BP神经网络和支持向量机相结合的电容器介损角辨识

针对电力电容器介质损耗的计算方法稳定性较差，频率波动对介损角的辨识有较大影响的问题，提出了BP神经网络和支持向量机（support vector machine， SVM）相结合（BP-SVM）的辨识方法，并且首次应用于电容器介损角的辨识。在辨识过程中，首先，对电容器工作一段时间的信号进行采样和预处理，预处理后的信号作为训练集训练BP-SVM模型；然后，使用训练好的BP-SVM模型对预处理后新的采样信号进行辨识，判断介损角的变化量。此外，给出了基于BP-SVM模型的介损角表示信号D_δ（t）的计算过程，同时分析了在讨论域内信号D_δ（t）的幅值即是介损角δ。仿真分析结果表明，提出的BP神经网络和SVM相结合的电容器介损角辨识方法比基于深度学习的辨识方法具有更高的辨识准确率，并且频率变化对BP-SVM方法的辨识性能无明显影响。

Capacitor dielectric loss angle identification based on a BP neural network and SVM

The stability of the calculation method for dielectric capacitor loss is poor, and the frequency fluctuation has a great influence on the identification of dielectric loss angle. To overcome this limitation, an identification method in combination with a back propagating (BP) neural network and support vector machine (SVM), BP-SVM, is proposed. For the first time, BP-SVM is applied to the identification of capacitor dielectric loss angle. In the identification process, first, the signal of a capacitor working for a period of time was sampled and preprocessed, and these signals were used as a training set to train the BP-SVM model. Then, the trained BP-SVM model was used to preprocess the newly sampled signal. The sampled signal was identified to determine the amount of change in the dielectric loss angle. In addition, the calculation process of the dielectric loss angle representation signal, D_δ(t), based on the BP-SVM model, is given. At the same time, the amplitude of the signal, D_δ(t), in the discussion section, is the dielectric loss angle δ. The simulation analysis results showed that the proposed method for identifying the dielectric loss angle of capacitors combined with a BP neural network and SVM had a higher recognition accuracy than the deep learning-based identification method, and the frequency variation had no significant effect on the identification performance of BP-SVM.