In the wake of the swift progression of scientific and technological advancements, in order to cope with the explosive growth of computing power demand, the performance of computing power scheduling methods should be improved. This paper proposes a multi-objective optimization heterogeneous computing power scheduling method based on a multi-dimensional fine-grained computing power measurement model. A fine-grained computing power measurement index system is constructed through four computing power dimensions, and the comprehensive performance of nodes is quantified by the hierarchical analysis method and the multi-attribute comprehensive evaluation model (Technique for Order Preference by Similarity to Ideal Solution, TOPSIS); and the two-archive evolutionary algorithm for constrained multiobjective (C-TAEA) is used for optimization. Experiments show that when the number of tasks is 50, the execution time of the research method is only 130 s, the computing power stability is 92%, the task mobilization efficiency reaches 98%, and the computing power utilization rate is 98.5%; when the number of tasks is 300, the computing power stability of the research method is 90%, and the task mobilization efficiency and computing power utilization rate are both maintained above 95%. The results show that the research method provides a solution that takes into account both efficiency and stability for the dynamic scheduling of heterogeneous resources in high-performance computing scenarios, and has important theoretical and application value.