摘要
利用油中溶解气体对变压器进行故障有无以及故障类别判断时,为抑制冗余信息的干扰,提取与分类模式密切相关的特征作为每层诊断模型的输入;增量学习算法通过提取模型的支持向量和误判样本,逐步积累样本的空间分布知识,提高诊断模型的精度与训练速度,同时剔除对构建模型无贡献的样本以节约存储空间。为提升算法的收敛速度,采用参数自适应优化算法动态搜索模糊支持向量机的模型参数。最后,通过实例将该算法与普通的多分类支持向量机以及多分类模糊支持向量机相比,得出该算法具有相对较好的收敛性和诊断效果。
When the oil dissolved gases are used to detect and classify the transformer faults,in order to suppress the interferences of redundant information,the features closely related to the classification structure are extracted as the inputs of the diagnostic model at each layer.By extracting the support vectors and false samples,the incremental learning algorithm gradually accumulates the information of sample spatial distribution,improves the accuracy and training speed of diagnosis model,and removes the useless samples to save storage space.The adaptive parameter optimization algorithm is applied to dynamically search the parameters of FSVM（Fuzzy Support Vector Machine） to enhance the convergence speed.Compared with multi-class SVM（Support Vector Machine） and multi-class FSVM,examples show that,the incremental learning algorithm has better convergence property and diagnosis results.
出处
《电力自动化设备》
EI
CSCD
北大核心
2010年第11期48-52,共5页
Electric Power Automation Equipment
关键词
模糊支持向量机
增量算法
隶属度
自适应
油中溶解气体
fuzzy support vector machine
incremental algorithm
membership
adaptation
oil dissolved gases
