Search

Sistemi za distribuciju električne energije često se modeluju grafovima gde korenski čvorovi simbolizuju transformatorske stanice koje dobijaju električnu energiju iz prenosnog sistema i isporučuju je potrošačima preko izvoda sastojanih od transformatora, razvodnih postrojenja, kablova i vodova. Visoko-kvalitetni dijagrami koji prikazuju ove sisteme nazivaju se jednopolne šeme i ključni su elementi za efikasan nadzor i upravljanje tim sistemima. Postojeća rešenja za automatsko generisanje dijagrama obično su ograničena na prikaz pojedinačnih izvoda ili sadrže značajno pojednostavljen model distributivnog sistema, koji predstavlja više izvoda sa uklonjenim međusobnim vezama. Cilj ovog istraživanja bio je da se razvije algoritam sposoban da generiše vizualno pogodne jednopolne šeme područja više transformatorskih stanica međusobno spojenih koji se sastoje od više korenskih čvorova i složene mreže od više hiljada međusobno povezanih elemenata. Rezultujući algoritam je definisan kao niz transformacija koji uzima model EES-a u CIM formatu kao ulaz i generiše jednopolnu šemu kao izlaz, takođe modelovanu u skladu sa CIM standardom. Korišćeni su različiti pristupi za optimizaciju algoritma, obezbeđujući efikasno izvršenje na standardnim personalnim računarima. Takođe, razvijena je nova procedura ocene kvaliteta dijagrama kako bi se automatski izmerio i uporedio kvalitet svakog generisanog dijagrama. Algoritam je testiran na 2510 različitih oblasti transformatorskih stanica različite složenosti izdvojenih iz mrežnih modela realnih evropskih distributivnih sistema. Kvalitet generisanih jednopolnih šema je izmeren novom procedurom evaluacije i verifikovan sa stručnjacima iz oblasti na podskupu generisanih dijagrama. Na osnovu ovih analiza je zaključeno da su generisane šeme pogodne za integraciju u novoizgrađene sisteme nadzora i upravljanja uz minimalnu intervenciju čoveka., Electric power distribution systems are often represented with graphs where hub nodes symbolize substations receiving power from the transmission system and distributing it to consumers through feeders, containing of transformers, switching kiosks, cables, and power lines. Creating high-quality diagrams that accurately portray these systems is crucial for effective monitoring and control. However, current automatic one-line diagram generation solutions often have limitations, such as only drawing single feeders or oversimplifying the distribution system by excluding interconnections between multiple feeders. This research aimed to develop an algorithm capable of generating visually appealing diagrams for multiple substation areas, consisting of multiple hub nodes and a complex network of thousands of interconnected elements. The algorithm was conceptualized as a sequence of transformations, taking an EPS model in CIM format as input and producing one-line diagram as output, also modeled in compliance with CIM standard. Various innovative approaches were incorporated to optimize the algorithm, ensuring efficient execution on standard desktop computers. Additionally, a novel diagram quality evaluation procedure was created to automatically assess and compare the usability of each generated diagram. The algorithm underwent testing on 2510 distinct substation areas with varying complexities, extracted from real-life European distribution system network models. The quality of the generated diagrams was quantified using the new evaluation procedure and validated by human domain experts who assessed a subset of the resulting diagrams. The analyses concluded that the generated diagrams are suitable for integration into newly constructed supervisory and control systems with minimal human intervention.

The research subject encompasses the automation of paper test grading through the utilization of artificial intelligence: machine learning and computer vision. Computer vision is concerned with enhancing the capabilities of the human visual system and concentrates on processing digital images and videos. Despite the various categories within computer vision, the majority is centered on detecting objects and recognizing patterns. As algorithms advance, specialized systems are crafted to undertake tasks akin to humans, and technological advancements empower the formation of systems that, in certain aspects, surpass human capabilities. While numerous e-learning and assessment platforms such as Moodle exist, paper tests have not waned in popularity. Diverse reasons contribute to this, often tied to the constraints of computer resources. Consequently, assessing a significant number of candidates is conducted by hand, utilizing standard or ballpoint pens. Given the volume of applicants and the urgency for prompt outcomes, manual grading poses a considerable challenge. The study also delves into analyzing frequently employed question types on paper tests and their categorization based on the core concepts these questions examine. Additionally, prevailing solutions within this domain undergo scrutiny and classification through a fresh issue-based system. The evaluation of the proposed system occurs via a comparison against prevailing systems within the same question category. The research objective is the development of a system for automated paper test assessment through the fusion of artificial intelligence: machine learning and computer vision. The system processes digital images of test pages and employs algorithms for identifying key regions and scoring questions. Furthermore, the system processes a configurable file containing the test's structure, using it to verify the accuracy of the detection process. The actualized system not only assesses questions but also iden

