Abstract: The optimization of site selection and capacity determination of urban electric taxi charging stations and reduction of the inefficiency of charging station usage are of great significance in reducing drivers' time costs and promoting green travel. First, the operation process of electric taxis is simulated to obtain random charging demand points. Then, the Dijkstra algorithm and queuing simulation model are used to simulate the taxis' searching process and waiting process for charging. Next, an optimization model for charging station location selection is established with the objective of minimizing the searching time, queuing time and loss costs. Experimental results show that the improved ALNS algorithm achieves optimal performance and minimizes the costs with reasonable station locations and charging pile allocation.