To further improve the accuracy and reliability of transient stability assessment(TSA)

a feature selection method(Powershap) based on the combination of statistics and Shapley values is proposed

and a power system transient stability assessment model is established. Firstly

the input feature set is constructed based on the steady-state components during the operation of the power system. Powershap is used to divide the dataset into multiple subsets for training

and key feature sets are selected. Then

multiple CatBoost models are trained using key feature sets and transient stability assessments are conduct to generate transient stability assessment models.Finally

simulation experiments are conducted on the New England 10-machine 39-node system and the New England 54-machine 118-node system with the addition of new energy generation

and evaluation results are provided. The experiments show that

in the 10-machine 39-node system in New England

using the Powershap feature selection method for classification can achieve an accuracy of 99.79%. On the improved New England 54-machine 118-node system

its accuracy can reach 99.49%

indicating that the method can effectively perform transient stability assessment of power systems. It is verified that the proposed TSA model has good robustness and generalization ability.