基于单传感器快速独立分量分析电弧故障检测的研究

为了准确识别电弧故障特征信号,提出在单个传感器的情况下运用FastICA算法。通过对不同负载分离后的噪声信号进行高阶累积量分析,发现故障电弧发生时,电路中的噪声信号的三阶累积量发生明显的负偏移状态,相比负载正常工作时电路中的噪声信号为近似白噪声,其三阶累积量近似为0。大量试验表明,通过检测电路中噪声信号的三阶累积量变化可以有效识别电弧故障。

Research of Arc Fault Detection Based on Fast Independent Component Analysis Using Single Sensor

In order to accurately identify arc fault characteristic signal, this paper proposed an algorithm of using FastICA in the case of single sensor. The arc fault experimental platform based on the UL1699 standard separates the reconstructed current signal in the circuit using FastICA algorithm. Through high-level statistical analysis of the separated noise signal of different loads,it is found that the third-order cumulant of the noise signal in the circuit has obvious negative deviation when arc fault takes place. In comparison with the noise signal which is close to white noise in the circuit of normal load, the third-order cumulant is close to zero. Therefore, arc fault can be identified by detecting third-order cumulant change of noise signal in the circuit. It is shown in a large number of experiments that arc can be identified effectively by the proposed method.