Short-term electricity price forecasting is now crucial practice in deregulated electricity markets, as it forms the basis for maximizing the profits of the market participants. In this thesis, short-term electricity prices are forecast using three different predictor schemes, Artificial Neural Networks (ANNs), Support Vector Machine (SVM) and a hybrid scheme, respectively. ANNs are the very popular and successful tools for practical forecasting. In this thesis, a hidden-layered feed-forward neural network with back-propagation has been adopted for detailed comparison with other forecasting models. SVM is a newly developed technique that has many attractive features and good performance in terms of prediction. In order to overcome the limitations of individual forecasting models, a hybrid technique that combines Fuzzy-C-Means (FCM) clustering and SVM regression algorithms is proposed to forecast the half-hour electricity prices in the UK electricity markets. According to the value of their power prices, thousands of the training data are classified by the unsupervised learning method of FCM clustering. SVM regression model is then applied to each cluster by taking advantage of the aggregated data information, which reduces the noise for each training program. In order to demonstrate the predictive capability of the proposed model, ANNs and SVM models are presented and compared with the hybrid technique based on the same training and testing data sets in the case studies by using real electricity market data. The data was obtained upon request from APX Power UK for the year 2007. Mean Absolute Percentage Error (MAPE) is used to analyze the forecasting errors of different models and the results presented clearly show that the proposed hybrid technique considerably improves the electricity price forecasting.

[pt] Apesar do crescimento exponencial da instalação de novas usinas eólicas nos últimos anos, a matriz energética Brasileira é composta, principalmente, por usinas hidrelétricas. Uma das principais características dos sistemas de geração com predominância hidráulica é a forte dependência dos regimes hidrológicos. Atualmente, o setor elétrico brasileiro utiliza a Energia Natural Afluente para gerar cenários hidrológicos a partir de um modelo PAR. Tal modelo é ajustado a partir dos parâmetros estimados do histórico da série temporal, isto é, não considera quaisquer informações exógenas que possam afetar os regimes hidrológicos e, consequentemente, a produção de energia. Estudos recentes identificaram que o uso de variáveis climáticas na modelagem de séries de afluências nas bacias brasileiras pode servir como fator de diminuição de incertezas devido a existência de correlação entre essas variáveis. Também foram identificados benefícios ao decompor as séries hidrológicas em sinal e ruído e utilizar somente o sinal para a modelagem. Neste contexto, o desenvolvimento de modelos híbridos que combinem técnicas de de composição das séries hidrológicas e modelos de séries temporais com variáveis exógenas são objetos de estudo deste trabalho,assim como o desenvolvimento de modelos que associem tais variáveis de formação-linear e periódica. Essas novas abordagens contemplam o uso das técnicas de decomposição SSA e MSSA em combinação com PAR, a aplicação do modelo PARX e o desenvolvimento do modelo PGAM. Como conclusão tem-se que os modelos aplicados se mostraram eficientes para os objetivos propostos e também apresentaram melhor performance, em alguns casos, quando comparados com modelos já publicados na literatura. / [en] Despite the exponential growth of wind farms in recent years, the Brazilian energy matrix is mainly composed of hydroelectric plants.One of the main characteristics of hydroelectric generation systems is the strong dependence on hydrological regimes. Currently, the Brazilian electric sector uses the Natural Energy In flow to generate hydrological scenarios from a PAR model.Such model is adjusted from the estimated parameters of the time series history, that is, it does not consider any exogenous information that could affect the hydrological regimes and, consequently, the energy production. Recent studies indicate that the use of climatic variables in the modeling of inflow series in the Brazilian basins may serve as a factor to reduce uncertainties due to the existence of correlation between these variables. It was also identified benefits by decomposing hydrological series into signal and noise and using only the signal for modeling. In this context, the development of hybrid models that combine techniques of decomposition of the hydrological series and time series models with exogenous variable are study objects of this work, as well as the development of models that associate such variables in a non-linear and periodic way. These new approaches contemplate the use of SSA and MSSA decomposition techniques in combination with PAR, the application of the PARX and the development of the PGAM model. As conclusion, the applied models were efficient for the proposed objectives and also presented better performance, in some cases, when compared with models already published in the literature.

