摘要
配电网台区分类研究对"煤改电"工程挖掘台区用电规律、规划电网等工作具有重要的指导意义。为了对台区进行有效处理分类,针对具有日趋复杂、大数据量、高维、非线性等特征的配电网台区变压器运行数据,文章采用基于数据驱动方式的稀疏降噪自编码器网络对台区变压器负载率数据无监督训练,自主提取特征,学习去除数据噪声,降低序列维度,然后利用K-means算法对特征序列聚类分析,得到台区类型。该模型能够很好地提取高维无标签数据特征,降低数据维度,提高聚类效率。实验结果表明,该模型在配电网台区分类应用中效果明显,且具有良好的抗噪性和泛化能力。
The research on the classification of distribution transformer area has important guiding significance to the work of "coal to electricity" project,such as power consumption rules,distribution network planning and so on.In order to effectively classify distribution transformer areas,aiming at the distribution transformer area operation data with increasingly complex,large data volume,high-dimensional,nonlinear and other characteristics,this paper uses a Sparse De-noising Auto-Encoder(SDAE) network based on data-driven mode to conduct unsupervised training of the load rate data in the distribution transformer area,extracting the features autonomously,learning to remove noise,reducing the numerical dimension of the sequence,and then using k-means algorithm to conduct clustering analysis of the feature sequence and obtain the transformer area type.The model can extract high-dimensional unlabeled data features,reduce data dimensions and improve clustering efficiency.Experimental results show,the model has obvious effects in the classification application of distribution transformer area,and has good anti-noise and generalization ability.
作者
张潇龙
齐林海
ZHANG Xiaolong;QI Linhai(School of Control and Computer Engineering,North China Electric Power University,Beijing 102206,China)
出处
《电力信息与通信技术》
2019年第12期15-23,共9页
Electric Power Information and Communication Technology
基金
国家电网有限公司科技项目资助“城市电网电能质量大数据深化分析及应用技术研究”(52094018001C)
关键词
稀疏降噪自编码器网络
序列降维
K-MEANS聚类
台区分类
Sparse De-noising Auto-Encoder
sequence dimension reduction
K-means clustering
distribution transformer area classification
