FBE- Elektrik Mühendisliği Lisansüstü Programı - Yüksek Lisans
Bu koleksiyon için kalıcı URI
Gözat
Son Başvurular
1 - 5 / 440
-
ÖgeDağıtım şebekesinde kısa dönem yük tahmini(Fen Bilimleri Enstitüsü, 2020)Günümüzde enerji ihtiyacı teknolojik gelişmeler ile birlikte sürekli olarak artmaktadır. Artan enerji ihtiyacının güvenilir, ekonomik, kaliteli ve sürdürülebilir bir şekilde sağlanabilmesi konusu gün geçtikçe önem kazanmaktadır. Enerjinin ekonomik, kaliteli ve sürdürülebilir bir şekilde sağlanabilmesi enerji sistem planlamasının optimizasyonu ve başarısı ile doğrudan ilişkilidir. Yük tahmini enerji sistem planlamasının ilk ve en önemli aşamalarından biridir. Elektrik yük tahmini geçmiş ve bugünkü elektrik yüklerini analiz ederek, yük tahminine etki eden faktörlerin belirlenmesi ve bunların analizi sonucunda farklı yöntemler, algoritmalar kullanılarak gelecekte olabilecek yükü belirlemek olarak tanımlanabilir. Yük tahmini farklı amaçlar doğrultusunda zaman kıstasına göre çok kısa dönem, kısa dönem, orta dönem ve uzun dönem olarak sınıflandırılmaktadır. Bir yıldan uzun yük tahminleri uzun dönem, bir haftadan bir yıla kadar olan yük tahminleri orta dönem, bir saatten bir haftaya kadar olan zamanları kapsayan tahminler kısa dönem ve bir saatten daha kısa yük tahminleri çok kısa dönem yük tahmini olarak sınıflandırılmaktadır. Elektrik yük tahmini yapmak için kullanılan giriş parametreleri değişkenlik göstermektedir. Uzun dönem yük tahmininde gayrisafi milli hasıla, nüfus gibi parametreler ağırlıklı olarak kullanılırken kısa dönem yük tahminlerinde meteorolojik veriler ağırlıklı olarak kullanılmaktadır. Elektrik tüketimi gün içerisinde saatlik bazda değişkenlik göstermektedir. Gün içerisinde değişen tüketim miktarını, devreye girecek ve devreden çıkacak enerji santrallerini belirlemek kısa dönem yük tahmini ile başlar. Kısa dönem yük tahmininin bir başka fonksiyonu ise enerjinin ekonomik olmasını sağlamaktır. Gün geçtikçe artan enerji ihtiyacını karşılamak ve enerjinin sürdürülebilirliğini sağlamak yük tahmini ile başlar. Bu sebeple yük tahmini uygulamaları gün geçtikçe önem kazanmaktadır. Literatürde, yük tahmini için şu ana kadar kullanılan birçok farklı algoritma ve metodoloji bulunmaktadır. Bu çalışmada literatürde kullanılan analitik yöntemler ve yapay zekâ yöntemleriyle Marmara bölgesinde bulunan bir ile ait 2016-2017 yılları arası veriler kullanılarak saatlik kısa dönem yük tahmini yapılmıştır. Çalışmada kullanılan modeller de temel olarak hafta günleri 3 farklı yöntem kullanılarak ayrılmıştır. Tahmin çalışmalarında kullanılan bağımsız değişken ve bağımlı değişkenler arasında pozitif ve/veya negatif ilişki ne kadar yüksekse model o kadar yüksek tahmin sonuçları vermiştir. Bu sebeple literatürde bazı çalışmalar hafta günleri arasında herhangi bir ayrım yapmadan, bazı çalışmalar hafta içi ve hafta sonu olarak ayrım yapıp iki farklı eğitim ve test verisi kullanılarak, bazı modeller ise hafta içi ve hafta sonu modellerini ayrı eğitim ve test verisi olarak kullanmamak için giriş parametrelerine hafta içi ve hafta sonu ayrımı yapmak için kukla değişken olarak isimlendirilen bağımsız değişken kullanılarak tahmin işlemi gerçekleştirilmiştir. Bu çalışmada gün ayrımlarının tahmin sonucuna etkisini gözlemlemek amacıyla 3 farklı gün ayrımı kullanılmıştır, kullanılan gün ayrımları hafta içi ve hafta sonu için günxx ayrımı yapılmadan, gün ayrımı yapılarak ve hafta içi ve hafta sonu için kukla değişken kullanılarak yapılan tahmin olarak en genel haliyle ifade edilebilir. Bu çalışmada gün modelleri 5 farklı bağımsız değişken seti kullanılarak tahmin işlemi