Low throughput has long been a bottleneck of blockchain systems, and many studies have made efforts to address this problem. Most of these studies focus on optimizing different aspects of the blockchain system architecture to improve system performance, while overlooking the impact of block size configuration. By reviewing experimental results in existing literature, this study finds that simply adjusting the block size can increase node throughput by 2 to 8 times, suggesting that fully exploiting performance potential requires optimizing block size. However, existing research on block size optimization is overly simplistic, neglecting the potential security issues caused by block size adjustment and failing to consider the impact of heterogeneous hardware resources across nodes. The proposed methods and their conclusions are limited to experimental settings and cannot be directly applied in real-world production environments. To address this problem, this study proposes an adaptive block size optimization algorithm. By integrating the optimization algorithm into the blockchain consensus protocol, all nodes are enabled to dynamically adjust block sizes to achieve better performance. Experimental results demonstrate that the proposed algorithm can adjust block sizes to be close to the optimal value, resulting in a twofold increase in node throughput compared to the initial settings.