<p>In the electricity market, collusive behavior among power generation companies is one of the primary causes of market failure. The characteristics of collusion between market participants vary under different scenarios, and the lack of actual operational data further hinders effective collusion analysis. Therefore, this paper adopts a multi-agent simulation modeling approach to construct a collusion behavior simulation framework based on the multi-agent deep deterministic policy gradient algorithm. The competitive and collusive relationships between market participants are modeled by designing differentiated observation values and reward functions. Further, the pricing strategies of different market participants are simulated. Finally, based on simulation results, the collusive market characteristics are analyzed from two aspects: the alliance pricing strategies and market clearing conditions. The findings show that the market clearing price is close to the marginal price under a perfectly competitive model. In contrast, under collusive scenarios, the price spread between peak and off-peak times increases, with peak prices significantly rising. Furthermore, the bidding behaviors among colluding participants show high similarity, with a tendency to explore high prices within the market's price cap. These findings are significant for understanding and preventing collusive behavior in electricity markets and for maintaining fair competition in the market.</p>