In order to achieve accurate prediction and optimization of milling surface residual stress. In this paper, Al7075-T6 aluminum alloy was taken as the research object. Firstly, a surface residual stress prediction model was proposed by analyzing the forces of orthogonal milling model and considering the machining parameters, milling force and heat. Compared with the traditional exponential empirical model, the superiority of the improved prediction model was proved. Taguchi algorithm and average signal-to-noise ratio were used to analyze the significant effect of process parameters on residual stress on workpiece surface, and the optimal process parameters for residual compressive stress generation were obtained. The optimized process parameters were verified by experiment and prediction model. The results show that compared with the traditional exponential model, the prediction accuracy of surface residual stress in X direction and Y direction is improved by 18.5% and 8.2%, respectively. After parameter optimization, the residual compressive stress in X and Y direction increases by 16.0% and 6.3%. The surface roughness value did not decrease obviously, which provides a theoretical basis for active control of residual stress of aerospace thin-walled parts.