A presente tese tem como objetivo propor um modelo de previsão para estimar o lucro médio esperado na concessão de crédito para pessoas físicas em empresas comerciais, obtendo assim uma medida monetária para dar suporte à tomada de decisão. O modelo proposto foi desenvolvido em três grandes etapas: 1) pré-processamento; 2) modelos de classificação; e 3) modelo de previsão do risco monetário. A primeira etapa inclui três passos: (i) delimitação da população, (ii) seleção da amostra, e (iii) análise preliminar. Na segunda etapa mais dois passos são necessários: (i) construção dos modelos, e (ii) qualidade dos modelos. Por fim, a última etapa trata das definições para construção do modelo de previsão do risco monetário propriamente dito, que utilizou os seguintes métodos: (i) ensemble, (ii) hybrid, e (iii) regressão linear múltipla. A exequibilidade do modelo proposto foi testada em dados reais de concessão de crédito. São avaliados os resultados de utilização do modelo de previsão, de forma a verificar o potencial aumento nos ganhos a partir da concessão do crédito, comparando quatro cenários: (i) sem utilizar nenhum modelo de previsão de risco de crédito; (ii) utilizando o modelo de classificação obtido com a regressão logística; (iii) utilizando o modelo de classificação obtido com a rede neural; e (iv) utilizando o modelo proposto para previsão do risco monetário. O modelo construído demonstrou resultados promissores na previsão do lucro médio esperado, apresentando um aumento estimado de 94,97% em comparação com o cenário sem uso de modelo de previsão, e um aumento de 26,08% quando comparado com o cenário de uso do modelo de classificação obtido com regressão logística. Uma análise de sensibilidade dos resultados com variações na margem de lucro por transação também foi realizada, evidenciando sua robustez. Nesse sentido, o modelo proposto se mostra efetivo como ferramenta de apoio para gestão no processo de decisão de concessão de crédito. / This thesis aims to propose a forecasting model to estimate the expected average profit in lending to individuals in commercial companies, thus obtaining a monetary measure to support decision making. The proposed model was developed in three major stages: 1) preprocessing, 2) classification models, and 3) model to forecast the currency risk. The first stage includes three steps: (i) delimitation of the population, (ii) sample selection, and (iii) preliminary analysis. In the second stage two more steps are necessary: (i) construction of models, and (ii) quality of the models. Finally, the last stage is regarding to the definitions for the construction of model prediction of the currency risk itself, which used the following methods: (i) ensemble, (ii) hybrid, and (iii) multiple linear regressions. The feasibility of the proposed model was tested on real data of grant credit. Results are evaluated using the prediction model in order to verify the potential increase in profits from the grant credit, comparing four scenarios: (i) without using any prevision model of credit risk, (ii) using the classification model obtained by logistic regression, (iii) using the classification model obtained with the neural network, and (iv) using the model to forecast the currency risk. The constructed model showed promising results in predicting the expected average profits, with an estimated increase of 94.97% compared to the scenario without the use of forecasting model, and an increase of 26.08% compared with the scenario of the classification model obtained by logistic regression. A sensitivity analysis of the results with variations in the profit margin per transaction was also performed, demonstrating its robustness. Accordingly, the proposed model proved effective as a support tool for management in the decision to grant credit.

