An Enhanced Cuckoo Search Algorithm and Its Application in Transformer Fault Diagnosis

Background: Dissolved Gas Analysis (DGA) is an effective method to detect the potential fault of oil immersed transformer, while the ratio method based on DGA technology has some defects. Objective: This study aims to timely comprehend the transformer condition and accurately predict the incipient fault. Methods: By investigating the recently published papers and patents relating to transformer fault diagnosis, a novel diagnosis model based on the enhanced cuckoo search algorithm combined with BP neural network (ECSBP) is designed. In the proposed approach, the adaptive adjustment strategy of step size is employed to enhance the convergence rate and solution quality. Furthermore, the dimension by dimension improvement mechanism is also utilized to balance the exploration and exploitation. Subsequently, the transformer fault diagnosis model using ECSBP approach is built. Results: By evaluating the optimization performance on a set of benchmark functions, the enhanced cuckoo search algorithm is apparently superior to the original CS and its 2 variants. Simultaneously, compared with the other 4 transformer fault diagnosis models, ECSBP method has the highest diagnostic accuracy. Conclusion: The proposed ECSBP model integrates the advantages of artificial intelligence and DGA technology, and then can be used in the fault diagnosis of power transformer.