Reactive Power Management For Loss Reduction In Electrical Distribution Systems Using Genetic Algorithm

Tez (Yüksek Lisans) -- İstanbul Teknik Üniversitesi, Fen Bilimleri Enstitüsü, 2015Thesis (M.Sc.) -- İstanbul Technical University, Instıtute of Science and Technology, 2015Bu tez çalışması, elektrik dağıtım sistemlerinin kayıplarının azaltılmasında reaktif güç yönetimi için genetik algoritma temeline dayanan ve hesaplanabilen bir çözüm bulunmasını amaçlamıştır. Son yıllarda elektrik sistemlerinin her alanda kullanılmaya başlanması ve yaygınlaşması ile elektrik enerjisine olan ihtiyaç giderek artmaktadır. Bu büyüyen ihtiyaç bizleri yeni enerji kaynakları bulmaya ve varolan enerji kaynaklarını daha verimli şekilde kullanmaya yöneltmektedir. Günümüz teknolojisinde bile elektrik dağıtım sistemlerinde bertaraf edemediğimiz kayıplar mevcuttur. Bu kayıplar, sisteme bağlı olan yüklerin çalışması için ihtiyaç duyduğu ancak herhangi bir işe yaramayan reaktif gücün getirmiş olduğu kayıplardır. İşte bu kayıpların azaltılması için birçok yöntem geliştirilmiş ve bu yöntemleri dahada ileriye taşımak için çalışmalar devam etmektedir. Bu yöntemlerden bir tanesi olan reaktif güç yönetiminde kapasite yerleşimi bu tez çalışmasında ele alınmıştır. Kapasite yerleşimi, kapasite yerleşiminin yapılacağı noktaların bulunması veya bu kapasitelerin güçlerinin bulunması olarak ifade edilebilir. Kapasite yerleşimi sistemdeli kayıpların azaltılmasında, gerilim kararlılığının sağlanmasında, daha güvenilir bir elektrik sistemine sahip olmamıza, sistemden çekilen akımların azalması sebebi ile yük dengelemelerinin daha kolay yapılmasına ve sistemdeki koruma ekipmanlarının gereğinden büyük seçilmesini önlemede bize faydalı olmaktadır. Dağıtım sistemlerinin büyüklüğünün artması ile karmaşıklaşan kapasite yerleşimi problemlerinin çözümünde yapay zekaya dayalı optimizasyon yöntemleri kullanılmaya başlanmıştır. Bu çalışmada yapay zekaya dayalı optimizasyon yöntemlerinden genetik algoritma tercih edilmiştir. Genetik algoritma belirlenmiş olan kısıtlamara uyarak, çözüm kümeleri oluşturmakta ve her çözüm kümesini birbiri ile karşılaştırarak en iyi olan yaşar mantığına göre karar vermektedir. Bir yandan en iyi çözümleri bir yukarı taşırken diğer yandan sonlandırma şartını kontrol etmekte ve bu şartlar sağlandığında algoritmayı sonlandırıp en iyi sonucu ortaya koymaktadır. Üç fider 16 baradan oluşan test sistemi ilk olarak kapasitesiz sonrasında ise yayınlarda verilmiş olan kapasite değelerine göre kapasiteli olarak incelenmiştir. Örnek verilen hatta yeniden yapılandırma sonrası sistemin almış olduğu yeni hal için genetik algoritmaya dayalı optimizasyon ile MATLAB R2011b ve Opendss programlarının yardımı ile çözülmüş ve kayıpların daha ne kadar azaltılabileceği ortaya koyulmuştur.This study proposes an efficient and computationally feasible solution approach based on genetic algorithm to the distribution system reactive power management for loss reduction. In recent years, demand of electric energy gradually increase because of the enlargement and development of electrical power systems. As we know electrical energy can not be stored. Because of this it has to be transmitted and distributed to supplier. Enlargements and developments of electical power system guide us to find new energy sources and using energy sources that are already using nowadays more efficently. Even in todays technology, we can not eliminate some losses and distortions in electrical distribution systems. These losses and distortions effects electrical energy quality. In electrical energy there are some kind of different losses. One cause of the losses is reactive power. Althought reactive power is known as useless power, reactive power is necessary for machines start-up and because of this there will be losses in electrical systems. Some researchers had been introduces some methods for reducing the losses in distribution systems and they are still working to improve these methods. Most knows are capacitor allocation or VAR management and reconfiguration of the system. Capacitor allocation in reactive power management, one of those methods are discussed in this thesis. Reactive power management can be explained to managing reactive needs of distribution system. This management can be done by finding the best capacitor placement and best capacitor sizing for loss reduction in distribution system. Furthermore this management has to provide quick response to power changes. Reactive power management helps us to have system more reliable, more stable and more quality. For reactive power management all systems have capacitor banks. Systems satisfy its reactive power needs from capacitor banks. Hereby system currents and voltage drop rate of the electrical lines will decrease. Along with this reducement, load balancing could be easier and system protection equipments could select more efficiently. Capacitor allocation and sizing in reactive power management problems are getting more