Applications of computational methods in smart grids

In this thesis study, different computational methods used in information technologies are combined with the current conceptual model of electric networks, and applied to problems encountered in the field of smart grids. New methods have been developed and effectively implemented for popular problems of smart grids such as demand-side management, preserving privacy and electric vehicle charging. Demand-side management models in the literature are under discussion about how they can be effective when the computational complexity is increasing with the presence of a vast number of users as it happens in electric grids. Instead of trying to examine all demands completely, demand analysis can be done with incomplete information when the complexity is high. In this thesis study, a novel computational method is developed. This method is not too much affected by the incomplete information or communication delays. The proposed method makes an agent-based system to process incomplete information possible. It is the first study in the literature, which uses Pythagorean fuzzy sets in the action mechanisms of agents in Bayesian games. A time-series data aggregation method has been developed to preserve the privacy of smart meter readings. The method developed in this thesis is suitable for devices with limited hardware such as smart meters due to its low computational complexity and resistance to different security attack types. Different than the other data aggregation methods in the literature, the cost of data aggregation, and privacy preserving is low in this study. Furthermore, reconstruction of the data takes place in a lossless fashion. As a result, it helps to reduce infrastructure installation costs. Inspired by the scheduling algorithms found in modern operating systems, a lottery-based scheduler model has been developed to solve the problem of peak loads in the grid during simultaneous charging of electric vehicles. There are not many studies using lottery-based scheduling strategy for this purpose. The proposed method shows better performance than a genetic algorithm-based solution and provided an innovative contribution by limiting the total energy allocation amount to decrease peak-load values. Finally, the fluctuation problem arising from the integration of renewable energy sources to the grid has been examined and analyzed economically with the consideration of a vehicle to grid system. For this purpose, a simulation study was carried out. Different than the other studies in the literature, this simulation study takes into account both vehicle battery life and driver behavior.