Objectives. In this study we review previous attempts in forecasting country seaborne container throughput, analyze them and then classify in form of table to provide a concrete base for researchers in this field. Another aim of this study is to provide a Decision Support System (DSS) to assist experts in port management and forecast their country seaborne container demand. It will lead to reasonable decisions so as to provide sufficient supply which handles containers demand. This DSS, is a global forecasting model which can be applied to every country, independently of their specific parameters. Methods. In theoretical phase a number of scientific databases such as: Google Scholar, ACM, SCOPUS, IEEE, SpringerLink and some other are used to collect previous studies. After review and analysis, selected papers are classified in a form of table to provide a complete resource for us as well as future researchers in this field. In order to provide appropriate model, we combine System Dynamics modeling with Genetic Programming to provide an accurate and reliable model. This model is the result of the analysis of previous studies and applied in this study for the first time. Results. Our final model was applied to two cases (Sweden and China) and provides provided reliable results for both countries. To analyze the uncertain variables in the model, Monte Carlo simulation was used to assess the sensitivity of our model. In order to compare with other methods, we conducted a case study with Artificial Neural Network (ANN) and compared the results of our model and ANN. The results show the disadvantages of statistical methods to system dynamics. Additionally to compare with other attempts, our model was confronted with another study which provided a model for Finland. By comparing and considering their advantages and disadvantages we found out that our simplified model could be applied as a global model to other countries. Conclusions. We conclude that our model is an appropriate DSS to assist experts, forecast their country throughput and make appropriate decisions so as to invest, extending their ports in right time. The application of Genetic Programming in our model provides accurate mathematical equations for the influencing variables which even may not need to calibrate the model. It is a global model which can be applied to different countries but still requires more experiments to prove this claim. / This research aims to provide a decision support system to assist experts in port management to forecast future trends of cargo demand. By forecasting the future demand, decision makers will be able to decide on sufficient supply. For example, in case of necessity, based on forecasting results, the infrastructure can be expanded and also the capacity of ports can be managed. This will help not only to invest in right place and time, but also to balance their demand between ports in a country. The majority of previous researches considered only statistical methods to forecast the future cargo demand. Sometimes the previous research studies applied only one method and then compared it with others and provided advantages and disadvantages of each methods. In some other cases the previous research studies were combining statistical methods to analyze linear and non linear behavior of influencing parameters in cargo demand to conduct a forecast later and its future demand. All the research studies that were collected were analyzed and then classified into a table (c.f., chapter 4). Recently, some studies applied system dynamics to analyze all interactions in the system and forecast the future cargo demand like (Ruutu 2008) and (E. Suryani et al. 2012). In this research we combined system dynamics with genetic programming to benefit from the advantages of each method. By using the system dynamics modeling technique, we defined all influencing parameters and their interactions in the system. By use of genetic programming we provided accurate equations between different parameters and country demand. In Genetic Programming, all the equations can be fitted into data. At last, even we do not need to calibrate the equations to fit into historical data. This will provide a reliable model to forecast demand and align the supply with it. To validate our model, it is applied on two different countries and the results from the analysis indicate that the simplified model provides an acceptable model and it follows the trend of historical data. To compare our model with previous statistical methods the results of our model in Sweden and China were compared with the result of neural network in another case study with the same data. To compare our model with other similar studies, it turned out that it is closely related to the model for Finland. After comparison and analysis of their advantages and disadvantages, we concluded that our simplified model can apply as a global model to other countries, but it needs to prove with a number of different case studies (different countries with different situations). To analyze the uncertain variables, which can affect the model, we used Monte Carlo simulation. It assesses the sensitivity of our model to changes in input variables. The final model is applicable to every country, but it needs to apply the local econometric parameters, which affect the country throughput. By considering the share of each port in total demand of the country, we can apply the model to each port and forecast the future trends in order to find the right date to invest and extend the capacity to handle Demand.

Background   By the beginning of the 20th century, the men’s suit entered the menswear market as one the most important fashion garments everdevised. At the same time, fashion became mainly a female engagement, resulting in an under representation of men’s fashion through out the past decade. Relating to the textile and apparel industry, fashion forecasting has become an increasingly important business activity. But the nature of fashion forecasting and the historical neglecting of the men’s suit has created complications when performing this activity. Purpose   The purpose of this thesis is to examine the men’s suit and its development from the given starting point in the 20th century until today, in order to derive a fashion forecasting model suggesting its development by 2029. Design/methodology/approach   This thesis uses an abductive research approach and qualitative multi-methods to answer the research questions. The usage of an intermediate research project answers the first research question. The second research question is answered through the synthesis ofa literature study and semi-structured interviews. The third research question is answered through the derived forecasting model, accomplished through theory matching. Findings   By carrying out a historical investigation of the men’s suit, and then applying this to the derived forecasting model, the men’s suit is expected to be found in both single- and double-breast styles. The suit will have classical features represented through the length, canvas structure, and shoulder construction. Originality/value   This paper carries out a historical investigation of the men’s suit never been done before. It introduces an evaluation framework to categorise and classify the men’s suit, as well as a forecasting model followed by an actual fashion forecast.

