Accurately predicting and effectively reducing sonic boom levels is one of the key issues in the development of the new generation of green supersonic civil aircraft. In order to improve the efficiency of low-boom optimal design for supersonic civil aircraft, a multi-fidelity optimal design program for low-boom was developed based on the CoKriging surrogate model combined with fast sonic boom prediction method and high fidelity sonic boom prediction method. The sonic boom prediction results of the TU-144 model are basically consistent with the experimental results, verifying the reliability of the two prediction methods. A parameter sensitivity analysis and optimal design were conducted on the wing shape of a certain supersonic civil aircraft model. The results showed that Stevens’ loudness level of the ground sonic boom was more sensitive to three parameters: the half span length of the outer wing, leading edge sweep angle of the outer wing, the half span length of the inner wing. After optimization, the maximum ground sonic boom overpressure was reduced by about 4Pa, and the Stevens’ loudness level was reduced by 4.26dB. Compared with the Kriging model that only uses high fidelity sample data, the CoKriging model integrates high and low fidelity sample data, saving about 43% of time cost while ensuring a certain prediction accuracy.