In this thesis a tool to be used in experimental design for batch processes is presented. Specifically, this method is to aid in the development of a process model. Currently, experimental design methods are either empirical in nature which need very little understanding of the underlying phenomena and without the objective of more fundamental understanding of the process. Other methods are model based which assume the model is correct and attempt to better define the model parameters or discriminate between models. This new paradigm for experimental design allows for process optimization and process model development to occur simultaneously. The methodology specifically evaluates multiple models as a check to evaluate whether the models are capturing the trend in the experimental data. A new tool for experimental design developed here is called the grid algorithm which is designed to constrain the experimental region to potential optimal points of the user defined objective function for the process. It accomplishes this by using the confidence interval on the objective function value. The objective function value is calculated using the model prediction of the best performing model among a set of models at the predicted optimal point. This new experimental design methodology is tested first on simulated data. The first simulation fits a model to data generated by the modified Himmelblau function (MHF). The second simulation fits multiple models to data generated to simulate a film growth process. In both simulations the grid algorithm leads to improved prediction at the optimal point and better sampling of the region around the optimal point. This experimental design method was then applied to an actual chemical vapor deposition system. The films were analyzed using atomic force microscopy (AFM) to find the resulting film roughness. The methodology was then applied to design experiments using models to predict roughness. The resulting experiments were designed in a region constrained by the grid algorithm and were close to the predicted optimum of the process. We found that the roughness of a thin film depended on the substrate temperature but also showed a relationship to the nucleation density of the thin film.

In this thesis we study three different modelling frameworks for biological systems of dispersal and combinations thereof. The three frameworks involved are individual-based models, group-level models in the form of partial differential equations (PDEs) and robot swarms. In the first two chapters of the thesis, we present ways of coupling individual based models with PDEs in so-called hybrid models, with the aim of achieving improved performance of simulations. Two classes of such hybrid models are discussed that allow an efficient simulation of multi-species systems of dispersal with reactions, but involve individual resolution for certain species and in certain parts of a computational domain if desired. We generally consider two types of example systems: bacterial chemotaxis and reaction-diffusion systems, and present results in the respective application area as well as general methods. The third chapter of this thesis introduces swarm robotic experiments as an additional tool to study systems of dispersal. In general, those experiments can be used to mimic animal behaviour and to study the impact of local interactions on the group-level dynamics. We concentrate on a target finding problem for groups of robots. We present how PDE descriptions can be adjusted to incorporate the finite turning times observed in the robotic system and that the adjusted models match well with experimental data. In the fourth and last chapter, we consider interactions between robots in the form of hard-sphere collisions and again derive adjusted PDE descriptions. We show that collisions have a significant impact on the speed with which the group spreads across a domain. Throughout these two chapters, we apply a combination of experiments, individual-based simulations and PDE descriptions to improve our understanding of interactions in systems of dispersal.

In order to respond to the changing demands of the business environment, organizations need to develop a higher degree of flexibility in relation to the market and prevailing competition. One way for organizations to become more adjustable is to adopt flexible ways of working that promote innovation. A dominant approach for this is agile methods. Although the basic values and principles of agile methods were initially formulated for work processes linked to software development, an interest in integrating the work method into other functions and industries beyond IT has emerged. An integration and usage of agile methods can, however, give rise to paradoxical tensions as it is challenging to combine organizational needs for stability with agile methods' emphasis on flexibility. The balance between stability and flexibility is highlighted as important and is described as organizational ambidexterity. This study aimed to investigate the integration and usage of agile methods in an organizational context and how to balance stability and flexibility to manage paradoxical tensions. To answer the study's purpose, a qualitative method has been conducted with semi-structured interviews within the chosen organization. The data has been analyzed with James March's (1991) theory Exploration and Exploitation in Organizational Learning and Tushman and O'reilly's (1996) theoretical framework Ambidextrous Organizations. The result shows that there are both external and internal motives for integrating agile methods. Furthermore, the study has found central challenges and opportunities linked to the integration and usage of agile methods. Prominent findings show the importance of reaching an organizational consensus and understanding when agile methods are being integrated, making a selection of what agile values and principles have the possibility to actually generate improvements and benefits, as well as applying traditional and agile methods in combination through hybrid models. The result also highlights the importance of balancing organizational stability and flexibility in order to manage paradoxical tensions and achieve organizational ambidexterity.

