<p>As an electromagnetic mechanism that converts electric energy, magnetic energy and mechanical energy, the accurate solution of its electromagnetic behavior is the key to evaluate the transmission capacity and design optimization of the motor. The authors combine the subdomain continuity theory of accurate subdomain method with physics-informed neural network (PINN-SCT) to calculate the electromagnetic characteristics of permanent magnet vernier motor (PMVM) with magnetic field modulation function. The dielectric interface continuity in accurate subdomain method is added to PINN, and the electromagnetic behavior of PMVM is calculated by applying physical constraints through the Maxwell’s equation and special solution in neural network (NN) loss. The method can accurately determine the electromagnetic characteristics of PMVM without discrete grid or complex formula derivation. Compared with the finite element method (FEM), the error is less than 3%, and the time consumption is shorter and without fine mesh generation. The authors analyze the influence of each neural network parameter on the electromagnetic behavior solution. The authors use transfer learning (TL) to solve similar problems or models, such as the electromagnetic behavior of split teeth, to reduce the training cost. This method reduces the training cost by 50% and enhances the generalization ability of PINN. Finally, the feasibility and accuracy of the proposed method are verified by the prototype test.</p>