Abstract: Accurate real-time topology identification of distribution networks is the basis for safe and stable operation of power systems, but with the access of renewable energy and the increasing scale of distribution network, the dynamic changes of the distribution network topology are more frequent and difficult to identify. However, the historical data used by the existing topology identification algorithms need to be labeled manually, and the topology identification time is long. So it is difficult to achieve realtime topology identification for distribution networks. Therefore, a real-time two-stage topology identification algorithm for distribution networks based on field programmable logic gate array（FPAG） is proposed. The proposed algorithm does not need to know the number of distribution network topology categories in advance, and can label and classify the existing historical data, and realize the real-time topology identification for distribution networks based on FPGA. The algorithm is divided into two stages. The first stage uses the variational Bayesian Gaussian mixture model to label and classify the existing historical data. In the second stage, the sparrow search algorithm is used to make the support vector machine converge quickly to get the optimal parameters, so as to realize the accurate topology identification for distribution networks. Based on this algorithm, a real-time topology identification platform is established by using FPGA parallel architecture and high speed and high density characteristics. Finally, the effectiveness and superiority of the proposed identification method are verified by case analysis.