<p>Aiming at the multi-objective and multi-variable problem of power cables of cylindrical floating production oil storage vessels, an intelligent system using an improved Deep Deterministic Policy Gradient (DDPG) algorithm is innovatively proposed to optimise the power cables. By interacting with the dynamic cable simulated by ANSYS, optimised the buoyancy module configuration to improve the Load-bearing capacity and durability of the cable. The overall performance of the method in terms of displacement, bending moment and axial force was improved by 67.8%. Displacement was reduced by 59.6%, bending moment by 45.7% and axial force by 32.5%. The number of buoyancy modules was reduced from 28 to 19, the position of the first buoyancy module was moved from 200 to 275 metres, and the spacing was increased from 5 to 10 metres. The number of clump weights was reduced from 8 to 1, the spacing between clump weight blocks was 3.5 metres, and the distance from the first clump weight to the suspension point was 100 metres. This method integrates reinforcement learning with high-precision structural simulation in a closed-loop manner, enabling iterative learning in a physical feedback environment and improving the efficiency and reliability of multi-objective optimisation. Meanwhile, the constructed intelligent agent system is able to adapt to the dynamic nonlinear response under wave current perturbation and simulate the optimisation behaviour that approximates the engineering hierarchical control logic. The study provides a novel numerical methodology for the adaptive design of marine power cable systems, which has strong potential for engineering application and promotion potential.</p>