To　improve　the　generalization　and　robustness　of　the　stochastic　configuration　network　(SCN)　model,　an　improved　SCN　algorithm　is　proposed.　Attenuated　L2　regularization　is　added　to　the　iterative　solution　process　of　the　output　weight,　which　not　only　reduces　the　output　weight　of　the　hidden　layer　nodes　in　the　early　iteration,　but　also　ensures　that　the　new　nodes　after　multiple　iterations　are　not　affected　by　the　larger　regularization　coefficient,　forming　the　constraint　effect　of　first　tight　and　then　loose.　In　addition,　the　multi-kernel　learning　method　is　introduced　to　obtain　the　characteristics　of　the　sample　data,　and　the　penalty　weight　matrix　of　the　training　samples　is　determined,　so　as　to　re-evaluate　the　output　weight　of　the　SCN　model　to　enhance　the　robustness　of　the　model.　The　performance　of　the　proposed　method　was　tested　on　four　standard　datasets　and　historical　data　from　municipal　solid　waste　incineration　processes.　The　experimental　results　show　that　the　modeling　method　proposed　in　this　paper　has　certain　advantages　in　improving　the　model＇s　generalization　and　robustness.　©　2024　IEEE.