User-generated content on Web forums is added much more often than it is deleted or changed, so its targeting during incremental crawling differs from the Web site pages crawling. Adding new content to a forum can result in moving existing content to new or existing pages. Incremental forum crawling is not a trivial task, because ignoring in which way the content is presented, distributed and sorted can lead to the transfer of posts that have already been indexed in the previous crawl cycles. On the other hand, there is a wide spectrum of forum technologies that allow different navigational paths to its latest posts, as well as different ways of presenting and sorting user generated content. This thesis presents Structure-driven Incremental Forum crawler (SInFo) that specializes in targeting the latest content in incremental forum crawling using advanced optimization techniques and machine learning. The main goal of the presented system is to avoid already indexed content in new crawling cycles regardless of its technology. In order to achieve this, the following Web Forum features have been used: (1) the sort method on the index and thread pages and (2) the available navigation paths between the pages that the current Web Forum technology offers. Since the date of content creation plays an important role in determining the type of sort, their detection and normalization is not a trivial task. Machine learning models were used for this task, because the generated dates can be in different formats and in different languages. On the other hand, the detection of navigational paths is achieved by interpreting the URL format and scanning the pages they target. It has been shown that using the proposed methods and techniques while targeting pages with the latest content can achieve a minimum number of duplicate content downloads and maximize the utilization of the navigational structure and paths of the current forum technology. The experiments were performed on a wide ra

The progress of CMOS technology very large scale integration has resulted in the intensive development of sensor technologies, wireless communications, and energy efficient processors, which together make sensor nodes capable of perceiving our environment, processing data and exchanging data with other devices and users on the Internet. Such technological development has led to the emergence of the Internet of Things (IoT) vision, with the goal of providing information of the real world that surrounds us to the Internet users. As one of the prerequisites for the realization of imaginary IoT services and products, the horizontal integration of deployed sensor networks is required, by implementing platforms that are able to serve a large number of users and would allow the integration of heterogeneous sensor networks consisted of different sensor devices, using different communication protocols and message formats. The subject of this dissertation is the architectures for the integration of sensor networks, which generally support IoT applications with high performance requirements. Among the available generic types, an architecture based on the data semantics is selected and the aim of this dissertation is the analysis of the architectures that enable semantic integration of sensor networks using Semantic Web technologies in order to achieve the interoperability of sensor data and networks. The basic generic architecture types were first identified and their key characteristics and method of implementation were given. For architectures based on the gateway and the message broker, the performance evaluation was performed in the scenario of real-time sensor messages delivery by creating a simulation environment. Then the architectures with semantic-based integration of sensor networks are classified by identifying typical design approaches. Two groups of approaches are identified, approaches oriented to sensor networks and applicationoriented approaches, whereas each g

The aim of this doctoral thesis is to present a software system for learning and application of the artificial intelligence algorithms SAIL (System for Artificial Intelligence Learning). Within the system 30 standard algorithms have been realised: search and game theory algorithms, reasoning and knowledge representation algorithms, problem solving algorithms, uncertain environment algorithms and machine learning algorithms. The idea to realise this system was generated during the teaching of the course Intelligent Systems, where students had difficulties to understand these algorithms and related concepts due to their complexity. The importance of this system is in its applications – in addition to its educational purposes, the system can be applied in solving practical and real problems. Its applications in finding free parking spot within the smart parking with wireless sensor network, within the system for decision-making for medical applications and within the big data processing system have been demonstrated. This research work included the review of the existing software systems that are used in this area at universities across the world, application of such systems in electronic and mobile learning, individual and distance learning, requirements definition for the development of completely new software system following the analysis of the existing systems, realisation of the new multiplatform system of common user interface for all the simulations and applications...

