O perfil de consumo de energia elétrica dos consumidores necessita ser conhecido em detalhe para muitos estudos, sejam eles técnicos ou comerciais. Esse conhecimento pode ser alcançado através da obtenção das curvas de carga de todos os clientes da empresa, porém, devido ao grande número de medidores necessários, essa prática é inviável. A alternativa utilizada atualmente nas revisões tarifárias do sistema elétrico brasileiro emprega a teoria de amostragem associada a técnicas de análise de dados. Após a obtenção das informações, são calculadas as tipologias de carga que representam cada cliente ou transformador, através de etapas de caracterização da carga. Os resultados obtidos permitem uma análise mais precisa do mercado de energia elétrica e, principalmente, o conhecimento da forma como cada classe de consumidores utiliza a rede. Este trabalho envolve parte do estudo mencionado sobre a análise dos dados coletados nas campanhas de medição, propondo e avaliando metodologias alternativas para duas etapas do processo de caracterização de tipologias de carga, a seleção de curvas típicas e a classificação de dados, adequadas às necessidades das revisões tarifárias e com base em métodos heurísticos e nas práticas do setor. Após o desenvolvimento e implementação das metodologias, foram realizados testes entre os processos propostos, comparando e avaliando suas particularidades para duas situações: a semelhança entre as tipologias encontradas para os transformadores e consumidores de um mesmo nível de tensão e o impacto nos custos marginais de capacidade. A análise das comparações realizadas permitiu a identificação dos impactos e características das metodologias desenvolvidas, para cada etapa estudada. / Knowing the way consumers use the energy is necessary for many studies, either commercial or technical. This knowledge can be reached by obtaining the load curves from all the customers of the company. However, given the great number of measurers necessary, this practice is not viable. The alternative used currently in tariff review in the Brazilian electrical system is based on the sampling theory associated with data analysis techniques. After obtaining the information, the load typologies that represent each transformer or customer are calculated through stages of load characterization. The results obtained allow a more precise analysis of the electric energy market and, specially, the knowledge of how each consumer class uses the electricity network. This research involves part of the previously mentioned study on the analysis of the data collected in the measurement campaigns, considering and evaluating alternative methodologies for two stages of the load typologies characterization process, the election of typical curves and the data classification, adjusted to the necessities of the tariff revisions and on the basis of heuristical methods and electricity sector practices. After the development and implementation of the methodologies, tests have been carried between the considered processes, comparing and evaluating their particularitities for two situations: the similarity between the typologies found for transformers and consumers on the same tension level and the impact in the marginal capacity costs. The analysis of the comparisons carried through allowed the identification of the impacts and characteristics of the developed methodologies, for each studied stage.

O perfil de consumo de energia elétrica dos consumidores necessita ser conhecido em detalhe para muitos estudos, sejam eles técnicos ou comerciais. Esse conhecimento pode ser alcançado através da obtenção das curvas de carga de todos os clientes da empresa, porém, devido ao grande número de medidores necessários, essa prática é inviável. A alternativa utilizada atualmente nas revisões tarifárias do sistema elétrico brasileiro emprega a teoria de amostragem associada a técnicas de análise de dados. Após a obtenção das informações, são calculadas as tipologias de carga que representam cada cliente ou transformador, através de etapas de caracterização da carga. Os resultados obtidos permitem uma análise mais precisa do mercado de energia elétrica e, principalmente, o conhecimento da forma como cada classe de consumidores utiliza a rede. Este trabalho envolve parte do estudo mencionado sobre a análise dos dados coletados nas campanhas de medição, propondo e avaliando metodologias alternativas para duas etapas do processo de caracterização de tipologias de carga, a seleção de curvas típicas e a classificação de dados, adequadas às necessidades das revisões tarifárias e com base em métodos heurísticos e nas práticas do setor. Após o desenvolvimento e implementação das metodologias, foram realizados testes entre os processos propostos, comparando e avaliando suas particularidades para duas situações: a semelhança entre as tipologias encontradas para os transformadores e consumidores de um mesmo nível de tensão e o impacto nos custos marginais de capacidade. A análise das comparações realizadas permitiu a identificação dos impactos e características das metodologias desenvolvidas, para cada etapa estudada. / Knowing the way consumers use the energy is necessary for many studies, either commercial or technical. This knowledge can be reached by obtaining the load curves from all the customers of the company. However, given the great number of measurers necessary, this practice is not viable. The alternative used currently in tariff review in the Brazilian electrical system is based on the sampling theory associated with data analysis techniques. After obtaining the information, the load typologies that represent each transformer or customer are calculated through stages of load characterization. The results obtained allow a more precise analysis of the electric energy market and, specially, the knowledge of how each consumer class uses the electricity network. This research involves part of the previously mentioned study on the analysis of the data collected in the measurement campaigns, considering and evaluating alternative methodologies for two stages of the load typologies characterization process, the election of typical curves and the data classification, adjusted to the necessities of the tariff revisions and on the basis of heuristical methods and electricity sector practices. After the development and implementation of the methodologies, tests have been carried between the considered processes, comparing and evaluating their particularitities for two situations: the similarity between the typologies found for transformers and consumers on the same tension level and the impact in the marginal capacity costs. The analysis of the comparisons carried through allowed the identification of the impacts and characteristics of the developed methodologies, for each studied stage.

