Maximum supply capacity calculation is an important issue in grid planning, and with the large amount of renewable energy sources connected to the grid, the voltage instability problem becomes more and more prominent. In this paper, the maximum power supply capacity (TSC) of partitioned flexible interconnected grids under multi-temporal and spatial scales is dynamically modeled, and the TSC model is solved by using the deep deterministic policy gradient (DDPG) method to achieve quantitative assessment of the TSC of the grid. Meanwhile, the effectiveness of the model and algorithm is verified through simulation experiments. The stepwise approximation method and DDPG algorithm without considering the transient voltage stabilization constraints and the obtained TSC in which the node voltages are less than 0.80 p.u. are all greater than 1.0 s, and the transient voltages are destabilized. While the DDPG algorithm considering transient voltage stabilization, the obtained node voltage is greater than 0.80p.u., and the transient voltage is in a stable state, which indicates that the algorithm can effectively reduce the risk of transient voltage instability in the power grid. The sum of the TSCs of A and B divisions after the zonal flexible interconnection is 9348 MW, which is higher than the sum of the TSCs of 8696 MW during the zonal open-loop operation, indicating that the zonal flexible interconnection can improve the overall TSC level of the power grid. In addition, compared with the traditional algorithm and other reinforcement learning algorithms, the TSC calculation based on the DDPG algorithm is more efficient and accurate. This paper provides methodological guidance for evaluating the power supply capacity of power grids at multi-temporal and spatial scales.