A　key　factor　of　the　electric　vehicle　power　battery　management　is　the　accurate　estimation　of　SOC(state　of　charge).　Considering　the　measurement　error　and　noise　of　the　battery　data　acquisition,　both　the　accuracy　and　the　convergence　speed　of　the　SOC　estimation　are　limited.　In　this　paper,　a　modified　estimation　method　for　lithium　battery　SOC　has　been　proposed,　which　is　an　online　identification　method　based　on　BCRLS　(bias　compensation　recursive　least　square)　and　combined　with　the　modified　AFEKF　(adaptive　fading　extended　Kalman　filter).　A　set　of　HPPC　(hybrid　pulse　characteristic)　containing　noise　data　has　been　tested　separately　by　EKF(extended　Kalman　filter),　FEKF(fading　extended　Kalman　filter)　and　AFEKF　algorithms.　In　terms　of　anti-interference　and　convergence　speed,　the　modified　AFEKF　algorithm　combined　with　BCRLS　has　obvious　advantages.　At　the　same　time,　the　estimation　of　SOC　absolute　error　of　the　proposed　algorithm　can　be　kept　within　2.36%.　©　2022　Technical　Committee　on　Control　Theory,　Chinese　Association　of　Automation.