Efficient　and　accurate　structural　parameter　identification　is　critical　for　the　practical　application　of　structural　health　monitoring.　In　this　paper,　a　novel　algorithm　named　refracted　salp　swarm　algorithm　(RSSA)　is　proposed　and　applied　to　identify　structural　parameters.　Firstly,　the　basic　salp　swarm　algorithm　is　improved　by　refracted　opposition-based　learning　strategy,　multi-leader　mechanism　and　adaptive　conversion　parameter　strategy.　The　superiority　of　the　proposed　algorithm　is　verified　by　experiments　of　eight　benchmark　functions　of　various　types　and　dimensions.　Secondly,　a　new　type　of　structural　parameter　identification　(SPI)　model　is　established　by　combining　RSSA　and　the　Newmark　integration　method,　which　is　mainly　used　to　solve　the　optimization　problem　based　on　structural　acceleration,　thereby　identifying　structural　parameters　such　as　stiffness,　mass　and　damping　ratio.　Numerical　simulation　test　of　seven-floor　frame　proves　that　the　new　proposed　RSSA　could　be　successfully　applied　in　the　SPI　model.　Compared　with　other　heuristic　algorithms,　RSSA　can　obtain　more　accurate　recognition　results　under　the　circumstances　incomplete　measurement　data　and　low　signal-to-noise　ratio.