In a more competitive environment, load forecast serves two different applications. First, load forecast results can be used by the retailers of power to study their opportunities and plan their business strategies. Second, accurate projections of load are useful for T&D operators in performing system operation and expansion studies. Several key elements in their market and system planning studies have strong location factors that the spatial load forecast can address. In this dissertation, a package that integrates a Geographic Information System (GIS) used for automatic mapping and facility management (AM/FM) and a spatial load forecast module is presented. The interface functions and the procedure of the fuzzy logic based spatial load forecast module are described. Simulation studies are performed on a metropolitan area of Kaohsiung, Taiwan. The conventional fuzzy modeling has a drawback in that the fuzzy rules or the fuzzy membership functions are determined by trial and error. In this dissertation an automatic model identification procedure is proposed to construct the fuzzy model for short-term load forecast. In this method an analysis of variance is used to identify the influential variables on the system load. To setup the fuzzy rules, a cluster estimation method is adopted to determine the number of rules and the membership functions of variables involved in the premises of the rules. A recursive least square method is then used to determine the coefficients in the conclusion parts of the rules. None of these steps involves nonlinear optimization and all steps have well-bounded computation time.

The forecasting of Eskom transmission electrical network demands is a complex task. The lack of historical data on some of the network components complicates this task even further. In this dissertation a model is suggested which will address all the requirements of the transmission system expansion engineers in terms of future loads and market trends. Suggestions are made with respect to ways of overcoming the lack of historical data, especially on the point loads, which is a key factor in modelling the electrical networks. A brief overview of the transmission electrical network layout is included to provide a better understanding of what is required from such a forecast. Lastly, some theory on multiple regression, neural networks and qualitative forecasting techniques is included, which will be of value for further model developments. / Computing / M. Sc. (Operations Research)

The forecasting of Eskom transmission electrical network demands is a complex task. The lack of historical data on some of the network components complicates this task even further. In this dissertation a model is suggested which will address all the requirements of the transmission system expansion engineers in terms of future loads and market trends. Suggestions are made with respect to ways of overcoming the lack of historical data, especially on the point loads, which is a key factor in modelling the electrical networks. A brief overview of the transmission electrical network layout is included to provide a better understanding of what is required from such a forecast. Lastly, some theory on multiple regression, neural networks and qualitative forecasting techniques is included, which will be of value for further model developments. / Computing / M. Sc. (Operations Research)

Electric power systems are experiencing dramatic changes with respect to structure, operation and regulation and are facing increasing pressure due to environmental and societal constraints. Bulk electric system reliability is an important consideration in power system planning, design and operation particularly in the new competitive environment. A wide range of methods have been developed to perform bulk electric system reliability evaluation. Theoretically, sequential Monte Carlo simulation can include all aspects and contingencies in a power system and can be used to produce an informative set of reliability indices. It has become a practical and viable tool for large system reliability assessment technique due to the development of computing power and is used in the studies described in this thesis. The well-being approach used in this research provides the opportunity to integrate an accepted deterministic criterion into a probabilistic framework. This research work includes the investigation of important factors that impact bulk electric system adequacy evaluation and security constrained adequacy assessment using the well-being analysis framework.<p> Load forecast uncertainty is an important consideration in an electrical power system. This research includes load forecast uncertainty considerations in bulk electric system reliability assessment and the effects on system, load point and well-being indices and reliability index probability distributions are examined. There has been increasing worldwide interest in the utilization of wind power as a renewable energy source over the last two decades due to enhanced public awareness of the environment. Increasing penetration of wind power has significant impacts on power system reliability, and security analyses become more uncertain due to the unpredictable nature of wind power. The effects of wind power additions in generating and bulk electric system reliability assessment considering site wind speed correlations and the interactive effects of wind power and load forecast uncertainty on system reliability are examined. The concept of the security cost associated with operating in the marginal state in the well-being framework is incorporated in the economic analyses associated with system expansion planning including wind power and load forecast uncertainty. Overall reliability cost/worth analyses including security cost concepts are applied to select an optimal wind power injection strategy in a bulk electric system. The effects of the various demand side management measures on system reliability are illustrated using the system, load point, and well-being indices, and the reliability index probability distributions. The reliability effects of demand side management procedures in a bulk electric system including wind power and load forecast uncertainty considerations are also investigated. The system reliability effects due to specific demand side management programs are quantified and examined in terms of their reliability benefits.

