L'effacement des consommations électriques a été identifié comme l'une des solutions pour pallier les problèmes liés aux pics de consommation électrique, à l’intermittence des énergies renouvelables et à la congestion des réseaux. Ces travaux de recherche s’intéressent à l’intégration de la consommation industrielle en tant que ressources effaçables à court terme dans le contexte de la réserve rapide du mécanisme d’ajustement français. Parmi les différents secteurs, le secteur industriel présente un intérêt particulier en raison de l’importance de sa consommation. Afin d'intégrer ce type de consommation dans l’équilibre offre-demande, il est nécessaire de prévoir le comportement de ces consommations à court terme ainsi que d’évaluer la fiabilité de ces prévisions. Ainsi, différentes méthodes de prévision à très court-terme adaptées aux données et au contexte ont été déployées sur différents consommations disponibles à deux niveaux d’agrégation différents : site et usage industriel. Des indicateurs de performance adaptés aux contraintes opérationnelles, appelés "taux de fiabilité", sont proposés et calculés pour évaluer la performance des méthodes de prévision. Ce taux de fiabilité est estimé pour différentes heures de la journée pour les différents sites et usages industriels étudiés. Les taux de fiabilité estimés permettent d'évaluer le risque pour une consommation spécifique (au niveau du site ou au niveau de chaque usage industriel) de ne pas respecter des contraintes opérationnelles imposées à chaque instant de simulation. / Demand response has been identified as one of the solutions to overcome the problems associated with peaks in electricity consumption, intermittency of renewable energy and network congestion. This thesis focuses on the integration of industrial electricity consumptions as short-term demand response resources in the context of a supply-demand balancing mechanism in France. Among the various sectors, industrial electricity consumptions are of particular interest because of their orders of magnitude. In order to integrate these consumptions to the supply-demand balance, it is necessary forecast their behavior in the short term and to evaluate the reliability of these forecasts. Thus, different short-term load forecasting methods adapted to the data and to the operational context are implemented on different sets of industrial consumptions data at two different consumption levels: the industrial site and the end-point equipment consumption. Performance indicators adapted to operational constraints, called "trust factors" are proposed and calculated to evaluate the performance of the forecasting methods. These trust factors are estimated for different hours of the day for all the different studied industrial sites and workshops. The estimated trust are used to assess the risks for a specific consumption to not to respect the operational constraints at a moment a forecast is simulated. Demand response is considered to become one of the elements to be implemented in order to achieve a successful energy transition through a more flexible power system.

The aim of this study is to develop novel forecasting methodologies. The applications of our proposed models lie in two different areas: macroeconomics and energy. Though we consider two very different applications, the common underlying theme of this thesis is to develop novel methodologies that are not only accurate, but are also parsimonious. For macroeconomic time series, we focus on generating forecasts for the US Gross National Product (GNP). The contribution of our study on macroeconomic forecasting lies in proposing a novel nonlinear and nonparametric method, called weighted random analogue prediction (WRAP) method. The out-of-sample forecasting ability of WRAP is evaluated by employing a range of different performance scores, which measure its accuracy in generating both point and density forecasts. We show that WRAP outperforms some of the most commonly used models for forecasting the GNP time series. For energy, we focus on two different applications: (1) Generating accurate short-term forecasts for the total electricity demand (load) for Great Britain. (2) Modelling Irish electricity smart meter data (consumption) for both residential consumers and small and medium-sized enterprises (SMEs), using methods based on kernel density (KD) and conditional kernel density (CKD) estimation. To model load, we propose methods based on a commonly used statistical dimension reduction technique, called singular value decomposition (SVD). Specifically, we propose two novel methods, namely, discount weighted (DW) intraday and DW intraweek SVD-based exponential smoothing methods. We show that the proposed methods are competitive with some of the most commonly used models for load forecasting, and also lead to a substantial reduction in the dimension of the model. The load time series exhibits a prominent intraday, intraweek and intrayear seasonality. However, most existing studies accommodate the ‘double seasonality’ while modelling short-term load, focussing only on the intraday and intraweek seasonal effects. The methods considered in this study accommodate the ‘triple seasonality’ in load, by capturing not only intraday and intraweek seasonal cycles, but also intrayear seasonality. For modelling load, we also propose a novel rule-based approach, with emphasis on special days. The load observed on special days, e.g. public holidays, is substantially lower compared to load observed on normal working days. Special day effects have often been ignored during the modelling process, which leads to large forecast errors on special days, and also on normal working days that lie in the vicinity of special days. The contribution of this study lies in adapting some of the most commonly used seasonal methods to model load for both normal and special days in a coherent and unified framework, using a rule-based approach. We show that the post-sample error across special days for the rule-based methods are less than half, compared to their original counterparts that ignore special day effects. For modelling electricity smart meter data, we investigate a range of different methods based on KD and CKD estimation. Over the coming decade, electricity smart meters are scheduled to replace the conventional electronic meters, in both US and Europe. Future estimates of consumption can help the consumer identify and reduce excess consumption, while such estimates can help the supplier devise innovative tariff strategies. To the best of our knowledge, there are no existing studies which focus on generating density forecasts of electricity consumption from smart meter data. In this study, we evaluate the density, quantile and point forecast accuracy of different methods across one thousand consumption time series, recorded from both residential consumers and SMEs. We show that the KD and CKD methods accommodate the seasonality in consumption, and correctly distinguish weekdays from weekends. For each application, our comprehensive empirical comparison of the existing and proposed methods was undertaken using multiple performance scores. The results show strong potential for the models proposed in this thesis.

