This　paper　describes　a　non-linear　information　dynamics　model　for　integrated　risk　assessment　of　complex　disaster　system　from　an　evolution　perspective.　According　to　the　occurrence　and　evolution　of　natural　disaster　system　with　complicated　and　nonlinear　characteristics,　a　non-linear　information　dynamics　mode　is　introduced　based　on　the　maximum　flux　principle　during　modeling　process　to　study　the　integrated　risk　assessment　of　complex　disaster　system.　Based　on　the　non-equilibrium　statistical　mechanics　method,　a　stochastic　evolution　equation　of　this　system　is　established.　The　integrated　risk　assessment　of　complex　disaster　system　can　be　achieved　by　giving　reasonable　weights　of　each　evaluation　index　to　stabilize　the　system.　The　new　model　reveals　the　formation　pattern　of　risk　grade　and　the　dynamics　law　of　evolution.　Meanwhile,　a　method　is　developed　to　solve　the　dynamics　evolution　equations　of　complex　system　through　the　self-organization　feature　map　algorithm.　The　proposed　method　has　been　used　in　complex　disaster　integrated　risk　assessment　for　31　provinces,　cities　and　autonomous　regions　in　China　mainland.　The　results　have　indicated　that　the　model　is　objective　and　effective.