coefficient is very close to 1, so it can be seen that the feature set is used to accurately analyze the mass disturbance of all kinds of dc energy. The test finds that 28 of the 30 test samples are correctly classified, only 1 set of voltage fluctuation and 1 set of voltage ripple are misclassified. The classification model achieves an accuracy rate of 92.1%. Through algorithm comparison, it can be concluded that the classification accuracy is lower when BP neural network algorithm is used for processing, and more training samples are required than SVM. The algorithm proposed in this paper can make the classification process achieve higher accuracy and significantly improve the optimization effect of PSO. However, it is weaker than multi-classification SVM in the term of training time.
Key words:distribution network;voltage fluctuation;SVM classifier;accuracy
0 引言
目前,多種电力电子技术呈现快速发展的趋势,由此产生了更明显的源储荷直流特征,这使得直流配用电技术也引起众多学者与企业的密切关注。根据前期相关研究资料可知,直流配电系统可以实现更强的输电性能,同时可以显著减小线路损耗,并表现出优异的新能源消纳能力,这使其成为了配电系统的一个重要发展方向[1-5]。由于直流配电网具备多源多变换的特征,因此在电能质量方面具有独特的性能,并且随着敏感负荷的占比增大,要求电网系统具备更高的电能质量,因此需在初期规划阶段对直流配电网电能质量进行深入分析[6-8]。……
