A Power Transformer Fault Diagnosis Method Based on Improved Sand Cat Swarm Optimization Algorithm and Bidirectional Gated Recurrent Unit

The bidirectional gated recurrent unit (BiGRU) method based on dissolved gas analysis (DGA) has been studied in the field of power transformer fault diagnosis. However, there are still some shortcomings such as fuzzy boundaries DGA data, and BiGRU parameters difficult to determine. Therefore, this paper proposes a diagnosis landmark isometric mapping (L-Isomap) Improved Sand Cat Swarm Optimization (ISCSO) optimize (ISCSO-BiGRU). Firstly, L-Isomap is used extract features from feature quantities. In addition, ISCSO further proposed build an optimal model BiGRU. For ISCSO, four improvement methods proposed. traditional sand cat swarm algorithm improved using logistic chaotic mapping, water wave dynamic factor, adaptive weighting, golden sine strategy. Then, benchmarking functions test optimization performance algorithms, results show that best accuracy convergence speed. Finally, ISCSO-BiGRU obtained. Using for diagnosis, example simulation L-ISOMP filter downscale inputs can better improve performance. compared with SCSO-BiGRU, WOA-BiGRU, GWO-BiGRU, PSO-BiGRU models. rate 94.8%, which 11.69%, 10.39%, 7.14%, 5.9% higher than PSO-BiGRU, respectively, validate effectively transformers.