基于惯性传感器协同1D CNN的输电线路覆冰情况识别方法

Identification Method of Transmission Line Icing Based on Inertial Sensor and 1D CNN

输电线路覆冰是最常见影响电网系统安全稳定运行的自然灾害。针对输电线路覆冰情况具有非线性增长、影响因素复杂等特点,提出了基于惯性传感器协同1D CNN的输电线路覆冰情况识别方法。该方法基于一维卷积神经网络,网络可自动从原始数据中进行特征提取和分类。采用集中质量法进行了输电线路覆冰的模拟实验,搭建了基于惯性传感器的输电线路模拟覆冰的三轴加速度的数据采集平台,构建了由未覆冰、轻度覆冰以及重度覆冰等三种线路覆冰情况的21824个样本组成的数据集。实验结果表明,基于惯性传感器协同1D CNN的线路覆冰情况识别的方法性能优秀,平均准确率可达到91%。

Transmission line icing is the most common natural disaster that affects the safe and stable operation of power grid system.Aiming at the characteristics of nonlinear growth and having complex influencing factors of transmission line icing,a method of transmission line icing situation identification based on inertial sensor and 1D CNN is proposed.The method is based on one-dimensional convolutional neural network,which can automatically extract and classify features from the original data.The concentrated mass method is used to simulate the icing of transmission lines,and a data acquisition platform to simulate the triaxial acceleration of the transmission line icing based on inertial sensors is built.A data set composed of 21824 samples of three kinds of icing conditions,i.e.,no icing,light icing and heavy icing is constructed.The experimental results show that the method based on inertial sensor and 1D CNN for line icing identification has excellent performance,and the average accuracy has reached 91%.