Electric power systems are experiencing dramatic changes with respect to structure, operation and regulation and are facing increasing pressure due to environmental and societal constraints. Bulk electric system reliability is an important consideration in power system planning, design and operation particularly in the new competitive environment. A wide range of methods have been developed to perform bulk electric system reliability evaluation. Theoretically, sequential Monte Carlo simulation can include all aspects and contingencies in a power system and can be used to produce an informative set of reliability indices. It has become a practical and viable tool for large system reliability assessment technique due to the development of computing power and is used in the studies described in this thesis. The well-being approach used in this research provides the opportunity to integrate an accepted deterministic criterion into a probabilistic framework. This research work includes the investigation of important factors that impact bulk electric system adequacy evaluation and security constrained adequacy assessment using the well-being analysis framework.<p> Load forecast uncertainty is an important consideration in an electrical power system. This research includes load forecast uncertainty considerations in bulk electric system reliability assessment and the effects on system, load point and well-being indices and reliability index probability distributions are examined. There has been increasing worldwide interest in the utilization of wind power as a renewable energy source over the last two decades due to enhanced public awareness of the environment. Increasing penetration of wind power has significant impacts on power system reliability, and security analyses become more uncertain due to the unpredictable nature of wind power. The effects of wind power additions in generating and bulk electric system reliability assessment considering site wind speed correlations and the interactive effects of wind power and load forecast uncertainty on system reliability are examined. The concept of the security cost associated with operating in the marginal state in the well-being framework is incorporated in the economic analyses associated with system expansion planning including wind power and load forecast uncertainty. Overall reliability cost/worth analyses including security cost concepts are applied to select an optimal wind power injection strategy in a bulk electric system. The effects of the various demand side management measures on system reliability are illustrated using the system, load point, and well-being indices, and the reliability index probability distributions. The reliability effects of demand side management procedures in a bulk electric system including wind power and load forecast uncertainty considerations are also investigated. The system reliability effects due to specific demand side management programs are quantified and examined in terms of their reliability benefits.

abstract: On-line dynamic security assessment (DSA) analysis has been developed and applied in several power dispatching control centers. Existing applications of DSA systems are limited by the assumption of the present system operating conditions and computational speeds. To overcome these obstacles, this research developed a novel two-stage DSA system to provide periodic security prediction in real time. The major contribution of this research is to develop an open source on-line DSA system incorporated with Phasor Measurement Unit (PMU) data and forecast load. The pre-fault prediction of the system can provide more accurate assessment of the system and minimize the disadvantage of a low computational speed of time domain simulation. This Thesis describes the development of the novel two-stage on-line DSA scheme using phasor measurement and load forecasting data. The computational scheme of the new system determines the steady state stability and identifies endangerments in a small time frame near real time. The new on-line DSA system will periodically examine system status and predict system endangerments in the near future every 30 minutes. System real-time operating conditions will be determined by state estimation using phasor measurement data. The assessment of transient stability is carried out by running the time-domain simulation using a forecast working point as the initial condition. The forecast operating point is calculated by DC optimal power flow based on forecast load. / Dissertation/Thesis / Masters Thesis Electrical Engineering 2017

Prior to the year 2007, Eskom Distribution followed a method of load forecasting (now referred to as legacy method in this report) that was based on collecting customer applications, historical load trending, and relied on the planner’s knowledge of the area to a large extent. It was based in a conventional Microsoft Excel spreadsheet. On seeking to improve its load forecasting approach, the utility adopted a technique that was based on spatial forecasting. This new technique was called a geographical based load forecasting (GLF) technique which was performed by using a custom based tool, called PowerGLF. The aim of this research was to assess any improvements (or lack thereof) that were brought about by adopting the GLF method as compared to the legacy method that was used previously. The hypothesis to be tested was declared as: “The use of the GLF method that was introduced to Eskom Distribution Planning brings about the improvement on the planning process of infrastructure that is adequate, reliable and economic, when compared to the legacy method that was used before it.” To carry out this assessment, a case study method was followed. Real network studies that were compiled in 2006 and 2007 were used. These network studies were based on GLF method and the legacy method. The load forecasts from the case studies were evaluated on forecast accuracy, how they influenced the planning of adequate, reliable and economic (ARE) network infrastructure and their impact on the procurement and construction of the network infrastructure (which represent the actual utility expenditure on infrastructure). The statistical comparative analysis was done. The research results revealed that the legacy method was more accurate than the GLF method in both the case studies that were evaluated. However, regarding the ability of a load forecast method to support the planning process, the GLF method showed to be supporting the planning of adequate, reliable and economic infrastructure better than the legacy method. It was found that the forecast error for the GLF and legacy method do not affect the utility infrastructure procurement and construction. Based on the test results, the study reached a conclusion that the use of the GLF method that was introduced to Eskom Distribution Planning brings about the improvement in the planning process of infrastructure that is adequate, economic and reliable when compared to the legacy method that was used before it. The author wishes to express that the results of this study must not be taken as a generic conclusive finding regarding the evaluated load forecasting methods; they are applicable to the tested case studies. To get to a general conclusive result, more case studies would need to be carried out where clear and consistent evidence on performance of these load forecasting methods will be seen. The findings of this study can be used as part of a larger sample if such a larger population of case studies was to be evaluated. The methodology followed in this research can be repeated and followed when similar assessments are done in future.