This thesis develops a method for automatically constructing, visualizing and describing a large class of models, useful for forecasting and finding structure in domains such as time series, geological formations, and physical dynamics. These models, based on Gaussian processes, can capture many types of statistical structure, such as periodicity, changepoints, additivity, and symmetries. Such structure can be encoded through kernels, which have historically been hand-chosen by experts. We show how to automate this task, creating a system that explores an open-ended space of models and reports the structures discovered. To automatically construct Gaussian process models, we search over sums and products of kernels, maximizing the approximate marginal likelihood. We show how any model in this class can be automatically decomposed into qualitatively different parts, and how each component can be visualized and described through text. We combine these results into a procedure that, given a dataset, automatically constructs a model along with a detailed report containing plots and generated text that illustrate the structure discovered in the data. The introductory chapters contain a tutorial showing how to express many types of structure through kernels, and how adding and multiplying different kernels combines their properties. Examples also show how symmetric kernels can produce priors over topological manifolds such as cylinders, toruses, and Möbius strips, as well as their higher-dimensional generalizations. This thesis also explores several extensions to Gaussian process models. First, building on existing work that relates Gaussian processes and neural nets, we analyze natural extensions of these models to deep kernels and deep Gaussian processes. Second, we examine additive Gaussian processes, showing their relation to the regularization method of dropout. Third, we combine Gaussian processes with the Dirichlet process to produce the warped mixture model: a Bayesian clustering model having nonparametric cluster shapes, and a corresponding latent space in which each cluster has an interpretable parametric form.

<p>Denna uppsats undersöker vilka faktorer som påverkar förändringar i elspotpriset på Nord Pool. Avsikten är att resultatet skall ligga till grund för en prisuppskattningsmodell för Lunds Energikoncernen AB. Faktorerna bestämdes genom en förstudie där viktig litteratur om elmarknaden studerades samt samtal med Lunds Energikoncernen AB. De faktorer som undersöks i denna uppsats är priset på utsläppsrätter, nettoexport till Tysk-land, temperatur, nederbörd, priset på kol och villaolja samt konjunkturutveckling i Sve-rige.</p><p>Undersökningen av faktorerna bestod av en multipel regressionsanalys med undersökta faktorer som oberoende variabler och elspotpriset på Nord Pool som den beroende va-riabeln. Faktorerna blev indelade i två grupper dagsgruppen och månadsgruppen, grunden till uppdelningen är som namnen antyder att statistiken var observerad dygnsvis och må-nadsvis. I månadsgruppen ingick nettoexport till Tyskland, priset på kol, villaolja samt konjunktur och ur denna grupp visade sig endast nettoexport till Tyskland ha statistisk signifikans.</p><p>I dagsgruppen ingick de faktorer som oftast omnämns i litteraturen som prispåverkande, nämligen temperatur, nederbörd och priset på utsläppsrätter. Dock visade sig nederbörd inte ha någon statistisk signifikant påverkan på elpriset varvid ett nytt test på ett nytt sta-tistiskt underlag gjordes för nederbörden vilket gav samma resultat, vilket var förvånan-de. Både temperatur och priset på utsläppsrätter visade sig dock ha statistisk signifikans och detta intygades genom ytterligare test.</p><p>Härefter gjordes en regressionsanalys med de faktorer visat sig ha statistisk signifikans som oberoende variabler, det vill säga nettoexport till Tyskland, temperatur och priset på utsläppsrätter gentemot elpriset som beroende variabel. Denna enkla prognosmodell kunde förklara så mycket som 70 procent av förändringarna i elpriset.</p><p>Slutligen diskuteras prognosmodellen av författarna, en brist är bland annat att den inte kan förutse hastiga förändringar i priset och att den behöver kalibreras om när den nya handelsperioden för utsläppsrätter sätter i gång 2008. Dock gav analysen positiva signa-ler om att det kan vara möjligt att basera en prismodell på el med de faktorer som har störst inverkan på den dyraste produktionsteknologi som oftast används i elproduktio-nen, då elmarknaden i praktiken tillämpar marginalprissättning.