The three-dimensional garage standby position strategy has been demonstrated to improve the efficiency of the vehicle out of the warehouse and reduce customer waiting times. A zone-based standby position strategy was designed by predicting vehicle departure time points within the parking system. To address the strong sample dependency of back propagation(BP) neural networks, a k-nearest neighbor(KNN) -BP neural network prediction model was proposed. Features such as workdays, specific time periods on workdays (e. g., morning and evening peak hours), temperature, and precipitation were utilized as feature vectors. These features were converted into binary sets to construct feature vector sets, and KNN was employed to group anomalous data, thereby enhancing the prediction accuracy of the BP neural network. Based on the predicted departure times, a departure probability distribution was established, and regional priority parameters along with an rail guided vehicle (RGV) standby position strategy were set. A simulation program was developed and validated using operational data from a commercial automated parking system in Xiaozhai, Xi'an. The simulation results demonstrate that the KNN-BP neural network predictive model increases by 20.17% compared to the traditional BP neural network. The average waiting time is reduced by 35.82%. The average service time of RGV is reduced by 39.51%, and the energy consumption of RGV operating is reduced by 38.32%. The average waiting time of the strategy cited strategy is reduced by 14.18%, the average service time of the RGV is 13.29%, and the energy consumption of RGV operating energy consumption is reduced 20.89%. The research results provide reference for improving the efficiency of stereo garage vehicles.