gerçekleştirilmiştir. Kullanılan bütün metodolojilerde kullanılan yük, sıcaklık veri seti gibi veriler aynı veri seti kullanılarak yapılmıştır. Kullanılan farklı metodolojilerin amacı gün ayrımı yapılmasının ve yapılan ayrımda kullanılan giriş parametrelerinin kullanılan yöntemlerle birlikte tahmin doğruluğunun üzerindeki etkisini araştırmaktır. Veri setinde görülebileceği üzere yük ve meteorolojik veriler mevsimsellik gösterdiği için mevsim parametresi kış, ilkbahar, yaz ve sonbahar sırasıyla (0, 1, 2, 3) dönüştürülerek tüm modeller için bağımsız değişken olarak kullanılmıştır. Bu çalışmada kısa dönem yük tahmini saatlik veya bir diğer ifadeyle bir saat arayla yapıldığı için kullanılan algoritmaların veri seti saatlik olarak tüm modellere tanıtılmıştır. Modellerde kullanılan meteorolojik verilerden nem ve sıcaklık verileri 1 saat önceki nem ve sıcaklık verileridir. Tahmin yöntemleri olarak ARIMA (Otoregregresif bütünleşik hareketli ortalama), destek vektör makineleri, destek vektör regresyonu, KNN (En yakın K Komşu) ve karar ağacı regresyonu algoritmaları kullanılmıştır.
-
ÖgeImaging of multiple slices with single lateral scan(Graduate School of Science Engineering and Technology, 2020)Bessel beams or so-called non-diffracting beams have received great attention from the biomedical optics community, especially in 3D imaging researches, with its long propagation distance in diffractive mediums (see section 1.1). This work presents an architecture to create higher-order Bessel-like optical beams, through off-axis coupling of light into a multimode fiber. Utilizing the aberration that is inherent in spherical surfaces, through directing the Bessel-like beam comprising multiple concentric rings onto a low-cost lens, we created multiple foci at different depths. Moreover, scanning of the optical fiber with a piezoelectric actuator in the lateral direction, we were able to acquire data from different targets located at different lateral planes, therefore proposing a faster confocal system. After coupling back into the fiber, the reflected portions of light from different targets could be spatially differentiated and directed to different photo-detection units. The proposed architecture is particularly appealing for laser scanning endoscopy applications, as it does not require additional passive or active optical components for generating Bessel beams (axicons or spatial light modulators), therefore offering both miniaturized realization and high optical transmission. Secondly, a two-dimensional piezo-fiber actuator geometry is proposed to achieve three different scanning pattern with single device. Having multiple scan options, the operator of the actuator is free to choose between improved uniformity, high frame rate to image abrupt biological event, or circular field-of-view to better image cylindrical cavities (gastrointestinal tract, esophagus, etc.) that are cylindrical in nature. The capabilities of the device has been demonstrated through first presenting its mechanical frequency behavior, then offering an actuation scheme to achieve different scan patterns, and finally generating 256 pixel width or diameter raster, spiral, and Lissajous patterns at ≥ 20 frames per second. With further development, the presented actuator design and drive scheme could serve as a useful tool as part of a minimally invasive optomedical device in the clinic Finally, the compact 3D imaging system utilizing both proposed methods will allow addressing multiple slices at a single lateral scan, thus offering improved speed and different scanning pattern options for endoscopic applications.
