Aiming at the problem that the unordered charging of large-scale electric truck (ET) increases the peak load of the grid and affects the power quality, a cluster division and two-tier optimal scheduling strategy for peak load balancing scenarios were proposed. Firstly, the demand response model of ET participating in power grid peak regulation was established considering real-time road flow and multi-energy consumption factors. With logistics factors as characteristic quantities, ET was divided into day-ahead clusters by an improved fuzzy clustering algorithm. Secondly, based on the clustering results, combined with the different interests of power grid dispatching and enterprise users, a two-tier scheduling model was established under the framework of master-slave game considering the flexible time window to solve the charging and discharging power of pure electric heavy duty card in the cluster in real time. Finally, particle swarm optimization based on Kriging model was used to speed up the solving of the model. The simulation results of ET data in a logistics area show that the two-tier scheduling strategy based on cluster division and flexible time window can better smooth the load curve and reduce the scheduling deviation of clusters. At the same time, Kriging optimization algorithm is more fast in solving the two-tier optimization model.