M60 M63

M30 M33

M60 M63

Paralelni testovi znanja su testovi koji sadrže različite skupove stavki (skupove pitanja ili zadataka različitih vrsta) sa približno jednakim mernim osobinama, što ih čini međusobno zamenljivim. Oni imaju primenu kada više ispitanika rešava test, bez obzira da li to čine u isto ili u različito vreme. Rezultati dobijeni testiranjem moraju da budu poredivi. Automatizam u procesu sastavljanja paralelnih testova znanja je praktično neophodan da bi se postigla objektivnost ispitivanja dobrim balansiranjem mernih osobina različitih primeraka testa i smanjila mogućnost grešaka ispitivača u merenju znanja. Što je testiranje masovnije, ušteda uloženog vremena ispitivača za pripremu paralelnih testova, primenom automatizacije sastavljanja, postaje značajnija. Pošto problem automatskog sastavljanja paralelnih testova (ASPT) spada u NP-teške probleme kombinatorne optimizacije, korišćenje heuristika je očekivano i neophodno. Problem ASPT je izomorfan sa nekim postojećim problemima kombinatorne optimizacije, pa se heuristike za rešavanje tih problema mogu koristiti i za rešavanje problema ASPT. U disertaciji je dat opsežan pregled postojećih formulacija ovog problema, odnosno njihovih matematičkih modela, a zatim pregled heuristika koje se koriste za rešavanje različitih formulacija problema. Najpre su formulacije problema i heuristička rešenja nezavisno klasifikovani, a zatim se njihovom unakrsnom klasifikacijom ukazalo na otvoren prostor za istraživanje kroz nove kombinacije formulacija problema i heuristika za njihovo rešavanje. Heuristike koje se koriste za rešavanje problema ASPT su uglavnom heuristike poboljšanja, čije se izvršenje prekida ili kada dođe do unapred zadatog vremena izvršavanja ili kada je dostignut željeni kvalitet rešenja. Konstruktivne heuristike se uglavnom koriste za kreiranje početnog rešenja za heuristike poboljšanja. Fokus ove disertacije je na predlogu novog efikasnog metoda polinomijalne kompleksnosti za kompletno rešavanje problema ASPT. Predloženi metod rešava jednu formulaciju Problema pakovanja u korpe prilagođavanjem konstruktivne heuristike koju su predložili Nawaz, Enscore i Ham, a koja do sada nije primenjivana za rešavanje problema ASPT. Među karakteristikama novog metoda ističu se predvidivo vreme njegovog izvršenja, efikasnost i jednostavnost implementacije. Predvidivo vreme i efikasno sastavljanje testova je od posebnog značaja u obrazovnim ustanovama u kojima se vrše česta testiranja relativno malog broja ispitanika, a gde su raspoloživi računarski resursi za ASPT ograničeni. Predloženi metod ASPT je implementiran i upoređen sa efikasnim poredivim metodima koji su zasnovani na heuristikama poboljšanja, kakve su simulirano kaljenje (eng. Simulated Annealing, SA) i pretraga promenljivih okolina (eng. Variable Neighborhood Search, VNS). U otvorenoj literaturi, VNS metod daje rezultate najboljeg kvaliteta za problem ASPT formulisan kao Problem pakovanja u korpe. Vreme izvršenja predloženog algoritma je značajno kraće u poređenju sa SA algoritmom, dok je kvalitet zanemarljivo lošiji. U poređenju sa VNS algoritmom, prosečno vreme izvršavanja je značajno kraće, dok za isto vreme izvršavanja (postignuto prekidanjem izvršavanja VNS algoritma), predloženi novi algoritam postiže bolji kvalitet rešenja za slučaj relativno malog broja paralelnih testova (20, 30 i 60) i banke stavki relativno male veličine (300 stavki), kakvi se često sreću kod testova znanja u obrazovnim institucijama. Parallel knowledge tests are tests which contain different sets of items (sets of questions or assignments of different types) with approximately equal measurement properties, that make them interchangeable. They are applicable when more examinees are solving the test, regardless of whether they do it at the same time or at different times. The results of the testing must be comparable. Automation of the parallel test assembly process is practically necessary in order to achieve objectivity of the testing by well balancing the measurement properties of different test forms and to reduce the possibility of subjective errors by the examiner. Also, the more massive the testing, the more significant is saving up the invested examiner’s time for preparation of parallel tests, by automation of assembly process. Since the automated parallel test assembly problem (APTA) belongs to NP-hard combinatorial optimization problems, the use of heuristics is expected and necessary. The APTA problem is isomorphic with some existing combinatorial optimization problems, therefore heuristics for solving those problems can also be used for solving the APTA problem. In the dissertation an extensive review of the existing formulations of this problem is presented, its mathematical models, then an overview of the heuristics used for solving different problem formulations. Firstly, formulations of the problem and heuristic solutions are independently classified, then their crossclassification pointed to an open space for exploration through new combinations of formulations of the problem and heuristics for their solving. The heuristics used to solve ASPT problems are mainly improvement heuristics, where the execution is interrupted either when a predetermined execution time occurs or when the desired quality of the solution is reached. Constructive heuristics are mainly used to create an initial solution for improvement heuristics. The focus of this dissertation is on the proposal of a new efficient method of polynomial complexity for the complete solving the APTA problem. The proposed method solves one formulation of the bin packing problem by adapting the constructive heuristic proposed by Nawaz, Enschore and Ham, which has not been used for solving the APTA problem so far. Among the characteristics of the new method, predictable execution time, efficiency and simplicity of implementation stand out. Predictable time and efficient tests assembly is of special importance in educational institutions where frequent testing of a relatively small number of examinees is performed, and where the available computer resources for ASPT are limited. The proposed APTA method has been implemented and compared to the efficient comparable methods based on improvement heuristics, such as simulated annealing (SA) and variable neighbourhood search (VNS). In open literature, the VNS method gives the results of the best quality for the APTA problem formulated as the bin packing problem. The execution time of the proposed algorithm is significantly shorter compared to the SA algorithm, while the quality is negligibly worse. Compared to the VNS algorithm, the average execution time is significally shorter, while for the same execution time (achieved by interrupting the execution of the VNS algorithm), the proposed new algorithm achieves a better quality of result for the case of a relatively small number of parallel tests (20, 30 and 60) and relatively small item banks (300 items), which are often found in knowledge tests in education institutions.