Orientador: Anna Diva Plasencia Lotufo / Banca: Percival Bueno Araújo / Banca: Walmir de Freitas Filho / Resumo: Neste trabalho, dá-se ênfase à previsão de carga multinodal, também conhecida como previsão de carga por barramento. Para realizar esta demanda, há necessidade de dispor de uma técnica que proporcione a precisão desejada, seja confiável e de baixo tempo de processamento. O conhecimento prévio das cargas locais é de extrema importância para o planejamento e operação dos sistemas de energia elétrica. Para realizar a previsão de carga multinodal foram empregadas duas metodologias, uma que prevê as cargas individualmente e outra que utiliza as previsões dos fatores de participação e a previsão de carga global. O principal objetivo deste trabalho é elaborar um modelo de previsor de carga de curto prazo, genérico e que pode ser aplicado na previsão de carga multinodal. Para tanto, utilizou-se redes neurais de regressão generalizada (GRNN), cujas entradas são compostas de variáveis exógenas globais e de cargas locais, sem a necessidade da inclusão de variáveis exógenas locais. Ainda, projetou-se uma nova arquitetura de rede neural artificial, baseada na GRNN, além de propor um procedimento para a redução do número de entradas da GRNN e um filtro para o pré-processamento do banco de dados de treinamento. Os dados, para testar as metodologias e as redes neurais artificiais, são referentes a um subsistema de distribuição de energia elétrica da Nova Zelândia composto por nove subestações / Abstract: In this work, it is emphasized the multi-nodal load forecast, also known as bus load forecast. To perform this demand, there it is necessary a technique that is precise, trustable and has a short-time processing. The previous knowledge of the local loads is of extreme importance to the planning and operation of the electrical power and energy systems. To perform the multi-nodal load forecast is employed two different methodologies, one that forecast the loads individually and another that uses the participation factors forecasts and the global load forecast. The main objective of this work is to elaborate a generic model of a short-term load forecaster, which can be applied to the multi-nodal load forecast. For this, it was used general regression neural networks (GRNN), with inputs based on external global factors and local loads, without the need of external local factors. Still, it was developed a new architecture of an artificial neural network based on a GRNN and proposed a procedure to reduce the number of input variables of the GRNN and a filter for preprocessing the training data. The dataset, to test the methodologies and the artificial neural networks, refers to a New Zealand electrical distribution subsystem composed of nine substations / Mestre

In Electric Power Systems understand the future behavior of electric loads is crucial to make a decision. The long, medium and short term load forecast is essential power systems. Considering the gradual transformation of the traditional distribution systems to smart grids, where the electric system automation and online communication are effective, the forecast in very short term gets new challenges. Based on these facts, the objective of this thesis is to identify the most appropriate methods to accomplish these forecasts to contribute to decision-making in distribution systems operation. Various techniques of forecasting and simulations in different methods were studied, in order to identify which of these offers the best results regarding demand the forecast in the very short term. The quantities considered to make predictions and, which have more relevance to the horizon under study are electrical and climate. The methods used in the simulations were the Artificial Neural Networks (ANN) type recurrent Elman e NARX and Neurofuzzy. / Em Sistemas Elétricos de Potência conhecer o comportamento futuro das cargas elétricas é de fundamental importância para tomada de decisões. A previsão de cargas elétricas é essencial nos horizontes de longo, médio, curto e curtíssimo prazo. Tendo em vista a gradual transformação dos sistemas de distribuição tradicionais para sistemas inteligentes de distribuição, onde a automação do sistema elétrico e a comunicação online estejam efetivas, a previsão no curtíssimo prazo ganha novos desafios. Com base nesses fatos esta dissertação busca identificar os métodos mais adequados para realizar essas previsões de forma a contribuir com a tomada de decisões na operação dos sistemas de distribuição de energia elétrica. Foram estudadas várias técnicas de previsão e realizadas simulações em diferentes métodos de forma a identificar qual desses apresenta melhor resultado com relação à previsão de demanda no curtíssimo prazo. As grandezas consideradas para realizar as previsões e que apresentam maior relevância para o horizonte em estudo são elétricas e climáticas. Os métodos utilizados nas simulações foram as Redes Neurais Artificiais (RNAs) do tipo recorrente Elman e NARX e Neurofuzzy.

