How to design a settlement mechanism adapted to the electricity market according to local conditions is a key theoretical and practical problem to be solved in the construction of the electricity market. In this paper, considering the competitive nature of the power market environment, a strategy based on the combination of deep reinforcement learning approach and game theory is proposed to analyze the complex power market competition in an equilibrium manner. The SFE model is used to model the bidding process of power producers and construct a power market clearing model, and the model is solved iteratively using the MADDPG method, on the basis of which a pervasive clearing mechanism for the power market is designed. Based on the IEEE33-node distribution network system, we carry out an arithmetic case analysis, and the results show that compared with constant/time-sharing tariffs, the game bargaining model not only improves the profits of the various subjects of interest participating in the market transaction under the premise of guaranteeing power supply security and power supply quality, but also mobilizes the demand-side resources to participate in the market, reduces the operating costs of the distribution network, significantly improves the equilibrium of benefit distribution, and reduces the new energy consumption of distribution network risk of new energy consumption in the distribution network. In addition, the generator’s offer strategy learned under the MADDPG algorithm has smaller variance, more stable results, and obtains a higher total return of the round, reaching a mean value of $2206.52. This validates the effectiveness of the MADDPG deep reinforcement learning algorithm proposed in this paper.