Accurate　estimation　of　battery　state　of　health　(SOH)　parameters　is　crucial　for　enhancing　the　reliability　of　battery　management　systems.　It　depends　on　the　quality　of　health　features　(HFs)　extraction　and　the　precision　of　the　estimation　algorithms.　Typically,　the　process　of　extracting　battery　HFs　is　cumbersome　and　heavily　relies　on　expert　knowledge,　with　few　studies　focused　on　improving　estimation　methods.　To　solve　this　problem,　a　novel　high-accuracy　intelligent　estimation　method　for　battery　SOH　has　been　proposed　in　this　work.　This　method　includes　the　fractional-order　refined　time-shift　multiscale　fuzzy　entropy　feature　extraction　(FRTSMFE)　algorithm,　and　the　least　square　support　vector　machine　(LSSVM)　estimation　algorithm,　improved　by　the　energy　valley　optimization　(EVO)　algorithm.　By　combining　the　fractional-order　derivatives,　the　refined　time-shift　multiscale　method,　and　fuzzy　entropy　theory,　the　FRTSMFE　algorithm　was　derived　to　extract　HFs　from　voltage　and　current　data　during　charging.　The　EVO　algorithm　was　used　to　optimize　the　hyperparameters　of　the　LSSVM　algorithm,　enhancing　its　estimation　accuracy.　The　EVO-LSSVM　algorithm　was　then　utilized　to　establish　the　correlation　between　the　HFs　and　SOH　of　each　battery.　Finally,　comparative　and　ablation　experiments　were　conducted　on　the　NASA,　CALCE　and　FB　datasets　to　validate　the　effectiveness　and　accuracy　of　the　proposed　high-accuracy　intelligent　estimation　method　for　battery　SOH.　Additionally,　the　robustness　of　the　proposed　algorithm　was　validated　using　the　MIT,　XJTU,　and　TJU　datasets.