In order to shorten the design cycle of compressor blades and improve the performance evaluation and efficiency of compressor blade profiles, a blade database using CFD based on the key parameters of the arc in the CDA blade profile and the inlet conditions of the blade profile is constructed. Based on this database, a model for the total pressure loss coefficient surrogate that can predict variable operating conditions is proposed. By combining the total pressure coefficient surrogate model with genetic algorithms, efficient optimization of CDA blade profiles can be achieved. By comparing the prediction results of the total pressure loss coefficient surrogate model with the CFD results, it was found that the mean square error of the prediction was between 0.03% and 0.04%, with a correlation coefficient higher than 0.97. The optimization results show that using this method for optimization reduces the total pressure loss coefficient of the original blade profile by 15% while significantly reducing optimization time and greatly improving optimization efficiency.