teaching – learning-based optimization algorithm

This paper investigates the accuracy and convergence rate of different metaheuristic algorithms in determining the stiffness of structural elements using structural modal parameters and defining a suitable objective function. To achieve this purpose, three different structures, including a three-story one-dimensional frame, a six-story one-dimensional frame and a two-dimensional truss, were investigated. The metaheuristic algorithms, employed in this study, were Genetic Algorithm, Particle Swarm Optimization, and Teaching–learning-based Optimization. The objective function utilized in this study consists of two terms; the first part involves the squared difference between the first frequency of the structure obtained from the responses of the investigated structure and the first frequency obtained from the hypothetical stiffness matrix in each generation of algorithms. The second part measures the norm of the difference between the first mode shape of the structure obtained from the responses of the investigated structure and the first mode shape obtained from the hypothetical stiffness matrix in each generation of algorithms. By minimizing the objective function, the Genetic Algorithm, Particle Swarm Optimization, and Teaching–learning-based Optimization determined the element stiffness of the three-story, six-story and truss structures, thus demonstrating the high efficiency of metaheuristic algorithms in resolving unknown parameters of structures. The average run time for the Genetic Algorithm was 3.38 seconds, 4.47 seconds, and 15.73 seconds for the three respective problems. For Particle Swarm Optimization, the times were 3.76 seconds, 6.47 seconds, and 16.76 seconds. The Teaching–learning-based Optimization achieved times of 1.92 seconds, 4.51 seconds, and 12.76 seconds. The Teaching–learning-based Optimization exhibited the highest convergence rate and the lowest error compared to the Genetic Algorithm and Particle Swarm Optimization. For example, in the two-dimensional truss, the values of the objective function in the last iteration of the Genetic Algorithm, Particle Swarm Optimization, and Teaching–learning-based Optimization were , and , respectively. The Particle Swarm Optimization demonstrated an acceptable convergence rate and error compared to the Genetic algorithm. The Genetic Algorithm, however, displayed a significant error rate in determining the stiffness of structural elements compared to the other two algorithms.

Today, issues such as restructuring, environmental issues, problems and limitations in transmission and distribution lines have led to the increasing use of distributed generation (DG) systems. DG units, according to the characteristics, technology and location of connection to the network, can cause positive effects such as improving the voltage profile, reducing power losses and reliability in distribution networks; Therefore, with the increase in the use of scattered production resources as well as the technical and financial issues of these technologies, new issues such as determining the capacity and location of connecting these equipments to the network have been investigated. The problem of rearranging the distribution network is another optimization problem in distribution networks, so that it optimizes the network for the intended purpose without the need for new equipment. In this article, using an improved model of the learning and teaching-based optimization algorithm (TLBO), simultaneous simulation of distributed generation location problem and distribution network reconfiguration problem, in order to optimize the voltage profile, increase the voltage stability index and reducing network power losses. The proposed method has been implemented with different scenarios on radial networks of 33 buses and 69 buses, and the results obtained after comparing with other methods show the efficiency of the proposed method.

Because the Internet of Things (IoT) deals with large amounts of data, it is not easy to process and store this amount of data. However, many of its applications suffer from cloud computing challenges such as latency, location awareness and real-time mobility support. Fog calculations help provide solutions to these challenges. This paper includes an IoT network simulation for allocating optimal shared resources in fog computing to solve the mix integer nonlinear programming (MINLP) problem, which aims to maximize the profitability of cloud service providers through fog computing. The network architecture consists of three layers: cloud service providers, fog nodes, and users. In this paper, the three-layer network is simulated and the algorithm used in this problem is the Teaching–Learning-Based Optimization (TLBO) algorithm, which uses two phases of learning and teaching for the three parameters of cloud service providers’ revenue, average delay and user satisfaction for selecting the best node with the aim of allocating shared resources. This algorithm is implemented on the model and compared with a random method. This model and algorithm increases the profit of service providers compared to the algorithms used to solve similar models.

Due to the simple structure of DC motors, these motors have found many applications in industry.Therefore, in this paper, the speed control of DC motor is investigated by considering the dynamics of drive and chopper with Sugeno-type fuzzy controller. A chopper is used to control the voltage applied to the DC motor armature. Considering the dynamics of the chopper drive increases the complexity of the system. After designing the fuzzy controller to increase the performance of the control system, the fuzzy controller parameters are adjusted using a teaching-learning-based optimization algorithm. This algorithm is new and one of its features is its small number of parameters. The results show that the fuzzy controller has better performance against changes in system parameters and uncertainties compared to the classic PID controller. Considering the appropriate criterion function, the value of the cost function for the proposed method is 0.2. But with the optimized PID controller about 0.31 which shows a 55% superiority of the proposed method.

To meet the energy demand, the expansion of transmission networks has caused low frequency oscillations in the power system. These fluctuations, if not attenuated, reduce the capability of the transmission lines and sometimes cause the whole system to be unstable. Proper selection of stabilizing parameters have great importance and increase the optimum performance during unwanted disturbances. In this article, a new algorithm based on teaching and learning is used to increase the stability of the multi-machine power system and thus increase the performance of the power system. In fact, since the optimal tuning of the control parameters has a direct effect on the low frequency stability, it becomes an optimization problem and is accomplished with a new algorithm based on the optimized training and learning algorithm of the power system stabilizer. In addition, the results are compared with the colonial competition algorithm. A four-Machin test system is used to evaluate the effectiveness of the proposed method. By examining the results in different modes, it is shown that the proposed method using the learning-based optimization algorithm is much more efficient than the proposed method using the colonial competition-based algorithm.

This paper presents a modified method based on teaching learning based optimization algorithm to solve the problem of the single- and multi-objective optimal location of distributed generation units to cope up the load growth in the distribution network .Minimizing losses, voltage deviation, energy cost and improved voltage stability are the objective functions in this problem. Load growth in a system is a natural phenomenon. As the value of distribution system indices depend on load growth, any change in load causes changes in distribution system indices. Therefore, for assessing the current studies the level of validity and the adequacy of system in future, load growth in future are discussed. The proposed algorithm integrates an adaptive teaching factor, and a crossover and mutation strategy. In the proposed algorithm, an external repository is considered to save non-dominated (Pareto) solutions. Since the objective functions are not the same, a fuzzy clustering technique is used to select the best compromise solution from the repository. The proposed algorithm is tested on the 33-bus IEEE test system. To demonstrate the effectiveness of the proposed approach, simulation results are compared with the results obtained by other methods.

In this paper , a new mathematical model for the electricity generation expansion planning problem are presented. The model, in terms of planning, is among the long-term planning in their literature. However, the steps of model are on a monthly basis. Considering the soft and hard constraints together and fits with other features is the model concept in the real world. Classifying plants according to construction areas should be added to other model features. Then the exact solution to small - scale of the model is presented by the software (GAMS). Then, due to the complexity of the problem, two meta-heuristic algorithms, genetic and traininig – learninig based optimization, was presented. Genetic algorithm parameters are set using RSM. The 22 series of numerical examples by each of these two algorithms are solved and then the results were compared with each other. The results show that the genetic algorithm has the advantage.