Ultraviolet–visible　(UV-Vis)　spectroscopy　is　widely　used　for　water　quality　parameter　analysis　due　to　its　advantages　of　real-time,　rapid　measurement,　simple　equipment,　and　no　secondary　pollution.　It　serves　as　a　critical　tool　for　online　water　quality　monitoring.　Water　quality　parameters　exhibit　unique　absorption　characteristics　in　the　UV-Vis　spectrum,　allowing　for　accurate　measurement　of　individual　parameters　using　spectral　analysis　and　regression　models.　However,　the　accuracy　of　traditional　linear　regression　models　decreases　when　detecting　multiple　parameters,　as　their　absorption　peaks　are　concentrated　in　the　190–320-nm　range.　This　study　employs　a　deep　learning　algorithm　based　on　convolutional　neural　networks　(CNNs)　to　perform　quantitative　analysis　on　mixed　spectra　of　multiple　parameters.　By　building　the　spectral　model　of　mixed　water　quality　parameters　and　training　the　network　with　appropriate　amount　of　spectral　data,　accurate　concentrations　of　multiple　water　quality　parameters　can　be　simultaneously　obtained.　This　research　provides　a　significant　theoretical　foundation　for　advancing　the　application　of　UV-Vis　spectroscopy　in　online　multiparameter　water　quality　monitoring.　©　1963-2012　IEEE.