In　this　article,　a　dynamic　nonlinear　generalized　predictive　control　method　based　on　exponential　function　to　adjust　predictive　period　is　proposed　for　the　problem　of　spacecraft　attitude　tracking　with　model　uncertainties　and　unknown　disturbances.　First,　a　virtual　generalized　predictive　control　law　is　designed　on　the　basis　of　invariable　predictive　period　for　attitude　subsystem.　Subsequently,　introducing　predictive　period　into　attitude,　a　closed-loop　error　equation　for　fusion　of　attitude　and　predictive　period　is　established.　Meanwhile,　a　dynamic　adaptive　law　of　predictive　period　and　a　disturbance　estimator　are　built　in　the　light　of　Lyapunov　theory.　Second,　through　a　similar　approach,　an　actual　control　input　is　designed　for　the　attitude　angular　velocity　subsystem.　In　order　to　simplify　the　calculation,　the　derivatives　of　virtual　control　and　unknown　disturbances　in　the　actual　control　are　dealt　with　adaptive　estimators　by　considering　them　as　uncertainties.　Finally,　it　is　proved　that　the　spacecraft　attitude　can　converge　to　the　desired　attitude　of　bounded　small　neighborhood.　The　simulation　results　of　comparative　analysis　demonstrate　that　the　predictive　period　can　be　adjusted　to　improve　the　control　performance　and　speed　up　the　attitude　convergence.