Abstract: Under the context of new power system construction, a large number of new energy sources and flexible loads are integrated into the distribution network, leading to a continuous differentiation in the structure and source-load characteristics of the distribution network. Consequently, maintaining the local power balance becomes increasingly difficult. To address this challenge, a hierarchical partition classification power balance system for distribution networks is proposed to enhance the partition balance capability of the distribution network through flexible resource collaborative optimization. Firstly, a balanced system combining the entire network and the partition balance of the distribution network is established, and the source-load uncertainty is modeled based on Copula theory and the improved K-means algorithm. Secondly, a bi-layer optimization model for regional power balance is constructed. Based on regional optimization the upper model optimizes the profit of regional operators by adjusting the time-of-use price, while the lower model enhances the local renewable energy consumption through the optimization of energy storage capacity. Therefore, the power balancing capability of distribution network regions is enhanced through the coordinated optimization of demand response and energy storage, thereby improving the power balance capacity of the entire network. Finally, the effectiveness of the proposed method is validated through the actual data from the distribution network.