In view of a solution of the minimizing scheduling length under energy constraints in heterogeneous cloud systems, a novel budget level (BL) energy preallocation strategy has thus been proposed, with a scheduling length minimization algorithm designed under energy constraints (BLMSL). The BLMSL algorithm consists of three stages: establishment of task priority queues, preallocation of task energy consumption constraints, and selection of the optimal processor and frequency combination. The results of experiments conducted on two types of scientific workflows, Epigenomics and LIGO, show that under the premise of meeting energy consumption constraints, the BLMSL algorithm is characterized with significant advantages in achieving a smaller scheduling length compared to the most advanced heuristic algorithms.