Submitted by Geandra Rodrigues (geandrar@gmail.com) on 2018-01-17T11:13:15Z No. of bitstreams: 1 samuelbelinidefilippo.pdf: 2610291 bytes, checksum: 6c4f48d00a0649b56977f6c8a7ada4e0 (MD5) / Approved for entry into archive by Adriana Oliveira (adriana.oliveira@ufjf.edu.br) on 2018-01-22T16:33:53Z (GMT) No. of bitstreams: 1 samuelbelinidefilippo.pdf: 2610291 bytes, checksum: 6c4f48d00a0649b56977f6c8a7ada4e0 (MD5) / Made available in DSpace on 2018-01-22T16:33:53Z (GMT). No. of bitstreams: 1 samuelbelinidefilippo.pdf: 2610291 bytes, checksum: 6c4f48d00a0649b56977f6c8a7ada4e0 (MD5) Previous issue date: 2017-09-20 / Todo a produção, transmissão e distribuição de energia elétrica ocorre concomitantemente com o consumo da energia. Isso é necessário porque ainda não existe hoje uma maneira viável de se estocar energia em grandes quantidades. Dessa forma, a energia gerada precisa ser consumida quase que instantaneamente. Isso faz com que as previsões de demanda sejam fundamentais para uma boa gestão dos sistemas de energia. Esse trabalho focaliza métodos de previsão de demanda a curto prazo, até um dia à frente. Nos métodos mais simples, as previsões são feitas por modelos lineares que utilizam dados históricos da demanda de energia. Contudo, modelos baseados em inteligência computacional têm sido estudados para este fim, por explorarem a relação não-linear entre a demanda de energia e as variáveis climáticas. Em geral, estes modelos conseguem melhores previsões do que os métodos lineares. Seus resultados, porém, são instáveis e sensíveis a erros de medição, gerando erros de previsão discrepantes, que podem ter graves consequências para o processo de produção. Neste estudo, empregamos redes neurais artificiais e algoritmos genéticos para modelar dados históricos de carga e de clima, e combinamos estes modelos com métodos lineares tradicionais. O objetivo é conseguir previsões que não apenas sejam mais acuradas em termos médios, mas que também menos sensíveis aos erros de medição. / The production, transmission and distribution of electric energy occurs concomitantly with its consumption. This is necessary because there is yet no feasible way to store energy in large quantities. Therefore, the energy generated must be consumed almost instantaneously. This makes forecasting essential for the proper management of energy systems. This thesis focuses on short-term demand forecasting methods up to one day ahead. In simpler methods, the forecasts are made by linear models, which use of historical data on energy demand. However, computer intelligence-based models have been studied for this end, exploring the nonlinear relationship between energy demand and climatic variables. In general, these models achieve better forecasts than linear methods. Their results, however, are unstable and sensitive to measurement errors, leading to outliers in forecasting errors, which can have serious consequences for the production process. In this thesis, we use artificial neural networks and genetic algorithms for modelling historical load and climate data, and combined these models with traditional linear methods. The aim is to achieve forecasts that are not only more accurate in mean terms, but also less sensitive to measurement errors.

