基于特高频信号的GIS局部放电诊断方法OA

To improve the discriminability of feature extraction and the accuracy of recognition in ultra-high frequency signal-based partial discharge diagnosis for gas insulated switchgear,a partial discharge diagnosis method for gas insulated switchgear is proposed by integrating improved complete ensemble empirical mode de-composition with a Transformer model.First,the improved complete ensemble empirical mode decomposition is used to decompose ultra-high frequency partial discharge signals,where an animated oat optimization algo-rithm is introduced to optimize key parameters and improve decomposition performance.Second,the Pearson correlation coefficient is adopted to further screen the decomposition results,and key feature indicators are ex-tracted to construct the shallow feature matrix of partial discharge signals.Then,a Transformer model im-proved by the multi-head causal self-attention mechanism is introduced to extract and fuse deep features of par-tial discharge signals.Finally,a gas insulated switchgear partial discharge fault simulation platform is built to simulate four typical fault types,and the collected ultra-high frequency partial discharge signals are subjected to feature extraction and fusion using the proposed method.A radial basis function support vector machine model optimized by the animated oat optimization algorithm is then constructed for fault diagnosis.Experimen-tal results show that the proposed method can effectively extract features from ultrahigh frequency partial dis-charge signals and improve the diagnosis accuracy.