The focus of this doctoral dissertation is on exploring the potentials of virtual worlds, for applications in research and education. Regarding this, there are two central aspects that are explored in the dissertation. The first one considers the concept of autonomous agents, and agent theory in general, in the context of virtual worlds. The second aspect is related to the educational applications of virtual worlds, while especially focusing on the concept of virtual laboratories. An introduction to basic terminology related to the subject is given at the start of the dissertation. After that, a thorough analysis of the role of agents in virtual worlds is presented. This, among others, includes the analysis of the techniques that shape the agent’s behavior. The development of the virtual gamified educational system, specially dedicated to agents is then presented in the dissertation, along with a thorough description. While, in the end, analysis of the concept of virtual laboratories in STE (Science, Technology, and Engineering) disciplines is performed, and existing solutions are evaluated according to the criteria defined in the dissertation. Фокус ове докторске дисертације је на истраживању потенцијала виртуелних светова за примене у истраживањима и образовању. У вези са тим, постоје два главна аспекта која су обрађена у дисертацији. Први аспект се тиче концепта аутономних агената, као и теорије агената у целини, а у контексту виртуелних светова. Други аспект је везан за примену виртуелних светова у образовању, при чему је посебан акценат стављен на виртуелне лабораторије. На почетку дисертације је дат кратак увод који се тиче терминологије и појединих појмова везаних за област којом се ова дисертција бави. Након тога је представљена систематична и темељна анализа улоге агената у виртуелним световима. Између осталог, ово укључује и анализу техника потребних за обликовање понашања агената. Потом је у дисертацији детаљно представљен развој оригиналног виртуелног образовног система посвећеног агентима. На крају, анализиран је концепт виртуелних лабораторија у НТИ (наука, технологија, инжењерство) дисциплинама и извршена је евалуација постојећих решења у складу са критеријумима који су дефинисани у дисертацији.

