18
Archived Theses
0
DOIs Assigned
0%
DOI Rate
Discipline
İkinci dereceden yöntemler ile ardışık bağlanım teknikleri
Sequential regression problem is one of the widely investigated topics in the machine learning and the signal processing literatures. In order to adequately model the underlying structure of the real life data sequences, many regression methods employ nonlinear modeling approaches. In this context, in the first chapter, we introduce highly efficient sequential nonlinear regression algorithms that are suitable for real life applications. We process the data in a truly online manner such that no storage is needed. For nonlinear modeling we use a hierarchical piecewise linear approach based on the notion of decision trees where the space of the regressor vectors is adaptively partitioned. As the first time in the literature, we learn both the piecewise linear partitioning of the regressor space as well as the linear models in each region using highly effective second order methods, i.e., Newton-Raphson Methods. Hence, we avoid the well-known over fi tting issues by using piecewise linear models and achieve substantial performance compared to the state of the art. In the second chapter, we investigate the problem of sequential prediction for real life big data applications. The second order Newton-Raphson methods asymptotically achieve the performance of the "best" possible predictor much faster compared to the fi rst order algorithms. However, their usage in real life big data applications is prohibited because of the extremely high computational needs. To this end, in order to enjoy the outstanding performance of the second order methods, we introduce a highly efficient implementation where the computational complexity is reduced from quadratic to linear scale. For both chapters, we demonstrate our gains over the well-known benchmark and real life data sets and provide performance results in an individual sequence manner guaranteed to hold without any statistical assumptions.
MPEG-4 ve H.264 video şifrelemesi için AES tabanlı algoritma
Nowadays data security has become the most global issue after global warming. Expectations about a world war in the field of data seem on the horizon. The internet connects all the people through many ways, such as e-mails or voice massages or even video calls. However, the flow of data between the source and destination which is the basic operation of the Internet, can by interrupted by another user or intruder. In order to prevent this intervention there are security algorithms meant to encrypt all the bits from starting point of flow up to the end point. The most widely used flow of data in the Internet is video streaming which takes about 80% of the data flow. In this thesis AES algorithm is used to encrypt the video streaming data. The present research aims to maintain a secure path for a video between the source and the destination by modifying and optimizing the existing AES data securtity algorithm. Amongst the analysed parameters are, PSNR, encryption time, and overhead bit through encryption with compression.
AES algoritmalarını kullanarak güvenli ses
This thesis study encryption \decryption text and voice by using AES algorithm, then comparing between them to know which one is better by using three standards. Speed, time duration and security. In this thesis we used five operation modes first ECB - Electronic Code Book. However the second one CBC - Cipher Block Chaining. The third one OFB - Output Feedback. The forth operation mode CFB - Cipher Feedback. The last operation mode is, CTR - Counter Mode. We cannot say which mode operation is the best. Because each of them have advantages and disadvantages but for encryption \decryption text counter mode is better than other mode operation. Because it is faster and simple for using more than other mode operations. After encrypted and decrypted audio by using the AES algorithm. Obtained different results however needed more time than text for encryption and decryption .in order to audio have more data than text Therefore, need more time than text for encryption and decryption (AbdelrahmanAltigani, Bazara Barry, 2013). According to speed for encrypt and decrypt text is faster than audio in order to for encryption and decryption audio have many data (AbdelrahmanAltigani, Bazara Barry, 2013).But text has less data if compared to voice. Then less data need less time. According to security audio is more secure than text and also more difficult to hack by hackers. Utilized Matlab code in this thesis in order to it has the capacity to plot graphical inclusive. use MATLAB is more appropriate than other packages like java and C#,in order to it has a large a collection of tools and libraries; moreover, for programming, it have not necessary for beginning from zero in order to MATLAB previously has functions for performing limited missions whilst some programs like C# should needs for beginning from zero.
