Abstract:Aiming at the situation of pesticide residues and less spraying under the conventional field quantitative spraying method, at the same time, in order to improve the timeliness of the existing variable spray control system and solve the lag of fuzzy decisionmaking, the BP neural network PID variable spray system was designed. Based on the multisensor realtime monitoring of speed, flow rate and pressure, the system used neural network selflearning ability to modify PID parameters toprecisely control the return flow of the liquid. It solved the technical problems such as large overshoot, large steadystate error and long response time in the existing variable application control algorithm, and realized the purpose of constant application amount per unit area in the field. In order to verify the superiority of the system algorithm in accurate variable application, the conventional PID, fuzzy PID and neural network control mode were modeled and simulated under the Simulink platform. Through comparison, it can be seen that the neural network PID control was superior to the other two control modes in terms of rising time, overshoot and steadystate error. The field experiment showed that the standard deviation of droplet deposition number was less than 1.4 per square centimetre, the coefficient of variation of longitudinal uniformity was less than 6%, and when the speed varied randomly in the range of 4~11km/h, the average adjustment time of the system was 0.72s, the average overshoot was 2.1%, and the difference between the dosage and the setting was 1.3%.