Potreba da se minimalizuje vreme otkaza usluge distribucije električne energije motivisala je češću upotrebu kompozitne prekidačke i privremene opreme u elektrodistributivnoj mreži. To je dovelo do pojave kompleksnijih topoloških slučajeva koje su postojeća rešenja za topološku analizu imali poteškoće da procesiraju. U ovom radu predloženo je jedno sveobuhvatno rešenje za topološku analizu uravnoteženih i neuravnoteženih elektrodistributivnih mreža. U osnovi rešenja je topološki model koji predstavlja matematičku apstrakciju elektrodistributivne mreže u vidu grafa, gde su svi relevanti fizički uređaji direktno modelovani koristeći čvorove i grane. Graf je memorisan pomoću matrične strukture koja definiše povezanost čvorova pojedinačnim fazama, modelujuću pored statičke konektivnosti i dinamičke podatke kao što su trenutno stanje prekidačke opreme i prisutnost privremene opreme. Tako unapred pripremljeni podaci omogućavaju efikasnu topološku analizu koja vrši pretragu grafa po širini. Rezultat topološke analize je široki skup topoloških informacija (formiranje ostrva, slojevite strukture mreže, stanja energizacije, aktivne faznosti, itd.) koje se mogu direktno primeniti na fizičku opremu. Prodloženo rešenje verifikovano je na različitim primerima stvarnih i testnih elektrodistributivnih mreža. Robusnost rešenja demonstrirana je na topološkim slučajevima s kojim su postojeća rešenja imala problema pri procesiranju., A need to minimize the duration of outage in the distribution network has motivated the more frequent use of composite switching equipment and temporary equipment. This has led to the creation of complex topological cases that existing solutions have a problem to process. A comprehensive solution for topological analysis of balanced and unbalanced distribution networks is proposed in this paper. The solution is based on a topological model that represents a mathematical abstraction of the distribution network in the form of a graph, where all relevant physical devices are directly modeled using buses and branches. The graph is memorized using a matrix structure that defines the logical connection between buses per single phase, taking into account static connectivity, current state of switching equipment and the presence of temporary equipment. The network data prepared in this way enable an efficient topological analysis that searches the graph per layers. The result of topological analysis is a wide set of topological data (islands, layered network structure, energization states, active phases, loop detection etc.) that can be directly applied to real physical equipment. The proposed solution was verified on various examples of real and test distribution networks. The robustness of the solution is demonstrated on topological cases which existing have a problem to process.

Oblast računarstva u klaudu se veoma intenzivno razvijala tokom prethodne decenije. Klaud okruženja i servisi koji takva okruženja pružaju postoje u različitim oblicima koji su i diskutovani u ovom radu. Prilikom njihove upotrebe uočeni su različiti problemi koji predstavljaju velike izazove za korisnike. Administratori i korisnici relacionih baza podataka su morali da rešavaju različite probleme i pre postojanja klaud platformi. Ažuriranje softvera, kao operativno zahtevan problem, pojavom relacionih baza podataka kao platforme u klaud okruženju postao je isključiva odgovornost pružaoca usluge. Nasuprot tome, problemi koji zahtevaju unapređivanje rada i optimizovanje relacione baze podataka postali su još izraženiji usled višeg nivoa apstrakcije koji donosi klaud okruženje i većeg broja relacionih baza koje je potrebno istovremeno održavati. S obzirom da se u klaud okruženju prikupljaju i čuvaju detaljne informacije o upotrebi servisa, u ovom radu je realizovan sistem koji na osnovu analize prikupljenih podataka olakšava korisnicima razumevanje funkcionisanja relacione baze podataka i pronalaženje uzroka problema koji se u radu sa njima mogu pojaviti. Nakon detaljnog pregleda uže i šire naučne oblasti, predložen je i definisan sistem koji uključuje dve vrste statističkih modela da bi se obezbedila i sveobuhvatnost i preciznost. Za donošenje konačnih odluka nad dobijenim podacima od strane statističkih modela o tome šta je uzrok a šta je posledica definisan je i ekspertski sistem. Opisan je i izgled infrastrukture koja je zasnovana na konceptu mikroservisa. Pored definisanog sistema, predstavljen je način organizacije tima sačinjenog od različitih aktera sa različitim odgovornostima. Konkretna implementacija sistema je izvršena u Azure platformi kompanije Microsoft. Implementirani sistem je potom podrobno testiran i evaluiran upotrebom realnog radnog opterecenja iz produkcionog okruženja Azure SQL relacione baze podataka tokom perioda od 6 meseci. Dobijeni rezultati su pokazali značajno unapređenje u pogledu performansi izvršavanja upita. Od pojedinačnih korisnika je dobijena i eksplicitna usmena i pismena potvrda o tome. Izvršena je i analiza dobijenih podataka o unapređenju korišćenja relacionih baza podataka svih korisnika platforme koji su se prijavili na ovaj sistem. Zaključak rada sadrži pravce i mogućnosti budućih istraživanja u ovoj oblasti. Development of the cloud computing area has grown immensely in the past decade. This work evaluates various types of cloud environments and services provided to clients. Various problems have been found in the use of the cloud and these present big challenges for the users. Users and administrators of the relational databases have encountered various problems even in times before the cloud existed. Problems such as software updates of relational databases services in s a cloud platform became the responsibility of the service provider. This was a significant improvement that reduced operational costs. However, problems with service improvement and query optimizations scaled to a higher level due to the number of the relational databases and the higher level of abstraction introduced by the cloud environment. In the cloud environment very detailed information about service usage are accumulated constantly. Here is proposed a system that, based on these data, allows users to understand how the relation database works and detects the source of the problem much easier. After a detailed analysis of related work, the system is carefully designed and elaborated. It includes two types of statistical data models to provide both recall and precision, and an expert system for making final decisions. The appropriate infrastructure is based on a microservice architecture. The project team organization was composed of several actors with different skillsets. The system is implemented within the Microsoft Azure platform. Some specific details of this implementation are also presented. The system was fully tested and evaluated using real data workload from the production environment of the Azure SQL relation database in a period of 6 months. The results have shown a significant improvement in the query execution performance. A response from the customers who used this service has shown that the user experience was significantly improved. The conclusion contains an overview of the project, suggests the ideas for improvement of the system, and discusses how the similar approach can be used in scientific areas.

