In　this　study,　a　battery　test　platform　is　built　to　study　and　analyze　the　battery　model.　A　fractional-order　model　(FOM)　of　battery　is　developed　by　combining　the　second-order　equivalent　circuit　model　with　fractional　calculus　theory.　By　combining　the　multi-innovation　with　the　unscented　Kalman　filter　algorithm　based　on　FOM,　a　fractional-order　multi　innovation　unscented　Kalman　filter　algorithm　is　proposed.　Under　different　temperatures　and　dynamic　conditions,　the　tracking　of　terminal　voltage　and　the　estimation　results　of　SOC　using　the　proposed　algorithm　are　compared　and　analyzed.　In　order　to　verify　the　accuracy　and　robustness　of　the　proposed　algorithm,　the　tracking　effect　of　the　terminal　voltage　and　the　estimation　results　of　SOC　under　different　initial　SOC　values　are　investigated.　The　experimental　results　indicate　that　the　FOM　is　more　accurate　than　the　integer-order　model　in　reflecting　the　battery　characteristics.　Under　different　temperatures,　dynamic　conditions,　and　initial　SOC　values,　the　values　of　mean　absolute　error　(MAE)　and　root　mean　squared　error　(RMSE)　of　the　proposed　algorithm　are　the　lowest,　which　verifies　the　effectiveness　of　the　proposed　algorithm.　The　maximum　values　of　MAE　and　RMSE　of　SOC　estimation　results　of　the　proposed　algorithm　are　1.04　%　and　0.9122　%,　respectively,　indicating　that　the　proposed　algorithm　has　high　accuracy　and　good　robustness.