Aggregating large-scale electric vehicles (EVs) can enable their effective participation in grid interactions. However

the existing inner and outer approximation methods for the aggregate feasible region (AFR) struggle to strike a balance between accuracy and computational efficiency. To address the highly random nature of charging periods of EVs

we propose an inner-outer combined approximation method for the AFR of EVs. First

we introduce a maximum inner approximation method based on dimension trimming

utilizing charging periods

to enhance the conservativeness of the inner approximation. Then

we perform outer approximation at the cluster level to reduce the infeasible region of the aggregation result while maintaining the AFR as a virtual battery model. To quantitatively evaluate the effects of different AFR approximation methods

we propose indicators for feasibility and conservativeness

along with an improved Hit-and-Run-based sampling method for EV feasible regions to calculate the evaluation metrics. Finally

based on actual charging data

we design case studies to compare typical AFR approximation methods in terms of feasibility

conservativeness

and computational efficiency

thus validating the effectiveness of the proposed method.