Carbonate reservoirs in the Shunnan area are the key target for natural gas exploration and development in the Tarim Basin. Fluid identification is an effective method to improve exploration efficiency. Therefore, this paper combines a multi-scale fractured rock physical model and frequency-dependent amplitude vs offset (AVO) inversion method to establish a set of fluid detection technologies. Firstly, based on the drilling, logging, seismic, and geology data, we use the Chapman equivalent model to simulate the fractured reservoirs in the Shunnan area, and use the reflectivity method for forward modeling the seismic wave response feature of reservoirs saturated with different types of fluids. Secondly, it is necessary to carry out seismic attribute sensitivity analysis of different types of reservoir to determine which  seismic dispersion is more sensitive to fluid. Finally, based on the situation of seismic data in the study area, we establish a technical process based on the petrophysical model and the frequency-dependent AVO inversion, and apply it in the Shunnan area. The results of fluid detection and real drilling verification show that the method proposed in this paper is practical and provides a method for fluid detection in this type of oil and gas reservoir.