The integration technology of magnetic components has been widely used as the converters tend to be lighter and more integrated nowadays. The design of high-frequency magnetic components is extremely important in DC-AC converters. The paper proposes a surrogate-assisted multiobjective optimization method for the integrated transformer in series resonant DC-AC converters, which can quickly give the Pareto front of the loss and volume of the transformer, avoiding the drawbacks of long traversal time and tedious process in conventional design methods. It is achieved by integrating Latin hypercube sampling (LHS) and extreme learning machine (ELM). The optimization variables and objectives of the integrated transformer are first identified. Then, a trained surrogate-assisted model is established by LHS and ELM. Subsequently, the Pareto front is obtained quickly with the help of a multiobjective optimization algorithm non-dominated sorting genetic algorithm-II (NSGA-II). Finally, the accuracy of the surrogate-assisted model and the results of the multiobjective optimization are well verified by simulation. Compared with conventional methods, the optimization time is shortened by 99%.