Blokzincir teknolojisi temelli dijital oylama sistemi
In recent years, blockchain technology has, to a large extent, affected all aspects of life, however, until now, elections have not received their share of blockchain support and paper-based elections have been implemented. It is time to upgrade the election scenario using modern technology such as blockchain and advanced cryptography methods. Both Estonia and the New South Wales state of Australia have been using e-Voting systems, but an example software of their systems produced for analysis was discovered to have weaknesses against many kinds of attacks, such as malware, network attacks, and servers' attacks. The fact that the Blockchain technology has demonstrated infinite immutability and resistance against hacking illustrates that its use to secure election results from fraud by saving every single piece of data, record or transaction with unchangeable history is possible. In this thesis, we propose and test implement a secure online voting system based on blockchain that facilitates citizens' election process. The essence of our work consists in abandoning the traditional database and compensating it with two private-blockchains instead of one with the blind signature to ensure voter/vote privacy and the results can be safeguarded from manipulation. Also, using Blockchain's distributed network reduces the load on the net. Finally, solutions to problems of impersonation and vote-selling are suggested. The technology behind the digital voting system design is explained in terms of the processes involved, such as ID creation, authentication, voting, and vote tallying.
İtmeli bir elektrikli araç sisteminin modellemesi ve simülasyonu
Electric Vehicles (EVs) are a scientific revolution in the development of cars and clean energy concerning the use of electricity as a source of energy instead of fuel. Raise in global warming, polluting the environment and loss of petroleum products make the EVs the perfect choice to overcome these problems. The main purposes of the electric vehicles are powerful structure, less of maintenance and less of power. Each one of these characteristics originated from the drive system because the drive system signifies the backbone of the EVs. In this particular thesis, there is two synchronous reluctance motor (SynRM) which have been used to operate the vehicle. In this system, the space vector pulse width modulation inverter for voltage source (VS-SVPWMI) has been employed to convert the DC battery voltage to three-phase AC voltage that feeds the electric motor. The optimized cascaded PID controller is used to regulate the SRM speed. Particle Swarm Optimization (PSO) algorithm has been applied to reach the value of the optimal parameter for the controller. Electronic differential controller (EDC) and electric vehicle system model were simulated. The EDC regulates the electric vehicle performance balance under the various procedure and road conditions. The proposed electric vehicle gives a stable and suitable performance along the different road conditions for each driving cycle. Matlab/Simulink program was used to control the propulsion drive electric vehicle system in this study.
Intra-Corneal Lens'i̇n post-operati̇f parametreleri̇ni̇ koruma i̇çi̇n akilli bi̇r algori̇tma
.
Kırılma indisi yapı parametresinin istatistiksel analizi
It is possible to model turbulence, which is one of the biggest problems encountered when designing an optical wireless communication system, with only the refractive index structure parameter and good estimation of the rest of the parameters. Which is why knowing the probability distribution function of the refractive index structure parameter is very important. Although different models have been proposed in this regard, none have been proved theoretically. The main topic of this study is to examine whether the probability distribution of the refractive index structure parameter is bounded or semi-infinite. At the start, Pearson and Johnson distribution systems and Extreme Value Theory were applied and it was concluded that the refractive index structure parameter could be better modelled with bounded distributions. After that, beta distribution and lognormal distribution, which belong to bounded and semi-infinite distributions families respectively and the more known and used distributions as a model for refractive index structure parameter, are compared with each other and it was found out that beta distribution gives better results. Finally, it had been shown that the method of moments, which is one of the methods for estimating the beta distribution parameters, is not enough and after the estimations are made by the method of moments, it is necessary to improve them with the least squares method. Keywords: Wireless optical communication, Optical turbulence, Refractive index structure parameter, Pearson distribution system, Johnson distribution system, Extreme Value Theory, Mean excess function, QQ plot, Method of moments, Least squares method, Chi-Square test, K-S test.
