基于GMM聚类和SVM的用户负荷行业分类辨识方法

摘要

用户行业分类是进行需求侧响应能力评估的前提,为此提出一种利用负荷数据进行用户行业判别的方法.对每一个用户的负荷数据采用高斯混合模型聚类算法提取其典型日负荷曲线,并利用支持向量机算法在训练集上学习用户类别与其典型日负荷曲线之间的关系,建立分类模型,并据此对新的用户进行行业分类.该方法实现了用户行业类型的高准确度辨识,为调度部门进行需求侧管理提供支撑.通过分析我国某省290个用户负荷数据,验证了该方法的有效性.%User industry classification is the precondition for assessing demand side response capability,the paper presents a kind of identification method for user industry classification by making use of load data.The method adopts Gaussian mixture model (GMM) clustering algorithm to extract typical daily load curve for load data of every user,and uses support vector machine (SVM) algorithm to study relationship between user classification and its typical daily load curve based on training data and build the classification model for industry classification for new users.The method has realized high accuracy in identification for user industry classification and provided support for demand side management for the dispatching department.Analysis on load data about 290 users in one province in China verifies effectiveness of this method.