To analyze the flow characteristics of the air-cooled cooling medium of the system， the flow field analysis is carried out by the Fluent software. The simulation of the inlet static pressure is compared with the experimental results. The error between the test and simulation results is less than 5.5%. The accuracy of the simulation model is verfied. Further， an optimal design method of air-cooled flow paths of the system based on Isight is proposed to rationally distribute the cooling flow of parallel air-cooled paths and reduce the static pressure of the system inlet. Firstly， the optimization objective is determined to achieve the theoretical air volume distribution ratio and the minimum static pressure of the cooling inlet， and the experimental design and parameter sensitivity analysis are carried out. Secondly， the approximate model of flow simulation is constructed based on radial basis neural network algorithm. Finally， the multi-objective optimization design of the approximation model is carried out by multi-island genetic algorithm， and the optimal structural parameters are obtained. Results show that the static pressure of the cooling inlet is reduced by 25.2%， and both the main channel and the inner channel reach the theoretical design goal.