Ni　particle　coarsening　is　an　important　factor　in　deteriorating　the　durability　of　solid　oxide　fuel　cell　(SOFC)　operations.　In　order　to　investigate　the　influence　of　Ni　coarsening　on　SOFC　performance,　the　transient　multi-physical　field　model　of　SOFC　was　developed　in　this　paper.　The　high　operating　temperature　accelerates　Ni　particle　growth　and　increases　the　attenuation　rate　of　SOFC　current　density　from　0.23%/kh　at　650　°C　to　2.6%/kh　at　800　°C.　The　increase　in　the　ratio　of　steam　to　carbon　also　intensifies　the　Ni　particle　coarsening　process　and　deteriorates　the　transient　performance　of　SOFC.　Increasing　YSZ　particle　diameter　could　hinder　the　growth　of　Ni　particles　and　slowing　down　the　increase　rate　of　Ni　particle　diameter.　Within　the　range　of　preset　YSZ　diameter　dYSZ,　increasing　dYSZ　reduces　the　attenuation　rate　and　increases　the　average　current　density.　Improving　Ni　phase　fraction　helps　to　reduce　the　attenuation　rate　of　current　density.　Since　multi-physical　field　(MPF)　simulation　needs　long　calculation　time　and　it　is　difficult　to　achieve　fast　prediction,　artificial　neural　network　(ANN)　is　trained　by　the　database　generated　by　MPF.　The　mapping　relationship　between　operating　parameters,　structural　parameters　and　attenuation　indexes　is　obtained.　Finally,　the　attenuation　performance　of　SOFC　is　optimized　by　genetic　algorithm　(GA)　through　data-driven　method.　The　absolute　average　relative　errors　of　all　parameters　in　predicting　attenuation　rate　and　average　current　density　are　as　low　as　0.767%　and　0.248%,　which　indicates　the　reliability　of　the　ANN　prediction.　After　optimization,　the　maximum　current　density　is　5848　A·m−2　under　operating　voltage　at　0.6　V　when　the　attenuation　rate　requirement　not　exceeding　1%　is　satisfied.　The　combination　of　MPF　simulation,　ANN　and　GA　provides　a　framework　for　fast　performance　prediction　and　optimization　of　strong　nonlinear　system.　©　2022　Elsevier　Ltd