</p> / <p>This thesis examines which factors that drive changes in the electricity spot price on the Nordic energy exchange Nord Pool. The intention with this thesis is to support Lunds Energikoncernen AB to create a pricing model. The factors were determined though a pre-study in which important literature on the electricity market were studied and inter-views with Lunds Energikoncernen AB. The examined factors in this thesis are; the price of emission allowances, net export to Germany, temperature, precipitation, the prices of coal and burning oil and Sweden’s business cycle.</p><p>The factor study was a multiple regression analysis with the above factors as independ-ent variables and the spot price of electricity on Nord Pool as the dependent. The fac-tors were divided in two groups, the day group and the month group, the two groups were decided due to statistical observations. The factors from the former group had daily ob-served data and the latter monthly data. The month group included net exchange with Germany, oil and coal prices and the business cycle which are measured in GDP. In the month group only the net exchange with Germany had statistical significance and was used in further studies.</p><p>In the day group the factors that are mostly discussed in the literature to impact on the electricity price namely, temperature, precipitation, and the price of emission allowances. As it, some what unexpected, turned out the precipitation did not have a statistical affect on the electricity price. The authors chose to carry out another analysis with precipita-tion from another area, neither this result had statistical significance. However, both the temperature and the price of emission allowances did have a statistical significant effect on the electricity price, the result were verified through one more round of analysis.</p><p>After the two initial analyses, a regression analysis with the three factors that had statis-tical significance and the electricity price were used in a final analysis. The factors in-cluded in this regression were net exchange with Germany, temperature and the price of emission allowances. This, somewhat, simple forecasting model explained as much as 70 percent of the changes in the electricity spot price.</p><p>At last the forecasting model were discussed by the authors who identified two major weaknesses, first the model may not explain sudden changes in the electricity price, and second the model has to be re-calibrated when the next trading period for emission al-lowances starts in early 2008. However the analysis did indicate that it might be possible to base an electricity price forecasting model on the factors that affects the most expen-sive production facility that are used to create energy, since the electricity market prac-tice marginal pricing.</p>

Eastern Kentucky is a 35-county region that is a part of the Cumberland Plateau of the Appalachian Mountains. With mountaintop removal and associated land cover change (LCC) (primarily deforestation), it is hypothesized that there would be changes in various atmospheric boundary layer parameters and precipitation. In this research, we have conducted sensitivity experiments of atmospheric response of a significant flash flood-producing rainfall event by modifying land cover and topography. These reflect recent LCC, including mountaintop removal (MTR). We have used the Weather Research and Forecasting (WRF) model for this purpose. The study found changes in amount, location, and timing of precipitation. LCC also modified various surface fluxes, moist static energy, planetary boundary layer height, and local-scale circulation wind circulation. The key findings were the modification in fluxes and precipitation totals. With respect to sensible heat flux (H), there was an increase to bare soil (post-MTR) in comparison to pre-MTR conditions (increased elevation with no altered land cover). Allowing for growth of vegetation, the grass simulation resulted in a decrease in H. H increased when permitting the growth of forest land cover (LC) but not to the degree of bare soil. In regards to latent heat flux (LE), there was a dramatic decrease transitioning from pre-MTR to post-MTR simulations. Then with the subsequent grass and forest simulations, there was an increase in LE comparable to the pre-MTR simulation. Under pre-MTR conditions, the total precipitation was at its highest level overall. Then with the simulated loss of vegetation and elevation, there was a dramatic decrease in precipitation. With the grass LC, the precipitation increased in all areas of interest. Then forest LC was simulated allowing overall slightly higher precipitation than grass.

