University of Macau / Faculty of Science and Technology / Department of Mathematics

Made available in DSpace on 2014-06-11T19:22:32Z (GMT). No. of bitstreams: 0 Previous issue date: 2010-11-26Bitstream added on 2014-06-13T19:28:04Z : No. of bitstreams: 1 campos_jr_me_ilha.pdf: 1235138 bytes, checksum: 9965ccc979ea59bf6f2a7e8558692b7b (MD5) / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / A previsão de carga, considerada essencial no planejamento da operação energética e nos estudos de ampliação e reforços da rede básica, assume importância estratégica na extensão comercial, valorizando os processos de armazenamento desses dados e da extração de conhecimentos através de técnicas computacionais. Nos últimos anos, diversos trabalhos foram publicados sobre sistemas de previsão de cargas (demanda) elétricas. Nos horizontes de curto, médio e longo prazo, os modelos neurais, estão entre os mais explorados. O objetivo deste trabalho é apresentar um sistema previsor de cargas elétricas de forma simples e eficiente através de sistemas baseados em redes neurais artificiais com treinamento realizado pelo algoritmo back-propagation. Para isto, optou-se pelo desenvolvimento de um software utilizando os paradigmas de programação orientada a objetos para criar um modelo neural de fácil manipulação, e que de certa forma, consiga corrigir o problema dos mínimos locais. Em geral, o sistema desenvolvido é capaz de atribuir os parâmetros da rede neural de forma automática através de processos exaustivos. Os resultados apresentados foram comparados utilizando outros trabalhos em que também se usaram-se os dados da mesma companhia elétrica. Este trabalho apresentou um ganho de desempenho bem satisfatório em relação a outros trabalhos encontrados na literatura para a mesma classe de problemas / Load Forecasting is essential in planning and operation of power systems, in enlarging and reinforcing the basic network, is also very important commercially, valorizing the filing process of these data and extracting knowledge by computational techniques. Lately, several works have been published about electrical load forecasting. Short term, medium term and long term horizons are equally studied. The objective of this work is to present an electrical load forecasting system, which is simple and efficient and based on artificial neural networks whose training is with the back-propagation algorithm. Therefore, a software is developed using the paradigms of the object oriented programming technique to create a neural model which is ease to manipulate, and able to correct the local minimum problem. This system attributes the neural parameters automatically by exhaustive procedures. Results are compared with other works that have used the same data and this work presents a satisfactory performance when compared with those and others found in the literature

