Abstract: Super-luminous supernovae have been studied intensively for nearly two decades． The high luminosity makes them of great significance for studying first-generation stars and early universe． However, the underlying energy mechanism remains unclear．In this study, 91 hydrogen-poor super-luminous supernovae with well-sampled light curves were examined to extract their characteristics for statistical analysis．Markov Chain Monte Carlo simulations were carried out to reproduce these observables considering three power mechanisms．The constraints combining observations on their energy mechanism were systematically explored．The average peak absolute magnitude at g band was found to be (−21.34 ± 0.8) mag, the rising and declining time scales were distributed in 7~79 and 14~114 d, where a strong correlation existed．Simulations indicated that both the magnetar and the interaction mechanism power．Under the interaction framework, the shell-like scenario is preferred based on comparison with observations．In the future, domestic telescopes and all-sky surveys (such as CSST and SiTian) will detect more super-luminous supernovae with high red shift, providing important information for a more comprehensive understanding of massive star evolution and explosions．