With the large-scale integration of distributed energy resources (DERs) and the gradual introduction of carbon trading mechanisms, low-carbon scheduling of distribution net-works faces challenges in terms of model scale expansion and high computational complexity. To address these issues, this paper proposes a low-carbon scheduling method for distribution networks considering the aggregated flexibility of DERs. First, a low-carbon scheduling model is established, incorporating dis-tributed generators, photovoltaic units, and DER clusters, where a stepped carbon pricing mechanism is introduced to characterize carbon trading costs. Then, to tackle the difficulty in accurately characterizing the aggregated feasible region (AFR) of DERs, an improved inner approximation method is proposed. By con-structing a basis set based on an energy variation model and integrating a boundary contraction strategy, a high-accuracy inner approximation of the AFR is achieved. Based on this, the con-structed AFR is embedded into the scheduling model, signifi-cantly reducing the model size while maintaining scheduling accuracy. Numerical results demonstrate that the proposed method outperforms existing inner approximation approaches in terms of total cost, approximation accuracy, and flexibility utili-zation of DERs, while achieving a close approximation to the scheduling results based on the complete DERs cluster mode. Moreover, it exhibits superior computational efficiency and scalability under different system scales.