Abstract: Simulation of large-scale interconnected power grids is relatively complex. A power system reduction method based on spectral clustering and PTDF is proposed in this paper to reduce complexity of power network and build large-scale equivalence mode of power grid due to that traditional equivalence method of power grid is not suitable for constructing equivalence model of whole system. Firstly, the power transfer distribution matrix is selected as the similarity criteria, and normalized spectral clustering method is used to partition the system by considering connectivity between buses. Then based on the clustering result, the power network is equivalence through bus aggregation technique by considering the power flow limits of branches. In addition, the equivalence model is built, and its parameters are calculated with aim of guaranteeing the correctness of power flow transfer between regions. Finally, a model for calculating marginal zonal price is presented based on equivalence model, and the calculation accuracy is verified through case analysis.