La sécurité des installations industrielles implique une gestion intégrée de tous les facteurs pouvant causer des incidents. La gestion d’alarmes est une condition qui peut être formulée comme un problème de reconnaissance de motifs pour lequel les motifs temporels sont utilisés pour caractériser différentes situations typiques, en particulier liées au phases de démarrage et d'arrêt. Dans cette thèse, nous proposons une nouvelle approche de gestion des alarmes basée sur un processus de diagnostic. En considérant les alarmes et les actions des procédures d'exploitation standard comme des événements discrets, le diagnostic repose sur la reconnaissance de situation pour fournir aux opérateurs des informations pertinentes sur les défauts induisant les flux d'alarmes. La reconnaissance de situation est basée sur des chroniques qui sont apprises pour chaque situation. Nous proposons d'utiliser un modèle causal hybride du système et des simulations pour générer les séquences d'événements représentatives à partir desquelles les chroniques sont apprises automatiquement en utilisant l'algorithme « Heuristic Chronicle Discovery Algorithm Modified » (HCDAM). Une extension de cet algorithme est présentée dans cette thèse où les connaissances d'experts sont prises en compte comme des restrictions temporelles qui constituent une nouvelle entrée pour HCDAM. Deux cas d’étude illustratifs dans le domaine des procédés pétrochimiques sont présentés. / Industrial plant safety involves integrated management of all the factors that may cause incidents. Process alarm management is a requisite that can be formulated as a pattern recognition problem in which temporal patterns are used to characterize different typical situations, particularly at startup and shutdown stages. In this thesis, we propose a new approach of alarm management based on a diagnosis process. Assuming the alarms and the actions of the standard operating procedures as discrete events, diagnosis relies on situation recognition to provide the operators with relevant information about the faults inducing the alarm flows. Situation recognition is based on chronicles that are learned for every situation. We propose to use the hybrid causal model of the system and simulations to generate the representative event sequences from which the chronicles are learned using the Heuristic Chronicle Discovery Algorithm Modified (HCDAM). An extension of this algorithm is presented in this thesis where expert knowledge is included as temporal restrictions which are a new input to HCDAM. Two illustrative case studies in the field of petrochemical plants are presented.

