In this paper, we extend a previous algorithm mL-SHADE by running the evolutionary process through multiple populations and adding dynamic control of mutation intensity and hyper-parameters. The whole population is partitioned into subpopulations by a random clustering method. Mutation intensity and hyper-parameters are adjusted based on the consumption of fitness function evaluations. Performance of the proposed algorithm is verified by ten benchmark functions in the CEC2020 Competition on Single Objective Bound Constrained optimization. The results show the competitiveness of the proposed algorithm.