Abstract: Due to the lack of partition processing for DC microgrids, the stability margin of DC microgrids is low. Therefore, a high-precision matrix decomposition algorithm based regional stability control method for DC microgrids is proposed. Using correlation matrix to describe the topology structure of DC microgrid DC network, using high-precision matrix factorization algorithm to decompose the voltage increment matrix and sensitivity matrix, achieving DC microgrid partitioning. Taking load optimization as the control objective, a regional stability control objective function for DC microgrids is constructed. Combined with reinforcement learning algorithms, the objective function is solved, and the optimal control strategy is obtained by inputting state and action strategies. The experimental results show that after applying the proposed method for stable control of DC microgrids, the stability margin of DC microgrids is higher, and they have relatively ideal control performance.