基于SVR的开关柜局部放电估计模型的研究

摘要

The SVR-based switchgear partial discharge estimation model was built. Making use of both ultra-sonic wave and high-frequency impulse current measured by switchgear partial discharge monitoring system to train SVR model and verify its accuracy was implemented. Simulating it with Matlab shows that, the mean square error of training set stays at 0. 015 2 together with a 0. 890 determination coefficient;as for a test set’s mean square error of 0 . 026 5 , the determination coefficient can reach 0 . 849 . This model can effectively esti-mate the switchgear partial discharge quantity and provides early warning for it.%构建了一种基于支持向量回归机( SVR)的开关柜局部放电估计模型,利用开关柜局部放电在线监测系统所测得的超声波和高频脉冲电流数据对SVR模型进行训练,并验证其准确度。 Matlab仿真结果为：训练集的均方差为0.0152,决定系数为0.890,测试集的均方差为0.0265,决定系数为0.849。该模型有效地对开关柜局部放电量做出了估计,为开关柜出现局部放电提供预警支持。