The　accurate　control　of　oxygen　content　in　flue　gas　is　of　great　significance　to　the　stable　and　efficient　operation　of　the　municipal　solid　waste　incineration　plant.　However,　it　is　difficult　to　achieve　effective　control　performance　of　oxygen　content　in　flue　gas　due　to　the　inherent　nonlinearity　and　uncertainty　of　the　municipal　solid　waste　incineration　process.　Therefore,　a　data-driven　predictive　control　scheme　of　oxygen　content　in　flue　gas　is　proposed　for　municipal　solid　waste　incineration　process.　Firstly,　the　prediction　model　based　on　the　self-organizing　long　short-term　memory　(SOLSTM)　network　is　designed.　The　structure　of　the　hidden　layer　is　dynamically　adjusted　by　integrating　the　activity　and　significance　of　neurons,　and　then　the　prediction　accuracy　of　oxygen　content　in　flue　gas　is　improved.　Secondly,　the　gradient　descent　method　is　utilized　to　obtain　the　control　law,　and　the　optimization　efficiency　is　guaranteed.　Thirdly,　the　stability　of　the　proposed　control　scheme　is　analyzed　based　on　the　Lyapunov　theory.　Finally,　the　effectiveness　of　the　proposed　control　method　is　verified　based　on　the　industrial　data.　Compared　with　other　methods,　the　proposed　method　achieves　stable　and　efficient　control　performance　for　oxygen　content　in　flue　gas.　©　2024　South　China　University　of　Technology.　All　rights　reserved.