No description available.

Recently meta-heuristics have become a popular solution methodology, in terms of both research and application, for solving combinatorial optimization problems. Meta-heuristic methods guide simple heuristics or priority rules designed to solve a particular problem. Meta-heuristics enhance these simple heuristics by using a higher level strategy. The advantage of using meta-heuristics over conventional optimization methods is meta-heuristics are able to find good (near optimal) solutions within a reasonable computation time. Investigating this line of research is justified because in most practical cases with medium to large scale problems, the use of meta-heuristics is necessary to be able to find a solution in a reasonable time. The specific meta-heuristic studied in this research is, Meta-RaPS; Meta-heuristic for Randomized Priority Search which is developed by DePuy and Whitehouse in 2001. Meta-RaPS is a generic, high level strategy used to modify greedy algorithms based on the insertion of a random element (Moraga, 2002). To date, Meta-RaPS had been applied to different types of combinatorial optimization problems and achieved comparable solution performance to other meta-heuristic techniques. The specific problem studied in this dissertation is parameter setting of Meta-RaPS. The topic of parameter setting for meta-heuristics has not been extensively studied in the literature. Although the parameter setting method devised in this dissertation is used primarily on Meta-RaPS, it is applicable to any meta-heuristic's parameter setting problem. This dissertation not only enhances the power of Meta-RaPS by parameter tuning but also it introduces a robust parameter selection technique with wide-spread utility for many meta-heuristics. Because the distribution of solution values generated by meta-heuristics for combinatorial optimization problems is not normal, the current parameter setting techniques which employ a parametric approach based on the assumption of normality may not be appropriate. The proposed method is Non-parametric Based Genetic Algorithms. Based on statistical tests, the Non-parametric Based Genetic Algorithms (NPGA) is able to enhance the solution quality of Meta-RaPS more than any other parameter setting procedures benchmarked in this research. NPGA sets the best parameter settings, of all the methods studied, for 38 of the 41 Early/Tardy Single Machine Scheduling with Common Due Date and Sequence-Dependent Setup Time (ETP) problems and 50 of the 54 0-1 Multidimensional Knapsack Problems (0-1 MKP). In addition to the parameter setting procedure discussed, this dissertation provides two Meta-RaPS combinatorial optimization problem applications, the 0-1 MKP, and the ETP. For the ETP problem, the Meta-RaPS application in this dissertation currently gives the best meta-heuristic solution performance so far in the literature for common ETP test sets. For the large ETP test set, Meta-RaPS provided better solution performance than Simulated Annealing (SA) for 55 of the 60 problems. For the small test set, in all four different small problem sets, the Meta-RaPS solution performance outperformed exiting algorithms in terms of average percent deviation from the optimal solution value. For the 0-1 MKP, the present Meta-RaPS application performs better than the earlier Meta-RaPS applications by other researchers on this problem. The Meta-RaPS 0-1 MKP application presented here has better solution quality than the existing Meta-RaPS application (Moraga, 2005) found in the literature. Meta-RaPS gives 0.75% average percent deviation, from the best known solutions, for the 270 0-1 MKP test problems.

<p>Over the last years, the emission and on board diagnostics legislations for heavy duty trucks are getting more and more strict. An accurate engine model that is possible to execute in the engine control system enables both better diagnosis and lowered emissions by better control strategies. </p><p>The objective of this thesis is to extend an existing mean value diesel engine model, to include turbo compound. The model should be physical, accurate, modular and it should be possible to execute in real time. The calibration procedure should be systematic, with some degree of automatization. </p><p>Four different turbo compound models have been evaluated and two models were selected for further evaluation by integration with the existing model. The extended model showed to be quite insensitive to small errors in the compound turbine speed and hence, the small difference in accuracy of the tested models did not affect the other output signals significantly. The extended models had better accuracy and could be executed with longer step length than the existing model, despite that more complexity were added to the model. For example, the mean error of the intake manifold pressure at mixed driving was approximately 3.0%, compared to 5.8% for the existing model. The reasons for the improvements are probably the good performance of the added submodels and the systematic and partly automatized calibration procedure including optimization.</p>

Over the last years, the emission and on board diagnostics legislations for heavy duty trucks are getting more and more strict. An accurate engine model that is possible to execute in the engine control system enables both better diagnosis and lowered emissions by better control strategies. The objective of this thesis is to extend an existing mean value diesel engine model, to include turbo compound. The model should be physical, accurate, modular and it should be possible to execute in real time. The calibration procedure should be systematic, with some degree of automatization. Four different turbo compound models have been evaluated and two models were selected for further evaluation by integration with the existing model. The extended model showed to be quite insensitive to small errors in the compound turbine speed and hence, the small difference in accuracy of the tested models did not affect the other output signals significantly. The extended models had better accuracy and could be executed with longer step length than the existing model, despite that more complexity were added to the model. For example, the mean error of the intake manifold pressure at mixed driving was approximately 3.0%, compared to 5.8% for the existing model. The reasons for the improvements are probably the good performance of the added submodels and the systematic and partly automatized calibration procedure including optimization.