Güç sistemi korunması tabanlı facts kontolörlerinin gelişmiş tepkisi
The Flexible AC Transmission System (FACTS) controllers have been essentially utilized for determining different power systems steady-state, voltage improvement and performance of system protection. The features and advantages of the FACTS controllers it is suitable and necessary to be applied in the electrical power system. In this thesis, the influence of FACTS controllers will be investigated therefore tow grid system will be designed with (six-buses), the first circuit has tow step-down transformer (132/33kV, 66/33kV), the SVC control and static synchronous compensator will be added to different location of grid, such as (bus-3, bus-5and bus-6) to improve the voltage profile, angle, real power and reactive power of grid buses in power flow circuit. This study presents the effect of three cases after adding compensator and before of indictive reactance (0.20) p.u and capacitive reactance(0.50,0.60 and 0.70) p.u on results of state variable of SVC by added to(bus-3, bus-5and bus-6). The load values of the circuit will be changed gradually to determine generation and loss of power. The second circuit (132/33kV ) will be using SVC Simulink modal with 1 TCR/3TCS, to demonstrate Waveform output of voltage, current, reactive power before and after adding SVC and see the effect of three-phase of fault among lines and ground also we can say action of protection that will improve the power transfer throw transmission line. The results will be evaluated by Power System Analysis Toolbox (PSAT) also MATLAB program. Keywords: Power system protection, FACTS, Static var compensator (SVC), Static synchronous compensator, Power flow.
3 serbestlik derecesine sahip (RRR) robotun yörünge planlaması ve uygulaması
In this study, the path planning of the RRR robot with 3 degrees of freedom has been made and has been applied on the model robot arm produced with 3d printer. Time optimization has been made by using Genetic Algorithms during path planning. Since the first three joints affect the position in the Cartesian space, the robot arm with the RRR structure which has the three rotating joints, was used as the model. Firstly, the forward kinematic analysis of the robot arm was performed. Denativ-Hartenberg method, which is the most widely used analysis method, was chosen as the forward kinematic analysis method. Inverse kinematic analysis was performed with the information obtained from the forward kinematic analysis. As a result of inverse kinematic analysis, the joint variables required for the model robot to reach a point in Cartesian space were obtained. Kinematic analyzes were performed to planning path in the joint space of the robot. Time optimization was performed by using Genetic Algorithms in order to complete the movements of the robot arm as soon as possible while planning path. The velocity and acceleration equations obtained for each joint as a result of path planning in joint space were used as objective functions in optimization. The limits of the objective function are the velocity and acceleration values of the servo motors used in the joints of the robot arm. As a result of optimization, it was found that each joint can complete the movement as soon as possible. The model robot arm was produced with the 3d printer and the experimental set was created. The data obtained as the result of optimization were compared with the data obtained from experimental studies. At the end of this study, time optimization was made by using Genetic Algorithms and a model which completed the movement in the shortest time was created that could be applied to any robot.
Fotovoltaik panel hatalarının termal kamera ve UV led ile tespiti
In this thesis, hot spot failures, panel surface fractures, and yellowing problems on 40 Watt polycrystalline photovoltaic (PV) panel were analysed through indoor laboratory experiments and outdoor fieldwork. The outdoor field tests were performed with a 40 Watt PV panel on a clear and sunny day in the garden of Ortaköy Vocational School of Aksaray University. It is observed that the output power value of the panel decreases in the open circuit fault which was created by disconnecting the series-connected cells within the panel. In addition to the observed change in the power value, temperature increase value due to broken glass surfaces and disconnected cells on the panel was observed with a thermal camera. In the indoor laboratory experiments, artificial shading and yellowing parts detection studies were performed with UV LED and photodiode which is designed as a rectangular prism and placed at 45-degree angles. To perform these works on the photovoltaic (PV) panel, the user interface was developed and the graphical values of the manually scanned panel were plotted on the screen. Besides, colour changes on the PV panel were tried to be determined at different spectrum values. Blue, green, yellow and red LED have been used to do this. It is understood from the figure that the blue led spectrum is most effective method to obtain PV diagnosis. According to the results obtained, colour changes on the PV panel could not be obtained in colours other than blue. Additionally, common faults in PV systems which are shading situation, line to line faults, panel soiling problem, increasing resistance value problem between panels, problems caused by bird droppings or tree leaves and branches, and the effect of temperatures on the photovoltaic (PV) panels were simulated, analysed, and P-V and I-V graphics of the panels were drawn.
