Abstarct: The increasing population and human on the one hand and the development of science and technology on the other hand have led to the increasing use of underground spaces. Caves, as an important part of the underground space, are used in hydroelectric power plants and hydrocarbon storage tanks. Water leakage into these areas is one of the major issues that have been analyzed and investigated by researchers to prevent related damage. The purpose of this study was to study the water leakage into the cavern of the pumped storage power plant of Rudbar dam in Lorestan using three-dimensional anisotropic seepage-stress coupling analysis. For this purpose, at first, the insitu data were classified into three categories: semi-horizontal, semi-vertical and background. Using Dips software for background joints of a joint batch, 2 batch-like and horizontal batch-3 joints were found to be significant. R statistical software was used to determine the appropriate probability distribution function by the length of the joints harvested in the area and Powerlaw as the most appropriate distribution was then obtained. baxsed on the above information, a 3D model of DFN with dimensions of 100m×100m×100m was constructed. The model was validated by using the Watson-Williams test with Oriana software. By using the DFN model, the REV block model was constructed. The 3D block model of REV with dimensions of 7 m and 10 m breakage filter was selected as the optimal model. Also, if the mesh size is 0.4 meters, the model had network independence. Finally, the hydraulic conductivity tensor was obtained using the optimal REV model. The resulting tensor is then equilibrated to examine the coupling leakage and stress in the continuous environment. For modeling in an equivalent continuous environment, the software was used. The finer mesh is intended for independence from the model. In order to solve the coupling problem, the initial hydraulic conductivity values were first introduced into the model and then the resulting stress was obtained. From the stresses obtained by the Louice equation, the new hydraulic conductivity values and the new hydraulic conductivity values were equated to the continuous environment model to obtain new stress values. This process continued until the tensor of the hydraulic conductivity coefficients reached steady state, which occurs from the third cycle onwards. The effect of coupling stress on the amount of leakage compared to the non-stress state was 4.8 times. Also, if the hydraulic conductivity is applied isotropically, it can increase or decrease by several times the amount of leakage. Hydraulic conductivity data were validated using field data of the Lugeon experiment with a 17% error rate. The results of this dissertation can be used to predict the amount of caverns to be used for injection operations and related issues.