In this paper, an improved self-adaptive particle swarm optimization (ISAPSO) algorithm is developed for estimating the best set of surge arrester model parameters. The purpose is to minimize the relative error between the calculated and manufacturer's measured residual voltage peak values for lightning, switching, and steep-front impulses. An objective function, which generalizes the model parameters for all impulse current types with different peak levels, is proposed. This objective function is the summation of three subfunctions that in each of them one type of impulse current with different current peak levels is considered. An efficient multiobjective approach is also applied along with the presented ISAPSO method. The proposed algorithm finds different sets of Pareto-optimal nondominated solutions for the objective functions. A fuzzy clustering technique is used to restrain the size of the repository within the desired limit. The proposed method is applied for a 150-kV TRIDELTA metal-oxide surge arrester in order to demonstrate its accuracy and effectiveness.