In electric power distribution systems, the major determinant in electricity supply strategy is the quantity of demand. Customers need to be accurately represented using updated nodal load information as a requirement for efficient control and operation of the distribution network. In Distribution Load Estimation (DLE), two major categories of data are utilized: historical data and direct real-time measured data. In this thesis, a comprehensive survey on the state-of-the-art methods for estimating loads in distribution networks is presented. Then, a novel method for representing historical data in the form of Representative Load Curves (RLCs) for use in realtime DLE is also described. Adaptive Neuro-Fuzzy Inference Systems (ANFIS) is used in this regard to determine RLCs. An RLC is a curve that represents the behavior of the load during a specified time span; typically daily, weekly or monthly based on historical data. Although RLCs provide insight about the variation of load, it is not accurate enough for estimating real-time load. This therefore, should be used along with real-time measurements to estimate the load more accurately. It is notable that more accurate RLCs lead to better real-time load estimation in distribution networks. This thesis addresses the need to obtain accurate RLCs to assist in the decision-making process pertaining to Radial Distribution Networks (RDNs).This thesis proposes a method based on Adaptive Neuro-Fuzzy Inference Systems (ANFIS) architecture to estimate the RLCs for Distribution Networks. The performance of the method is demonstrated and simulated, on a test 11kV Radial Distribution Network using the MATLAB software. The Mean Absolute Percent Error (MAPE) criterion is used to justify the accuracy of the RLCs.

Nesta dissertação é proposto e avaliado um modelo de caracterização da carga para utilização no planejamento operacional dos sistemas de distribuição baseado na caracterização dos consumidores através de curvas típicas de carga. A identificação dos padrões de curvas típicas baseou-se na aplicação da rede neural tipo mapa auto organizável, sobre a grande massa de dados de medições de clientes da campanha de medidas realizada pela distribuidora no processo de revisão tarifária o que permitiu a identificação dos padrões de consumo de energia ativa e fator de potência para os consumidores ao longo do dia. O módulo de agrupamento baseado no mapa auto organizável associado a técnica clássica de agrupamento das k-médias mostrou-se uma ferramenta extremamente robusta e eficaz na identificação de padrões para grandes bases de dados. A comparação dos resultados das estimativas de carga para cerca de 200 alimentadores de distribuição medidos através do sistema SCADA complementa e valida a aplicação desta metodologia, que culmina com a proposição de um modelo de otimização, que com base nas medições, possibilita melhorias significativas na estimativa de carga dos alimentadores estudados. A metodologia proposta neste trabalho demonstra ser uma ferramenta eficaz para que a distribuidora de energia elétrica possa constantemente realimentar os dados sobre os hábitos de consumo de seus clientes, garantindo assim a manutenção de estimativas consistentes para o planejamento operacional de seu sistema de distribuição. / This dissertation presents and validate a load characterization framework for the operational planning of electric distribution networks based on characterization of customer typical load curves. The pattern recognition of typical load curves was based on the usage of self organizing maps, a type of neural network, over the huge database of customer field measurements performed by the electric energy utility for tariff review process, allowing the characterization of daily active energy consumption and power factor behaviors. The grouping module is based on self organizing map technique along with classic k-means technique, which proved to be an extremely effective tool for pattern recognition over large databases. The results of load estimation for 200 distribution feeders measured by SCADA system ensures the quality and accuracy of this framework which presents also an optimization model based on such measures, resulting significant improvement on load estimation for these feeders. This framework proves to be an effective tool for electric energy utilities for constant evaluation of customer’s power consumption behavior, allowing the maintenance of accurate estimations for operational planning of distribution networks.

