Large internal gears are widely utilized in wind turbine gearboxes, with form milling being the primary processing method. Unlike end milling, gear form milling employs gear milling cutters with indexable carbide inserts that feature complex geometries, resulting in significant variations in cutting angles at different points along the cutter axis during machining. This complexity leads to a convoluted mapping relationship between the undeformed chip thickness (UCT) and cutting parameters, significantly reducing the predictive accuracy of traditional cutting force models in gear form milling. To address this challenge, this study developed a cutting force prediction model for large internal gear form milling. The model presents a partitioned approach for calculating UCT, establishing corresponding shear cutting models for each region to effectively address challenges posed by the unique geometry of the cutting edge in gear form milling. Subsequently, this study cleverly utilized the structural characteristics of the internal gear milling machine to establish a mapping relationship between cutting force and spindle box strain, employing the Artificial Hummingbird Algorithm (AHA) to estimate cutting coefficients. This optimized method overcomes the difficulties of inverse calculations due to model complexity and alleviates measurement challenges during large gear machining caused by system rigidity and clamping limitations. Experimental validation shows that the model's average error is below 7%. Hence, this research significantly contributes to the analysis of cutting forces in large internal gear form milling.