Blockchain, a revolutionary technology, is widely used in cryptocurrencies to record transactions in chronological order. With its rapid development and decentralization, blockchain has expanded to applications in smart factories, edge computing, supply chains, and smart cities. Consensus algorithms are crucial for achieving agreement on transaction results among participants. However, as network complexity and user numbers grow, current popular consensus algorithms face challenges like poor scalability, unfair node election, low message transmission efficiency, high communication overhead, and lack of dynamic support. To enhance consensus efficiency and security, we propose AD-BFT (Adaptive Dynamic Byzantine Fault Tolerance), a dynamic algorithm that improves fault tolerance by using partial node participation under normal conditions and switches to full participation when needed. A Markov chain-based committee switching strategy optimizes network performance, ensuring scalability and high throughput in dynamic environments.