Submitted by Renata Lopes (renatasil82@gmail.com) on 2017-03-07T14:25:21Z No. of bitstreams: 1 victorteixeirademelomayrink.pdf: 2587774 bytes, checksum: 1319cc37a15480796050b618b4d7e5f7 (MD5) / Approved for entry into archive by Adriana Oliveira (adriana.oliveira@ufjf.edu.br) on 2017-03-07T15:06:57Z (GMT) No. of bitstreams: 1 victorteixeirademelomayrink.pdf: 2587774 bytes, checksum: 1319cc37a15480796050b618b4d7e5f7 (MD5) / Made available in DSpace on 2017-03-07T15:06:57Z (GMT). No. of bitstreams: 1 victorteixeirademelomayrink.pdf: 2587774 bytes, checksum: 1319cc37a15480796050b618b4d7e5f7 (MD5) Previous issue date: 2016-08-31 / FAPEMIG - Fundação de Amparo à Pesquisa do Estado de Minas Gerais / O armazenamento de energia elétrica em larga escala ainda não é viável devido a restrições técnicas e econômicas. Portanto, toda energia consumida deve ser produzida instantaneamente; não é possível armazenar o excesso de produção, ou tampouco cobrir eventuais faltas de oferta com estoques de segurança, mesmo que por um curto período de tempo. Consequentemente, um dos principais desafios do planejamento energético consiste em realizar previsões acuradas para as demandas futuras. Neste trabalho, apresentamos um modelo de previsão para o consumo de energia elétrica a curto prazo. A metodologia utilizada compreende a construção de um comitê de previsão, por meio da aplicação do algoritmo Gradient Boosting em combinação com modelos de árvores de decisão e a técnica de amortecimento exponencial. Esta estratégia compreende um método de aprendizado supervisionado que ajusta o modelo de previsão com base em dados históricos do consumo de energia, das temperaturas registradas e de variáveis de calendário. Os modelos propostos foram testados em duas bases de dados distintas e demonstraram um ótimo desempenho quando comparados com resultados publicados em outros trabalhos recentes. / The storage of electrical energy is still not feasible on a large scale due to technical and economic issues. Therefore, all energy to be consumed must be produced instantly; it is not possible to store the production leftover, or either to cover any supply shortages with safety stocks, even for a short period of time. Thus, one of the main challenges of energy planning consists in computing accurate forecasts for the future demand. In this paper, we present a model for short-term load forecasting. The methodology consists in composing a prediction comitee by applying the Gradient Boosting algorithm in combination with decision tree models and the exponential smoothing technique. This strategy comprises a supervised learning method that adjusts the forecasting model based on historical energy consumption data, the recorded temperatures and calendar variables. The proposed models were tested in two di erent datasets and showed a good performance when compared with results published in recent papers.

MSc (Statistics) / Department of Statistics / Generally, short term load forecasting is essential for any power generating utility. In this dissertation the main objective was to develop short term load forecasting models for the peak demand periods (i.e. from 18:00 to 20:00 hours) in South Africa using. Quantile semi-parametric additive models were proposed and used to forecast electricity demand during peak hours. In addition to this, forecasts obtained were then used to nd an optimal number of generating units to commit (switch on or o ) daily in order to produce the required electricity demand at minimal costs. A mixed integer linear programming technique was used to nd an optimal number of units to commit. Driving factors such as calendar e ects, temperature, etc. were used as predictors in building these models. Variable selection was done using the least absolute shrinkage and selection operator (Lasso). A feasible solution to the unit commitment problem will help utilities meet the demand at minimal costs. This information will be helpful to South Africa's national power utility, Eskom. / NRF

This degree project undertakes a detailed examination of various algorithms used in Very Short-Term Load Forecasting (VSTLF) within network control systems, prioritizing forecasting accuracy and computational efficiency as critical evaluation criteria. The research comprehensively assesses a range of forecasting methods, including statistical models, machine learning algorithms, and advanced deep learning techniques, aiming to highlight their respective advantages, limitations, and suitability for different operational contexts. The study conducts a detailed analysis by comparing essential performance metrics such as Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE), and execution time, before and after implementing adjustments to the formulations. This approach highlights how optimization strategies enhance the effectiveness of the models. Notably, the study identifies Support Vector Machine (SVM) and Multiple Linear Regression as frontrunners in terms of balancing accuracy with computational demand, making them particularly suitable for real-time forecasting needs. Meanwhile, Long Short-Term Memory (LSTM) networks demonstrate a commendable ability to capture complex, non-linear data patterns, albeit at a higher computational cost. The degree project further explores the sensitivity of these forecasting models to parameter adjustments, revealing a nuanced landscape where strategic modifications can significantly enhance model performance. This degree project not only contributes to the ongoing discourse on optimizing VSTLF algorithms but also provides actionable insights for stakeholders in the energy sector, aiming to facilitate the development of more reliable, efficient, and sustainable power system operations.