Made available in DSpace on 2017-07-21T14:19:24Z (GMT). No. of bitstreams: 1 Melo, Geisla Albuquerque.pdf: 2986772 bytes, checksum: 02494f1ef68a9df48a1184c0a3e81dce (MD5) Previous issue date: 2015-05-25 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / According to Embrapa (2013), Brazil is the world's second largest soy producer just after the United States. Season after season, the production and planted area in Brazil is growing, however, climatic factors and crop diseases are affecting plantation, preventing further growth, and causing losses to farmers. Asian rust caused by Phakopsora pachyrhizi, is a foliar disease, considered one of the most important diseases at present, because of the potential for loss. Asian rust can be mistaken for other diseases in soybeans, such as Bacterial Blight, a Stain Brown and Bacterial Pustule, due to similar visual appearances. Thus, the present study aimed to develop an application for mobile devices using the Android platform to perform automatic recognition of the Asian soybean rust severity indices to assist in the early diagnosis and therefore assist in decision-making as the management and control of the disease. For this, was used techniques of digital image processing (DIP) and Artificial Neural Networks (ANN). First, around 3.000 soybean leaves were collected in the field, where about 2.000 were harnessed. Then it were separated by severity index, photographed in a controlled environment, and after that were processed in order to eliminate noise and background images. Filtering preprocessing phase consisted of median filter, Gaussian filter processing for gray scale, Canny edge detector, expansion, find and drawcontours, and finally the cut of leaf. After this was extracted color and texture features of the images, which were the average R, G and B Variant also for the three channels R, G and B according angular momentum, entropy, contrast, homogeneity, and finally correlation the severity degree previously known. With these data, the training was performed an ANN through the neural network simulator BrNeural. During training, parameters such as number of severity levels and number of neurons of the hidden layer have changed. After training, was chosen network architecture that gave better results, with 78.86% accuracy for Resilient-propagation algorithm. This network was saved in an object and inserted into the application, ready to be used with new data. Thus, the application takes the soybean leaf picture and filters the acquired image. After this, it extracts the features and commands internally to the trained neural network, which analyzes and reports the severity. Still, it is optionally possible to see a georeferenced map of the property, with the severities identified by small colored squares, each representing a different index. / Segundo a Embrapa (2013), o Brasil é o segundo maior produtor de soja do mundo, atrás apenas nos Estados Unidos. Safra após safra, a produção e a área plantada do Brasil vem crescendo, entretanto, fatores climáticos e doenças da cultura vêm afetando as lavouras, impedindo um crescimento ainda maior, e causando perdas para os agricultores. A ferrugem asiática, causada pelo fungo Phakopsora pachyrhizi, é uma doença foliar, considerada uma das doenças de maior importância na atualidade, devido ao grande potencial de perdas. A ferrugem asiática pode ser confundida com outras doenças na soja, como o Crestamento Bacteriano, a Mancha Parda e a Pústula Bacteriana, devido às aparências visuais semelhantes. Deste modo, O presente estudo teve por objetivo desenvolver um aplicativo para dispositivos móveis que utilizam a plataforma Android, para realizar o reconhecimento automático dos índices de severidade da ferrugem asiática da soja, para auxiliar no diagnóstico precoce e por consequência, auxiliar na tomada de decisão quanto ao manejo e controle da doença. Para isto, foram utilizadas técnicas de Processamento Digital de Imagens (PDI) e Redes Neurais Artificiais (RNA). Primeiramente, foram coletadas aproximadamente 3 mil folhas de soja em campo, onde cerca de 2 mil foram aproveitadas. Então elas foram separadas por índices de severidade, fotografadas em ambiente controlado, e após isto foram processadas com o objetivo de eliminar ruídos e o fundo das imagens. A fase de filtragem do pré-processamento consistiu nos filtros da mediana, filtro Gaussiano, transformação para escala de cinza, detector de bordas Canny, dilatação, find e drawcontours, e por fim o recorte da folha. Após isto, foram extraídas as características de cor e textura das imagens, que foram as médias R, G e B, Variância também para os três canais R, G e B, Segundo Momento Angular, Entropia, Contraste, Homogeneidade, Correlação e por fim, o Grau de Severidade previamente sabido. Com estes dados, foi realizado o treinamento de uma RNA através do simulador de redes neurais BrNeural. Durante o treinamento, parâmetros como quantidade de níveis de severidade e quantidade de neurônios da camada oculta foram alterados. Após o treinamento, foi escolhida a arquitetura de rede que deu melhor resultado, com 78,86% de acerto para o algoritmo Resilient-propagation. Esta rede foi salva em um objeto e inserida no aplicativo, pronta para ser utilizada com dados novos. Assim, o aplicativo tira a foto da folha de soja e faz a filtragem da imagem adquirida. Após isto, extrai as características e manda internamente para a rede neural treinada, que analisa e informa a severidade. Ainda, opcionalmente é possível ver um mapa georreferenciado da propriedade, com as severidades identificadas por pequenos quadrados coloridos, representando cada um, um índice diferente.

In this thesis, the primary powertrain components of a power split hybrid electric vehicle are modeled. In particular, the dynamic model of the energy storage element (i.e., traction battery) is exactly linearized through an input transformation method to take advantage of the proposed optimal control algorithm. A lipschitz continuous and nondecreasing cost function is formulated in order to minimize the net amount of consumed fuel. The globally optimal solution is obtained using a dynamic programming routine that produces the optimal input based on the current state of charge and the future power demand. It is shown that the global optimal control solution can be expressed in closed form for a time invariant and convex incremental cost function utilizing the interval back propagation approach. The global optimality of both time varying and invariant solutions are rigorously proved. The optimal closed form solution is further shown to be applicable to the time varying case provided that the time variations of the incremental cost function are sufficiently small. The real time implementation of this algorithm in Simulink is discussed and a 32.84 % improvement in fuel economy is observed compared to existing rule based methods.

Now a days, Closed Circuit Television (CCTV) cameras are installedeverywhere in public places to monitor illegal activities like armedrobberies. Mostly CCTV footages are used as post evidence after theoccurrence of crime. In many cases a person might be monitoringthe scene from CCTV but the attention can easily drift on prolongedobservation. Eciency of CCTV surveillance can be improved by in-corporation of image processing and object detection algorithms intomonitoring process.The object detection algorithms, previously implemented in CCTVvideo analysis detect pedestrians, animals and vehicles. These algo-rithms can be extended further to detect a person holding weaponslike rearms or sharp objects like knives in public or restricted places.In this work the detection of weapon from CCTV frame is acquiredby using Histogram of Oriented Gradients (HOG) as feature vector andarticial neural networks performing back-propagation algorithm forclassication.As a weapon in the hands of a human is considered to be greaterthreat as compared to a weapon alone, in this work the detection ofhuman in an image prior to a weapon detection has been found advan-tageous. Weapon detection has been performed using three methods.In the rst method, the weapon in the image is detected directly with-out human detection. Second and third methods use HOG and back-ground subtraction methods for detection of human prior to detectionof a weapon. A knife and a gun are considered as weapons of inter-est in this work. The performance of the proposed detection methodswas analysed on test image dataset containing knives, guns and im-ages without weapon. The accuracy rate 84:6% has been achievedby a single-class classier for knife detection. A gun and a knife havebeen detected by the three-class classier with an accuracy rate 83:0%.

