Otimização de um modelo de propagação com múltiplos obstáculos na troposfera utilizando algoritmo genético

• Autor(a) principal: Vilanova, Antonio Carlos
• Data de Publicação: 2013
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Repositório Institucional da UFU
• Texto Completo: https://repositorio.ufu.br/handle/123456789/14317; https://doi.org/10.14393/ufu.te.2013.7
• topic: Algoritmos genéticos; Equações parabólicas; Divisor de passos de Fourier; Custo computacional; Ondas eletromagnéticas; Algoritmos genéticos; Genetic algorithms; Parabolic equations; Split-step Fourier; Computational cost; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA

This thesis presents an evaluation methodology to optimize parameters in a model of propagation of electromagnetic waves in the troposphere. The propagation model is based on parabolic equations solved by Split-Step Fourier. This propagation model shows good efficiency and rough terrain situations where the refractivity varies with distance. The search for optimal parameters in models involving electromagnetic waves requires a large computational cost, especially in large search spaces. Aiming to reduce the computational cost in determining the parameter values that maximize the field strength at a given position of the observer was developed an application called EP-AG. The application has two main modules. The first is the propagation module that estimates the value of the electric field in the area of a given terrain irregularities and varying with the refractivity with distance. The second is the optimization module which finds the optimum antenna height and frequency of operation that lead the field to the maximum value of the land in a certain position. Initially performed only the propagation module using different profiles of land and refractivity. The results shown by contours and profile field shown the efficiency of the model. Subsequently to evaluate the optimization by genetic algorithms were used two different settings as well as the irregularity of the terrain, refractivity profile and size of the search space. In each of these settings picked up a point observation in which the value of the electric field served as a metric for comparison. At this point, we determined the optimal values of the parameters by the brute force method and the genetic algorithm optimization. The results showed that for small search spaces virtually no reduction of the computational cost, however for large search spaces, the decrease was very significant and relative errors much smaller than those obtained by the method of brute force.