Dissolved Gas Analysis for Transformer Fault Based on Learning Spiking Neural P System with Belief AdaBoost
Xihai Zhang, Gexiang Zhang, Pirthwineel Paul, Jinquan Zhang, Tianbao Wu, Songhai Fan and Xingzhong Xiong
This paper proposes a bio-inspired learning approach, fault diagnosis method based on learning spiking neural P system with belief AdaBoost, for oil-immersed power transformer. The learning spiking neural P system is used for identification of the fault in the transformer under the framework of ensemble learning. To test the robustness of learning spiking neural P system with belief AdaBoost, the experiment is required to repeat many times to get average accuracy. The results of experiment show that the learning spiking neural P system with belief AdaBoost is effective in diagnosing faults in transformer for thermal and electrical fault situations with dissolved gas data and is superior to other methods, like Improved Three-Ratio Method, Back-Propagation Neural Network (BPNN), Support Vector Machine (SVM), Deep Belief Network (DBN), Learning Spiking Neural P system (LSN P system), in terms of the correctness of diagnosis results.
Keywords: Spiking neural P system, learning spiking neural P system, belief adaBoost, transformer fault diagnosis, dissolved gas analysis