Abstract: Atmospheric water vapor absorption and line-of-sight path delays impose severe constraints on ground-based millimeter and submillimeter astronomical observations. High-precision monitoring of Precipitable Water Vapor (PWV) with high temporal resolution is therefore crucial for ensuring observation quality and enhancing interferometric imaging accuracy. To address the challenges of extremely low water vapor environments at high-altitude sites, such as the Tibetan Plateau, we developed a high-sensitivity 183 GHz Superconducting Water Vapor Radiometer (SWVR) utilizing Superconductor-Insulator-Superconductor (SIS) mixing technology. This paper details the system’s hardware architecture and the design of a dual-temperature blackbody calibration unit. Thermal simulations and laboratory measurements demonstrate that the high-temperature blackbody calibration source achieves a thermal stability of 0.01 K, even under extreme ambient temperatures as low as -40\mathrm^\ \circC , thereby ensuring high radiometric calibration accuracy. Regarding retrieval algorithms, we evaluated and compared the performance of Linear Regression (LR), Quadratic Polynomial Regression (QPR), and Deep Neural Network (DNN) models based on a site-specific atmospheric radiative transfer model. The results indicate that the QPR algorithm outperforms both LR and DNN in terms of retrieval accuracy (achieving an RMSE of 22 μm), computational efficiency, and robustness against noise, particularly given the significant non-linear saturation effects of the 183 GHz line over a wide dynamic range. Consequently, QPR is identified as the optimal algorithm for the SWVR system. These findings validate the SWVR’s capability to monitor PWV with precision on the order of tens of microns in extremely dry environments, providing reliable technical support for radiometric imaging and phase correction in ground-based submillimeter telescopes.