O aumento da demanda por energia registrado nos últimos anos preocupa, pois a construção de novas fontes geradoras é barrada, muitas vezes, por restrições ambientais. Assim, o governo e as empresas de energias estão investindo em um melhor planejamento do sistema. No entanto, para haver uma proposta mais consistente para os consumidores residenciais se faz necessário conhecer melhor o perfil de cada consumidor, que é uma tarefa um tanto quanto difícil, visto que cada consumidor possui o livre arbítrio para consumir a energia de acordo com o conforto que ele deseja, de acordo com seus padrões econômicos e conforme aspectos culturais e sociais do ambiente em que ele vive. Neste contexto, a proposta desta tese foi analisar os impactos que os aspectos socioeconômicos tinham sobre o consumo de energia da classe residencial, sendo desenvolvido um algoritmo que gera curvas de carga virtuais baseadas em dados estatísticos do IBGE e da ANEEL. A partir dados de curvas virtuais, as ferramentas computacionais inteligentes, mais especificamente, as Redes Bayesianas e a Floresta de Caminhos Ótimos, foram treinadas com intuito de avaliar a possibilidade de criação de perfis e classificação dos consumidores e de suas características. Os resultados alcançados demonstram que a consideração dos aspectos socioeconômicos em avaliação de curvas de carga são pertinentes e que devem fazer parte do planejamento do sistema. Outra constatação é que as ferramentas computacionais inteligentes estudadas podem ser exploradas para auxiliar na previsão de consumo e criação de padrões e perfis dos consumidores. / The need for energy has increased in the past years, thus requiring the design of new power plants. However, the project of such new constructions has been considerably neglected, mainly due to environment constraints. However, the whole government and companies are now focusing on a better management of the national energy grid. Despite that new policy, there is a need for a better knowledge concerning the user\'s behavior in order to present proposals that really take into account the consumers, since each them has the freedom to use the energy the way he wants to, as well as according to his socioeconomic habits. In this context, this thesis proposes to analyze the socioeconomic impacts of the energy consumption concerning residential consumers, being also developed an algorithm that generates virtual load curves based on statistical data from both IBGE and ANEEL. With that data on hand, the intelligent tools, e.g., Bayesian Networks and Optimum-Path Forest, were trained aiming at evaluating the possibility to create profiles for the further identification of their classes according to that information. The results obtained highlighted the importance of the socioeconomic information when evaluating the load curves, which should be part of the whole system. Another conclusion concerns the intelligent tools, which can be further used for consumer forecasting, as well as to create patterns related to the consumers\' profiles.

O aumento da demanda por energia registrado nos últimos anos preocupa, pois a construção de novas fontes geradoras é barrada, muitas vezes, por restrições ambientais. Assim, o governo e as empresas de energias estão investindo em um melhor planejamento do sistema. No entanto, para haver uma proposta mais consistente para os consumidores residenciais se faz necessário conhecer melhor o perfil de cada consumidor, que é uma tarefa um tanto quanto difícil, visto que cada consumidor possui o livre arbítrio para consumir a energia de acordo com o conforto que ele deseja, de acordo com seus padrões econômicos e conforme aspectos culturais e sociais do ambiente em que ele vive. Neste contexto, a proposta desta tese foi analisar os impactos que os aspectos socioeconômicos tinham sobre o consumo de energia da classe residencial, sendo desenvolvido um algoritmo que gera curvas de carga virtuais baseadas em dados estatísticos do IBGE e da ANEEL. A partir dados de curvas virtuais, as ferramentas computacionais inteligentes, mais especificamente, as Redes Bayesianas e a Floresta de Caminhos Ótimos, foram treinadas com intuito de avaliar a possibilidade de criação de perfis e classificação dos consumidores e de suas características. Os resultados alcançados demonstram que a consideração dos aspectos socioeconômicos em avaliação de curvas de carga são pertinentes e que devem fazer parte do planejamento do sistema. Outra constatação é que as ferramentas computacionais inteligentes estudadas podem ser exploradas para auxiliar na previsão de consumo e criação de padrões e perfis dos consumidores. / The need for energy has increased in the past years, thus requiring the design of new power plants. However, the project of such new constructions has been considerably neglected, mainly due to environment constraints. However, the whole government and companies are now focusing on a better management of the national energy grid. Despite that new policy, there is a need for a better knowledge concerning the user\'s behavior in order to present proposals that really take into account the consumers, since each them has the freedom to use the energy the way he wants to, as well as according to his socioeconomic habits. In this context, this thesis proposes to analyze the socioeconomic impacts of the energy consumption concerning residential consumers, being also developed an algorithm that generates virtual load curves based on statistical data from both IBGE and ANEEL. With that data on hand, the intelligent tools, e.g., Bayesian Networks and Optimum-Path Forest, were trained aiming at evaluating the possibility to create profiles for the further identification of their classes according to that information. The results obtained highlighted the importance of the socioeconomic information when evaluating the load curves, which should be part of the whole system. Another conclusion concerns the intelligent tools, which can be further used for consumer forecasting, as well as to create patterns related to the consumers\' profiles.