Classification is one-out-of several applications in the neural network (NN) world. Multilayer perceptron (MLP) is the common neural network architecture which is used for classification tasks. It is famous for its error back propagation (EBP) algorithm, which opened the new way for solving classification problems given a set of empirical data. In the thesis, we performed experiments by using three different NN structures in order to find the best MLP neural network structure for performing the nonlinear classification of multiclass data sets. A developed learning algorithm used here is the batch EBP algorithm which uses all the data as a single batch while updating the NN weights. The batch EBP speeds up training significantly and this is also why the title of the thesis is dubbed 'fast NN …'. In the batch EBP, and when in the output layer a linear neurons are used, one implements the pseudo-inverse algorithm to calculate the output layer weights. In this way one always finds the local minimum of a cost function for a given hidden layer weights. Three different MLP neural network structures have been investigated while solving classification problems having K classes: one model/K output layer neurons, K separate models/One output layer neuron, and K joint models/One output layer neuron. The extensive series of experiments performed within the thesis proved that the best structure for solving multiclass classification problems is a K joint models/One output layer neuron structure.

This thesis describes the implementation of a Radial Basis Function (RBF) network to be used in predicting the effectiveness of various strategies for reducing the Vehicle Trip Rate (VTR) of a worksite. Three methods of learning were utilized in training the Gaussian hidden units of the network, those being a) output weight adjustment using the Delta rule b) adjustable reference vectors in conjunction with weight adjustment, and c) a combination of adjustable centers and adjustable sigma values for each RBF neuron with the Delta rule. The justification for utilizing each of the more advanced levels of training is provided using a series of tests and performance comparisons. The network architecture is then selected based upon a series of initial trials for an optimum number of hidden Radial Basis neurons. In a similar manner, the training time is determined after finding a maximum number of epochs during which there is a significant change in the SSE. The network was compared for effectiveness against each of the following methods of data analysis: force-entered regression, backward regression, forward regression, stepwise regression, and two types of back-propagation networks based upon the attributes discovered to be most predictive by these regression techniques. A comparison of the learning methods used on the Radial Basis network shows the third learning strategy to be the most efficient for training, yielding the lowest sum of squared errors (SSE) in the shortest number of training epochs. The result of comparing the RBF implementation against the other methods mentions shows the superiority of the Radial Basis method for predictive ability.

In the design and plan of the coast engineering, long-term and continual tidal database represent the indispensable role. This paper collect the tidal database, their locations include the ocean around the Taiwan and the South China Sea. Use the artificial neural networks (ANN) to build model and find the relationship between neighbor tidal observation stations. There are many reasons to cause the tide phenomenon, include the tide generating force, season, coastal geography, geography of sea floor, resonance of gulf or estuary, change depth of sea, and so on, it will be determined by local environment. The tide analysis and prediction usually use the harmonic analysis method. This method need long-term and continual tidal record, and the theory depend on the tide generating force, it has limit about accuracy. The application of artificial neural networks is used in nonlinear science problems in general cases. The back propagation (BP) networks is the one model of the artificial neural networks, this paper use ANN-BP model to build the relationship from different tide observed stations, and verify the quality of model. From the result of verified models, the ANN-BP model can predict and supplement the tide record very well. The items of research include: ¡i1¡j the relationship between two neighbor tide observed stations. (one station input, one station output) ¡i2¡jthe relationship between three neighbor tide observed stations. (two station input, one station output) ¡i3¡j input several tide observed stations and output one station. ¡i4¡j the correlation of connected weight and threshold between different models. ¡i5¡j change the parameters of ANN-BP model and discus the affect of model¡¦s quality. ¡i6¡j application of truly case. From the result of this paper, in the design and plan of the coast engineering, the long-term tide observed record can be predict from the ANN-BP model and tide record of neighbor observed stations. When the tide record has miss or lost cause by machine or other reasons, the ANN-BP model can supplement the lost tide record well. This paper show the ANN-BP model can be apply to predict and supplement the tide record very well, and will be possible applied method.

Thesis (M.S.) -- Mississippi State University. Department of Computational Engineering. / Title from title screen. Includes bibliographical references.

The need and desire to create robust and accurate joining of materials has been one of up most importance throughout the course of history. Many forms have often been employed, but none exhibit the strength or durability as the weld. This study endeavors to explore some of the aspects of welding, more specifically relating to the Double Sided Arc Welding process and how best to model the dynamic non-linear response of such a system. Concepts of the Volterra series, NARMAX approximation and neural networks are explored. Fundamental methods of the neural network, including radial basis functions, and Back-propagation are investigated.