A Novel Transfer Learning Approach to Detect the Location of Transformers in Distribution Network
Muhammad Ali, Carlos Andres Macana, Krishneel Prakash, Bill Tarlinton, Robi Islam, Hemanshu Roy Pota · 2020
One of the key requirements for designing reliable and scalable electrical distribution networks is the identification of the correct location of its electrical components. The location of transformers in the distribution network is crucial as they are the backbone for transforming one form of electricity into another. In this study, a transfer learning approach based on artificial intelligence is presented to determine the location of transformers in distribution networks. To implement the concept, a dataset of transformer images is collected from public platforms such as Google images. The collected dataset is then used to train the transfer learning model developed in Python. A custom object detection model for transformers is developed using the YOLOv3 architecture. To evaluate the potential of proposed approach, the developed model is tested on real-world images of transformers. The experimental results show that the transfer learning method can achieve a high detection accuracy for the detection of transformer locations. The presented research would be useful in the educational and practical fields as it could be used by electricity companies to develop, extend, and prepare modern distribution networks.