基于相关向量机的GIS气体绝缘状态评估  

Assessment of Gas Insulation State in GIS Equipment Based on Relevance Vector Machine

作　　者：宋人杰[1] 赵萌[1] 关潇卓[2] 杜轲琼 刘静[2] 

机构地区：[1]东北电力大学信息工程学院,吉林132012 [2]国网吉林省电力有限公司吉林供电公司信息通信分公司,吉林132012

出　　处：《高压电器》2016年第12期237-243,共7页High Voltage Apparatus

基　　金：国家自然科学基金资助项目(61271115)~~

摘　　要：文中提出了基于多参量的GIS设备气体绝缘状态评估方法。该方法首先对影响GIS设备绝缘状况的因素进行了详细的分析,确定影响该气体绝缘状态的主要因素。然后利用气体组成成分、温度、湿度、压力、纯度作为因素集,对其进行绝缘评估。最后,采用一对一的分类方法建立基于相关向量机的GIS设备气体绝缘状态评估模型,并将GIS设备气体的绝缘状态分为"正常"、"注意"、"异常"和"故障"4类,利用训练样本对评估模型进行训练,进而实现GIS设备气体绝缘状态的评估。实例分析表明,相对传统方法,该方法在GIS设备气体的绝缘状态分类中具有较高的准确率。A multi-parameter assessment method of the gas insulation state in GIS equipment is proposed. Factors affecting the insulation conditions of GIS equipment are analyzed in detail, and the most influencing factors determined. Assessment of insulation using a parameter set of gas composition, temperature, humidity, pressure and purity is carried out. Finally, a one-on-one （ORO） classification method is used to establish a GIS insulation state model based on a relevance vector machine（ RVM ）. The GIS gas insulation condition can be divided into four categories corresponding to ＂normal＂, ＂attention＂, ＂abnormal＂ and ＂malfunction＂. The model is trained with training samples before it is used for the evaluation of gas insulated equipment state. A use case analysis shows that the method has better accuracy in the classification of insulation state of the GIS equipment.

关 键 词：GIS 相关向量机 绝缘状态评估模型 一对一分类方法 六氟化硫 

分 类 号：TM855[电气工程—高电压与绝缘技术]