-
ÖgeInvestigation of timbral qualities of guitar using wavelet analysis(Institute of Science and Technology, 2020)This study presents a practical and accurate technique using the power of wavelet analysis for investigating timbral qualities of classical guitars. Work includes complete work frame from the experimental procedure through capturing the data with piezoelectric film sensors, which have not been widely applied for this type of application, to the analysis of captured information. Aim is to present a study for more objective evaluation of guitars that is suitable for performing by the different parties of the field; from instrument makers and designers to the quality control engineers and end users. It is believed that with the right improvements and modifications, system presented in this thesis can be used for detailed cataloging of different guitars and their timbral characteristics, which in turn can be used in range of areas such as music information retrieval, guitar research and development and probably even for genuity testing of the guitars. Structure of this study as follows: first understanding the concept that is being investigated, timbre is explored after the introduction. Properties of timbre and the studies that discovered these properties are reviewed. In the following chapter, wavelet theory, which is the backbone and most important aspect of this study is explained in a level that provides intuitive grasping of the concept, because the area is vast and ever improving. Then subject of the study that is guitar is presented through the studies of guitar analysis literature. Then, naturally comes the chapter that the experimental setup and procedure is explained as well as performed analysis and its results are presented with their interpretation and comments. Finally, conclusion of the study with a discussion is given. Experiments was performed in the main studio of Istanbul Technical University/Center for Advance Studies in Music. A mass produced guitar that is described as a learning guitar by its production company and a luthier-made guitar were used. Procedure is done while a guitar player holding the guitar in a conventional playing position and plucking the strings with a pluck. A minor diatonic scale with addition of A minor chord at the end, both in open position and 5th position is played. Choice of the played scale was rather arbitrary whereas the choice of playing the scale in two position on the fingerboard is made to observe the effect of excitation epicenter on the finger board. Piezofilm sensors was used because of their wide frequency range, near-flat frequency response and low mass. Analysis were performed in MATLAB environment using Wavelet Toolbox. First Continuous Wavelet Transform is applied. Continuous Wavelet Transform is a highly computation heavy method for general applicability and the resolution of the results can be far more detailed than necessary for some applications in the area. For this reason, Wavelet Packet Transform is performed to present a more computationally inexpensive solution that provides adequate detail. Results show detailed presentation of harmonic and non-harmonic partials as well as time envelopes of them in a resolution that cannot be obtainable using Short Time Fourier Transform. Resolution of the results allow for objective analysis on the timbre of a guitar as well as making comparison between the timbral properties of the two guitars.
-
ÖgeFlexible load management in active distribution system(Institute of Science and Technology, 2020)Power networks have been developing since they appeared, and this development has accelerated because of developed technology. Along with the changing living conditions, the amount of electricity consumed in the networks has increased regularly every year. On the other hand, industrialization has caused the more big-amount loads to be added to the network, especially industrialization has been a major factor in the increase of electricity demand of developing countries such as southeast Asia and in this direction, their increase in electricity generation capacities or towards them to the import of electricity. As a result of these changes and developments in the networks, the power systems have grown over time and interconnected or wide area synchronous networks have emerged. Sometimes these networks have emerged as sub-networks of a national network, and sometimes by connecting the national networks of different countries. Interconnected networks were created in order to meet the consumption in the network more comfortably, especially in networks where power plants and load weights are distant, in order to carry the generated power to other regions more easily. Moreover, another reason for the emergence of these networks was to increase network stability, security, and reliability. In fact, some international interconnected grids were also helped to improve the operating standards and monitoring mechanisms of some of the countries they include, since the main principle of these networks is that they are fully synchronous in some factors such as frequency. As a result of the synchronous drift that would occur for any reason in the existing sub-networks of the interconnected network, the sub-network is disconnected and immediately separated from the interconnected network. Failure of disconnection or late separation of the network that has synchronous slippage may drag the entire network to a collapse. In addition to the increase in electricity consumption, the nature of the network and type of elements such as load and generators have changed in recent years. Moreover, the inclusion of renewable and distribution generators, which are producing on a small scale, the fact that some loads enter and exit to the network rapidly, and the emergence of a fact like electricity market, in spite of the benefits they add to the network, can be caused some problems if not properly managed. For instance, there