基于 QPSO-SVR 算法的 SF₆ 断路器触头烧蚀状态评估

The prediction of contact resistance and mass loss of contacts plays an important role in the condition assessment of high-voltage SF₆ circuit breakers. In this paper, a method based on quantum particle swarms optimization and support vector regression (QPSO-SVR) is proposed to effectively predict the incremental contact resistance and mass loss of arc contacts under different arc ablation conditions. The optimum training parameters for the SVR algorithm are obtained by combining the experimental data. The QPSO-SVR method has shown good prediction capability for different arc ablation conditions when compared with other prediction methods. The relative error of prediction for the incremental contact resistance is 3.023%, while the relative error of prediction for the mass loss is 4.61%, both of which show good robustness. Finally, the mass loss and contact resistance increments obtained from QPSO-SVR prediction and the monitored accumulated arc energy are subjected to fuzzy logic inference to construct a contact ablation state assessment system based on QPSO-SVR algorithm, which classifies the contact ablation state into four classes: class O, class I, class II and class III. The method can provide reference for high-voltage SF₆ circuit breaker maintenance.