A key task in cancer genomics research is the identification of cancer driver genes. In recent years, control methods have been applied to identify cancer drivers with notable success. However, these methods primarily focus on exploring the biological network topology, potentially overlooking the critical influence of node and edge weights derived from genomic data. Moreover, many methods have overly simplistic evaluation metrics and do not perform well on large networks. In our previous work, we have proposed a novel control theory-based method, along with a comprehensive evaluation metric to handle those problems. Building upon this foundation, this work further proposes an enhanced approach that incorporates a local optimal control set selection method based on the greedy algorithm to refine the identification of key drivers. Numerical experiments have been conducted to identify both coding and miRNA drivers, as well as drivers across different cancer subtypes. The experimental results demonstrate the effectiveness and robustness of the weighted network control method, highlighting its superior performance in detecting cancer driver genes.