Abstract: To address the stability challenges posed by the high proportion of renewable energy grid connection, this paper proposes a power dispatch control strategy that integrates power forecasting and stability threshold identification. This strategy employs deep learning for probabilistic power forecasting of generation and load, and utilizes machine learning to identify system dynamic stability thresholds. Based on these, a multi-objective optimization dispatch model incorporating preventive stability constraints is constructed. Simulation results demonstrate that this strategy effectively enhances the system's operational resilience and stability under uncertain disturbances.