Postojeći pravci razvoja računarstva imaju za cilj da se performanse računarskih sistema podignu na što viši nivo, da bi se zadovoljile potrebe za obradom velike količine podataka... Existing trends in computer design aim to raise the performance of computer systems to the highest possible level in order to meet the needs for processing large amounts of data. Attention is focused on the design of a processor as the main actor in the data processing process. Improvement trends in processor performance predicted by Moore’s Law has been slowing down recently due to physical limitations of semiconductor technology and increasing performance is getting harder and harder. This problem is attempted to be compensated by various techniques aimed at improving performance without increasing transistor and power consumption. In this thesis, asymmetric multicore processors with support for transactional memory are considered. Two new techniques have been proposed to increase the performance of such processors. One technique aims to reduce transaction congestion due to high parallelism by migrating transactions to a faster core. The transactions that contribute the most to an occurrence of congestion are selected for migration. Executing them on a faster core reduces their chances of conflict with other transactions and thus increases the chance of avoiding congestion. Another technique adjusts the cache subsystem to reduce caches’ access latency and to reduce the chances of false conflicts while reducing the number of transistors required to implement the cache. This can be achieved by using small and simple caches. Detailed implementation proposals are given for both techniques. Prototypes of these techniques were made in the Gem5 simulator, which models processor’s system in detail. Using prototypes, the proposed techniques were evaluated by simulating a large number of applications from a standard benchmark set for transactional memory. The analysis of the simulation results gave suggestions on how and when the proposed techniques should be used.

