The distributed optical fiber sensing system based on Brillouin optical time domain analysis (BOTDA) can be used to measure temperature, stress, and other information in the external environment. In order to further improve the measurement accuracy and speed of distributed optical fiber sensing system, a Brillouin frequency shift extraction method based on adaptive gradient descent algorithm is proposed, and a 24.4 km distributed temperature sensing system based on BOTDA is setup. The results show that, compared with the traditional Levenberg-Marquardt Lorentz fitting algorithm, the proposed method can extract Brillouin frequency shift spectrum quickly and accurately, which is of great significance to improve the measurement accuracy of BOTDA distributed temperature sensing system.