Permeability is a crucial parameter for characterizing unconventional reservoirs, yet predicting it in shale oil reservoirs remains challenging due to their extreme heterogeneity and the multifactorial influences on permeability. In this study, a novel analytical permeability prediction model based on fractal theory is provided. This model integrates porosity, maximum pore radius, the fractal dimension of pore size distribution, and tortuosity. The model is validated using tight core samples from shale reservoirs in the Jimsar Sag, Junggar Basin, NW China, with data obtained from high-pressure mercury intrusion measurements. Furthermore, a rock typing method based on a maximum pore radius is introduced, which enhances the accuracy of permeability predictions across different reservoir types, particularly when the model is simplified. Comparative analysis with classical models, including Pittman, Swanson, and Winland, demonstrates that the simplified model consistently provides a higher prediction accuracy for reservoir permeability.