Yol kenarı ünitesi kullanılarak AD-HOC tabanlı gezgin araç performansinin iyileştirilmesi
VANET is used basically to exchange safety signaling between vehicles to alert drivers of accidents existence; so, they can change their root or to take other precautions to prevent car crashing near accident zone. The motivations of using this kind of networks are their lower cost as compared to other mobile networks and their ease of deployment. In this study, mobile vehicles are randomly moving in the high ways and hence in case on car crashing, car will assume stopped and the place of crashing is susceptible to intake more than one vehicle. As vehicles are being droved in high speed and drivers; due to their unawareness of accident zone, they can get into it and hence problem is enlarging. On the other hand, ad-hoc network is attempting to share safety norms to the inward drivers to avoid the crashing in that particular location. For some reasons more likely due to car limited radio converge and speed of the cars running on the high way, the reachability of this message is critical and hence, network further development was mandatory. The presence of RSU is helped network payloads to be extended to other nodes which is already far from recurrent node coverage. We proposed using the mobile node to act as RSU and perform data packet routing as RSU does. The main problem of using large number of RSUs is performance degradation be consumption of large time for data delivery. In our study, we attempted using different number of mobile nodes as RSUs which is differs from classical RSUs as the last is static and other is dynamic (in motion). Outcomes of this study shown that large number of mobile nodes as RSUs may enhance the connectivity by reducing the link re-healing time and maintaining the connection for longer time.
Tavsiye sistemi için veri madenciliği tekniklerini kullanarak veri seyrekliğini ve soğuk başlatma sorunlarını hafifletmek için çeşitli modeller
The development of Web 2.0 and the rapid growth of available data have led to the evolution of multiple systems among them the Recommendation Systems (RSs) which can handle the information overload. Due to the fact that, RSs performance is substantially limited by sparsity and cold-start problems; this research study takes the route for a major objective to attenuate these problems. To realize this objective, four data mining techniques are proposed namely: multi-steps resource allocation- singular value decomposition (MSRA-SVD), MSRA-SVD++, clustering community detection, and overlapping community detection. The first two models are dedicated to tackle the data sparsity problem whereas the rest models are adopted to alleviate cold-start users' problem. The core strategy of the first two models is to use the (MSRA) method to identify hidden relations in social network. the MSRA method is applied to determine the probability of their relation. If the probability exceeds a threshold, a new relationship will be established. For the second model (MSRA-SVD++), an implicit feedback source is exploited as an additional source of information, which can be extracted via rating information. Additionally, clustering and overlapping models are adopted to overcome cold-start user problem. The main idea of these models is to apply a clustering technique to group users into several communities. In order to attain that, explicit and implicit social relations with the confidence values are integrated to compute distance values. Additionally, the partitioning around medoids (PAM) clustering algorithm is adopted. Later, for the last model, the average distances of all clusters are computed. For all users, the distance between users and the center node of a particular cluster is computed. If the distance is less than the average of this cluster, a new user for this cluster will be added. Moreover, the SVD++ method is employed for each cluster to compute the prediction value. The proposed models are evaluated through the usage of three real-world datasets. Ultimately, findings exhibited a great deal of insights on how the proposed models outperformed a number of the state-of-the-art studies in terms of prediction accuracy.
