الفهرس 
ACKNOWLEDGMENTOnly 14 pages are availabe for public view
 |  | from | 141 |  |  |
| | | from | 141 | | |
Abstract
An attempt has been made to study metaheuristic optimization techniques, the background from nature by which they were inspired, and the applicability of such methods in real-world industry problems. The research conducted in this thesis does not only focus on the immunological metaphors and Artificial Immune Systems (AIS) built on them but also a number of algorithms in the Evolutionary Computations package were also reviewed and some were actually used within experimentations for comparison aspects.This thesis reviews different methods to maintain diversity of populations in population-based optimization algorithms which can hugely impact their performance. A novel operator for diversity maintenance is then introduced which was studied and assessed using two different algorithms (an immune algorithm and a genetic algorithm). The implemented algorithms with the use of the operator were tested on a set of widely-used benchmark functions from the literature. The proposed operator was able to improve the results of the algorithms by leading them out of local optima and guiding them to keep exploring the search space and eventually to find better solutions.Real-world industry problems are far more complex than benchmarks and are usually challenging with an arsenal of constraints and often being multi- objective. Along with reviewing some of the important optimization problems in industry, the research in this thesis continued by attempting to use artificial immune systems to solve the Wind Farm Layout Optimization (WIND-FLO) problem using GECCO’s WIND-FLO competition API. This allowed to test the proposed diversity operator to assess how it can be of help in a more realistic environment. Using the immune algorithm aided with the diversity operator for the experiments, the results were similar (and sometimes better) than the published results of the competitions in which the participated algorithms are all tweaked specifically for such problem unlike the one we used which was for general optimization with no specific instructions for the supplied API.Finally, a few areas for future research were identified to improve the optimality of the algorithm, such as automating the parameters of the proposed operator. Furthermore, and since many real-world problems have multi- objective nature, it looks helpful to study implementing multiobjectivisation by converting the proposed diversity operator from an injected operator to an auxiliary objective on its own.