经验模态分解和近似熵下的隔离开关失效预警

Isolation Switch Failure Warning under Empirical Mode Decomposition and Approximate Entropy

拥有隔离开关的设备由于大多具有封闭性,难以直观地确认其隔离开关和接地开关的位置状态,通常通过辅助开关发出的二次分合闸信号来判断其内部是否失效。一旦辅助开关位置信号出现误差,影响对隔离开关失效状态的判断。提出一种考虑信号特征的隔离开关失效自动化预警方法。通过改进小波去噪阈值函数来处理隔离开关振动信号,以降低噪声影响。通过集合经验模态分解和近似熵提取振动信号特征,生成多维特征向量。为了降低维度并保留有效信息,使用核主成分分析对多维特征向量进行降维。采用支持向量机建立失效自动化预警模型,根据特征向量预测传动链的失效情况。实现隔离开关失效自动化预警。实验结果表明,所提方法预警正确率高、漏检率和虚警率低且计算时间短。

Due to the fact that most equipment with isolation switches are enclosed,it is difficult to visually confirm the position and status of their isolation switches and grounding switches.Usually,the internal failure is judged by the secondary opening and closing signal sent by the auxiliary switch.Once there is an error in the position signal of the auxiliary switch,it will affect the judgment of the failure status of the isolation switch.Propose an automated warning method for isolation switch failure considering signal characteristics.By improving the wavelet denoising threshold function to process the isolation switch vibration signal,the impact of noise is reduced.Extracting vibration signal features through ensemble empirical mode decomposition and approximate entropy to generate multidimensional feature vectors.In order to reduce dimensionality and preserve effective information,kernel principal component analysis is used to reduce dimensionality of multidimensional feature vectors.Establish an automated failure warning model using support vector machines,and predict the failure of the transmission chain based on feature vectors.Realize automatic warning of isolation switch failure.The experimental results show that the proposed method has high accuracy in early warning,low missed detection and false alarm rates,and short calculation time.