Abstract: To address issues of traditional soybean seeder fertilizer systems, such as inability to dynamically adjust fertilizer rates according to soil conditions and low fertilizer utilization efficiency, an intelligent soybean fertilization system integrating fuzzy PID control with PSO-BP neural network has been designed．Soil pH, electrical conductivity, volumetric water content, and operating speed were introduced as control variables．A predictive model taking multi-source soil data as input was constructed to achieve real-time soil state perception and dynamic fertilizer application rate regulation along seeding path．Controller was implemented on an STM32 platform with integrated hardware and software, adopted a predictive-feedback dual-loop architecture verified through Matlab/Simulink simulations．Results demonstrated that system exhibited excellent robustness and adaptive adjustment capability, providing an effective approach for precision agricultural fertilization．Under conditions of simulation step size of 0.1 s and total duration of 1000 s, model maintained mean fertilization error within±0.03 units, with short response time and minimal fluctuation amplitude, significantly improving system's stability and control accuracy．