Distribution Network Topology Identification Based on Co-Training of ResNet and Conformal Prediction

With the development of new power systems, the topology of distribution networks has become increasingly complex. Changes in topology occur more frequently. Accurate topology is the foundation for power flow calculations and state estimation in power systems. Therefore, accurate topology identification is an important and challenging research topic. Existing topology identification methods often only provide one possible operating topology with no guarantee of confidence. Thus, this paper proposes a distribution network topology identification method based on co-training of ResNet and conformal prediction. The Resnet-Based classification model is constructed by establishing a mapping relationship between distribution network measurement data and operating topology. The training process of the ResNet is linked to the calibration process of conformal prediction, which is established by incorporating the performance metric for distribution network topology identification into the loss function of ResNet. The proposed method is validated using the IEEE 123-node distribution network test case. The experimental results demonstrate that the proposed model can improve the performance of distribution network topology identification.