Nesta dissertação é proposto e avaliado um modelo de caracterização da carga para utilização no planejamento operacional dos sistemas de distribuição baseado na caracterização dos consumidores através de curvas típicas de carga. A identificação dos padrões de curvas típicas baseou-se na aplicação da rede neural tipo mapa auto organizável, sobre a grande massa de dados de medições de clientes da campanha de medidas realizada pela distribuidora no processo de revisão tarifária o que permitiu a identificação dos padrões de consumo de energia ativa e fator de potência para os consumidores ao longo do dia. O módulo de agrupamento baseado no mapa auto organizável associado a técnica clássica de agrupamento das k-médias mostrou-se uma ferramenta extremamente robusta e eficaz na identificação de padrões para grandes bases de dados. A comparação dos resultados das estimativas de carga para cerca de 200 alimentadores de distribuição medidos através do sistema SCADA complementa e valida a aplicação desta metodologia, que culmina com a proposição de um modelo de otimização, que com base nas medições, possibilita melhorias significativas na estimativa de carga dos alimentadores estudados. A metodologia proposta neste trabalho demonstra ser uma ferramenta eficaz para que a distribuidora de energia elétrica possa constantemente realimentar os dados sobre os hábitos de consumo de seus clientes, garantindo assim a manutenção de estimativas consistentes para o planejamento operacional de seu sistema de distribuição. / This dissertation presents and validate a load characterization framework for the operational planning of electric distribution networks based on characterization of customer typical load curves. The pattern recognition of typical load curves was based on the usage of self organizing maps, a type of neural network, over the huge database of customer field measurements performed by the electric energy utility for tariff review process, allowing the characterization of daily active energy consumption and power factor behaviors. The grouping module is based on self organizing map technique along with classic k-means technique, which proved to be an extremely effective tool for pattern recognition over large databases. The results of load estimation for 200 distribution feeders measured by SCADA system ensures the quality and accuracy of this framework which presents also an optimization model based on such measures, resulting significant improvement on load estimation for these feeders. This framework proves to be an effective tool for electric energy utilities for constant evaluation of customer’s power consumption behavior, allowing the maintenance of accurate estimations for operational planning of distribution networks.

Les travaux de recherche présentés dans ce mémoire ont pour objet d’apporter une vision stratégique d’intégration des productions distribuées (PD) dans les réseaux électriques. Ces travaux concernent la localisation optimale du point de raccordement, le dimensionnement et le type de production dans l’objectif de maximiser les bénéfices de la PD et de minimiser les pertes dans les réseaux. Les travaux de cette thèse concernent également la prise en compte de la variabilité de la charge et de la production dans la planification et la gestion opérationnelle des réseaux électriques. Tout d’abord, des algorithmes ont été développés pour les études des flux de puissance dans les systèmes d'alimentation en utilisant la méthode du complément Schur et la méthode « Run Length Encoding ». Ensuite, les pertes ont été estimées dans le calcul de la production réelle en développant un modèle linéaire simple, efficace et flexible. Par la suite, des productions décentralisées connectées aux réseaux électriques ont été modélisées en utilisant une méthode qui fusionne les filtres de Kalman et la théorie des graphes dans le but d'estimer la taille optimale de la production décentralisée. Une méthode qui comporte deux étapes est proposée. Dans la première étape, la méthode graphique est utilisée pour générer la matrice incidente pour construire le modèle linéaire et dans la deuxième étape, un algorithme Kalman est appliqué pour obtenir la taille optimale de production décentralisée à chaque jeu de barres. Les défis de l'utilisation de productions décentralisées ont été abordés pour minimiser la fonction objective (pertes de puissance réelle) en tenant compte de la capacité des productions décentralisées, de la capacité de la ligne de transmission et des contraintes de profil de tension. L’algorithme génétique et de techniques d'optimisations comme la méthode de points intérieurs ont été proposés pour déterminer localement et globalement le dimensionnement optimal et l'emplacement optimal des productions décentralisées dans les réseaux électriques. Enfin, un modèle de charge active a été conçu pour étudier différents types de courbe de charge (résidentielle, commerciale et industrielle). Nous avons développé également des algorithmes de simulation pour étudier l'intégration des parcs éoliens dans les réseaux électriques. Nous avons conçu des méthodes analytiques pour sélectionner la taille et l’emplacement d’une ferme éolienne, basé sur la réduction des pertes de puissance active. Nous avons montré que les variations de la vitesse moyenne annuelle du vent pourraient avoir un effet important sur les calculs de pertes de puissance active. Les méthodes analytiques et les algorithmes de simulation ont été développés sous Matlab/Simulink. / The research presented in this thesis aims at providing a strategic vision for the integration of distributed generators (DGs) into grid networks. This work focuses the optimal location of the connection point, dimensioning and type of production in order