In order to solve the efficiency issues in distributed load responses to frequency regulation commands, an innovative strategy was introduced based on reinforcement learning for load aggregators' pricing incentives in response to frequency commands. Within this strategy, a game-theoretic model between the load aggregators and load clusters was constructed, and the load aggregators adjust incentive prices based on frequency commands and their pricing strategies, while loads adjust their power consumption based on their own electricity costs to flexibly respond to the frequency commands. The multi-agent soft actor-critic (MASAC) algorithm was used to investigate the solution. The results show that the pricing incentive method enables effective load response to frequency commands, and the use of the MASAC algorithm not only optimizes the decision-making process but also significantly reduces computational complexity, achieving efficient dynamic adjustment. It is concluded that this method provides an effective solution for frequency regulation in power systems, offering significant theoretical significance and practical value.