complex because of enlargement of the electrical distribution systems. These complex systems are very hard to solve with traditional mathematical methods. Because of this hardship, some new methods, which are based on artificial intelligence are appeared. Most known methods based on artificial intelligence are genetic algorithm, artificial neural network, expert system, simulating annealing and fuzz logics systems etc. xx Our test system, which is wellknown with its reconfiguration research by Civanlar and his colleague. In one chapter of our thesis, effects of capacitors at loss reduction are examined. Test system solved without capacitors by using POWERWORLD program and losses are calculated. Then the same system is solved and losses are calculated again. Results are compared with each other. According to result, benefits of capacitors are shown. We can understand VAR management importance with this comparation. In this thesis, genetic algorithm that based on artificial intellicence is used for solving the problem. Fundamental of genetic algorithm is based on evolution theory, which is presented out by Darwin. According to this thery best solutions live and worst solutions die. Genetic algorithm basically creates new solutions that calls generation in genetic algorithm. It has its own tools for creation new generation like selection, crossover, mutation, mating and elitizm. After creation of new generation, it compare new generation with initial generation. This comparison would be our best generation for solution. Hovewer, it is not enough to make just solution comparison. All systems have their own criteria. Important thing is while creating new generation, algorithm has to respect systems criteria. Genetic algorithm can work forever but it has to be stop at one point. The other important thing in genetic algorithm is termination conditions. Termination condition basically defines genetic algorithms worktime. If algorithm does not has a termination condition, it can work forever. Hovewer we are using genetic algorithm for finding quick solution. Therefore termination condition has to be choose according to our target. It should be generation number, time or fitness value boundary. In this thesis, genetic algorithm implemented to capacitor sizing problem for finding best solution in short time. Our genetic algorihm generates solution sets using selection, crossover, elitizm and mutation tools and compare the solution sets with each other, while respecting the identified constraints, and finally decide according to best one lives logic. Our constaints are minimum and maximum capacitor sizes and voltage profile of the system in our algorithm. Algorithm gives value to capacitors while solving the algorithm according to minimum and maximum capacitor sizes that defined in algoritm. On the one hand while moving the best solution set to new generation and on the other hand algorithm checks the termination conditions for ending the algorithms. In our algorithm, termination solution is generation number. When terminating conditions are provided, algorithm takes the solution set for solution and shows that as solution. In final chapter of thesis, same test system is solved with genetic algorithm for finding best capacitor sizing with OpenDSS and Matlab program. Solution compared with original test system’s result and improvement on the result is given. Then some changes applied to our test systems. As second application, system’s datas are increased % 25 and new datas are created. Genetic algorithm applied for finding capacitor sizing. Then this system is solved with POWERWORLD program with original capacitor sizes. Results are compared. Gain of capacitor sizing is shown. Then as third application, datas are decreased % 20 and and new datas are created. Genetic algorithm applied for finding capacitor sizing. Then this system is solved with POWERWORLD program with original capacitor sizes. Results are compared. These two application applied for testing response of the algorithm. One of the most important thing in reactive VAR management is response time of system and our proposed system gives us very quick response for solution. Result shows us capacitor sizing is decreases losses of distribution lines. xxi Then a new distribution system is created for testing the algorithm capacity as fourth application. In previous tests, there was 3 feeder and 16 bus, in new created system, we deleted two of three feeder for enlargement of the system. According to result of the new created systems solution, voltage profile was out of boundaries that we decided. Our proposal is adding a parallel line for main feeder lines and decrase the empedans of the systems. We applided four parallel line to four main feeder and decrease empedance of these lines to half of original. As last application, we solved this new system as proposal solution for enlargement systems and it decreased power losses more than expected.Yüksek LisansM.Sc