are some issues that should be managed well in the electricity market such as taking the offers for the next day and make the necessary preparations by making the day ahead load forecast very carefully. Furthermore, some power plants refrain from bidding for economic reasons, or they may prefer to stay out of the market by placing higher bids, so that the control authority must impose necessary preventive practices in this regard in order to avoid generation deficiencies. It is very important to schedule a planned maintenance in power plants, also small generation sources should be evaluated and planned very well. Even if the market conditions of the day before have been well evaluated, it is inevitable that some problems in the network will still be experienced. These problems may be encountered when the generation and consumption capacity balance is on the border, or when the reserves do not meet the total consumption. The balance of generation and consumption should always be equal in the network, that is, generation or consumption should not be less or more than one another, so the network operator tries to keep the frequency and voltage values in the network within the acceptable range by observing the balance between generation and consumption at any time. In case of severe disturbances in the network, there is a loss of generation generally, which starts to change the frequency and voltage value in the network. In this case, the network operator needs to drop the excess load to ensure balance, this operation is called Load Shedding (LS). Load shedding is generally divided into Under-Frequency Load Shedding (UFLS) and Under-Voltage Load Shedding (UVLS). The difference between these methods is the load shedding is performed by observing the frequency value in one and the load shedding done by observing the voltage value in the other. Conventional methods are available for both load shedding, these conventional methods often attempt to balance the network by removing the load in predetermined steps. These methods are generally decentralized, that is, they do not provide a specific solution for every situation, but they go towards reducing the load in certain buses and at certain levels. Of course, this situation can sometimes lead to under-shedding or an over-shedding of the desired amount, which in turn causes the network to collapse. In addition to the amount of load to be thrown, the time to apply the load shedding is also very important. If the operation is not carried out in time, the network goes to the collapse again and the load to be thrown is not important anymore. The buses where the load shedding is to be applied is another important aspect so that when conventional load shedding methods cannot produce a solution to the situation, they apply the load shedding process at the previously determined buses. In developing power networks, developing new methods for load shedding is at the forefront of experts for a long time. Almost all of the newly developed methods are centralized and they have the ability to generate solutions by observing all the buses in their own responsibility areas and collecting data, evaluating the status of the network in case of problems by using the communication infrastructures. When the grid condition is evaluated and processed with the help of computers, the amount of load, time and place to be thrown can be selected and applied more accurately. These methods proposed in recent years due to increase in the amount of data, limited response time, and most importantly, to achieve the best answer are generally developed with supporting of optimization algorithms. The performance and usability of optimization algorithms in power systems can be clearly seen in other studies topics such as protection systems. In this thesis, flexible load management optimization by using load shedding and optimal reconfiguration in active distribution power networks were studied. Hence, in the first chapter of this thesis, the development, structure, basic concepts, network values which were taken from different countries, and consumption data with real data were shown and compared. Furthermore, information and evaluations about previous studies and the purpose of the thesis were included in this chapter. In the second chapter, voltage stability, voltage collapse, load shedding definition, and classifications were discussed. In the third chapter, the amount, location and timing of load shedding, load shedding, and power flow formulations were discussed. Also, the definition of various optimization algorithms like Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Binary Particle Swarm Optimization (BPSO) were discussed. Finally, an explanation about Optimal Distribution Network Reconfiguration (ODNR), also explanation and formulation of voltage stability index which was used in the second study case are also available. In the fourth chapter, for the first study case, simulation and results of the applied method were given. At this stage, after evaluating a 69-bus and 33-bus IEEE distribution test system under normal conditions, the system was self-healing with optimal switching with the help of the BPSO optimization algorithm, and the results were shared. Then, the BPSO optimal switching and load shedding through the power flow process were performed together, and the results were given for each evaluated in three cases. A novel method was proposed in the second study case. In this study case, in the 33-bus IEEE distribution test system, a smart grid was considered. load management was carried out with controllable loads in smart homes by load shedding with the help of the PSO optimization algorithm. The results of this newly proposed method were given for two different analysis cases. In the last chapter, conclusions and recommendations regarding both study cases of the thesis were presented in the fifth chapter.
-
ÖgeVeri işleme prosedürleri ve hatalar(Fen Bilimleri Enstitüsü, 1984-06)