to maximize the benefits of DGs and minimize power losses in the networks. The work also concerns the impact of the variability of the load and the production in the planning and the operational management of the networks. First, algorithms have been developed for power flow studies in power systems using the Schur complement method and the "Run Length Encoding" method. Then, losses were estimated in the calculation of power output by developing a simple, efficient and flexible linear model. Subsequently, decentralized outputs connected to the electrical networks were modeled using a method that merges Kalman filters and graph theory in order to estimate the optimal size of decentralized production. A method which consists of two steps is proposed. In the first step, the graphical method is used to generate the incident matrix to construct the linear model and in the second step a Kalman algorithm is applied to obtain the optimal decentralized production size for each busbar. The challenges of using decentralized production have been addressed to minimize the objective function (real power losses) by taking into account the capacity of the decentralized productions, transmission line capacity and voltage profile constraints. The genetic algorithms and optimization techniques such as the method of interior points have been proposed to determine locally and globally the optimal dimensioning and the optimal location of the decentralized productions in the electrical networks. Finally, an active load model was designed to study different types of load curves (residential, commercial and industrial). We have also developed simulation algorithms to study the integration of wind farms in power grids. We have designed analytical methods to select the size and location of a wind farm, based on the reduction of active power losses. We have shown that variations in the mean annual wind speed could have a significant effect on the calculations of active power losses. Analytical methods and simulation algorithms were developed under Matlab / Simulink.

Electricity is the most preferred source of energy, because of its quality and convenience of usage. It is probably one of the most vital infrastructural inputs for economic development of a country. Indeed it is the fulcrum which can leverage the future pace of growth and development. These reasons have made the electric power industry one of the fastest growing sectors in most developing countries and particularly in India. Therefore it is not surprising to observe the economic growth of a country being related to the increase in electricity consumption. In India, the growth rate of demand for power is generally higher than that of Gross Domestic Product (GDP). However, to achieve this kind of growth in electricity supply, the capital investments required are very huge. Even though the electricity sector generally gels a major share in the budgetary allocations in India, this is inadequate to add the required quantum of new generation capacity to keep pace with the increase in demand for electricity. Additional constraints like capital scarcity in the public sector, lack of enthusiasm among the private and foreign investors, and strong opposition from the environmentalists have further contributed to this slow pace of new generating capacity addition. This has resulted in severely constrained systems in India. The main focus of the present research work is on the development of an integrated approach for electricity planning using a mathematical modeling framework in (he context of resource constrained systems. There are very few attempts in the literature to integrate short, medium and long term issues in electricity planning. This is understandable from the point of view of unconstrained electricity systems where this type of integration is unnecessary since such systems have a luxury of surplus capacity to meet the current demand and capacity additions are required only for meeting predicted future increase in demand. However, in the case of constrained electricity systems, which are characterized by shortages, this kind of integration is very essential. These systems have to manage with inadequate capacity in the present, plan capacity additions to bridge the existing gap and to meet future increase in demand, and always explore the possibility of adding capacity with short gestation period. The integrated approach is expected to achieve effective supply-demand matching on a continuous basis encompassing both the short term and long term horizons. To achieve this, we have considered three alternatives- existing supply, new supply and non-supply (rationing) of electricity. The electricity system of the state of Karnataka, which is severely constrained by both limited capital and energy resources, has been selected for this purpose. As a first step, the supply and demand situation has been studied in the context of resource constraints. In terms of supply, both existing and future additions are studied in detail with respect to the potential created, generation types, import potential, technical constraints, energy and power shortages, planned and proposed capacity additions by both public and private sectors, etc. The demand patterns have been studied by introducing a new concept of "Representative Load Curves (RLCs)". These RLCs are used to model the temporal and structural variations in demand for electricity. Also, appropriate non-supply options (rationing measures) for effective management of shortages are identified. Incorporating this information, an integrated mathematical model, which is expected to generate a target plan for a detailed generation scheduling exercises and a requirement plan for a regular generation expansion planning, has