<p>Consensus algorithms are one of the core components of blockchain technology and play a crucial role in the security and performance of blockchain systems. The Practical Byzantine Fault Tolerance (PBFT) algorithm is a widely used consensus algorithm. In order to improve the security of PBFT and reduce the communication overhead, this paper proposes an improved PBFT consensus algorithm based on a reputation evaluation model. The algorithm combines static and dynamic assessments to derive a comprehensive reputation score and improves the primary node election process by constructing a new node reputation evaluation model. By dynamically monitoring primary-node behavior and restricting primary-node selection to high-reputation groups, the algorithm reduces the likelihood that Byzantine nodes are selected as primary nodes. By grouping nodes based on reputation, the algorithm effectively reduces message redundancy and communication complexity during the consensus process. Furthermore, the BLS aggregate signature algorithm is used to optimize the consensus process, addressing the high communication overhead as the node scale increases. Experimental results demonstrate that, in large-scale consortium blockchain scenarios, the proposed algorithm significantly enhances system throughput, effectively reduces latency and communication overhead, and outperforms comparable consensus algorithms.</p>