基于参数优化的最小二乘支持向量机触电电流检测方法

发布时间：2018-05-05 11:18

本文选题：优化 + 电流　；参考：《农业工程学报》2014年23期

【摘要】：针对如何从低压电网总泄漏电流中检测出生物体触电电流信号的难题,提出了一种基于网格搜索和交叉验证的最小二乘支持向量机的触电电流信号检测方法。首先在剩余电流动作保护装置触电物理试验系统平台上通过故障录波器获得生物体在3个典型时刻(电源电压最大时刻、电源电压过零时刻及电源电压任意时刻)发生触电过程的总泄漏电流和触电电流波形,并截取触电前1个周期和触电后3个周期共800个采样点的信号数据作为触电试验样本数据;然后将触电试验样本数据进行滤波预处理,预处理后的多个样本采样点的总泄漏电流组合成特征向量输入最小二乘支持向量机(least square-support vector machine,LS-SVM),相应样本采样点的触电电流作为其输出,并通过网格搜索与交叉验证相结合的方法来优化最小二乘支持向量机参数,利用输出最优参数组合对触电电流与总泄漏电流的关系进行训练,从而建立了触电电流的检测模型;最后利用该方法对10组测试样本数据进行了检测,检测结果为:当训练样本数据为20组时,检测均方误差为14.0040,当训练样本数据为40组时,检测均方误差为11.7469,当训练试验数据为65组时,检测均方误差为11.1849。与径向基(radial basis function,RBF)神经网络方法相比,最小二乘支持向量机方法比径向基神经网络方法检测均方误差分别低3.7272、1.9132、0.1556,从而可较准确地从总泄漏电流中检测出生物体触电电流信号,为开发新一代基于生物体触电电流分量而动作的自适应型剩余电流保护装置提供理论依据。
[Abstract]:In order to solve the problem of how to detect the biological electric current signal from the total leakage current of the low-voltage power network, a method of detecting the electric shock current signal based on the least square support vector machine (LS-SVM) based on grid search and cross-validation is proposed. First, on the platform of electroshock physical test system platform of residual current action protection device, through fault recorder, the organism is obtained at three typical times (maximum power supply voltage, The total leakage current and the electric shock current waveform of the electric shock process occur at the zero-crossing time of the power supply voltage and at any time of the power supply voltage. The signal data of 800 sampling points in the first cycle and three cycles after the electric shock are taken as the sample data of the electric shock test, and then the sample data of the electric shock test are filtered and preprocessed. After preprocessing, the total leakage current of several sample sampling points is combined into eigenvector input least squares support vector machine (LS-SVM), and the electric shock current of the corresponding sample sampling point is taken as its output. The parameters of least squares support vector machine are optimized by the combination of grid search and cross validation, and the relationship between the total leakage current and the electric shock current is trained by the optimal output parameter combination, and the detection model of the electric shock current is established. Finally, the method is used to detect 10 groups of test data. The results are as follows: when the training sample data is 20 groups, the mean square error of detection is 14.0040, and when the training sample data is 40 groups, the detection mean square error is 14.0040 when the training sample data is 40 groups. The mean square error was 11.7469, and the mean square error was 11.1849 when the training data was 65 groups. Compared with radial basis function RBF-based neural network method, least square support vector machine method is lower than radial basis function neural network method in measuring mean square error of 3.7272 (1.9132) and 0.1556, respectively, which can accurately detect the electrical current signal of organism from total leakage current, and the mean square error of the least square support vector machine method is lower than that of the radial basis function neural network method, and the mean square error is lower than that of the radial basis function neural network method, respectively. It provides a theoretical basis for the development of a new generation of adaptive residual current protection devices based on the components of biological shock current.
【作者单位】：中国农业大学信息与电气工程学院;黑龙江八一农垦大学信息技术学院;
【基金】：国家自然科学基金项目(51177165) 中央高校基本科研业务费专项资金资助(2013YJ008)
【分类号】：TM774

【相似文献】