Supercritical CO2 power cycle, which has gained a tremendous interest in recent year owing to its compactness and enhanced efficiency, was investigated in this study. In the present study, leakage characteristics of the 500℃- class sub-miniature supercritical carbon dioxide turbine with Labyrinth seals were investigated for CSP (Concentrated Solar Power) generation. In contrast to previous studies, a new leakage management scheme using barrier gas was proposed and simulated. The modelling of the labyrinth seals was validated against the published results. Subsequently, multivariate optimisations based on the genetic algorithm were carried out to maximise the improvement of the leakage characteristics. Results show that the power demand for the recovery of leakage by the proposed scheme can be reduced by up to 15% compared to the power demand according to the conventional scheme under the simulation conditions considered in the present study.