Design of Online Health Monitoring System for Distribution Transformer Based on Deep Transfer Learning

In order to reduce the inaccuracy of health status monitoring of distribution transformer caused by low fault recognition ability, the design of on-line health status monitoring system of distribution transformer based on deep transfer learning is proposed. The system architecture based on data acquisition, web service and application is established. The hardware part designs the microcontroller, storage circuit, power supply and auxiliary circuit. Firstly, the collected data are processed by ARIMA on the microprocessor. Secondly, the deep learning migration algorithm is used to compare and judge the monitoring data. Finally, the TrAdaBoost algorithm is used to diagnose the health data. The experimental results show that the design system can improve the monitoring accuracy by setting a reasonable time series length, and the fault identification accuracy is as high as 97.26%. It has a high fault identification effect on the system, and it can get a better calculation efficiency by setting Bolt and Bolt concurrency number reasonably.