been developed. The other important alternative "Demand-Side-Management (DSM)", which could be considered as an effective option to achieve efficient supply-demand matching has not been included in the present research work. The major reason for not including the DSM alternatives is due to the difficulty in integrating these in the modelling approach adopted here. In the present approach we have used typical daily load curves (RLCs) to represent the demand for electricity. These are aggregate load curves and do not contain any sector-wise or end-use-wisc details. On the other hand, DSM alternatives are end-use focused. To incorporate DSM alternatives, we should have information on end-usc-wisc power demand (kW or MW), savings potential, time-of-use, etc. For this purpose it may be required to have end-use-wisc daily load curves. This information is not available and a separate detailed survey may be required to generate these load curves. This, we felt, is out of the scope of this present research work and a separate study may be required to do this. Therefore, we restricted our focus to supply planning alone. A detailed literature review is conducted to understand different types of modeling approaches to electricity planning. For the present study, however, the review of literature has been restricted to the methods of generation expansion planning and scheduling. In doing so, we attempted to bring out the differences in various approaches in terms of solution methods adopted, alternatives included and modifications suggested. Also, we briefly reviewed the literature on models for power and energy rationing, because management of shortages is an important aspect of the present study. Subsequently, a separate section is devoted to present an overview of the non-supply of electricity and its economic impacts on the consumers. We found that the low reliability of the electrical system is an indicator of the existence of severe shortages of power and energy, which cause non-supply of electricity to the consumers. The overview also presented a discussion on reasons for non-supply of electricity, and the types of non-supply options the utilities adopt to over come these shortages. We also attempted to explain what we mean by non-supply of electricity, what are its cost implications, and the methods available in the literature to estimate these costs. The first objective of the research pertains to the development of a new approach to model the varying demand for electricity. Using the concept of Representative Load Curves (RLCs) we model the hourly demand for a period of four years, 1993-94, 1994-95, 1995-96 and 1996-97, to understand the demand patterns of both unconstrained and constrained years. Multiple discriminant analysis has been used to cluster the 365 load curves into nine RLCs for each of the four years. The results show that these RLCs adequately model the variations in demand and bring out the distinctions in the demand patterns existed during the unconstrained and constrained years. The demand analysis using RLCs helped to study the differences in demand patterns with and without constraints, impacts of constraints on preferred pattern of electricity consumption, success of non-supply options in both reducing the demand levels and greatly disturbing the electricity usage patterns. Multifactor ANOVA analyses are performed to quantify the statistical significance of the ability of the logically obtained factors in explaining the overall variations in demand. The results of the ANOVA analysis clearly showed that the considered factors accounted for maximum variations in demand at very high significance levels. It also brought out the significant influence of rationing measures in explaining the variations in demand during the constrained years. Concerning the second objective, we explained in detail, the development of an integrated mixed integer-programming model, which we felt is appropriate for planning in the case of resource constrained electricity systems. Two types of integrations are attempted (i) existing supply, non-supply and new supply options for dynamically matching supply and demand, (ii) operational and strategic planning in terms of providing target plans for the former and requirement plans for the latter. Broadly, the approach addresses the effective management of existing capacity, optimal rationing plan to effectively manage shortages and rationally decide on the new capacity additions both to bridge the existing gap between supply and demand, and to meet the future increases in demand. There is also an attempt to arrive at an optimal mix of public and private capacity additions for a given situation. Finally, it has been attempted to verify the possibility of integration of captive generation capacity with the grid. Further, we discussed in detail about the data required for the model implementation. The model is validated through the development of a number of scenarios for the state of Karnataka. The base case scenario analyses are carried out for both the unconstrained and constrained years to compare the optimal allocations with actual allocations that were observed, and to find out how sensitive are the results for any change in the values of various parameters. For the constrained years, a few more scenarios are used to compare the optimal practice of managing shortages with to what has been actually followed by the utility. The optimal allocations of the predicted demand to various existing supply and non-supply options clearly showed that the actual practice, reflected by the actual RLCs, are highly ad hoc and sub-optimal. The unit cost comparisons among different scenarios show that the least cost choice of options by the utility does