In order to solve the problems of high operating cost and poor service quality of school bus due to the scattered distribution of bus stops in rural areas, multi-objective SBRP (School Bus Routing Problem) models were developed for the mixed-load and non-mixed-load scenarios.In the non-mixed-load scenario, a model of the SBRP was developed to optimize the students' travel cost and school operating cost, while in the mixed-load scenario, another model of the SBRP was developed to consider the input cost and operation cost of the school bus.Several heuristic algorithms were compared, based on which the simulated annealing algorithm was selected to solve the models, and the horizontal comparison of the solution results based on genetic algorithm were determined.Tests were conducted on an international bench mark case and the constructed models were solved by introducing different search operators into the simulated annealing algorithm, then the proposed approach was applied to the optimal design of school bus routes in Wulian county, Rizhao, Shandong province.The results showed that in the non-mixed-load scenario, compared with the original school bus operation mode, the school bus input, mileage and travel cost were reduced by 28.6%, 37.8% and 35.6%, respectively, and students' travel cost was reduced by 4.3% considering the students' perception of school bus service.While in the mixed-load scenario, the proposed approach reduced the school bus input, mileage and travel cost by 37.5%, 42.0% and 35.8%, respectively;due to the complexity of the mixed-load scenario, it is difficult to take the travel cost into account, thus the students' travel cost was increased by 0.5%.The proposed SBRP models were verified to be effective and the simulated annealing approach can optimize service quality and reduce operation cost of rural school bus to a greater extent than the genetic algorithm.