In　view　of　the　advantages　of　CFD　technology　in　the　simulation　of　small　and　medium-scale　chemical　hazard　diffusion,　one　near-field　prediction　model　of　chemical　hazard　diffusion　named　CHDNFP　is　constructed　based　on　CFD　technology,　and　the　accuracy　and　efficiency　of　the　model　are　improved　by　improved　differential　evolution　algorithm　(DEA)　and　fireworks　algorithm　(FWA).　Firstly,　based　on　the　component　conservation　equation,　momentum　conservation　equation,　and　turbulence　control　equation,　CHDNFP　model　is　constructed,　whose　basic　solution　process　is　proposed　from　three　aspects:　non-uniform　mesh　refinement　in　diffusion　space,　model　discretization,　and　iterative　solution　of　control　equation.　Secondly,　comprehensive　considering　the　global　search　ability,　local　search　ability,　and　convergence　characteristics　of　integrated　DEA　and　FWA,　a　hybrid　optimization　algorithm　IDEFWA　is　designed,　which　is　suitable　for　predictive　model　solving.　Finally,　the　CHDNFP　model　and　IDEFWA　are　verified　by　tracer　experiments.　The　result　shows　that:　IDEFWA　can　reduce　the　relative　root　mean　square　error　between　the　predicted　concentration　field　and　the　observed　concentration　field　to　about　25%,　with　the　calculation　accuracy　of　10–19　and　the　standard　deviation　accuracy　of　10–9;　compared　with　ABCA　and　GA,　IDEFWA　can　get　more　accurate　solutions　faster　under　the　same　algebraic　and　population　conditions;　the　calculation　accuracy　of　CHDNFP–IDEFWA　and　PISOFOAM　is　almost　the　same,　where　the　relative　difference　is　about　3%,　and　CHDNFP–IDEFWA　has　better　calculation　accuracy　than　PISOFOAM,　which　is　improved　by　about　26.05%.　©　2021,　The　Author(s),　under　exclusive　licence　to　Springer　Nature　Switzerland　AG.