Karst aquifers can be conceptualized as dual flow systems comprised of a low-conductive matrix with embedded high-conductive conduits / preferential flow zones. Discharge in conduits ranges from low-velocity laminar flow to high-velocity transitional and turbulent flow. Commonly employed continuum models do not account for the specific behavior of transitional and turbulent flow. In response to this limitation, enhancements have been made to MODFLOW, a commonly used groundwater flow model, by adding a discrete conduit network to the matrix continuum (hybrid model). The Conduit Flow Process (CFP) package is the latest realization of this model approach. CFP Mode 1 (CFPM1) computes laminar and turbulent flow in discrete conduits that are coupled to the laminar continuum model. CFP Mode 2 (CFPM2) accounts for turbulent flow in preferential flow layers by adapting the continuum model. Therefore, laminar hydraulic conduc-tivities are converted into turbulent hydraulic conductivities. CFPM2 was further modified to consider steady turbulent pipe flow. Karst models based on CFPM2 require potentially less input data and computational efforts than karst models based on CFPM1. Furthermore, CFPM2 integrates more easily into MODFLOW versions including e.g. transport models. Parameter studies for a synthetic catchment demonstrates that continuum models with turbulent flow representation and an additional flow barrier between conduits and matrix can represent karst systems similar to hybrid models. For simulation of highly transient flow processes in karst conduit systems, i.e. during flood events, it is crucial to consider dynamics such as free-surface flow, wave propagation, and changes between pressurized and non-pressurized conduit flow. The coupled overland- and groundwater flow model MODBRANCH was therefore enhanced to consider unsteady and non-uniform flow processes in karst conduits. Flow in discrete conduits is simulated using the Saint-Venant-equations for free-surface flow. Contrary to overland flow, the cross sectional area of karst conduits is finite. Accordingly, both pressurized and non-pressurized flow may occur within conduits. To simulate pressurized flow, a hypothetical, narrow, open-top slot (Preissmann slot) is added to the conduit crown, which allows the use of the free-surface flow equations for fully filled conduits. Beyond this, the model features a variable time step to consider wave speed variations, for example due to the transition from free-surface to pressurized flow. Parameter studies for a synthetic catchment demonstrate the significance of free-surface flow representation for variably filled conduits. / Karstgrundwasserleiter können als duale Fließsysteme konzeptionalisiert werden, bestehend aus einer geringdurchlässigen Matrix mit eingebundenen hochdurchlässigen Bereichen, z. B. Karströhren. Der Abfluss in den hochdurchlässigen Bereichen reicht von langsamer laminarer Strömung bis zu schneller turbulenter Strömung. Herkömmliche numerische Grundwasser-strömungsmodelle berücksichtigen nicht die spezifischen Eigenschaften von nicht-laminarer Strömung (Übergangsbereich laminar-turbulent bzw. turbulente Verhältnisse). Ein Ansatz um diese Einschränkung zu umgehen, ist die Erweiterung des laminaren Kontinuums um ein dis-kretes Röhrenmodell, das zustandsabhängig laminare und turbulente Strömung berücksichtigt (Hybridmodell). Eine aktuelle Umsetzung dieses Ansatzes ist Conduit Flow Process (CFP), ein Modul für das weitverbreitete Grundwasserströmungsmodell MODFLOW. CFP Mode 1 (CFPM1) berechnet laminare und turbulente Strömung in diskreten, mit dem Kontinuummodell gekoppelten Röhren. CFP Mode 2 (CFPM2) berücksichtigt nicht-laminare Strömung in hochdurchlässigen Schichten mit einer angepassten hydraulischen Leitfähigkeit des Kontinuummodells. CFPM2 wurde weiter modifiziert, so dass auch turbulente Strömung in Karströhren berechnet werden kann. Dadurch kann möglicherweise der Parameterbedarf sowie der Rechenaufwand gegenüber Hybrid¬modellen reduziert werden. CFPM2 lässt sich einfach in vorhandene MODFLOW Modelle einbinden, z. B. zur Berechnung von Transportprozessen. Parameterstudien für ein idealisiertes Karsteinzugsgebiet zeigen, dass Kontinuummodelle bei Berücksichtigung der turbulenten Strömung sowie des zusätzlichen hydraulischen Widerstand zwischen Röhren und Matrix, Karstsysteme ähnlich wie Hybridmodelle darstellen. Zur Simulation von instationären Prozessen in Karströhren, z. B. ausgeprägte Abflusssignale infolge pulsförmiger Grundwasserneubildung, ist es notwendig, dynamische Prozesse infolge Freispiegelabfluss, Wellenausbreitung sowie Wechsel zwischen Abfluss in teil- und vollgefüllten Röhren zu berücksichtigen. Aus diesem Grund wurde das numerische Modell MODBRANCH, welches ein diskretes Oberflächenwassermodell mit einem Kontinuummodell koppelt, so angepasst, dass instationäre und nichtgleichförmige Abflussprozesse in Karströhren berücksichtigt werden können. Der Abfluss in diskreten Röhren wird dabei mit den Saint-Venant-Gleichungen für Freispiegelabfluss berechnet. Im Gegensatz zu Oberflächengewässern ist der für den Abfluss zur Verfügung stehende Querschnitt in Karströhren limitiert, so dass sowohl Freispiegel- als auch Druckabfluss innerhalb der Röhren auftreten kann. Druckabfluss wird mit Hilfe eines schmalen virtuellen Schlitzes an der Röhrenoberkante simuliert (Preissmann Schlitz), der auch im Fall vollgefüllter Röhren die Anwendung der Gleichungen für Freispiegelabfluss erlaubt. Durch die Verwendung eines variablen Zeitschrittes kann die geänderte Dynamik beim Übergang von Freispiegel- zu Druckabfluss berücksichtigt werden. Parameterstudien für idealisierte, synthetische Karsteinzugsgebiete demonstrieren die Bedeutung der Berücksichtigung von Freispiegelabfluss in teilgefüllter Röhren.