Akıllı bina içi nitrojen dioksit sensorlerı ağına dayalı serbest uzay optik iletişimi
The Intelligent Wireless Sensor Network (IWSN) is one of the widely used applications in present time, whereas it is utilized for: 1) sensing of the natural phenomenon or environmental conditions that occur in our world, 2) processing the incoming data from their sensors, 3) presenting a appropriate decision according to the entered data. In this work, an Intelligent Wireless Sensor Network has been designed and implemented, which can be used for applications of carbon monoxide gas sensing, it involves three main parts, the intelligent system, the sensor interface unit, and the wireless communication system. The proposed intelligent system has been utilized to processing the incoming data from carbon monoxide sensors, then presenting the average value of these data. The sensor interface unit has been used for converting the incoming analog signals from the nitrogen dioxide sensors to binary data that should be driven to the intelligent system, where the last one should be saved in FPGA (Field Programmable Gate Array). The wireless communication system has been implemented for transferring the digital data between the sensor node and the intelligent system for remote distances. A Back-Propagation neural network has been utilized as an intelligent system for this work, three layers had been designed in this network, input, single hidden, and output layers. This network has been trained by several training functions, and has used two linear activation functions, the SATLINS function for the hidden layer, and the SATLIN for the output layer. Using TRAINPSO (Particle Swarm Optimization) training function, an optimal result has been also presented, but with reaching the MSE to zero value in 46 iteration using of only three neurons in the hidden layer. A laser FSO(Free-Space-Optical) system has been designed and implemented as a wireless communication unit for the proposed system, cause this technique possesses low power consumption and high bandwidth, directivity, immunity than other techniques for same range of transmission. The ON/OFF keying modulation has been used in this technique for modulating the digital data of the sensor units with an infra-red laser light carrier, which is a powerful widely used modulation technique in the laser communication systems.
CAD for corneal diseases based on topographical parameters to improve the clinical decision
Bilgisayar Destekli Tanı, tıbbi görüntüde önemli bir konudur. tıpta hekimlere tıbbi görüntülerin yorumlanmasında yardımcı olan sofistike bir prosedürdür. İnsan korneası, gözün önden şeffaf siperi. Görmeyi tetiklemek için ışığı retinaya yansıtır. Bu nedenle korneadaki herhangi bir kusur görme bozukluğuna neden olabilir. Bu eksiklik, oftalmologlar tarafından ölçülen ve değerlendirilen bir dizi topografik görüntü ile tahmin edilmektedir. Sonuç olarak, önemli bir öncelik, makine öğrenme algoritmaları kullanılarak kornea bütünlüğünü etkileyebilecek hastalıkların erken ve doğru teşhisidir. Bir Pentacam cihazı tarafından üretilen kornea görüntüleri, edinim sırasında rotasyona veya bazı bozulmalara maruz kalabilir; bu nedenle, doğru teşhis, görüntüdeki yerel özelliklerin kullanılmasını gerektirir. Buna göre, bu zorlukların üstesinden gelmek ve cornel koşullarını teşhis etmeyi iyileştirmek için bu çalışmada önerilen yeni