A dynamic correction system consisted of a deviation detection device， a correction execution device， a control system， etc. for the manure cleaning belt was designed to solve the deviation of the manure cleaning belt during the farming of duck with stacked cage. Firstly， the operation status of the manure cleaning belt was monitored in real-time through 6 pressure sensors arranged on both sides of the belt. When the manure cleaning belt deviated， its side pushed the anti-deviation device in the deviation detection device， causing the values of the 3 pressure sensors on the deviated side to change. The direction and offset amount of the cleaning belt was indirectly inferred through the changes. When the STM32 microcontroller determined the deviation of the manure cleaning belt based on the data collected from sensors， it automatically controlled the movement of the stepper motors on both sides， thereby adjusting the forward and backward movement of the drum along the sliding rail to achieve correction. At the same time， the Fuzzy PID control algorithm was introduced to achieve closed-loop control through feedback from displacement sensors to improve the stability and control accuracy of the control system. The results showed that the overall success rate of correction was 85%， with an average deviation of 17 mm and an average relative deviation of 16%. It is indicated that the dynamic correction system designed for the manure cleaning belt basically meets the requirements of the stacked cage-farming duck for automatic correction of the manure cleaning belt， greatly reducing the labor intensity of the farming personnel.