基于布里渊光纤传感的电网电压监测仪导线故障检测方法

A Wire Fault Detection Method for Power Grid Voltage Monitor Based on Brillouin Fiber Optic Sensing

电网电压监测仪导线长期受到自然环境影响和侵蚀,导线很容易劣化甚至发生断裂等故障。在此背景下,进行故障检测是十分必要的。针对以往检测方法准确性不足的问题,研究一种基于布里渊光纤传感的电网电压监测仪导线故障检测方法。该方法利用布里渊光纤传感技术采集电网电压监测仪导线的背向布里渊散射光电信号并对其实施去噪处理。提取布里渊光纤传感信号的两个时域特征,即峭度指标和峰值指标,两个频域指标,即频谱集中程度和频谱分散程度,由此组成特征向量。利用随机森林算法构建检测模型,以特征向量为输入,得出电网电压监测仪导线故障类型并借助脉冲激光开始发射到被采集这一过程中所花费的时间,计算出导线上故障发生位置,完成电网电压监测仪导线故障检测。结果表明:所研究方法应用下,检测出样本1发生了劣化故障,故障位置为距激光器2.63 m处,样本3发生了断裂故障,故障位置为距激光器7.08 m处,而样本2则没有发生故障。所研究检测方法的基尼系数相对更大,大于0.9,由此说明该检测方法的准确性更高。

The wires of the power grid voltage monitoring instrument have been affected and corroded by the natural environment for a long time,and the wires are prone to deterioration or even breakage and other faults.In this context,fault detection is essential.To address the issue of insufficient accuracy in previous detection methods,a wire fault detection method for power grid voltage monitoring instruments based on Brillouin fiber optic sensing is studied.This method utilizes Brillouin fiber optic sensing technology to collect the back Brillouin scattering photoelectric signal of the power grid voltage monitor wire and perform denoising processing on it.Extract two time-domain features of Brillouin fiber optic sensing signals,namely kurtosis index and peak index,and two frequencydomain indicators,namely spectral concentration degree and spectral dispersion degree,to form feature vectors.The random forest algorithm is used to build the detection model.With the eigenvector as the input,the fault type of the wire of the power grid voltage monitor is obtained.The time spent in the process of starting to transmit the pulse laser to the acquisition is used to calculate the location of the fault on the wire and complete the fault detection of the wire of the power grid voltage monitor.The results showed that under the application of the studied method,sample 1 was detected to have a deterioration fault,with the fault location being 2.63 m away from the laser,sample 3 had a fracture fault,with the fault location being 7.08m away from the laser,while sample 2 did not have a fault.The Gini coefficient of the studied detection method is relatively larger,greater than 0.9,indicating a higher accuracy of the detection method.