With the ongoing transformation of new-type power system

the extensive integration of renewable energy and the increasing complexity of loads are significantly altering the dynamic behaviors and operational characteristics of the power system. As data-driven "black-box" models

surrogate models offer a promising solution by approximating the responses of original models at a reduced computational cost

thereby replacing or complementing conventional analysis methods. However

different surrogate model techniques exhibit significant differences in data characteristics

prediction accuracy

and response speed. Therefore

the challenge lies in the meticulous selection of appropriate surrogate models for specific power system applications. This paper first reviews the current applications of surrogate models in the context of the new-type power system. Then it offers recommendations for selecting surrogate models based on the requirements of generalization ability

accuracy

and real-time performance in various operational scenarios. Subsequently

the paper explores potential improvements for sampling strategies

variable design

and parameter settings to enhance the performance of surrogate models for a practical power system engineering problem. Finally

a framework for power system optimization based on surrogate models is proposed. The paper also presents an outlook on the future applications and research directions of surrogate models in new-type power system. This study aims to provide strategic guidance for the application of surrogate model technology in the power system

fostering its efficient implementation and widespread adoption.