The Practical Byzantine Fault Tolerance consensus algorithm provides the blockchain with a reliable mechanism for reaching consensus in the event of node failures and malicious behavior, ensuring that malicious nodes are unable to manipulate the consensus outcome without exceeding the algorithm's fault tolerance thresholds. However, PBFT suffers from easy prediction of master nodes, insufficient node differentiation, and limited performance in large-scale networks. To solve these problems, this thesis proposes an Improved Practical Byzantine Fault Tolerance Consensus Algorithm Based on Scored and Dynamic Tenure. The algorithm limits the number of nodes directly involved in consensus by distinguishing between consensus nodes and ordinary nodes, evaluates and quantifies the consensus behavior of nodes through a point mechanism, and uses a point-based election mechanism to select consensus nodes and block-out nodes. In addition, the algorithm introduces a dynamic tenure mechanism to balance the impact of different nodes on the system efficiency. Comparison experiments of SDTPBFT with PBFT and RPBFT show that the improved algorithm improves the throughput by 63% and reduces the average delay of communication by 83%, which significantly improves performance and security.