To　achieve　real-time　control　in　random　driving　cycles　and　prolong　battery　life　for　plug-in　hybrid　electric　vehicles　(PHEV),　a　battery　aging-aware　energy　management　strategy　with　dual-state　feedback　control　is　proposed　based　on　multiple　neural　networks　(Multi-NN)　learning　algorithms　to　minimize　the　life　cycle　cost　(LCC).　First,　the　offline　optimal　control　results　prepared　for　real-time　strategy　knowledge　learning　are　obtained　by　an　improved　Pontryagin＇s　minimum　principle　(PMP).　Second,　the　determination　coefficient　is　introduced　to　determine　the　train　data　for　neural　network　learning.　Besides,　a　k-means　algorithm　is　used　to　cluster　the　offline　optimal　control　sequences　into　three　sub-data　clusters,　which　are　used　as　the　training　data　of　the　sub-neural　networks.　Then,　an　online　driving　pattern　recognition　method　based　on　generalized　regression　neural　network　is　trained　to　select　the　corresponding　sub-neural　network.　Finally,　dual-state　feedback　control　is　applied　to　the　energy　management　strategy　by　introducing　the　reference　SOC　and　reference　effective　Ah-throughput.　The　simulation　validations　show　that　the　LCC　of　the　proposed　strategy　is　similar　to　the　LCC　of　PMP　considering　battery　aging　under　three　random　driving　cycles,　and　the　LCC　is　reduced　by　20.97,　22.25　and　22.28%　compared　with　CD-CS.　©　2021　Elsevier　Ltd