Problems in the public policy decision-making environments are typically complex and continuously evolve. In a resource-constrained environment, several alternatives, criteria, and conflicting objectives must be considered. As a result, solutions to these types of problems cannot be modelled solely using single-criteria techniques. It has been observed that most techniques used to shape energy policy and planning either produce sub-optimal solutions or use strong assumptions about the preferences of decision-maker(s). This difficulty creates a compelling need to develop novel techniques that can handle several alternatives, multiple criteria and conflicting objectives to support public sector decision-making processes. First, the study presents a novel scenario-based multi-objective optimisation framework based on the augmented Chebychev goal programming (GP) technique linked to a value function for analysing a decision environment underlying energy choice among low-income households in isolated rural areas and informal urban settlements in South Africa. The framework developed includes a multi-objective optimisation technique that produced an approximation of a Pareto front linked to an a priori aggregation function and a value function to select the best alternatives. Second, the study used this model to demonstrate the benefits of applying the framework to a previously unknown subject in public policy: a dynamic multi-technology decision problem under uncertainty involving multiple stakeholders and conflicting objectives. The results obtained suggest that while it is cost-optimal to pursue electrification in conjunction with other short-term augmentation solutions to meet South Africa's universal electrification target, sustainable energy access rates among low-income households can be achieved by increasing the share of clean energy generation technologies in the energy mix. This study, therefore, challenges the South African government's position on pro-poor energy policies and an emphasis on grid-based electrification to increase energy access. Instead, the study calls for a portfolio-based intervention. The study advances interventions based on micro-grid electrification made up of solar photovoltaics (PV), solar with storage, combined cycle gas turbine (CCGT) and wind technologies combined with either bioethanol fuel or liquid petroleum gas (LPG). The study has demonstrated that the framework developed can benefit public sector decision-makers in providing a balanced regime of technical, financial, social, environmental, public health, political and economic aspects in the decision-making process for planning energy supply interventions for low-income households. The framework can be adapted to a wide range of energy access combinatorial problems and in countries grappling with similar energy access challenges.

Data/information fusion is an integral component of many existing and emerging applications; e.g., remote sensing, smart cars, Internet of Things (IoT), and Big Data, to name a few. While fusion aims to achieve better results than what any one individual input can provide, often the challenge is to determine the underlying mathematics for aggregation suitable for an application. In this dissertation, I focus on the following three aspects of aggregation: (i) efficient data-driven learning and optimization, (ii) extensions and new aggregation methods, and (iii) feature and decision level fusion for machine learning with applications to signal and image processing. The Choquet integral (ChI), a powerful nonlinear aggregation operator, is a parametric way (with respect to the fuzzy measure (FM)) to generate a wealth of aggregation operators. The FM has 2N variables and N(2N − 1) constraints for N inputs. As a result, learning the ChI parameters from data quickly becomes impractical for most applications. Herein, I propose a scalable learning procedure (which is linear with respect to training sample size) for the ChI that identifies and optimizes only data-supported variables. As such, the computational complexity of the learning algorithm is proportional to the complexity of the solver used. This method also includes an imputation framework to obtain scalar values for data-unsupported (aka missing) variables and a compression algorithm (lossy or losselss) of the learned variables. I also propose a genetic algorithm (GA) to optimize the ChI for non-convex, multi-modal, and/or analytical objective functions. This algorithm introduces two operators that automatically preserve the constraints; therefore there is no need to explicitly enforce the constraints as is required by traditional GA algorithms. In addition, this algorithm provides an efficient representation of the search space with the minimal set of vertices. Furthermore, I study different strategies for extending the fuzzy integral for missing data and I propose a GOAL programming framework to aggregate inputs from heterogeneous sources for the ChI learning. Last, my work in remote sensing involves visual clustering based band group selection and Lp-norm multiple kernel learning based feature level fusion in hyperspectral image processing to enhance pixel level classification.

Made available in DSpace on 2016-06-02T19:50:05Z (GMT). No. of bitstreams: 1 2520.pdf: 1207796 bytes, checksum: d7395bbee81744b10cc9081ce73234c4 (MD5) Previous issue date: 2009-06-22 / This work aim at developing models using linear programming, goal programming and robust optimization to support decision making in the frozen concentrated orange juice planning process. The proposed model includes orange harvesting plan, which takes into account oranges maturation curves. This fact leads to a model that incorporates a large portion of the supply chain involved in the frozen concentrated orange juice sector. Another point to highlight is the inclusion of the blending process of different types of juices to match ratio specification of the product. This study uses orange acidity to calculate ratio specification of the product. This study also explores the importance of data uncertainty incorporation to the aggregate production planning for this business and evaluate results from different approaches of robust optimization to this problem. This author is not aware of previous work in the literature with such approach to the orange juice industry. The problem modeling uses blending problem concepts and production planning with multiple products, stages and periods concepts as well. To solve the linear programming, goal programming and robust optimization models, an algebraic modeling language and a state of art optimization solver of mathematical programming problems is used. A case study was developed in an orange juice company located in the São Paulo State. This company has many facilities and a worldwide distribution system, similar to other companies in this sector. The results show that the proposed approach can be used in real situations. / O objetivo deste trabalho é desenvolver modelos de programação linear, programação por metas e otimização robusta para apoiar decisões no processo de planejamento agregado da produção de suco concentrado congelado de laranja. A modelagem proposta incorpora o planejamento de colheita da laranja, levando-se em consideração as curvas de maturação das laranjas. Esse fato conduz a um modelo que incorpora grande parte da cadeia de suprimento envolvida no setor de produção de suco concentrado congelado de laranja. Outro ponto a destacar é a consideração do processo de mistura de diferentes tipos de sucos para a obtenção da especificação de ratio do produto acabado. No caso desse estudo, utiliza-se a acidez da laranja como base de cálculo para a especificação de ratio do produto acabado. Este estudo também explora a importância da incorporação de incerteza a determinados parâmetros envolvidos no processo de planejamento de produção nesse setor e analisa os resultados das diferentes abordagens de otimização robusta para o problema. Sendo que, este autor desconhece trabalhos anteriores na literatura com esta abordagem para a indústria de suco de laranja. A modelagem do problema utiliza conceitos de problemas de mistura e planejamento de produção com múltiplos produtos, estágios e períodos. Para resolver os modelos de programação linear, programação por metas e otimização robusta, utilizou-se uma linguagem de modelagem algébrica e um aplicativo de última geração de solução de problemas de programação matemática. Um estudo de caso foi realizado em uma empresa de suco de laranja localizada no Estado de São Paulo, envolvendo várias plantas e com uma rede de distribuição internacional com características típicas de outras empresas do setor. Os resultados indicam que a abordagem aqui proposta pode ser aplicada em situações reais.