Markov　chain　technology　has　demonstrated　great　potential　in　predicting　the　transient　transport　of　pollutants　rapidly.　The　state　transfer　matrix　is　at　the　core　of　the　Markov　chain,　and　its　size　affects　the　computational　cost,　while　its　values　directly　impact　the　accuracy　of　the　calculations.　In　order　to　reduce　the　computational　cost　of　the　Markov　chain　model　while　ensuring　acceptable　calculation　accuracy,　we　propose　an　improved　method　based　on　velocity-based　non-uniform　state　division　for　the　combined　model　of　CFD　and　Markov　chain　technology.　We　compared　the　predictive　results　of　the　improved　model　with　experimental　measurement　data　and　CFD　simulation　data,　and　the　results　showed　that　the　model＇s　predictive　performance　for　particle　distribution　concentration　is　good.　Moreover,　the　prediction　accuracy　is　higher　when　using　the　non-uniform　state　division　method.　©　2023　IBPSA.All　rights　reserved.