本研究主要應用基於總體經驗模態分解法(EEMD)之倒傳遞類神經網路(BPNN)預測兩種不同的非線性時間序列數據，包括政大逐時用電量以及逐日歷史黃金價格。透過EEMD，這兩種資料會分別被拆解為數條具有不同物理意義的本徵模態函數(IMF)，而這讓我們可以將這些IMF視為各種影響資料的重要因子，並且可將拆解過後的IMF放入倒傳遞類神經網路中做訓練。 另外在本文中，我們也採用移動視窗法作為預測過程中的策略，另外也應用內插法和外插法於逐時用電量的預測。內插法主要是用於補點以及讓我們的數據變平滑，外插法則可以在某個範圍內準確預測後續的趨勢，此兩種方法皆對提升預測準確度占有重要的影響。 利用本文的方法，可在預測的結果上得到不錯的準確性，但為了進一步提升精確度，我們利用多次預測的結果加總平均，然後和只做一次預測的結果比較，結果發現多次加總平均後的精確度的確大幅提升，這是因為倒傳遞類神經網路訓練過程中其目標為尋找最小誤差函數的關係所致。 / In this paper, we applied the Ensemble Empirical Mode Decomposition (EEMD) based Back-propagation Neural Network (BPNN) learning paradigm to two different topics for forecasting: the hourly electricity consumption in NCCU and the historical daily gold price. The two data series are both non-linear and non-stationary. By applying EEMD, they were decomposed into a finite, small number of meaningful Intrinsic Mode Functions (IMFs). Depending on the physical meaning of IMFs, they can be regarded as important variables which are input into BPNN for training. We also use moving-window method in the prediction process. In addition, cubic spline interpolation as well as extrapolation as our strategy is applied to electricity consumption forecasting, these two methods are used for smoothing the data and finding local trend to improve accuracy of results. The prediction results using our methods and strategy resulted in good accuracy. However, for further accuracy, we used the ensemble average method, and compared the results with the data produced without applying the ensemble average method. By using the ensemble average, the outcome was more precise with a smaller error, it results from the procedure of finding minimum error function in the BPNN training.

This paper investigates correlation between energy consumption 24 hours ahead and features used for predicting energy consumption. The features originate from three categories: weather, time and previous energy. The correlations are calculated using Pearson correlation and mutual information. This resulted in the highest correlated features being those representing previous energy consumption, followed by temperature and month. Two identical feature sets containing all attributes1 were obtained by ranking the features according to correlation. Three feature sets were created manually. The first set contained seven attributes representing previous energy consumption over the course of seven days prior to the day of prediction. The second set consisted of weather and time attributes. The third set consisted of all attributes from the first and second set. These sets were then compared on different machine learning models. It was found the set containing all attributes and the set containing previous energy attributes yielded the best performance for each machine learning model. 1In this report, the words ”attribute” and ”feature” are used interchangeably. / I denna rapport undersöks korrelation och betydelsen av olika attribut för att förutspå energiförbrukning 24 timmar framåt. Attributen härstammar från tre kategorier: väder, tid och tidigare energiförbrukning. Korrelationerna tas fram genom att utföra Pearson Correlation och Mutual Information. Detta resulterade i att de högst korrelerade attributen var de som representerar tidigare energiförbrukning, följt av temperatur och månad. Två identiska attributmängder erhölls genom att ranka attributen över korrelation. Tre attributmängder skapades manuellt. Den första mängden innehåll sju attribut som representerade tidigare energiförbrukning, en för varje dag, sju dagar innan datumet för prognosen av energiförbrukning. Den andra mängden bestod av väderoch tidsattribut. Den tredje mängden bestod av alla attribut från den första och andra mängden. Dessa mängder jämfördes sedan med hjälp av olika maskininlärningsmodeller. Resultaten visade att mängden med alla attribut och den med tidigare energiförbrukning gav bäst resultat för samtliga modeller.