Abstract: To address the limitation in measurement accuracy of existing ultrasonic time difference anemometry caused by the inability of transducer installation positions to dynamically adapt to different roadway cross-sections, this study investigated two ultrasonic transducer installation positions for different roadway cross-sections (such as rectangle, trapezoid, arch, et al). Fluent numerical simulations and underground manual measurements were conducted to verify the existence of an average wind velocity line parallel to the roadway floor. On this basis, by combining turbulent stress decomposition theory with Prandtl's mixing length theory and introducing a modified coefficient, the height function of the average wind speed line from the roadway floor for different roadway sections was derived. Together with the determination of the velocity measurement interval, the installation positions of the two ultrasonic transducers were determined. Simulation and field test results showed that: ① compared with the traditional fixed-height transducer installation method, the proposed transducer installation position determination method achieved higher measurement accuracy under the same operating conditions. ② The installed transducers performed best in rectangular roadways, with a relative error of 4.486% compared with manual measurements, followed by arched roadways with a relative error of 4.935%, while trapezoidal roadways showed the largest relative error of 5.579%. ③ Using standard deviation as the evaluation criterion rsingstandard deviation as the evaluation criterion, the standard deviation of wind velocity measurements obtained by the installed ultrasonic transducers was smaller than that of manual measurements, indicating more stable data acquisition in underground environments.