DMS sistem je sistem koji povećava efikasnost, upravlja, nadgleda optimizuje rad i sprečava preopterećenja u elektroenergetskim mrežama. U ovoj doktorskoj disertaciji se proučava takav jedan sistem čije je vreme odziva od značaja sa svim njegovim elektroenergetskim proračunima, koristeći mikroservise. Na osnovu analize svih elektroenergetskih proračuna data je arhitektura jednog takvog sistema za prelazak sa monolitnog na mikroservisno okruženje. Arhitektura se oslanja na mikroservisnu arhitekturu, gde je jedan servis podeljen na više mikroservisa, koji svaki za sebe obavlja svoju poslovnu logiku. Nakon faze istraživanja i sagledavanja jedne takve arhitekture svih elektroenergetskih proračuna jednog DMS sistema na mikroservisno okruženje, napravljen je prototip na cloud-u i za platformu korišćen je Microsoft Service Fabric. Za testiranje takve arhitekture je birano dva elektroenergetska proračuna koji su najznačajniji u jednom DMS sistemu, a to su topološka analiza i tokovi snaga. Testovi su izvršeni u šest eksperimenata gde se vrši analiza i za monolitnu i za mikroservisnu arhitekturu. Hipoteze doktorske disertacije su adresirane kroz postupak primene cloud-a u jednom DMS sistemu, primene jedne takve arhitekture i dokazivanje da primena u mikroservisnom okruženju daje bolje rezultate nego u monolitnom okruženju., Distribution Management System (DMS) is a system that enhances efficiency, manages, supervises, optimizes, and prevents overloading of the electricity networks. This dissertation explores such a system based on microservices and whose response time is significant as it relates to its electric power calculations. Based on the analysis of these calculations, this dissertation also offers the architecture of such a system, and thus enables the transfer from a monolithic to a microservice-based environment. This microservice-based architecture is divided into multiple microservices, each of which performs its individual bussines logic. After relevant extensive research and exploration, a prototype of this architecture was created on the cloud, using Microsoft Service Fabric as a platform. In testing this prototype, two electric power calculations that are most significant in a DMS system were chosen: topological analysis and power flows. Six experiments were conducted in which the analyses for the monolithic and microservice-based architecture were performed. The thesis hypotheses are addressed through the cloud application process within a DMS, showing that the application of such an architecture yield better results in a microservice-based environment than in a monolithic one.

Pametna elektroenergetska mreža predstavlja mrežu nove generacije koja treba da bude efikasna, proširiva, pouzdana i jednostavna za upravljanje. Pametnu mrežu karakteriše veliki broj uređaja i dvosmerna komunikacija sa njima. Ovi uređaji će generisati ogromne količine podataka koje je potrebno pročitati i transportovati do kontrolnog centra, za šta je neophodna odgovarajuća komunikaciona infrastruktura koja obezbeđuje adekvatan kvalitet usluge. U ovoj disertaciji je prikazano rešenje za obezbeđivanje kvaliteta usluge sabraćaja sa dinamičkim promenama prioriteta i propusnog opsega bazirano na programabilnim računarskim mrežama. Takođe je razvijena platforma za evaluaciju komunikacione infrastrukture pametnih mreža kako bi se omogućilo jednostavnije emuliranje različitih mrežnih topologija za potrebe razvoja novih algoritama upravljanja. Performanse rešenja su potvrđene putem šest testnih scenarija i pokazano je da predstavljeno rešenje daje bolje rezultate za sve scenarije sa aspekta obezbeđivanja propusnog opsega i mrežnog kašnjenja., Smart grid represents the next generation power network which should be efficient, extensible, reliable and easy to manage. The smart grid will have a great number of devices with two-way communication. These devices will generate large amount of data that needs to be read and transported to utility control center, which further requires adequate communication infrastructure with appropriate quality of service. This dissertation presents a solution for providing quality of service for traffic with dynamic priority and bandwidth requirements, based on software defined networks. The platform for smart grid communication infrastructure evaluation is developed to enable easy emulation of different network topologies for the purpose of developing new control algorithms. Solution performance is verified using six test scenarios and it is shown that the proposed solution gives better results for all scenarios from the aspect of bandwidth provision and network latency.

Problem nedostajućih podataka je često prisutan prilikom razvojaprediktivnih modela. Umesto uklanjanja podataka koji sadrževrednosti koje nedostaju mogu se primeniti metode za njihovuimputaciju. Disertacija predlaže metodologiju za pristup analiziuspešnosti imputacija prilikom razvoja prediktivnih modela. Naosnovu iznete metodologije prikazuju se rezultati primene algoritamamašinskog učenja, kao metoda imputacije, prilikom razvoja određenih,konkretnih prediktivnih modela., The problem of missing data is often present when developing predictivemodels. Instead of removing data containing missing values, methods forimputation can be applied. The dissertation proposes a methodology foranalysis of imputation performance in the development of predictive models.Based on the proposed methodology, results of the application of machinelearning algorithms, as an imputation method in the development of specificmodels, are presented.