Este trabalho foi baseado na metodologia desenvolvida para modelagem computacional de equivalentes dinâmicos de sistemas elétricos de potência apresentada em BRITO, 2009 [5]. Sendo que nesse trabalho desenvolveu um método de modelagem de equivalentes dinâmicos de sistemas elétricos de potência com geradores, máquinas e seus controladores. O equivalente dinâmico consiste em uma substituição de uma parcela de um Sistema Elétrico de Potência, composta por linhas de transmissão, barras de carga e barras geradoras, de forma que estas serão sintetizadas e minimizadas de forma que o número de elementos seja diminuído, mas conservem o comportamento dinâmico elétrico da área original, sendo o mais próximo possível do encontrado no sistema elétrico de potência original. Foram realizados cálculos teóricos para obtenção dos equivalentes estáticos e dinâmicos do referido sistema de potência. Para isso utilizamos o programa ANAREDE para obtenção dos resultados dos fluxos de carga do sistema completo e do sistema reduzido. Estes resultados foram utilizados em um programa de análise de pequenas perturbações, denominado PACDYN, o qual verifica o comportamento dinâmico dos sistemas. Para possibilitar essa análise foi inclusa uma pequena perturbação de tensão no sistema, próxima à sua barra de referência angular. Os resultados encontrados pelo programa PACDYN foram exportados para um programa de análise matemática e simulações, denominado MATLAB, onde foram gerados os gráficos dos resultados obtidos e aplicada a técnica de mínimos quadrados, para ajuste dos resultados. Por meio da comparação dos fluxos de potência dos sistemas completo e equivalente, obteve-se a validação do trabalho, sendo utilizado o sistema New England, como sistema-teste. Neste trabalho aplicamos a metodologia desenvolvida em BRITO, (2009) [5] no sistema de potência-teste New England, complementando o trabalho original, em que a parte interna e externa do sistema, escolhida para construção do equivalente dinâmico eram interligadas por uma única barra de fronteira. Em nosso trabalho utilizamos uma área que nos demandou a utilização de quatro barras de fronteira, quatro linhas de transmissão e uma barra fictícia de acoplamento, onde foi conectado o gerador equivalente e cargas equivalentes. Comparando os resultados do sistema completo e reduzido obtivemos valores próximos o bastante para comprovar a robustez do sistema fictício e validar mais uma vez a metodologia. / This work was based on the methodology developed for computer modeling of dynamic of electric power systems equivalents presented in BRITO, 2009 [5]. And in this work we developed a method of modeling of dynamic equivalents of electric power systems with generators, machines and their controllers. The dynamic equivalent consists of a substitution of a portion of an Electric Power System, consisting of transmission lines, the load buses and generating bars, so that these will be synthesized and minimized so that the number of elements is decreased, but retain the electric dynamic behavior of the original area, with the closest to the found in the electrical system of original power. Theoretical calculations for obtaining the static and dynamic equivalents of said power system were performed. For that use the ANAREDE program to obtain the results of the full system load flows and reduced system. These results were used in a parsing program small disturbances, called PACDYN, which checks the dynamic behavior of the system. To enable this analysis has included a small voltage perturbation in the system, next to its angular reference bar. The results found by PACDYN program were exported to a mathematical analysis and simulation program called MATLAB, which graphs the results obtained and applied the technique of least squares were generated to adjust the results. By comparing the equivalent of complete systems and power flows obtained if the validation work, by using the New England system as test system. In this paper we apply the methodology developed in BRITO, (2009) [5] in the New England power test system, complementing the original work, in which the inside and outside of the system, chosen to build the dynamic equivalent were joined by a single border bar. In our work we use an area which required us to use four border bars, four transmission lines and a dummy coupling bar, which was connected to the equivalent generator and equivalent charges. Comparing the results of the full and reduced system obtained values close enough to demonstrate the robustness of the dummy system and to validate again the methodology.

This master's thesis describes the design of effective working artificial neural network in FPGA Virtex-5 series with the maximum use of the possibility of parallelization. The theoretical part contains basic information on artificial neural networks, FPGA and VHDL. The practical part describes the used format of the variables, creating non-linear function, the principle of calculation the single layers, or the possibility of parameter settings generated artificial neural networks.

