摘要
提出一种基于Pearson相关系数作为聚类判据的负荷曲线聚类算法——Pearson相关聚类(Pearson Correlation Clustering,PCC)。首先对负荷数据进行数据清理以及去噪处理,再选择合适的降维算法以降低数据处理的复杂性。提出利用Pearson相关系数阈值作为聚类中心选择依据的方法解决初始聚类中心选择的随机性;利用电力负荷曲线数据与聚类中心之间的Pearson相关系数进行聚类,以DBI指标作为聚类效果的评价标准,分析了不同系数对聚类效果的影响。算例结果表明,该算法相比传统算法运行时间短,鲁棒性强,聚类效果更好。
A Pearson Correlation Clustering (PCC) algorithm based on Pearson correlation coefficient as a clustering criterion is proposed. Firstly, the load data is cleaned and noiseeliminationis made, thenthe appropriate dimension- ality reduction algorithm is selected to reduce the complexity of data processing. The Pearson correlation coefficient threshold is proposed to be used as the method of clustering center selection to solve the stochasticity of the initial clustering center selection. Clusteringis madebyusing Pearson correlation coefficient between the power load curve data and the clustering center, and the influence of different coefficients on the clustering effect is analyzedby tak- ingthe DBI indexas the clustering effect. The results show that compared with the traditional algorithm, the algo- rithm has a short run time, strong robustness and better clustering effect.
出处
《黑龙江电力》
CAS
2017年第5期397-401,415,共6页
Heilongjiang Electric Power
关键词
Pearson相关系数
负荷曲线分类
降维
中值滤波
聚类有效性
Pearson correlation coefficient
load curve classification
dimensionality reduction
median filter
clus- tering validity