Metaheuristic　algorithms　are　extensively　employed　to　solve　high-dimensional　optimization　problems,　with　particle　swarm　optimization　(PSO)　garnering　considerable　attention　for　its　computational　efficiency　and　simplicity.　In　tackling　time-consuming　and　complex　engineering　optimization　tasks,　PSO　typically　utilizes　cluster　computing　techniques　to　aggregate　substantial　computing　resources,　thereby　accelerating　the　optimization　process.　However,　this　approach　may　face　computational　interruptions　due　to　power　failures,　program　crashes,　or　network　instability,　thereby　impeding　the　optimization　process.　Moreover,　the　dynamic　nature　of　cluster　computing　resources　necessitates　efficient　resource　utilization　methods,　such　as　adaptive　population　size　adjustment.　In　this　study,　we　propose　a　recoverable　PSO　to　address　interruptions　during　prolonged　optimization　processes.　Building　upon　this,　we　further　develop　an　enhanced　PSO　with　adaptive　swarm　size　reduction.　The　study　begins　by　reviewing　and　categorizing　existing　population　size　reduction　strategies　and　introducing　several　novel　approaches.　The　effectiveness　of　these　strategies　is　evaluated　using　the　CEC　benchmark　test　suite,　comparing　their　convergence　speed　and　accuracy.　Furthermore,　the　optimal　strategy　is　validated　through　three　real-world　engineering　optimization　problems　under　constrained　computing　resources.　The　results　demonstrate　that　the　proposed　method　significantly　enhances　PSO　performance,　offering　valuable　insights　for　future　research　on　population　size　control　in　PSO　and　its　engineering　applications.