基于集成学习的自适应提升分类模型的局部放电识别研究

Pattern Recognition for Partial Discharge Using Adaptive Boost Classification Model Based on Ensemble Method

根据3种不同单一特征集对于局部放电各类缺陷识别的差异性和互补性,提出了基于集成学习的自适应提升分类模型。首先,设计了8种类型局部放电的物理模型。然后,通过局部放电特高频检测试验系统,获取的每一缺陷在不同试验电压下稳定放电时的局部放电脉冲序列数据进行有效性与正确性的验证。通过相位相关脉冲序列数据获取的3种单一特征集两两组合以及3种联用构建新的特征集。通过比较分析,从单一特征集、组合特征集中选取最终的特征集作为分类模型的输入。最后,使用集成学习中提升算法处理训练数据集,以支持向量机为基分类器,通过使用基于信息熵的“非成对型”多样性指标,测量一个基分类器和其他基分类器的不一致度,得到一系列具有适度正确率的多样SVM基分类器,最终得到基于集成学习的自适应提升分类模型。对于每一种缺陷,在相同的测试电压水平下获得25个样本数据,6个电压水平总共获得150个本数据。结果表明,所提出的方法成功地识别了局部放电绝缘缺陷类型。

Based on the difference and complementarity of three different single feature sets for various types of defect recognition,this paper proposes an adaptive boost classification model based on the ensemble method.First,the eight types of partial discharge physical models are designed.Then,through the partial discharge detection and test system based on ultra-high frequency method,the partial discharge pulse sequence data sets for each defect in a stable discharge under different test voltages are obtained,which are used to verify the proposed method.After that,three kinds of single feature sets are extracted from the data sets.The combined feature set is made by these single feature sets with pairs and three.The final feature set is selected from these feature sets as the input of the classification model through comparative analysis.Finally,using the boosting algorithm and taking the support vector machine as the base classifier to obtain an adaptive boost classification model based on the multi-feature combination method.By using the "unpaired" diversity index based on information entropy to measure the inconsistency between one base classifier and other base classifiers,a series of diverse SVM base classifiers with moderate accuracy rates are obtained.For each defect,25 samples were obtained at the same test voltage level,and a total of 150 samples were obtained at 6 voltage levels through multiple experiments.The proposed method was compared with the traditional methods using these data sets.The results revealed that the proposed method successfully identified the types of partial discharge insulation defects.