The　steady　state　model　of　the　heat　pipe　heat-exchange　system　is　built　using　the　mechanism　analysis　approach.　For　the　unknown　key　parameter　estimation　in　modeling,　the　particle　swarm　optimization　algorithm　and　recursion　algorithm　with　restricted　memory　are　used　based　on　the　actual　running　data.　Because　of　the　change　of　model　parameter,　a　RBF　neural　network　is　used　to　build　a　quantitative　relationship　between　input　variable　and　model　parameter,　and　the　validity　of　this　method　is　verified　using　field　data.　Finally,　a　precise　steady-state　model　is　obtained.　It　lays　a　solid　foundation　for　further　system　analysis　and　real　time　optimization.