基于SVM-神经网络融合反馈的触电电流检测方法

Prediction Method of Electric Shock Current Based on SVM and Neural Network Fusion Feedback

在低压配电网剩余电流中检测出生物体触电电流信号并还原是一个典型的回归预测问题,然而单纯采用典型的回归预测方法如支持向量机(support vector machine,SVM)、神经网络(neural network,NN),效果并不理想。该文提出一种基于SVM-神经网络融合反馈的触电电流检测方法,该方法在标准SVM及神经网络的基础上进行融合判定,有效利用各个模型的优点进行融合分析。针对上述方法,该论文通过对实际动物、植物的触电实验获得相关训练数据和测试数据,实验表明基于SVM与神经网络的融合反馈方法可较大的提升生物体触电电流信号检测的准确性。

Identifying shock current signal and reverting from residual current in low voltage distribution network is a typical regression prediction problem. However, usually, the effect of mere utilization of simple regression forecasting methods such as support vector machine(SVM), neural network, is not ideal. In this paper, a fusion feedback method for electric shock current detection based on SVM and neural network is proposed. The method makes fusion determination, combining respective advantage of each model of SVM and neural network for integration analysis. Experiments are performed on real animals and plants to obtain training data and testing data. Results show that the fusion feedback method based on SVM and neural network can greatly improve accuracy of electric shock current signal detection.