Different indices are defined and calculated to assess the system adequacy in power system reliability studies. The role of different elements of the system in such indices are then quantitatively estimated using "component importance measures" from different perspectives. Furthermore, Bayesian Networks have become a powerful tool in modeling and analysis of engineering systems in recent years. Bayesian Networks enable the integration of data and information resources with human experiences and expert’s opinions and considering multivariate variables and the simultaneous use of discrete and continuous variables for modeling and analysis of the system. Such advantages are also applicable to reliability assessment of power systems. Using Bayesian Networks, various indices, components importance measures, system adequacy, and availability of electrical energy at various load points are simply and accurately calculated with much less computational burden compared to conventional methods. This thesis is focused on the construction of Bayesian Networks corresponding to a composite power system based on the concept of minimal cut-sets. This method is then applied to various reliability assessment indices, considering the effect of weather conditions and calculating the component importance measures. Accordingly, Bayesian Networks are firstly defined and their application in engineering and power system reliability assessment is reviewed. Consequently, different methods of structure identification, parameter determination, and Bayesian Network formation are introduced and pros and cons of each method is investigated. Then, a method for constructing the Bayesian Network associated with a proposed composite power system based on its minimal cut-sets is presented. Using this concept makes it possible to consider the effect of the outage of different elements of the system and provide a convenient and easy way to determine the structure and parameters of the Bayesian network with sufficient accuracy. As the application of Bayesian Networks in long-term reliability assessment of power system is the main goal of this research, it is essential to consider the annual load change in its formation. Therefore, this research estimates the hourly load curve with a multi-level load duration curve with sufficient number of load levels. Subsequently, power system cut-sets are determined in each load level and corresponding reliability indices are calculated. With the weighting of such calculated indices for each load level with its probability, the expected values of power system reliability indices are evaluated. K-means method is used to appropriately determine the number and the center of load levels in forming the multi-level load curve. In this thesis, the effect of weather conditions on the power system reliability indices formed through Bayesian Networks is also studied. Accordingly, parent nodes that are indicating the weather conditions of each area are added to the corresponding Bayesian Network. Thus, the impact of weather conditions on reliability indices are considered in the augmented forced outage rate of transmission lines in that area. The case studies of this research include construction of the Bayesian Network associated with the standard test system of RBTS using the concept of minimal cut-sets and using this to implement various studies. Evaluation of various reliability indices at each load point and in the complete system, evaluation of the impact of weather conditions on such indices, and calculation of the component importance measures of various system elements using the corresponding Bayesian Network are couple of assessments in this dissertation.