algoritmalar. İlk olarak, kornea görüntülerinden yerel özelliklerin çıkarılması için bir SWFT algoritması önerildi. Dalgacık dönüşümü, standart SIFT algoritmasında olduğu gibi Gauss Farkı (DoG) kullanmak yerine farklı ölçeklerde görüntüler üretmek için kullanılır. İkinci olarak, IG-GLCM algoritması, zaman alıcı bir kusur olarak bilinen GLCM algoritmasının dezavantajının üstesinden gelmeyi önerdi. IG-GLCM'de görüntü gradyanı farklı yönlerde ölçülür ve ardından GLCM'yi oluşturulan görüntülere uygular. Üçüncü olarak, SIFT'in dalgacık dönüşümünün çok ölçekli alt bantları ile kullanımını araştırın. Son olarak, Yerel Bilgi Modeli tanımlayıcısı adlı yeni algoritma, görüntüden bilgi kaybına neden olan yerel ikili model eksikliğinin üstesinden gelmeyi ve görüntü döndürme sorununu çözmeyi önerdi. LIP, hangi sözde kontrast Tabanlı Merkez (CBC) değerinin yanı sıra yerel paterni (LP) hesaplamak için kullanabilen komşuların ağırlıklarını tahmin etmek için alt görüntü merkez yoğunluğunu kullanmaya dayanır. Naive Bayes, KNN, karar ağacı ve SVM sınıflandırıcılar olarak kullanıldı. Önerilen model, farklı haritaların 4848 görüntüsünden oluşan toplanmış bir veri kümesi üzerinde eğitilmiş ve başarıyla test edilmiştir. Anahtar Kelimeler: Bilgisayar destekli teşhis, öznitelik çıkarma, makine öğrenimi, destek vektör makineleri, yerel İkili Model, GLCM
Robust idss design for IoT based on smart algorithms
Genel olarak Nesnelerin İnterneti'nin (IoT'ler) ve özellikle Hareketli Nesnelerin İnterneti'nin (IoMT'ler) güvenlik açıkları, araştırmacıları davetsiz misafirlere ve saldırılara karşı güvenlik sistemleriyle donatmaya motive eder. IoMT'ler için anormallik tespitinin izinsiz giriş tespiti ile entegrasyonu yeterince ele alınmamıştır. Bu çalışma, anormallik tespiti için bir Kalman ve Cauchy kümelemesi oluşturarak ve bunu Extreme Learning Machine (ELM) sınıflandırıcısını kullanarak IoMT'ler içindeki kimlik doğrulama düğümleri için kullanarak bu sorunu ele almaktadır. Algoritma, çeşitli bileşenlerden oluşur. Bunlardan ilki, WiFi'yi IMU verileriyle birleştirmeye dayalı bir kapalı ortam içindeki yayaların yörüngesini tahmin etmek için Kalman filtresi tabanlı model, ikincisi Kalman filtresini kullanarak tahmini yörüngeye dayalı olarak IoMT'deki anormallik davranışını tespit etmek için güvenilirlik değerlendirmesi, üçüncüsü de, bir Online Sequential Extreme öğrenme makinesi (OSELM) kullanarak saldırıların tanımlanması için anormallik algılamayı çevrimiçi öğrenme ile entegre ederek IoMT sistemleri için IDS modelidir. OSELM algoritması, WiFi parmak izi için TamperU veri seti ve izinsiz giriş tespiti için KDD99 kullanılarak uygulanmış ve değerlendirilmiştir. Ayrıca, izinsiz giriş tespiti ve anormallik tespiti için karşılaştırma ölçütleri ile yapılan bir karşılaştırma, önerilen tüm sınıflandırma ölçütleri açısından önerilen yaklaşımın üstünlüğünü kanıtlamaktadır. Geliştirilen algoritma, anormallik tespiti için mevcut iki modelle, yani gelişen veri akışı için bir çok yoğunluklu kümeleme algoritması (MUDI) ve gelişen veri akışlarının rastgele şekillendirilmiş kümeler halinde tamamen çevrimiçi kümelenmesi (CEDAS) ile karşılaştırıldı. Sonuçlar, bu çalışmada geliştirilen algoritmanın, kullanılan anormalliklerin farklı yüzdeleri, farklı sayıda yaya sayısı ve farklı ortalama yaya hızlarını içeren üç farklı senaryo altında anormallik ve izinsiz giriş tespiti açısından üstünlüğünü kanıtlamıştır.
