基于SVM增量学习算法的煤矿高压断路器故障模式识别方法

发布时间：2018-11-08 10:56

【摘要】：高压断路器故障模式的准确识别是矿井电网智能化发展过程中的重要支撑环节。针对高压断路器故障数据不易获取且故障样本较少的问题,提出了一种支持向量机与增量学习算法相结合的故障识别方法,确定了以断路器控制回路电流信号、电压信号以及分合闸振动信号为状态监测量,模拟了弹簧松动、铁芯卡涩、供电异常与线圈老化4种常见故障,提取了故障特征量并建立了故障数据样本与增量学习数据样本,采用支持向量机增量学习算法训练得到了故障识别模型,并利用新增数据样本对其进行了验证。结果表明:支持向量机增量学习算法可准确识别上述4种常见故障,并可以通过对新增样本的不断学习进一步提高识别精度。
[Abstract]:The accurate identification of fault mode of high voltage circuit breaker is an important supporting link in the process of intelligent development of mine power network. Aiming at the problem that the fault data of high voltage circuit breaker is difficult to obtain and the fault samples are few, a fault identification method combining support vector machine and incremental learning algorithm is proposed, and the current signal of circuit breaker control circuit is determined. Voltage signal and closing vibration signal are state monitoring quantity. Four common faults are simulated, such as spring loosening, iron core sticking, power supply anomaly and coil aging. Fault characteristic quantity is extracted and fault data sample and incremental learning data sample are established. The incremental learning algorithm of support vector machine is used to train the fault identification model, and the new data sample is used to verify the model. The results show that the support vector machine incremental learning algorithm can accurately identify the four common faults mentioned above and can further improve the recognition accuracy through the continuous learning of new samples.
【作者单位】：矿用智能电器技术国家地方联合工程实验室(太原理工大学);煤矿电气设备与智能控制山西省重点实验室(太原理工大学);中国煤炭科工集团太原研究院有限公司;
【基金】：国家自然科学基金资助项目(U1510112,51377113) 山西省科技重大专项资助项目(20131101029)
【分类号】：TM561

【相似文献】