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

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.