Abstract: In order to enhance the performance of surrogate-assisted evolutionary algorithms for solving high-dimensional expensive optimization problems, this paper proposed a dual model-driven differential evolution with multi-preference strategy adaption（SOEA-SS）. SOEA-SS relied on three multipreference evolutionary strategies supported by global and local surrogates. At each iteration, SOEA-SS adaptively adjusted the evolutionary strategies in a sequential manner to strike the global exploration and local exploitation equilibrium, according to the online feedback concerning the update of optimal solution.In order to promote the optimal information sharing among the population, an elites-driven differential perturbation strategy was developed to enrich the prior knowledge of the optimal regions. Experimental results show that SOEA-SS has significant superiority over four advanced algorithms on at least 17 out of 26 high-dimensional benchmark problems.