The　degree　of　polymerization　(DP)　of　insulating　paper　is　a　key　factor　to　measure　the　performance　of　power　transformer.　Near　infrared　spectroscopy　(NIRS)　can　test　the　degree　of　polymerization　of　insulating　paper　quickly　and　nondestructively,　however,　the　poor　generalization　ability　of　prediction　algorithm　limits　its　application.　In　this　paper,　we　carry　out　accelerated　aging　test　on　four　kinds　of　insulating　paper,　which　are　typical　insulating　paper　of　transformer　in　Hainan　power　grid.　The　insulating　paper　with　different　degrees　of　aging　were　detected　by　the　NIRS　and　the　DP　with　viscosity　method.　The　data　of　NIRS　and　DP　were　used　to　establish　the　quantitative　mathematical　relationship　applied　with　Back　Propagation　Neural　Networks　(BPNN)　model.　The　results　show　that　the　BPNN　model　has　a　high　accuracy,　which　is　more　suitable　in　Hainan　power　grid.　©　2021　IEEE.