A variable-rate fertilization control system was developed for high-speed rapeseed seeding operations, based on a dual-variable control sequence involving the rotational speed and opening of a variable-radius centrifugal centralized fertilizer applicator. A GA-MOPSO fertilization decision model was constructed by analyzing the mechanisms of the multi-objective particle swarm optimization (MOPSO) and genetic algorithm (GA), and formulating a bi-objective model with fertilization error and controller response time based on calibration experiments. An integral sliding mode control (ISMC) algorithm was designed for precise control of the fertilizer shaft speed. Simulation results showed that at varying target fertilization rates, the GA-MOPSO algorithm achieved hypervolume indicators of 1.004, 1.029, and 1.023 across 30 independent runs, with superior convergence and uniformity compared with that of MOPSO and DE-MOPSO. The ISMC algorithm achieved a steady-state time of 0.212s, a steady-state error of 0.013%, and zero overshoot, outperforming both PID and fuzzy PID controllers. Bench tests indicated that, with a weight vector of (0.9, 0.1), the proposed decision model reduced the mean relative fertilization error from 4.17% to 2.27% and the average response time from 0.92s to 0.83s. The ISMC algorithm achieved an average error of 2.73%, with row-to-row coefficient of variation under 5.62%, outperforming traditional controllers. Road tests showed a mean relative error of 3.56% and an average response time of 0.79s. Field trials demonstrated that at speed of 6~12km/h and application rates of 300~600kg/hm2, the error remained below 4.90% and the maximum response time was 1.08s.The research result can provide technical support for high-speed variable-rate fertilization with centralized fertilizer applicators.