Multi-point access of flexible loads and renewable distributed energy sources can lead to frequent changes in the magnitude and direction of distribution network flows. On the distribution side, prosumers often engage in peer-to-peer (P2P) energy trading with the sole goal of maximizing their own benefits, neglecting the impact of energy flows from different devices on the safe operation of the system. This makes transactions results easily lead to network congestion. This paper addresses the network congestion caused by the decentralized P2P trading model using continuous double auction mechanism and zero-intelligence enhanced bidding strategies, taking into account prosumer data privacy management and consumer psychology. A two-tier market dynamic coordination congestion management method is proposed. This aims to maximize the overall benefits of prosumers and minimize the adjustment of network congestion, using the Distflow model to construct a node flexibility ancillary service price calculation model based on node marginal electricity prices. Subsequently, the congestion cost is obtained through distributed optimization algorithms, and the cost is allocated to responsible prosumers based on the sequence of P2P transactions and the contribution of line congestion. This achieves elimination of congestion. A case study using an improved 11-bus distribution system based on the IEEE 33-bus system demonstrates that the proposed method can effectively address network congestion over specified time scales and computational accuracy conditions, reducing congestion costs and losses in decentralized P2P energy trading benefits.