A demanda e, consequentemente, a geração de energia elétrica são questões de suma importância para o desenvolvimento econômico e social dos países. Modelos para previsão destes parâmetros no longo e médio prazo são empregados com a finalidade de antever possíveis cenários e propor estratégias para a realização de um planejamento energético adequado. Neste contexto, o presente estudo tem como objetivo realizar a previsão da geração de energia elétrica no estado do Rio Grande do Sul (RS) em um horizonte de médio prazo (um ano), utilizando Redes Neurais Artificiais (RNA’s) do tipo feedforward com algoritmo de aprendizado supervisionado backpropagation. Para o desenvolvimento deste trabalho elaborou-se um script para executar as simulações necessárias, as quais foram realizadas através do software Matlab®. As variáveis de influência selecionadas como entradas do modelo de previsão referem-se à economia (estadual e nacional), ao balanço de energia elétrica e à meteorologia do estado, durante o período de janeiro de 2009 a março de 2016. Para realizar o treinamento da rede neural, adicionou-se a matriz de entrada este conjunto de dados, com frequência mensal, referentes a janeiro de 2009 a março de 2015 e para previsão foram inseridos dados de abril de 2015 a março de 2016. Por fim, depois de realizada a simulação completa da RNA, comparou-se o resultado observado da geração de energia elétrica do estado com o obtido através do modelo de previsão, indicando um erro percentual absoluto médio (MAPE) de 5,86% e um desvio absoluto médio (MAD) de 134,15 MW médio. Os resultados obtidos neste trabalho mostram-se promissores, além de semelhantes aos encontrados na literatura, demonstrando assim confiabilidade e eficácia do método empregado. / The demand and, consequently, the generation of electric power are very important issues for social and economic development of countries. Models to forecast these parameters in long and medium terms are used to anticipate possible sceneries and propose strategies for the energy planning of countries. In this context, the present study aims to forecast the generation of electric energy in Rio Grande do Sul State (RS) in a medium-term horizon (one year) using, Artificial Neural Networks (ANNs) of the feedforward type with algorithm of supervised learning backpropagation. For the development of this work, a script was elaborated in order to execute the necessary simulations, which were carried out through Matlab® software. The selected variables of influence as inputs of forecasting model refer to economy (State and National), to the electric energy balance and to the meteorology State, during the period from January, 2009 to March, 2016. In order to train the neural network, this data set was added to the entrance matrix, with monthly frequency, from January, 2009 to March, 2015 and for prediction, data were inserted from April, 2015 to March, 2016. Finally, after RNA complete simulation, the observed result of the electric power generation of the State was compared with the one obtained through the prediction model, indicating a mean absolute percent error (MAPE) of 5.86% and a mean absolute deviation (MAD) of 134.15 average MW. The obtained results in this work are promising, besides; they are similar to those found in literature, in this way demonstrating the reliability and efficacy of the using method.

Denna uppsats undersöker vilka faktorer som påverkar förändringar i elspotpriset på Nord Pool. Avsikten är att resultatet skall ligga till grund för en prisuppskattningsmodell för Lunds Energikoncernen AB. Faktorerna bestämdes genom en förstudie där viktig litteratur om elmarknaden studerades samt samtal med Lunds Energikoncernen AB. De faktorer som undersöks i denna uppsats är priset på utsläppsrätter, nettoexport till Tysk-land, temperatur, nederbörd, priset på kol och villaolja samt konjunkturutveckling i Sve-rige. Undersökningen av faktorerna bestod av en multipel regressionsanalys med undersökta faktorer som oberoende variabler och elspotpriset på Nord Pool som den beroende va-riabeln. Faktorerna blev indelade i två grupper dagsgruppen och månadsgruppen, grunden till uppdelningen är som namnen antyder att statistiken var observerad dygnsvis och må-nadsvis. I månadsgruppen ingick nettoexport till Tyskland, priset på kol, villaolja samt konjunktur och ur denna grupp visade sig endast nettoexport till Tyskland ha statistisk signifikans. I dagsgruppen ingick de faktorer som oftast omnämns i litteraturen som prispåverkande, nämligen temperatur, nederbörd och priset på utsläppsrätter. Dock visade sig nederbörd inte ha någon statistisk signifikant påverkan på elpriset varvid ett nytt test på ett nytt sta-tistiskt underlag gjordes för nederbörden vilket gav samma resultat, vilket var förvånan-de. Både temperatur och priset på utsläppsrätter visade sig dock ha statistisk signifikans och detta intygades genom ytterligare test. Härefter gjordes en regressionsanalys med de faktorer visat sig ha statistisk signifikans som oberoende variabler, det vill säga nettoexport till Tyskland, temperatur och priset på