abstract: This thesis focuses on the influence of a grounded back electrode on the breakdown characteristics. The back electrode is an electrode which attaches at the back side of solid insulation. Insulation with grounded back electrode is a common type of insulation which is adopted in many high voltage power devices. While most of the power equipment work under AC voltage, most of the research on back electrode is focused on the DC voltage. Therefore, it is necessary to deeply investigate the influence of the back electrode under AC applied voltage. To investigate the influence of back electrode, the research is separated into two phases, which are the experiment phase and the electric field analysis phase. In the experiments, the breakdown voltages for both with and without back electrode are obtained. The experimental results indicate that the grounded back electrode does have impact on the breakdown characteristics. Then with the breakdown voltage, based on real experiment model, the electric field is analyzed using computer software. From the field simulation result, it is found that the back electrode also influences the electric field distribution. The inter relationship between the electric field and breakdown voltage is the key to explain all the results and phenomena observed during the experiment. Additionally, the influence of insulation barrier on breakdown is also investigated. Compared to the case without ground electrode, inserting a barrier into the gap can more significantly improve breakdown voltage. / Dissertation/Thesis / M.S. Electrical Engineering 2014

Cette thèse de doctorat traite des méthodes d’évaluation de l’état des transformateurs de puissance à huile. Elle apporte une approche particulière de mise en oeuvre des méthodes de classification dans la fouille de données. Elle propose une stratégie qui met en oeuvre deux nouveaux indicateurs de santé de l’huile construit à partir d’un système neuro flou ANFIS (Adaptative Neuro-Fuzzy Inference System) et un classifieur ou prédicteur de défaut construit à partir des méthodes de classification supervisée, notamment le classifieur Bayésien naïf. Un organigramme simple et efficace d’évaluation de l’état des transformateurs y est proposé. Il permet de faire une analyse rapide des paramètres issus des analyses physico-chimiques de l’huile et de des gaz dissous. Une exploitation des méthodes de classification non supervisée, notamment les méthodes de k-moyennes et C-moyennes flous a permis de reconstruire les périodes de fonctionnement d’un transformateur marquées par des défauts particuliers. Il a aussi été démontré comment ces méthodes peuvent servir d’outil d’aide à l’organisation de la maintenance d’un groupe de transformateurs à partir des données d’analyses d’huile disponibles. / This PhD thesis deals the assessment method of the state of power transformers filled with oil. It brings a new approach by implementing classification methods and data mining dedicated to transformer maintenance. It proposes a strategy based on two new oil health indicators built from an adaptive Neuro-Fuzzy Inference System (ANFIS). Two classifiers were built on a labeled learning database. The Naive Bayes classifier was retained for the detection of fault from gases dissolved in oil. A simple and efficient flowchart for evaluating the condition of transformers is proposed. It allows a quick analysis of the parameters resulting from physicochemical analyzes of oil and dissolved gases. Using unsupervised classification techniques through the methods of kmeans and fuzzy C-means allowed to reconstruct operating periods of a transformer, with some particular faults. It has also been demonstrated how these methods can be used as tool to help the maintenance of a group of transformers from available oil analysis data.

In the present thesis an investigation is performed for the effect of moisture contenton the aging of the solid insulation for transformers that are dynamically loaded. Theinvestigation is based on a theoretical analysis and a model.First, a literature review is conducted on the basics of transformer operation, transformerinsulation and moisture in oil-paper systems. Furthermore, a model is developedbased on moisture equilibrium curves created by Oommen and MIT, moisture diffusionprocesses in oil-paper insulation systems and calculations for the aging of cellulose insulationfrom IEC 60076-7. The model represents an experimental system which is loadedon different load patterns that simulate dynamic loading. The aim of the model is toconclude whether the load patterns will cause the paper to age differently dependingon the frequency that the moisture migration phenomenon between paper and oil occurs.The result of the modeling part is that the aging process is affected by the load pattern,and that the higher the frequency the moisture migration phenomenon occurs within aloading cycle, the larger the impact on insulation degradation. This difference, though,is too small to be measured experimentally in terms of DP and it is suggested thatdifferent load patterns are used in the experiment than those used in the model, whichwill amplify the effect of moisture migration even further.Finally, experimental work is conducted in the thesis, which focuses on implementingthe LabVIEW design from previous work into hardware, debugging the system andpreparing the experimental set-up on practical matters that occurred in the lab. Somefinal work is required before the experiment is able to run, such as preparation of theexperimental units. / I detta examensarbete undersöks effekten av fuktinnehåll på åldring av fast isolationsmateriali transformatorer med dynamisk last. Arbetet är baserat på en teoretisk analysoch en modell.Först genomförs en litteraturstudie på grundläggande transformatorfunktion, transformatorisolationoch fukt i oljeimpregnerade papperssystem. Vidare utvecklas en modellbaserad på jämviktskurvor for fukt skapade av Oommen och MIT, fuktdifussionsprocesseri isolationssystem baserade på oljeimpregnerat papper och beräkning av åldringav cellulosaisolation från IEC 60076-7. Modellen representerar ett experimentellt systemsom lastas för att simulera dynamisk last. Målet med modellen är att avgörahuruvida lastprofilen påverkar åldrandet av pappret beroende på frekvensen av fuktmigrationenmellan papper och olja.Resultatet av modelleringen är att åldrandet påverkas av lastprofilen och desto oftarefuktmigrationen sker inom en lastcykel, desto större är effekten på isolationsdegraderingen.Skillnaden är dock för liten att mäta experimentellt med avseende på DP och andralastprofiler föreslås i framtida experiment, för att förstärka effekten av fuktmigration.Slutligen utförs experimentellt arbete som fokuserar på implementation av LabVIEWdesignenfrån tidigare arbete i hårdvara, felsökning av systemet samt förberedelse av denexperimentella installationen för praktiska bekymmer som uppstått i laboratoriemiljön.En liten mängd arbete återstår före experimentet kan utföras, såsom förberedelse av deexperimentella enheterna.