Pour résoudre des problèmes d'optimisation discret de type boîte noire, de nombreux algorithmes stochastiques tels que les algorithmes évolutionnaires ou les métaheuristiques existent et se révèlent particulièrement efficaces selon le problème à résoudre. En fonction des propriétés observées du problème, choisir l'algorithme le plus pertinent est un problème difficile. Dans le cadre original des environnements de calcul parallèle et distribué, nous proposons et analysons différentes stratégies adaptative de sélection d'algorithme d'optimisation. Ces stratégies de sélection reposent sur des méthodes d'apprentissage automatique par renforcement, issu du domaine de l'intelligence artificielle, et sur un partage d'information entre les noeuds de calcul. Nous comparons et analysons les stratégies de sélection dans différentes situations. Deux types d'environnement de calcul distribué synchrone sont abordés : le modèle en île et le modèle maître-esclave. Sur l'ensemble des noeuds de manière synchrone à chaque itération la stratégie de sélection adaptative choisit un algorithme selon l'état de la recherche de la solution. Dans une première partie, deux problèmes OneMax et NK, l'un unimodal et l'autre multimodal, sont utilisés comme banc d'essai de ces travaux. Ensuite, pour mieux saisir et améliorer la conception des stratégies de sélection adaptatives, nous proposons une modélisation du problème d'optimisation et de son opérateur de recherche locale. Dans cette modélisation, une caractéristique importante est le gain moyen d'un opérateur en fonction de la fitness de la solution candidate. Le modèle est utilisé dans le cadre synchrone du modèle maître-esclave. Une stratégie de sélection se décompose en trois composantes principales : l'agrégation des récompenses échangées, la technique d'apprentissage et la répartition des algorithmes sur les noeuds de calcul. Dans une dernière partie, nous étudions trois scénarios et nous donnons des clés de compréhension sur l'utilisation pertinente des stratégies de sélection adaptative par rapport aux stratégies naïves. Dans le cadre du modèle maître-esclave, nous étudions les différentes façons d'agréger les récompenses sur le noeud maître, la répartition des algorithmes d'optimisation sur les noeuds de calcul et le temps de communication. Cette thèse se termine par des perspectives pour le domaine de l'optimisation stochastique adaptative distribuée. / To solve discrete optimization problems of black box type, many stochastic algorithms such as evolutionary algorithms or metaheuristics exist and prove to be particularly effective according to the problem to be solved. Depending on the observed properties of the problem, choosing the most relevant algorithm is a difficult problem. In the original framework of parallel and distributed computing environments, we propose and analyze different adaptive optimization algorithm selection strategies. These selection strategies are based on reinforcement learning methods automatic, from the field of artificial intelligence, and on information sharing between computing nodes. We compare and analyze selection strategies in different situations. Two types of synchronous distributed computing environment are discussed : the island model and the master-slave model. On the set of nodes synchronously at each iteration, the adaptive selection strategy chooses an algorithm according to the state of the search for the solution. In the first part, two problems OneMax and NK, one unimodal and the other multimodal, are used as benchmarks for this work. Then, to better understand and improve the design of adaptive selection strategies, we propose a modeling of the optimization problem and its local search operator. In this modeling, an important characteristic is the average gain of an operator according to the fitness of the candidate solution. The model is used in the synchronous framework of the master-slave model. A selection strategy is broken down into three main components : the aggregation of the rewards exchanged, the learning scheme and the distribution of the algorithms on the computing nodes. In the final part, we study three scenarios, and we give keys to understanding the relevant use of adaptive selection strategies over naïve strategies. In the framework of the master-slave model, we study the different ways of aggregating the rewards on the master node, the distribution of the optimization algorithms of the nodes of computation and the time of communication. This thesis ends with perspectives in the field of distributed adaptive stochastic optimization.

An algorithm that is capable of estimating the parameters of non-stationary sinusoids in real-time lends application to various branches of engineering. Non-stationary sinusoids are sinusoidal signals with time-varying parameters. In this dissertation, a nonlinear filter is applied to power system applications to test its performance. The filter has a structure which renders it fully adaptive to tracking time variations in the parameters of the targeted sinusoid, including its phase and frequency. Mathematical properties of the differential equations which govern the proposed filter are presented. The performance of the proposed filter in the field of power systems is demonstrated with the aid of computer simulations and practical experimentations. The filter is applied to synchronous generator excitation control, voltage dip mitigation as well as the real-time estimation of symmetrical components. The parameter settings of the filter are tested and optimized for each of the applications. This dissertation demonstrates the simulation and experimental results of the filter when applied to the various power system applications. AFRIKAANS : 'n Filter wat bevoeglik is met die beraming van die parameters van beweeglike sinusoïdale in ware-tyd, kan bruikbaar aangewend word in verskeie takke van ingenieurswese. Beweeglike sinuskrommes is sinusoïdale seine met tyd-wisselende parameters. In hierdie verhandeling word `n nie-liniêre filter aangewend in verskeie kragstelseltoepassings om die werksverrigting van die filter te toets. Die filter het 'n struktuur wat dit toelaat om wisselende tydvariasies in die parameters van die teikensinusoïdaal op te spoor, insluitende die fase en frekwensie. Wiskundige eienskappe van die differensiaalvergelykings wat die voorgestelde filter beheer is ondersoek. Die werksverrigting van die voorgestelde filter in die veld van kragstelsels is gedemonstreer met die hulp van rekenaarsimulasies asook praktiese eksperimente. Die filter is toegepas tot opgewekte, sinkrone eksitasie-beheer, spanningsverlaging versagting, asook die ware tyd estimasie van simmetriese komponente. Die parameter verstellings van die filter is getoets en geoptimeer vir elk van die toepassings. Hierdie verhandeling demonstreer die simulering en eksperimentele resultate van die filter wat aangewend is vir verskeie kragstelseltoepassings. Copyright / Dissertation (MSc)--University of Pretoria, 2010. / Electrical, Electronic and Computer Engineering / unrestricted