not necessarily lead to good choices from the consumers’ perspective. Further, a number of future scenarios are developed to verify the ability of the model to achieve the overall objective of supply-demand matching both in the short and long term. For this purpose both the short horizon annual scenarios (1997-98 to 2000-01) and long horizon terminal year scenarios (2005-06 and 2010-11) are developed assuming capacity additions from only public sector. Overall, the results indicated that with marginal contributions from non-supply options and if the public sector generates enough resources to add the required capacity, optimal matching of supply and demand could be achieved. The scenario analyses also showed that it is more economical to have some level of planned rationing compared to having a more reliable system. The quantum of new capacity additions required and the level of investments associated with it clearly indicated the urgent need of private sector participation in capacity additions. Finally, we made an attempt to verify the applicability of the integrated model to analyse the implications of private sector participation in capacity additions. First, a number of scenarios are developed to study the optimal allocations of predicted hourly demand to private capacity under different situations. Secondly, the impacts of privatisation on the public utility and consumers are analysed. Both short term and long term scenarios are developed for this purpose. The results showed the advantage of marginal non-supply of electricity both in terms of achieving overall effective supply-demand matching and economic benefits that could be generated through cost savings. The results also showed the negative impacts of high guarantees offered to the private sector in terms of the opportunity costs of reduced utilization of both the existing and new public capacity. The estimates of unit cost of supply and effective cost of supply facilitated the relative comparison among various scenarios as well as finding out the merits and demerits of guarantees to private sector and non-supply of electricity. The unit cost estimates are also found to be useful in studying the relative increase in electricity prices for consumers on account of privatization, guarantees and reliable supply of electricity. Using the results of scenario analyses, likely generation expansion plans till the year 2010-11 are generated. The analyses have been useful in providing insights into fixing the availability and plant load factors for the private sector capacity. Based on the analysis, the recommended range for plant utilization factor is 72.88 - 80.57%. The estimated generation losses and the associated economic impacts of backing down of existing and new public capacity on account of guarantees offered to private sector are found to be significantly high. The analyses also showed that the backing down might take place mainly during nights and low demand periods of monsoon and winter seasons. Other impacts of privatization that studied are in terms of increased number of alternatives for the utility to buy electricity for distribution and the associated increase in its cost of purchase. Regarding the consumers, the major impact could be in terms of significant increase in expected tariffs. The major contributions of this thesis are summarized as follows: i. An integrated approach to electricity planning that is reported here, is unique in the sense that it considers options available under various alternatives, namely, existing supply, non-supply and new supply. This approach is most suited for severely constrained systems having to manage with both energy and capital resource shortages. ii. The integration of operational and strategic planning with coherent target plans for the former and requirement plans for the latter bridges the prevailing gap in electricity planning approaches. iii. The concept of Representative Load Curves (RLCs), which is introduced here, captures the hourly, daily and seasonal variations in demand. Together, all the RLCs developed for a given year are expected to model the hourly demand patterns of that year. These RLCs are useful for planning in resource constrained electricity systems and in situations where it is required to know the time variations in demand (e.g. supply-demand matching, seasonal scheduling of hydro plants and maintenance scheduling). RLCs are also useful in identifying the factors influencing variations in demand. This approach will overcome the limitations of current method of representation in the form of static and aggregate annual load duration curves. iv. A new term, "non-supply of electricity" has been introduced in this thesis. A brief overview of non-supply presented here includes reasons for non-supply, type of non-supply, methods to estimate cost of non-supply and factors influencing these estimates. v. The integrated mixed integer programming model developed in the study has been demonstrated as a planning tool for- • Optimal hourly and seasonal scheduling of various existing supply, non-supply and new supply options • Estimation of supply shortages on a representative hourly basis using the information on resource constraints • Effectively planning non-supply of electricity through appropriate power/energy rationing methods • Estimation of the need for the new capacity additions both to bridge the existing gap and to take care of increase in future demand levels • Optimal filling of gaps between demand and supply on a representative hourly basis through new supply of electricity • Optimally arriving at the judicious mix of public and private capacity additions • Studying the impacts of private capacity on the existing and new public sector capacity, and on