Frosting　is　a　critical　challenge　that　affects　the　energy　efficiency　of　air　source　heat　pumps　(ASHPs).　Defrosting　periodically　is　the　most　widely　used　method　to　solve　this　problem.　However,　the　rate　of　frosting　varies　considerably　for　ASHPs　due　to　their　configurations　and　operations　(CICO)　being　different.　This　leads　to　the　obvious　differences　in　their　defrosting　initiating　time.　Defrosting　too　early　or　too　late　will　both　cause　energy　loss.　To　ensure　the　efficient　operation　of　ASHPs,　the　variations　of　energy　loss　coefficient　caused　by　frosting-defrosting　(epsilon　NL)　and　the　defrosting　initiating　time　for　ASHPs　with　different　CICO　values　were　investigated.　Based　on　the　theory　of　optimal　defrosting　initiating　time　(Topt)　and　experimental　results　under　different　frosting　durations,　the　Topt　for　different　ASHPs　is　found.　Then,　a　prediction　model　for　ASHPs　with　different　CICO　values　is　established.　Results　show　that　there　is　a　Topt　for　ASHPs　with　different　CICO　values,　respectively.　Under　the　frosting　condition　of　2/1　degrees　C,　the　Topt　increases　with　the　rise　of　CICO　values.　When　the　CICO　values　increase　from　3.14　to　22.44,　the　corresponding　Topt　increases　from　22　min　to　148　min.　By　fitting　the　Topt　and　CICO　values,　a　cubic　mathematical　model　for　the　prediction　of　the　optimal　defrosting　time　of　ASHPs　has　been　developed.