Abstract: A differential evolutionary (DE) algorithm based on search preference knowledge — PKLSHADE (i.e. preference-knowledge-based LSHADE) is proposed for complex multimodal optimization. PKLSHADE applies prior search preference knowledge in the evolutionary process of the population and differentiates the diversity and convergence of the population at different evolutionary stages, i.e., the importance is attached to the perturbation in the early stages of evolution to enhance the global development of the algorithm, and more local searches centering around the current optimal solution are carried out in the later stage. At the same time, the mutation strategy based on search preference knowledge can realize the global development of differential evolutionary algorithm and the smooth adaptive transition of local search to avoid the direct switch of the two search stages. The experimental results on the complex multimodal functions of CEC2017 show that PKLSHADE is superior to recent excellent algorithms such as LSHADE, EBLSHADE, jSO and AMECoDEs in terms of the accuracy of the optimal solution and the stability of the algorithm.