The invention discloses an
intelligent decision-making method based on static game and
reinforcement learning. The method comprises the following steps: S1, constructing a static game model: designing a formalized revenue function; s2, solving static Nash equilibrium: carrying out multi-round strategy
simulation by adopting a virtual chess playing
algorithm, and approaching a Nash equilibrium strategy combination; s3, introducing
reinforcement learning: dynamically optimizing the strategy by adopting a near-end strategy optimization
algorithm; s4, dynamically correcting the revenue function according to environment feedback; and S5, re-calculating an
equilibrium solution of the updated revenue function, and feeding back the
equilibrium solution to the PPO training process to form a'modeling-solving-learning-feedback-re-modeling 'closed-
loop optimization architecture. The invention further discloses an
intelligent decision-making
system applied to the whole production and management chain of the tobacco enterprise. The
intelligent decision-making method based on deep fusion of static game and
reinforcement learning provided by the invention overcomes the technical problems of insufficient strategy modeling, unstable training, insufficient balance, weak adaptability and the like of the existing intelligent decision-making technology in a multi-party confrontation and dynamic evolution environment.