This　study,　which　focuses　on　the　strong　coupling　relationship　among　the　parameters　of　the　organic　Rankine　cycle　system,　proposed　a　key　parameter　set　identification　method　based　on　a　machine　learning　model　established　on　the　basis　of　experiments.　First,　system　performance　tests　under　different　heat　source　conditions　were　performed　on　a　set　of　10　kW　organic　Rankine　cycle(ORC)experimental　systems.　Then,　the　six　system　parameters　of　pressure　and　temperature　at　the　evaporator　outlet,　temperature　and　pressure　at　the　condenser　inlet,　working　fluid　pump　efficiency,　and　expander　shaft　efficiency　were　selected　as　the　initial　variables,　and　the　thermal　efficiency　of　the　ORC　system　was　selected　as　the　target　variable.　Multiple　linear,　artificial　neural　network,　and　support　vector　machine　learning　models　were　established　for　the　ORC　system.　Finally,　the　optimal　machine　learning　model　and　key　parameter　set　were　determined.　Results　showed　that　the　use　of　the　key　parameter　set　can　reduce　the　average　error　of　the　models　by　13.　36%　and　improve　the　accuracy　of　the　models.　A　highly　accurate　machine　learning　model　of　ORC　systems　can　improve　predictive　performance,　thereby　providing　support　for　the　high-efficiency　operation　and　control　of　ORC　systems.　©　2023　Editorial　Board　of　Journal　of　Harbin　Engineering.　All　rights　reserved.