Abstract: Existing scheduling methods for arch dam cable crane groups primarily focus on concrete placement tasks, with coarse decision granularity and separated decision-making for task allocation and execution sequence, which lead to low efficiency and potential risks in construction safety and concrete quality. To address these issues, this paper proposes a refined scheduling optimization method that further considers auxiliary lifting tasks beyond concrete placement tasks and operates at the level of individual lifting cycles. A single-cycle task micro-element is defined to represent the two types of lifting tasks uniformly. Considering the complex temporal-spatial constraints of parallel cable crane operation and the sequential decision-making characteristics of asynchronous multi-machine operation, a scheduling model based on Markov Decision Process is established to construct an integrated decision-making framework. A root-parallel Monte Carlo Tree Search algorithm, enhanced with constraint pruning and a standardized reward function, is adopted to efficiently search for optimal scheduling schemes. Verified through trial tests on the arch dam of the under-construction Yebatan Hydropower Station in Southwest China, the proposed method reduces the total makespan by 14.2% compared with manual scheduling while satisfying all construction constraints, and significantly improves concrete placement intensity and crane utilization. Moreover, it generates detailed and optimal scheduling schemes under varying numbers of cranes, task workloads, and auxiliary task execution modes, providing theoretical support for refined resource scheduling in intelligent arch dam construction.