In　order　to　effectively　use　the　cloud　data　of　connected　vehicle　to　estimate　the　battery　state　of　charge　(SOC),　an　estimation　method　based　on　noise　adaptive　particle　filter　(N-APF)　is　proposed　in　this　paper.　Firstly,　several　cells　are　connected　in　series　under　the　laboratory　environment　to　simulate　the　grouping　of　battery　packs　in　real　vehicle.　Besides,　the　federal　test　procedure　(FTP)　operating　current　for　battery　pack　is　obtained　through　software　simulation　combined　with　the　actual　vehicle　parameters.　Then,　the　Thevenin　equivalent　circuit　model　is　established　and　the　reliability　of　online　identification　of　model　parameters　based　on　10s　interval　data　is　verified.　Furthermore,　the　effectiveness　of　the　proposed　noise　adaptive　particle　filter　method　for　adjusting　the　process　noise　and　enhancing　the　stability　of　the　SOC　estimation　is　proved.　Finally,　the　reliability　of　the　improved　SOC　estimation　method　for　the　connected　vehicle　is　verified　based　on　the　10s　interval　cloud　data,　which　shows　the　proposed　noise　adaptive　particle　filter　estimation　method　can　stabilize　the　SOC　estimation　error　below　5%　except　for　some　high-current　discharge　phases.　©　2021　IEEE.