Solar　spectral　irradiance　has　a　crucial　impact　on　building　energy　conservation,　especially　on　photovoltaic　(PV)　generation.　However,　it　takes　a　high　cost　to　measure　and　predict　the　dynamic　solar　spectral　irradiance　for　various　atmosphere　conditions　and　sun　positions.　Combining　with　machine　learning,　this　paper　developed　a　novel　solar　spectral　irradiance　estimation　model　to　evaluate　the　annual　solar　spectral　property　in　a　region.　This　paper　employs　the　readily　accessible　subaerial　meteorology　as　model　input.　The　average　photon　energy　(APE)　serves　as　a　connection　between　the　normalized　solar　spectral　irradiance　and　the　meteorology　parameters.　Verification　showed　the　model　this　paper　proposed　estimated　the　normalized　solar　spectral　irradiance　well.　Further,　annual　simulation　of　solar　spectral　irradiance　was　conducted　by　inputting　typical　meteorology　year　(TMY)　dataset.　The　annual　difference　of　the　normalized　spectral　irradiance　reached　to　10.57　%,　which　reflects　the　great　importance　to　determine　the　practical　solar　spectral　irradiance.　A　typical　day　of　spectra　was　proposed　for　each　month　to　reveal　the　monthly　variation　in　solar　spectral　irradiance.　This　study　provides　a　convenient　technical　method　to　evaluate　the　solar　spectral　property　for　engineering　applications.　The　results　may　guide　industries　in　selecting　suitable　solar　cells　for　the　region,　thereby　prompting　the　development　of　solar　applications.