The　power-voltage　characteristic　curves　of　a　series-connected　photovoltaic　(PV)　system　exhibit　multiple　peaks　in　partial　shading　scenario　(PSS).　However,　there　lacks　a　feasible　model　that　obtains　the　capability　of　predicting　the　global　maximum　power　point　(GMPP)　locus.　This　paper　analyzes　the　GMPP　characterization　and　demonstrates　that　the　voltage　at　the　GMPP　is　located　over　a　large　range　in　various　PSS.　With　the　aim　of　improving　the　prediction　performance,　multiple　Gaussian　process　regression　(MGPR)　models　are　proposed　to　predict　the　GMPP　locus　by　using　shading　information,　such　as　the　shading　rate　and　shading　strength.　The　performance　of　the　method　is　evaluated　using　a　dataset　obtained　in　a　variety　of　environmental　conditions.　The　results　show　that　it　outperforms　the　existing　prediction　models　in　terms　of　accuracy.　With　its　high　prediction　capability,　the　proposed　method　can　be　used　to　increase　the　maximum　power　point　tracking　performance　in　PV　systems.　©　2019　IEEE.