To investigate the fatigue scatter in TA15 titanium alloy caused by microstructural variations, statistical volume elements? were generated based on microstructural characterization data. A thermally-activated crystal plasticity constitutive model considering dislocation slip mechanisms was established, and strain energy dissipation was introduced as the fatigue indicator parameter (FIPw). Generalized extreme value analysis was performed on the maximum FIPw samples to systematically study fatigue scatter under different loading conditions and grain characteristics. The results reveal that near the fatigue limit load, the maximum FIPw follows a heavy-tailed distribution, while at higher loads, the distribution shifts to the Gumbel type, accompanied by reduced fatigue life and decreased scatter. Increasing grain size significantly raises the extreme FIPw values and exacerbates fatigue scatter. Compared with equiaxed grains, elongated grain structures exhibit lower FIPw and reduced fatigue scatter. Specific crystallographic textures effectively lower both the extreme FIPw values and the associated scatter. Comparisons with experimental fatigue data and related studies validate the effectiveness of the proposed method in capturing the sensitivity of fatigue scatter to both loading and microstructure. This work provides a theoretical basis for the microstructural design and performance optimization of TA15 titanium alloy.