采用混合核函数支持向量机算法的大型发电机定子线棒绝缘状态评估方法  

An Evaluation Method for Insulation State of Large Generator Stator Bar Based on Support Vector Machine with Hybrid Kernel Function

作　　者：张跃 梁智明 胡波 张莹 李志成 刘凌[2] ZHANG Yue;LIANG Zhiming;HU Bo;ZHANG Ying;LI Zhicheng;LIU Ling(Dongfang Electric Machinery CoLtd., Deyang, Sichuan 618000, China;School of Electrical Engineering, Xi’an Jiaotong University, Xi’an 710049, China)

机构地区：[1]东方电气集团东方电机有限公司,四川德阳618000 [2]西安交通大学电气工程学院,西安710049

出　　处：《西安交通大学学报》2020年第6期44-50,共7页Journal of Xi'an Jiaotong University

基　　金：国家自然科学基金资助项目(51977173)。

摘　　要：为解决大容量发电机定子线棒的绝缘状态难以评估的问题,提出了一种采用混合核函数支持向量机算法的大型发电机定子线棒绝缘状态评估方法。该方法根据多因子老化试验平台,得到能表征定子线棒绝缘老化状态的非破坏性特征参量如吸收比、介质损耗、介损增量等在不同周期下的数值,并通过计算皮尔逊相关系数进一步验证上述非破坏性参量在数值上与剩余击穿场强的显著相关性;然后采用混合核函数支持向量机算法建立定子线棒非破坏性特征参量与剩余击穿场强的映射关系模型,预测出定子线棒剩余击穿场强;最后结合层次分析法及上述支持向量机算法的预测结果实现了对发电机定子主绝缘状态的评估。数值仿真结果表明:与传统单核函数支持向量机预测算法相比,基于混合核函数支持向量机和层次分析法的评估方法弥补了单核核函数预测结果不收敛的缺陷,对剩余击穿场强的识别准确率达到97%,该方法表现出了较强的状态评估能力。A method using a support vector machine algorithm with hybrid kernel function is proposed to solve the problem that it is difficult to evaluate the insulation status of stator bars of large-capacity generators.The method bases on a multi-factor aging test platform to obtain non-destructive characteristic parameters such as absorption ratio,dielectric loss,and dielectric loss increment at different periods that can characterize the insulation aging state of a stator bar.Pearson correlation coefficient is calculated to further verify the significant correlation between the above non-destructive parameters and the residual breakdown field strength.Then the support vector machine algorithm with hybrid kernel function is used to establish the mapping relationship between the non-destructive characteristic parameters of the stator bar and the residual breakdown field strength and to predict the residual breakdown field strength of the stator bar.Finally,the main insulation state of the generator stator is evaluated by the analytic hierarchy process and prediction results based on the support vector machine algorithm.Numerical simulation results and a comparison with the traditional single-kernel function support vector machine algorithm show that the algorithm based on the proposed support vector machine and the analytic hierarchy process makes up the shortcoming that the prediction results of the single-core kernel function do not converge,and the recognition accuracy of breakdown field strength reaches 97%.The method shows a strong state assessment ability.

关 键 词：发电机 定子线棒 非破坏性特征参量 支持向量机 状态评估 

分 类 号：TM31[电气工程—电机]