Abstract: Abstract: The wheel motor drive system is one of the structure transmission schemes in an electric tractor. The coordinated control of electronic differential speed and torque can significantly improve the energy utilization of tractors. The motor of the electric tractor can also be urgently required for the energy conversion device for higher energy, efficiency, and reliability, compared with the traditional fuel tractors. Among them, the finite element (FE) method can be used to accurately predict and evaluate the electromagnetic performance of the wheel drive motor in electric tractors and then to optimize the structure of the electromagnetic converter. However, traditional models cannot consider the bidirectional magnetic-thermal coupling between the electromagnetic and temperature field, leading to the low prediction accuracy of the model. In this work, the FE model was developed and then verified by a series of experiments. A kind of 12-slot, 7-pole brushless DC wheel drive motor with the reducer was also taken as the research object. Dynamic characteristics of the motor were then analyzed under no-, half- and rated loading. Simulation results show that there was no magnetic leakage, where the flux density was in the unsaturated state under no-load conditions. The loss of copper, iron, and eddy current in the motor rose dramatically with the increase of the load, where the copper loss increased the most. Specifically, the copper losses of the motor tended to be 0, 60, and 340 W, respectively, under no-, half- and rated loading. The average core losses were 9, 15, and 22 W, respectively, whereas, the average eddy current losses were 11, 32, and 58 W, respectively, under the three load conditions. The magnetothermal bidirectional coupling model of the motor was then constructed to simulate the distribution of the temperature field. The results demonstrated that the highest temperature of the motor occurred on the permanent magnets under three working conditions. Temperatures of permanent magnets were elevated by 54.61, 77.63, and 87.1 ℃, respectively. In addition, the air-gap width, stator-notch width, spacing, and the width of the permanent magnet were set as the variables, while the groove torque, torque density, and permanent magnet loss were taken as the optimization objectives. A two-layer multi-objective optimization was proposed to improve the optimization efficiency and accuracy using Taguchi and response surface. The optimal structural parameters were determined as follows: The optimized spacing between permanent magnets, the width of permanent magnets, the air-gap width, and stator-notch width were 0.4, 3.5, 1.1, and 4.75 mm, respectively. Simulation results showed that the motor torque density increased by 8% compared to that before optimization, whereas, the permanent magnet loss decreased by 18.6%. The bench test of the prototype showed that the groove torque of the motor was reduced by 20.9% after optimization, indicating the effectiveness of the optimization. This finding can provide a strong reference to promote motor performance.