Abstract: The conventional magnetic levitation belt conveyor adopts the electromagnetic structure composed of permanent magnets and electromagnets, which has the problems of easy heat generation and high current loss under the working conditions with high demand of magnetic levitation support force. To solve this problem, an electromagnetic structure based on Halbach array is proposed in this study. The mathematical model of electromagnetic structure optimization is established with the maximum magnetic induction intensity of electromagnetic structure as the objective function and the size of electromagnetic structure and the range of magnetic induction intensity distribution as the constraints. When solving the mathematical model of electromagnetic structure optimization, the Teaching and Learning Optimization (TLBO) algorithm is easily to fall into the local optimum. To solve this problem, an improved TLBO algorithm is proposed so as to enhance the diversity and search ability of the population by introducing new populations through screening and improving the learning methods in the teaching stage and mutual learning stage. The test results show that the accuracy and stability of the improved TLBO algorithm are better than the standard TLBO algorithm. The improved TLBO algorithm is used to solve the electromagnetic structure optimization mathematical model of the magnetic levitation belt conveyor. The optimal electromagnetic structure parameters are obtained as follows: the height of a single permanent magnet in Halbach array is 7 mm, the width is 9 mm, and the number of permanent magnets is 7. The experimental results show that under the same size conditions, the maximum magnetic induction intensity of the Halbach array-based electromagnetic structure is increased by 47.69% compared with the permanent magnet-based electromagnetic structure.