The　control　system,　as　an　important　part　in　biological　wastewater　treatment　system　(BWTS),　is　employed　to　meet　the　operational　goals　for　reaching　required　effluent　quality;　however,　the　control　performance　will　be　degraded　under　drastic　uncertainties　and　various　conditions.　In　this　paper,　a　self-learning　sliding　mode　controller　(SLSMC)　is　proposed　for　BWTS　without　the　knowledge　of　uncertainties.　First,　a　mathematical　kernel　function　(MKF)　is　established　to　estimate　the　bounds　of　uncertainties,　which　is　used　to　pursue　the　optimized　control　law　of　SLSMC.　Second,　a　self-learning　optimization　algorithm　is　designed　to　modify　the　parameters　of　MKF,　which　ensure　that　there　are　no　overestimation　parameters　of　SLSMC.　Third,　a　gain　adaptation　mechanism,　based　on　MKF　and　conventional　conditions　of　BWTS,　is　developed　to　suppress　the　chattering　and　maintain　control　accuracy　simultaneously.　Finally,　to　show　the　effectiveness　of　SLSMC,　it　is　applied　to　BWTS　under　uncertainties　and　different　conditions　in　comparison　with　other　existing　methods.　The　results　demonstrate　that　SLSMC　performs　favorably　in　terms　of　both　chattering　reduction　and　control　accuracy.　©　2019　IEEE.