Transformer Fault Detection Based on IAHA Optimized BP Neural Network

In view of the shortcomings of artificial hummingbird algorithm (AHA) and BP neural network algorithm in the process of fault detection, this paper improves the AHA by introducing the chaotic sequence of tent map, and proposes an improved AHA (IAHA), which makes the initialization of hummingbird population more uniform. To test the effectiveness of the IAHA, using IAHA to optimize the BP neural network model, and is applied to fault detection together with AHA-BP model, SSA-BP model and GWO-BP model. The results show that the fault detection accuracy of the proposed IAHA-BP model is 97.5%, AHA-BP model is 92.5%, SSA-BP model is 90%, GWO-BP model is 92.5%. It shows that the proposed IAHA-BP model has higher accuracy in the field of transformer fault detection.