With　the　development　of　information　technology　and　the　digital　transformation　of　the　power　system　and　chemical　industry,　such　industrial　control　systems　are　becoming　increasingly　vulnerable　to　network　attacks,　and　the　deployment　of　traditional　methods　has　shown　shortcomings,　especially　in　terms　of　resilience　and　adaptation　to　future　diversification　trends,　so　the　deployment　of　highly　efficient　and　accurate　intrusion　detection　technology　is　crucial　in　the　power　system.　In　this　paper,　an　intrusion　detection　model　scheme　based　on　machine　learning　optimization　algorithms　is　proposed　for　the　data　collected　by　SCADA　networks　and　network　communication　traffic　in　industrial　control　power　systems.　The　intrusion　detection　model　scheme　for　SCADA　network　mainly　includes　three　modules:　data　set　construction,　machine　learning　intrusion　detection　algorithm　and　algorithm　optimization:　after　the　process　of　data　collection　and　pre-processing,　OCSVM　is　adopted　as　the　main　intrusion　detection　algorithm,　which　is　suitable　for　the　anomaly　detection　task,　whereas　in　the　algorithm　optimization,　PSO　is　used　to　adjust　the　parameters　of　OCSVM,　which　is　used　to　improve　the　detection　performance　of　the　model.　The　final　experimental　data　shows　that　the　OCSVM-PSO　model　of　this　scheme　has　stronger　potential　and　advantages　in　the　field　of　power　system　intrusion　detection　compared　to　the　OCSVM　model.　©　2024　IEEE.