In the instrument suite of the Thirty Meter Telescope (TMT), the linear atmospheric dispersion corrector (LADC) is positioned at the forefront of the spectrometer to mitigate atmospheric dispersion. The LADC comprises two large aperture wedge prisms, each approximately 1.5 meters in diameter. These prisms, which are transmissive optical elements, are supported radially along their outer circumferences. However, due to the rotational asymmetry of the prisms relative to the optical axis, the support forces do not align with the plane of the center of gravity, leading to a bending moment caused by gravitational and reaction forces. This misalignment results in significant deformation of the prism surfaces. The structural-thermal-optical performance integrated model method is typically employed to evaluate the impact of the support system on prism surface deformation. This process often involves extensive manual iterations and operates in an open-loop manner, making it inefficient and heavily reliant on the analyst's expertise. To address these limitations, this paper proposes a closed-loop integrated model analysis optimization method aimed at enhancing optimization efficiency. By integrating interface programs between analysis tools and applying advanced optimization algorithms, the traditional open-loop integrated model analysis process is transformed into a closed-loop system, enabling more efficient and reliable optimization. The closed-loop integrated model method incorporates Optimal Latin Hypercube Design (Opt LHD) and Particle Swarm Optimization (PSO) algorithms to optimize the prism support structure, reducing the analysis time from one week to just two days. Compared to the original support structure, the root mean square (RMS) value of optical surface deformation decreased by 49.3%, from 155.0 nm to 78.6 nm, under gravity and 2°C temperature coupled conditions; and by 62.5%, from 289. 1 nm to 108. 4 nm, under gravity and 42°C temperature coupled conditions. These results demonstrate that the closed-loop integrated model optimization method not only improves efficiency but also achieves significantly better outcomes.