Excellent self-righting performance is important to guarantee the normal navigation of Unmanned Surface Vehicles (USVs) after overturning, and the restoring time is an important index in design requirements. Traditionally, the static stability method and experiments on full-scale vehicles were used to analyze the large-angle stability of the USV. However, when it comes to the analysis of self-righting performance, the traditional static stability method will cause improper integration, and experiments are not convenient. To solve these problems, an improved static stability method was proposed, and a whole self-righting process simulation of a physical model was finished. The numerical simulation method was used to predict the self-righting process of a USV under four working conditions. Firstly, a midpoint average method based on the static stability theory was adopted to compute the static restoring time, and the results were compared with the results of the references, which verifies the effectiveness of the midpoint average method. Also, the midpoint average method is convenient because it only needs the restoring arm curve, the width and the gravity center height. Then, a numerical simulation of a physical model in static water was finished, and an experiment for a physical model in a towing tank was conducted. Comparing the restoring time of the midpoint average method, the numerical simulation and the experiment, the results show that the numerical simulation has high accuracy. Moreover, the numerical simulation was used to predict the self-righting process and analyze the self-righting performance of a USV under four working conditions.