Large-scale ultrasonic positioning systems have recently shown substantial improvement, according to the current great interest concerning large-scale metrology applications in many different fields of manufacturing industry. However, the sound velocity is greatly influenced by the ambient temperature especially in large-scale applications. The traditional sound velocity compensation method mainly uses the average temperature in measurement space based on an assumption that the temperature field is uniform and stable. As the assumption is invalid in most cases, especially in industrial measurement environments, the traditional compensation method will bring obvious errors. To reduce these errors caused by sound velocity, this article proposes a novel compensation method by constructing a three-dimensional temperature field of measurement space through heat transfer theory as well as the finite element method, and then estimating the distance accurately using a couple iterating algorithm. Verification experiments demonstrate that the ranging deviation estimated through the proposed method may keep within 0.5 mm in 5.4x2 m measurement space, whereas the average distance measurement uncertainty is about 0.25 mm.