فصلنامه هوش محاسباتی در مهندسی برق
سال یازدهم شماره 4 (زمستان 1399)

One of the major issues in the planning stage for distribution networks is the optimal design of microgrids (MGs). One of the important goals in the design of MGs is supplying the load with the highest reliability and the lowest cost. Construction of MGs in the distribution systems may have many benefits such as local control strategy in order to minimize the interaction between different MGs, prevent the fault propagation and finally, increase the reliability of the system for consumers and distribution companies. Therefore, in this paper, with the aim of improving reliability and cost, three different approaches to constructing multi-microgrid (MMG) systems in the active distribution network (ADN) are proposed and compared. In this way, by determining the location and capacity of energy storage systems (ESSs) as well as clustering of ADN based on the mentioned objectives and proposed optimization algorithms, in different scenarios, using the Monte Carlo (MC) algorithm and the Forward-backward load flow, the objective function is calculated. Then, with the repetition and various suggestions made by the optimization algorithm, the optimal value of the objective function is obtained and the results of the designed approaches are compared. To simulate and demonstrate the effectiveness of the proposed method, the 119-bus test distribution system is used.

In this paper, a new method of data-driven controller (DDC) design using Simultaneous Perturbation Stochastic Approximation Algorithm (SPSA) and Neural Network (NN) training is presented. This method can be used to control a variety of linear and nonlinear systems. In the simultaneous perturbation stochastic approximation algorithm, the controller is assumed to have a fixed structure and its parameters must be estimated. In this paper, a Proportional, Integral, and Derivative controller (PID) is considered and the parameters that should be estimated by the proposed algorithm are proportional, integral and derivative terms of this controller. In the proposed method, the simultaneous perturbation stochastic approximation algorithm is quantified by using neural network training which increases the convergence speed and also improves the performance of the algorithm against system input changes. Simulations performed on cement grinding particle size distribution process and pitch angle control of aircraft show the high efficiency and potential of the proposed method.

Permanent magnet machines are widely used in wind turbines due to theirs prominent characteristics. The direct-drive wind generators have large size and high manufacturing cost because of low operational speed so that these disadvantages deteriorate when the rated output power increases. Therefore, the main objective of the optimal design proposed in this paper is to minimize the wind turbine system cost maintaining the high efficiency characteristic. Thus, a multiphysics model of the wind generation system is firstly presented to be the basis of the optimal design. One of the paramount feature of the current studyis to precisely calculate all of the energy production costs, including the manufacturing cost of generator, power converter, electrical subsystem, and also wind turbine tower cost. In order to obtain the optimum solution, the Genetic Algorithm (GA) optimization method is implemented with consideration of a predetermined proper range for optimal variables. By comparing the optimal design results with those attained by the inner-rotor permanent magnet generator, the superiority of the proposed method is justified in terms of the total volume, weight, and annual output energy. Eventually, the finite element analysis has also been carried out to validate the outcomes obtained from the proposed optimal design.