In this paper, we propose a graph-based kinetostatic state estimation refinement method for cable-driven parallel robots (CDPRs), addressing the challenges posed by cable sag in large-scale systems. CDPRs offer significant advantages, such as lower inertia, cost-effectiveness, and large workspaces; however, cable sag introduces inaccuracies that hinder precise control and pose estimation. Our method leverages a unified graph-based optimization framework to model and solve the coupled kinetostatic equations of CDPRs. To eliminate the need for force sensors, a dedicated force distribution optimizer estimates cable forces, which are used to refine the robot’s pose in subsequent iterations. Finite element (FE) simulations using RecurDyn validate the proposed approach, demonstrating its ability to achieve significant improvements in position and orientation accuracy while maintaining computational efficiency suitable for real-time applications.