A　bionic　operant　conditioning　reflex(OCR)　learning　control　scheme　is　proposed　based　on　the　thought　of　sliding　model　control(SMC)　and　Elman　network　for　a　class　of　SISO　higher-order　nonlinear　control　system.　In　this　method,　an　Elman　network　is　used　as　an　evaluation　function　of　sliding　surface　and　action　in　the　scheme.　Reward　signal　is　designed　according　to　the　change　of　sliding　surface,　and　then　the　evaluation　function　is　updated　through　the　reward　stimulation,　while　the　behavior　choice　probability　is　changed.　Through　the　definition　of　entropy　for　each　round,　a　quantitative　analysis　about　the　knowledge　change　in　the　learning　process　is　given.　The　results　of　the　simulation　experiment　in　the　walking　inverted　pendulum　system　show　that,　bionic　OCR　learning　control　is　used,　which　realizes　the　balancing　control　for　the　walking　inverted　pendulum　system.