[pt] O setor elétrico brasileiro é fortemente dependente da energia hidrelétrica e a predição acurada das séries de vazões é essencial para o planejamento e gestão de risco. Recentemente, os modelos híbridos, que combinam técnicas de previsão e pré-processamento de dados, têm se destacado. Entretanto, na literatura, não há consenso sobre a superioridade de previsão destes modelos em relação aos tradicionais (puros). Este trabalho visa contribuir para literatura com a avaliação de performance de previsão e a adequabilidade de modelos puros e híbridos para séries mensais estacionárias e não estacionárias de vazões. Para isso, foram construídos modelos usando as técnicas de previsão de Redes Neurais Artificiais e ARIMA acoplados com as técnicas de pré-processamento de dados Singular Spectrum Analysis (SSA) e Seasonal and Trend decomposition based on Loess (STL). Como resultado, este estudo mostra para a série de Belo Monte (estacionária) os modelos puros obtiveram um melhor desempenho, já para a série de Sobradinho (não estacionária) os modelos híbridos foram os melhores. / [en] The Brazilian electricity sector is strongly dependent on hydropower and the accurate prediction of streamflow series is essential for planning and risk management. Recently, hybrid models, which combine prediction and data preprocessing techniques, have stood out. However, in the literature there is no consensus on the predictive superiority of these hybrid models versus their pure version. This paper aims to contribute to the literature with the evaluation of prediction performance suitability of pure and hybrid models for monthly stationary and non - stationary series of streamflow. For this, models were constructed using Artificial Neural Network and ARIMA forecasting techniques coupled with the Singular Spectrum Analysis (SSA) and Seasonal and Trend decomposition based on Loess (STL) data pre-processing techniques. As a result, this study shows that pure models obtained a better performance for the Belo Monte (stationary series), already hybrid models were the best for the Sobradinho (non-stationary series).

To ease development and lower the entry barrier for new adopters many development environments offer visual means to edit complex data. Cyber-physical systems are a perfect candidate for such manipulations since they are usually described in the form of isolated, well defined components that can be manipulated individually. The physical parts of such systems often can be directly translated into real world objects and allowing the developer to interact with those in a familiar manner can greatly increase the usability and agility of the development process. In this thesis we focus on the exploration of interactive manipulation of hybrid system models. Our research examines a solution based on the Acumen simulation environment. We describe the tight integration between the textual model and 3D visualisation, go into detailed analysis of the implementation and use case-studies to illustrate concrete applications

Crude oil prices impact the world economy and are thus of interest to economic experts and politicians. Oil price’s volatile behaviour, which has moulded today’s world economy, society and politics, has motivated and continues to excite researchers for further study. This volatile behaviour is predicted to prompt more new and interesting research challenges. In the present research, machine learning and computational intelligence utilising historical quantitative data, with the linguistic element of online news services, are used to predict crude oil prices via five different models: (1) the Hierarchical Conceptual (HC) model; (2) the Artificial Neural Network-Quantitative (ANN-Q) model; (3) the Linguistic model; (4) the Rule-based Expert model; and, finally, (5) the Hybridisation of Linguistic and Quantitative (LQ) model. First, to understand the behaviour of the crude oil price market, the HC model functions as a platform to retrieve information that explains the behaviour of the market. This is retrieved from Google News articles using the keyword “Crude oil price”. Through a systematic approach, price data are classified into categories that explain the crude oil price’s level of impact on the market. The price data classification distinguishes crucial behaviour information contained in the articles. These distinguished data features ranked hierarchically according to the level of impact and used as reference to discover the numeric data implemented in model (2). Model (2) is developed to validate the features retrieved in model (1). It introduces the Back Propagation Neural Network (BPNN) technique as an alternative to conventional techniques used for forecasting the crude oil market. The BPNN technique is proven in model (2) to have produced more accurate and competitive results. Likewise, the features retrieved from model (1) are also validated and proven to cause market volatility. In model (3), a more systematic approach is introduced to extract the features from the news corpus. This approach applies a content utilisation technique to news articles and mines news sentiments by applying a fuzzy grammar fragment extraction. To extract the features from the news articles systematically, a domain-customised ‘dictionary’ containing grammar definitions is built beforehand. These retrieved features are used as the linguistic data to predict the market’s behaviour with crude oil price. A decision tree is also produced from this model which hierarchically delineates the events (i.e., the market’s rules) that made the market volatile, and later resulted in the production of model (4). Then, model (5) is built to complement the linguistic character performed in model (3) from the numeric prediction model made in model (2). To conclude, the hybridisation of these two models and the integration of models (1) to (5) in this research imitates the execution of crude oil market’s regulators in calculating their risk of actions before executing a price hedge in the market, wherein risk calculation is based on the ‘facts’ (quantitative data) and ‘rumours’ (linguistic data) collected. The hybridisation of quantitative and linguistic data in this study has shown promising accuracy outcomes, evidenced by the optimum value of directional accuracy and the minimum value of errors obtained.