utsläppsrätter gentemot elpriset som beroende variabel. Denna enkla prognosmodell kunde förklara så mycket som 70 procent av förändringarna i elpriset. Slutligen diskuteras prognosmodellen av författarna, en brist är bland annat att den inte kan förutse hastiga förändringar i priset och att den behöver kalibreras om när den nya handelsperioden för utsläppsrätter sätter i gång 2008. Dock gav analysen positiva signa-ler om att det kan vara möjligt att basera en prismodell på el med de faktorer som har störst inverkan på den dyraste produktionsteknologi som oftast används i elproduktio-nen, då elmarknaden i praktiken tillämpar marginalprissättning. / This thesis examines which factors that drive changes in the electricity spot price on the Nordic energy exchange Nord Pool. The intention with this thesis is to support Lunds Energikoncernen AB to create a pricing model. The factors were determined though a pre-study in which important literature on the electricity market were studied and inter-views with Lunds Energikoncernen AB. The examined factors in this thesis are; the price of emission allowances, net export to Germany, temperature, precipitation, the prices of coal and burning oil and Sweden’s business cycle. The factor study was a multiple regression analysis with the above factors as independ-ent variables and the spot price of electricity on Nord Pool as the dependent. The fac-tors were divided in two groups, the day group and the month group, the two groups were decided due to statistical observations. The factors from the former group had daily ob-served data and the latter monthly data. The month group included net exchange with Germany, oil and coal prices and the business cycle which are measured in GDP. In the month group only the net exchange with Germany had statistical significance and was used in further studies. In the day group the factors that are mostly discussed in the literature to impact on the electricity price namely, temperature, precipitation, and the price of emission allowances. As it, some what unexpected, turned out the precipitation did not have a statistical affect on the electricity price. The authors chose to carry out another analysis with precipita-tion from another area, neither this result had statistical significance. However, both the temperature and the price of emission allowances did have a statistical significant effect on the electricity price, the result were verified through one more round of analysis. After the two initial analyses, a regression analysis with the three factors that had statis-tical significance and the electricity price were used in a final analysis. The factors in-cluded in this regression were net exchange with Germany, temperature and the price of emission allowances. This, somewhat, simple forecasting model explained as much as 70 percent of the changes in the electricity spot price. At last the forecasting model were discussed by the authors who identified two major weaknesses, first the model may not explain sudden changes in the electricity price, and second the model has to be re-calibrated when the next trading period for emission al-lowances starts in early 2008. However the analysis did indicate that it might be possible to base an electricity price forecasting model on the factors that affects the most expen-sive production facility that are used to create energy, since the electricity market prac-tice marginal pricing.

A demanda e, consequentemente, a geração de energia elétrica são questões de suma importância para o desenvolvimento econômico e social dos países. Modelos para previsão destes parâmetros no longo e médio prazo são empregados com a finalidade de antever possíveis cenários e propor estratégias para a realização de um planejamento energético adequado. Neste contexto, o presente estudo tem como objetivo realizar a previsão da geração de energia elétrica no estado do Rio Grande do Sul (RS) em um horizonte de médio prazo (um ano), utilizando Redes Neurais Artificiais (RNA’s) do tipo feedforward com algoritmo de aprendizado supervisionado backpropagation. Para o desenvolvimento deste trabalho elaborou-se um script para executar as simulações necessárias, as quais foram realizadas através do software Matlab®. As variáveis de influência selecionadas como entradas do modelo de previsão referem-se à economia (estadual e nacional), ao balanço de energia elétrica e à meteorologia do estado, durante o período de janeiro de 2009 a março de 2016. Para realizar o treinamento da rede neural, adicionou-se a matriz de entrada este conjunto de dados, com frequência mensal, referentes a janeiro de 2009 a março de 2015 e para previsão foram inseridos dados de abril de 2015 a março de 2016. Por fim, depois de realizada a simulação completa da RNA, comparou-se o resultado observado da geração de energia elétrica do estado com o obtido através do modelo de previsão, indicando um erro percentual absoluto médio (MAPE) de 5,86% e um desvio absoluto médio (MAD) de 134,15 MW médio. Os resultados obtidos neste trabalho mostram-se promissores, além de semelhantes aos