ABSTRACT The popularity of mobile smart terminals has driven edge computing to become a research hotspot, and blockchain technology provides a new way to address data security 15 in edge computing; however, traditional consensus algorithms are difficult to adapt to IoT environments with limited storage and computational resources. For this reason, this paper proposes a Practical Byzantine Fault Tolerance (PBFT) consensus optimisation algorithm based on dynamic reputation and adaptive clustering (DRAC‐PBFT). First, a hierarchical regulation model based on dynamic reputation is proposed, where nodes are divided into upper‐layer and lower‐layer node sets according to their reputation values, which are responsible for consensus decision‐making and data validation and forwarding, respectively. A reputation assessment mechanism based on node roles and behaviours is proposed to dynamically adjust the node tiers and maintain incentives. The master node selection is randomly generated in the upper‐layer nodes using SHA‐256 hash to ensure security. Second, the weight factor and the number of clusters K are dynamically adjusted by comprehensively considering the node reputation and load condition, and the geographic coordinates and Euclidean distance of the edge gateway nodes are used for efficient clustering to optimise the data transmission path and reduce the latency. Finally, a two‐layer consensus mechanism with separation of duties is designed, where the lower‐layer nodes pre‐verify the transactions within the cluster and submit only qualified data to the cluster head to reduce redundant communication. Inter‐cluster interactions are coordinated by the upper consensus layer to ensure consistency and optimize data synchronisation. Experimental results show that DRAC‐PBFT improves the average throughput by 67% and reduces the average delay by 50% compared to PBFT. Compared with other PBFT improvement algorithms, DRAC‐PBFT shows better performance in large‐scale consortium chain environments, and is suitable for IoT or edge computing environments with limited computing resources and stringent requirements on latency and throughput.