Abstract: As a fourth-generation synchrotron radiation source, the Wuhan Advanced Light Source (WALS) adopts a compact multi-bend achromat (MBA) lattice design, which brings significant performance improvements while imposing more stringent requirements on beam orbit stability. This paper systematically investigated the closed-orbit correction in the storage ring of the WALS. First, focusing on the unique high-gradient compact structure of the 1.5 GeV storage ring, the study conducted a detailed analysis of its lattice design characteristics. The study reveals that the significantly reduced spacing between magnet units leads to a breakthrough in performance compared to conventional designs, but also increases the system’s sensitivity to various errors, posing substantial challenges to beam stability. In the study of closed-orbit correction methods, this paper thoroughly examined the limitations of the traditional singular value decomposition (SVD) method in the context of fourth-generation light sources. The strong focusing and high-gradient structure of the fourth-generation light sources leads to a significant increase in the condition number of the orbit response matrix, rendering that the conventional SVD method becomes highly sensitive to minor perturbations during matrix inversion. Even slight measurement errors are dramatically magnified, resulting in unstable correction outcomes. Particularly in the regions adjacent to super-bends, residual orbit distortions persist with amplitudes up to three times the average level across the entire ring. To address these issues, this paper proposed an improved correction strategy based on adaptive weighted least squares (AWLS). The AWLS method is an enhanced least-squares estimation approach primarily designed to handle heteroscedasticity or outliers. Its core principle lies in dynamically adjusting the weights of data points, thereby increasing the model’s sensitivity to high-quality data while reducing the influence of low-quality data (e.g., noise or outliers). Specifically, this method assigns adaptive weights to each singular value—higher weights for larger singular values and lower weights for smaller ones, effectively avoiding the information loss caused by manual truncation. By preserving the complete matrix information while suppressing noise, the approach successfully optimizes correction performance in high-field-strength regions. Through numerical simulations, this study compared the closed-orbit distortion correction results of both methods. The proposed improved algorithm significantly reduces the RMS residual errors from 86.3 μm and 53.7 μm (horizontal and vertical) with conventional SVD method to 46.7 μm and 48.8 μm. Furthermore, the modified algorithm demonstrates remarkable improvement in suppressing localized residual distortions, particularly in critical regions such as super-bends, confirming its superior performance for the fourth-generation light sources. This research not only provides crucial technical support for the construction of the WALS, but the proposed adaptive weighting correction strategy also establishes a novel approach for closed-orbit control in the fourth-generation synchrotron radiation facilities. These findings offer valuable insights for advancing technical development in accelerator physics.