encontrados na literatura, demonstrando assim confiabilidade e eficácia do método empregado. / The demand and, consequently, the generation of electric power are very important issues for social and economic development of countries. Models to forecast these parameters in long and medium terms are used to anticipate possible sceneries and propose strategies for the energy planning of countries. In this context, the present study aims to forecast the generation of electric energy in Rio Grande do Sul State (RS) in a medium-term horizon (one year) using, Artificial Neural Networks (ANNs) of the feedforward type with algorithm of supervised learning backpropagation. For the development of this work, a script was elaborated in order to execute the necessary simulations, which were carried out through Matlab® software. The selected variables of influence as inputs of forecasting model refer to economy (State and National), to the electric energy balance and to the meteorology State, during the period from January, 2009 to March, 2016. In order to train the neural network, this data set was added to the entrance matrix, with monthly frequency, from January, 2009 to March, 2015 and for prediction, data were inserted from April, 2015 to March, 2016. Finally, after RNA complete simulation, the observed result of the electric power generation of the State was compared with the one obtained through the prediction model, indicating a mean absolute percent error (MAPE) of 5.86% and a mean absolute deviation (MAD) of 134.15 average MW. The obtained results in this work are promising, besides; they are similar to those found in literature, in this way demonstrating the reliability and efficacy of the using method.

A demanda e, consequentemente, a geração de energia elétrica são questões de suma importância para o desenvolvimento econômico e social dos países. Modelos para previsão destes parâmetros no longo e médio prazo são empregados com a finalidade de antever possíveis cenários e propor estratégias para a realização de um planejamento energético adequado. Neste contexto, o presente estudo tem como objetivo realizar a previsão da geração de energia elétrica no estado do Rio Grande do Sul (RS) em um horizonte de médio prazo (um ano), utilizando Redes Neurais Artificiais (RNA’s) do tipo feedforward com algoritmo de aprendizado supervisionado backpropagation. Para o desenvolvimento deste trabalho elaborou-se um script para executar as simulações necessárias, as quais foram realizadas através do software Matlab®. As variáveis de influência selecionadas como entradas do modelo de previsão referem-se à economia (estadual e nacional), ao balanço de energia elétrica e à meteorologia do estado, durante o período de janeiro de 2009 a março de 2016. Para realizar o treinamento da rede neural, adicionou-se a matriz de entrada este conjunto de dados, com frequência mensal, referentes a janeiro de 2009 a março de 2015 e para previsão foram inseridos dados de abril de 2015 a março de 2016. Por fim, depois de realizada a simulação completa da RNA, comparou-se o resultado observado da geração de energia elétrica do estado com o obtido através do modelo de previsão, indicando um erro percentual absoluto médio (MAPE) de 5,86% e um desvio absoluto médio (MAD) de 134,15 MW médio. Os resultados obtidos neste trabalho mostram-se promissores, além de semelhantes aos encontrados na literatura, demonstrando assim confiabilidade e eficácia do método empregado. / The demand and, consequently, the generation of electric power are very important issues for social and economic development of countries. Models to forecast these parameters in long and medium terms are used to anticipate possible sceneries and propose strategies for the energy planning of countries. In this context, the present study aims to forecast the generation of electric energy in Rio Grande do Sul State (RS) in a medium-term horizon (one year) using, Artificial Neural Networks (ANNs) of the feedforward type with algorithm of supervised learning backpropagation. For the development of this work, a script was elaborated in order to execute the necessary simulations, which were carried out through Matlab® software. The selected variables of influence as inputs of forecasting model refer to economy (State and National), to the electric energy balance and to the meteorology State, during the period from January, 2009 to March, 2016. In order to train the neural network, this data set was added to the entrance matrix, with monthly frequency, from January, 2009 to March, 2015 and for prediction, data were inserted from April, 2015 to March, 2016. Finally, after RNA complete simulation, the observed result of the electric power generation of the State was compared with the one obtained through the prediction model, indicating a mean absolute percent error (MAPE) of 5.86% and a mean absolute deviation (MAD) of 134.15 average MW. The obtained results in this work are promising, besides; they are similar to those found in literature, in this way demonstrating the reliability and efficacy of the using method.