This thesis presented a new approach based on Maximum a Posterior probability (MaP) inference in Bayesian networks to identify and rank the weak components of the composite power system established on reliability indices. In this method, the necessary training data that models the behavior of the proposed power system is firstly simulated, and the corresponding Bayesian network is constructed. In this way, the Bayesian network structure is optimally designed by the simulated annealing method, and the Bayesian estimator algorithms determine its parameters. The constructed Bayesian network can be used with eminent speed and accuracy to analyze and predict the behavior of the power system in various studies. The proposed method identifies the weak components of the composite power systems, which substantially impact system unreliability. Based on the accuracy required, the number of these weak components can be limited only to the most important components. In this study, weak components are categorized into global and local groups, considering the status of load supply in the whole system and specific load points, respectively. After identifying the weak components, they are directly ranked by the MaP inference from the LOLP and LOLF indices points of view, based on the system peak load. Furthermore, from the LOLE and EENS indices points of view, which are affected by the duration and changes of the load; a new Bayesian-based method is used to form a composite load duration curve and eva‎luate their rank. By the proposed method, the Bayesian network corresponding to the power system is formed only once, and the necessary conditions in terms of different reliability indices are observed in the MaP inference and estimated by the Bayesian network. However, in the conventional state enumeration method, identification and ranking of weak components may be established only after examining almost all operating states of the power system, which is very time-consuming and has a heavy computational burden. Finally, due to the increasing importance of renewable energy sources, especially wind generation, the effect of these resources on the participation of weak components in the power system reliability indices are also investigated and analyzed using the proposed method. The proposed technique has several advantages compared with the conventional methods, such as: (i) requiring much fewer data, (ii) lower computational burden, (iii) the ability to identify and rank local weak components, (iv) the possibility of limiting the process based on the required accuracy, and (v) a significant increase in computation speed with almost the same accuracy.

اكبر ابراهيمي

زينب مالكي

محمد‌امين لطيفي، مهران صفاياني