the consumers • Optimally verifying the feasibility of integrating the captive generation with the total system vi. The demand analysis using RLCs helped to bring out the differences in demand patterns with and without constraints, impacts of constraints on preferred pattern of electricity consumption, success of non-supply options in both reducing the demand levels and greatly disturbing the electricity usage patterns. Multifactor ANOVA analyses results showed that the logically obtained factors accounted for maximum variations in demand at very high significance levels. vii. A comparison of optimal (represented by optimal predicted RLCs) and actual (reflected by actual RLCs) practices facilitated by the model showed that the actual practice during constrained years is highly ad hoc and sub-optimal. viii. The results of the scenario analyses showed that it is more economical to have some amount of planned rationing compared to having a more reliable system, which does not allow non-supply of electricity. ix. The scenarios, which analysed the impacts of high guarantees offered to the private sector, showed the negative impacts of these in terms of reduced utilization of both the existing and new public capacity. x. Generation expansion plans till the year 2010-11 are developed using the results of various kinds of scenario analyses. Two groups of year-wise generation expansion plans are generated, one with only public sector capacity additions and the other with private sector participation. xi. The impacts of privatization of capacity additions are studied from the point of view of the utility and consumers in terms of expected increase in cost of purchase of electricity and tariffs. xii. The analyses are also made for developing some insights into fixing the availability and plant load factors for the private capacity. Based on the analysis, the recommended range for plant utilization factor is 72.88 - 80.57%. We believe that the integrated approach presented and the results obtained in this thesis would help utilities (both suppliers and distributors of electricity) and governments in making rational choices in the context of resource constrained systems. The results reported here may also be used towards rationalization of Government policies vis-a-vis tariff structures in the supply of electricity, planning new generation capacity additions and effective rationing of electricity. It is also hoped that the fresh approach adopted in this thesis would attract further investigations in future research on resource constrained systems.

Este trabalho tem por objetivo propor uma nova metodologia para o cálculo das perdas por segmento do sistema de distribuição. As perdas técnicas são agrupadas nos seguintes segmentos: rede secundária, transformador de distribuição, rede primária e subestação de distribuição. Desenvolveu-se uma metodologia destinada ao cálculo das perdas de forma hierárquica: por exemplo, selecionada uma subestação específica, são calculadas as perdas na subestação e em seus componentes a jusante (redes primárias, transformadores de distribuição, redes secundárias). As perdas, inicialmente, são obtidas por meio de cálculo elétrico para os segmentos envolvidos, com a utilização dos parâmetros da rede, com os dados de faturamento e as curvas de carga típicas por classe de consumidor e seus tipos de atividade. Com os resultados desses cálculos, treinam-se redes neurais que irão calcular as perdas em sistemas genéricos utilizando os parâmetros e topologia do segmento e as curvas típicas de cargas dos consumidores e a energia mensal consumida. O trabalho apresenta um exemplo de aplicação, em sistema de distribuição existente, mostrando os resultados obtidos, e termina apresentando as principais vantagens da metodologia. Finalmente, os resultados obtidos com a nova metodologia são comparados com os resultados obtidos por métodos analíticos de cálculo intensivo. / In this work, a new methodology for the calculation of the energy technical losses in a distribution system, is presented. The proposed approach regards the segmentation of the distribution system, thus, the losses will be obtained for segments such as: the secondary network, distribution transformer, primary network and distribution substation. It was developed a computational system aimed to the calculation of the technical losses within specific distribution networks and usable in a microcomputer. Such a calculation is done in a hierarchical way. For instance, once selected a specific substation it is calculated the losses within the substation and in all the above cited components existing downstream the substation. The energy technical losses are calculated for each segment involved in the distribution system. This is done by using the network\'s recorded data, the energy consumption data and the typical load curves by class of consumer and type of activity developed. The outcome of these calculations are then used to train the neural networks, which in turn will calculate the losses in generic distribution systems where characteristics such as the circuit parameters and topology, the consumer\'s load curves and the monthly energy consumed, are known. By using the energy data available in the supplying points, the total energy billed per month as well as the loss indexes per segment, it will be obtained the total amount of the energy losses in each segment of the system. Likewise, this procedure will enable an evaluation of the non technical losses. The results of a case study related to an existing distribution system and the main advantages of the proposed methodology, are also presented herein. Finally, the results obtained with the new methodology are compared with those obtained through analytical methods.