In order to solve the optimization design problem of complex shape of underwater glider and improve the optimization efficiency of Kriging surrogate model while improving its lift drag ratio, an adaptive surrogate optimal algorithm is proposed. The proposed improved parallel expectation improvement（PEI） criterion and probability of improvement（PI） criterion are used for global exploration, and the minimizing surrogate prediction（MSP） criterion is used for local exploration. The switching between global exploration and local exploration is realized according to the relationship between new sample points and known sample points. The proposed algorithm, the maximum expected improvement（EI） criterion multiple expectation of improvement（q-EI） and the improved probability improvement（IPI） are combined through a mathematical example. The result shows that the proposed algorithm converges faster and has higher accuracy. Finally, the optimal algorithm is used to optimize the shape of the underwater glider, and the maximum lift drag ratio is increased by 19.37%.