基于支持向量机的微机保护装置状态评估的研究

Study on state evaluation for microprocessor protective device based on SVM

为了给微机保护装置的状态检修或计划检修提供科学的决策依据,提出了一种基于支持向量机的微机保护装置状态评估的方法。首先采用运行工况、定检信息作为支持向量机的输入特征向量,然后通过核函数将输入特征向量映射到高维特征空间,用支持向量机的模式识别方法来识别微机保护装置状态。实验结果表明SVM对微机保护装置进行状态评估是可行、有效的,在小样本情况下有较高的评估正确率和较好的稳定性,径向基核函数的SVM分类方法应用于微机保护装置状态评估最理想。同样条件下比人工神经网络的评估正确率高,速度快。

To provide the scientific decision basis of state maintenance or scheduled maintenance for microprocessor protective device, a method of state evaluation of microprocessor protective device is brought forward based on support vector machine（SVM）. The running and scheduled maintenance information is taken as the input vectors of SVM. Then they are mapped to high-dimensional feafure space through kernel function, and state evaluation of microprocessor protective device is recognized by SVM. The results show that the SVM method is feasible and effective for state maintenance of microprocessor protective device, superior classification and better stability with small training set of sample. The comparison of different kernel functions for SVM shows that RBF kemel function is most suitable for state evaluation of microprocessor protective device. It is more higher evaluation rate and speed compared with neural network for state evaluation as the same condition.