The integrated measurement and control gate, as a new type of measurement and control equipment, calculates the flow rate through ultrasonic probes arranged in layers inside the measurement box. However, when the flow pattern is turbulent or the flow pattern is rapidly changing, the individual probe will not be able to accurately obtain the path flow velocity, and the flow velocity area method will produce large errors when used for flow calculation. There is a lack of a method to rate the flow in the measurement box. A large number of indoor measuring box overflow tests were conducted on the box culvert type integrated measurement and control gate widely used in the Ningxia Yellow Diversion Irrigation Area. The effects of submerged water depth and probe shading on flow measurement accuracy were analyzed. Based on 450 sets of overflow test sample data and BP neural network and its optimization algorithm, a integrated measurement and control gate flow prediction model with adaptive flow state change capability was established, and the effectiveness and accuracy of the prediction model were verified. The results show that the submerged depth of the measuring box should not be less than 10 cm. To improve the anti-interference ability of the equipment, the number of probe layers should meet the requirement of at least 8 layers. Compared to the other three optimization algorithms, the back propagation (BP) model optimized by the sparrow search algorithm (SSA) or the particle swarm algorithm (PSO) has a higher prediction accuracy and a more concentrated distribution of the error. Comprehensively analyzing the three aspects of prediction accuracy, error distribution and running time, SSA-BP model can be selected when there is a high demand for prediction accuracy, and PSO-BP model can be selected when there is a high demand for running time. The prediction method can provide a reference for flow rate calibration and anomaly rejection of integrated gate for measurement and control in Ningxia Yellow Diversion Irrigation Area.