Akdeniz meyve sineğinin (Ceratitis capitata) akıllı sistem ile tespit edilmesi
Nowadays, the most critical agriculture-related problem is the harm caused to fruit, vegetable, nut, and flower crops by harmful pests, particularly the Mediterranean fruit fly, Ceratitis capitata, named Medfly. Medfly's existence in agricultural fields must be monitored systematically for effective combat against it. Special traps are utilised in the field to catch Medflies which will reveal their presence and applying pesticides at the right time will help reduce their population. A technologically supported automated remote monitoring system should eliminate frequent site visits as a more economical solution. This paper develops a deep learning system that can detect Medfly images on a picture and count their numbers. A particular trap equipped with an integrated camera that can take photos of the sticky band where Medflies are caught daily is utilised. Obtained pictures are then transmitted by an electronic circuit containing a SIM card to the central server where the object detection algorithim runs. This study employs a faster region-based convolutional neural network (Faster RCNN) model in identifying trapped Medflies. When Medflies or other insects stick on the trap's sticky band, they spend extraordinary effort trying to release themselves in a panic until they die. Therefore, their shape is badly distorted as their bodies, wings, and legs are buckled. The challenge is that the deep learning system should detect Medflies of distorted shape with high accuracy. Therefore, it is crucial to utilise pictures that contain trapped Medfly images with distorted shapes for training and validation. In this academical study, the success rate in identifying Medflies when other insects are also present is 94.05%, achieved by the deep learning system training process, owing to the considerable amount of purpose-specific photographic data. This rate may be seen as quite favourable when compared to the success rates provided in the literature.
Uçak motorlarının baroskopik incelemesinde derin öğrenme kullanılarak hasar tespiti
Aircraft engine inspection is a key pillar of aviation safety by maintaining adequate performance standards to ensure the airworthiness of an engine. In addition, it is vital for asset value retention. Borescope inspection is currently the most widely used aircraft engine visual inspection method. However, borescope inspection is a time consuming, subjective, and complex process which heavily depends on the experience of the inspector as well as on their concentration level during inspection. On the other hand, cost saving of airlines and maintenance, repair, and overhaul (MRO) centers expose pressure and workload on inspectors. These make an automated system to support damage detection during borescope inspection necessary to avoid potential risks. Deep learning has found very wide application and has proven to be very successful during the last 10-15 years in the image recognition domain. In this research, we suggest a deep learning based automated damage detection framework from aircraft engine borescope inspection images. Faster R-CNN based deep learning model with Inception v2 feature extractor is utilized for the architecture. Due to the limited number of images, data augmentation and other overfitting methods are employed. The framework supports crack, burn, nick and dent damage types across all modules of turbofan engines. It is trained and validated with moderate to complex borescope images obtained from the field. The framework achieves 92.05% accuracy for nick or dent, 92.64% for crack and 81.14% for burn damage classes. Moreover, it achieves 88.61% average accuracy.
Topluluk öğrenme teknikleri yoluyla maksimum güç noktası takipini geliştirme
Maximum Power Point Tracking (MPPT) is an essential method in photovoltaic (PV) solar systems for optimizing the extraction of available power. This technique enhances energy conversion efficiency and aligns with ongoing efforts to improve the effectiveness of renewable energy sources. This thesis presents a systematic investigation into the predictive modeling of solar energy phenomena, specifically focusing on solar power generation and solar radiation prediction. Through the comprehensive evaluation of various individual machine learning models, including Linear Regression (LR), Support Vector Regression (SVR), and XGBoost Regressor, as well as an Ensemble Learning (EL) approach, the study elucidates the complexities and nuances of modeling solar energy systems. The analysis was conducted on two distinct datasets: Solar Power Generation and Solar Radiation Prediction. The individual models were rigorously assessed using multiple statistical metrics, revealing varying degrees of accuracy, fit, and performance. A remarkable discovery was the efficacy of the Ensemble Learning model, employing techniques such as Bagging Regressor, which consistently outperformed the individual models across both datasets. By adeptly aggregating the predictions of multiple underlying models, the EL approach achieved superior predictive accuracy, explaining an impressive proportion of the variance in both solar power generation and solar radiation. The findings of this research contribute significantly to the understanding of solar energy modeling, endorsing ensemble learning as a potent and versatile tool for enhancing prediction accuracy. Moreover, the comparative analysis sheds light on the trade-offs between different modeling techniques, offering guidance for future research and practical applications within the renewable energy sector. This thesis not only sets a new benchmark in the field of solar energy forecasting but also aligns with the broader imperatives of sustainable energy management and climate stewardship.