Solar　arrays　are　the　main　source　of　energy　to　the　on-orbit　satellite,　whose　output　power　largely　determines　the　life　cycle　of　on-orbit　satellites.　Monitoring　and　further　forecasting　the　output　power　of　solar　arrays　by　using　the　real-time　observational　data　are　very　important　for　the　study　of　satellite　design　and　on-orbit　satellite　control.　In　this　paper,　we　firstly　describe　the　dynamical　model　of　output　power　with　summarizing　the　influencing　factors　of　attenuation　for　solar　arrays　and　elaborating　the　evolution　trend　of　influencing　factors　which　change　with　time.　Based　on　the　empirical　model,　a　particle　filtering　algorithm　is　formulated　to　predict　the　output　power　of　solar　arrays　and　update　the　model　parameters,　simultaneously.　Finally,　using　eight-year　observational　data　of　voltage　and　current　from　a　synchronous　on-orbit　satellite,　an　experiment　is　carried　out　to　illustrate　the　reliability　and　accuracy　of　the　particle　filtering　method.　Comparative　results　with　classical　curve　fitting　also　are　presented　with　statistical　root　mean　square　error　and　mean　relative　error　analysis.