Toppeffekter används av elnätsbolag för att dimensionera elnätet, vilket blir allt viktigare för varje år. Fler och fler invånare och företag ökar sin elkonsumtion och förväntar sig en driftsäker och stabil elförsörjning. Det finns två vanliga metoder att beräkna toppeffekter. Första sättet är Velanders formel som är en enkel metod för att uppskatta toppeffekter. Velanders formel behöver bara årsenergi och vetskap om kundkategori med tillhörande Velanderkonstanter för beräkning av uppskattad toppeffekt. Sedan finns den mer komplexa typkurvemetoden som behöver flera olika parametrar, t.ex. graddagtal, dygnsmedeltemperatur, gränssannolikhet och kundkategori. Detta examensarbete undersöker en enkel metod för att ta fram konstanter till Velanders formel för beräkning av toppeffekter. Detta genomfördes med hjälp av regressionsanalys av historiska elanvändningsdata från Mälarenergi Elnät AB:s (MEE) kunder från 12 olika kundkategorier. Detta på grund av att MEE önskade att utveckla en metod för att ta fram konstanter till Velanders formel baserad på historiska elanvändningsdata. Metoden för att ta fram konstanter till Velanders formel går ut på att med hjälp av MATLAB utföra en regressionsanalys på simulerade kundgrupper skapade från timvis historiska elanvändningsdata. En kurva baserad på Velanders formel tas sedan fram som beskriver den övre gränsen till toppeffekterna för de simulerade kundgrupperna. Från kurvan fås sedan de Velanderkonstanter som söks. Resultaten av den undersökta metoden presenteras i form av grafer och tabeller för tre utvalda kundkategorier. Alla kategorier och deras resultat finns som bilagor till rapporten. Valideringen av resultaten och metoden gjordes med hjälp av korsvalidering och jämförelse mot heterogena simulerade kundgrupper. Känslighetsanalysen visar att den undersökta metoden var känslig för flera faktorer såsom kategorisering av kunder, tidsspann för historiska elanvändningsdata, antal simulerade kundgrupper och kundantal. Med tillräcklig dimensionering av dessa faktorer bedömdes metoden vara användbar. Resultaten visade på att de framtagna Velanderkonstanterna gav en god uppskattning av toppeffekter för de kundkategorier som undersökts. Jämförelse av de uppskattade toppeffekterna och de observerade visade på att det fanns en viss differens mellan dem. Detta var dock förväntat eftersom de uppskattade toppeffekterna ska avspegla den övre toppeffektsgränsen. / This degree project has examined a simple method aiming to obtain coefficients for Velanders formula which purpose is to calculate peak loads. This was done by using regression analysis on historical data on consumption of electricity from 12 different customer categories acquired from Mälarenergi Elnät AB (MEE). The reason being that MEE wanted to examine a method which could obtain coefficients for Velanders formula based on hourly historical electricity consumption data. The method for obtaining Velander coefficients uses MATLAB to do regression analysis on simulated customer groups, created from hourly historical electricity consumption data. The Velander coefficients are then obtained from a regression curve based on Velanders formula. Results from the evaluation of the method is presented with the help of plots and tables for three chosen customer categories. Validation of the method was done by cross-validation and comparison against heterogeneous customer groups. Sensitivity analysis showed the examined method to be sensitive to several factors such as categorization of customers, the timespan of historical electricity consumption data, the number of simulated customer groups that were used and how many customers a category contained. By dimensioning these factors carefully, the method examined was assessed to be viable. The results indicated that the obtained Velander coefficients gave a good estimation of the peak loads for the chosen customer categories. Comparison between the estimated and observed peak loads indicated that there was a certain difference between them. This was to be expected since the estimated peak loads are to reflect the upper peak load limit.