Numerous coal-fired power plants in China have been progressively implementing ultra-low emission retrofits. However, the current zoned ammonia injection method still leads to problems such as uneven reactions and excessive ammonia injection, which affect denitrification performance. This research proposes a precise coordinated ammonia injection strategy for retrofitted zonal denitrification systems. First, a multi-objective evolutionary optimization method based on generative adversarial network and classification preselection is proposed. It addresses problems including dimensional curse, low computational efficiency, and operator randomness. Next, a fuzzy grey relation projection method based on analytic hierarchy process and inverse entropy weights is proposed. It introduces the membership degrees defined by fuzzy set theory to perform multi-attribute decisions and determine the optimal ammonia injection plans. Finally, the experiments are conducted on a 660 MW supercritical coal-fired power generation unit in China that has undergone the ultra-low emission retrofit. The results indicate that, compared with conventional methods, the proposed strategy improves the zonal reaction uniformity by an average of 34.76% and reduces denitrification costs and outlet nitrogen oxides (NOx) concentrations by an average of 36.94% and 32.74%, respectively. It not only resolves the problem of poor denitrification performance following the zonal retrofit but also significantly enhances the unit’s economic efficiency.