Experiences Learned from the On-line Internal Monitoring of the Behaviour of a Transformer zyx Miguel A. Sanz-Bobi, Aurelio Garcia-Cerrada,Rafael Palacios, Jose Villar Instituto de Investigaci6n Tecnol6gica, Universidad Pontificia Comillas, Albert0 Aguilera 23, 28015 Madrid, Spain Jose Rolin, Bernard0 Morin Uni6n ElCctrica Fenosa, S.A. Capitin Haya, 53,28020 Madrid, Spain Abstract zyxwvutsrqpon - Knowledge of the health of power transformers is important to prevent high costs caused by failures and to maintain the quality of the service. On-line monitoring and diagnosis seem to be the right way to reach this objective. In zyxwvutsrq this paper a set of experiences related to the installation of vibration and temperature sensors inside a test transformer will be described. Different types of sensors have been tested, and some of them discarded because of induced electromagnetical noise. Some of the main results and conclusiom are shown in this paper. These experiences are included in a more general project called TRAFES, whose main objective is the continuous monitoring and diagnosis of large power transformers. I. INTRODUCTION TRAFES [I] is a project developed to know the health of power transformers during its operation, suggest criteria for the design and construction of new ones, and more generally, improve the existing knowledge their internal behaviour. The Spanish electrical utility Uni6n ElBctrica Fenosa is the project leader, and supports it conjointly with OCIDE, manager of the electrical research program PIE. TRAFES is conceived as a tool for on-line diagnosis of power transformers using the signals coming from internal and external sensors. It is also able to interpret gases concentration and physico-chemical properties from the oil analysis. It is based on a fuzzy expert system and uses a set of models for the characterisation of the transformer behaviour to detect anomalies. The zyxwvutsrq aim of this paper is to present our experience about the location, measurement and testing of sensors inside a test transformer, which is one of the research areas of the project. Since there is not much bibliography about this kind of experiments [2]-[3], this paper tries to somehow contribute in this field. working conditions of the transformer. All the experiences were carried out in the power plant with the important and valuable help of its staff. In. FIRST EXPERIENCE The main objective of the first experience was to determine a feasible way to locate and install vibration and temperature sensors (taking into account possible insulation and electromagnetical problems). This would allow the analysis of the quality of the output signal of the installed sensors. Different types of sensors were used to compare their outputs. Figure 1 shows the sensor location. Accelerometer #I and #4 were ICP type, the others were charge type. Temperature sensor T2 was a PT-100, and the others were thermocouples. The winding accelerometers were fixed to an insulating support, adhered with epoxy cement to the winding varnish. Several materials were tested for the support, and fibber glass was finally chosen for its better insulating properties. Teflon was tested but no way to fix it to the windings was found. Every electrical contact was covered with silicone to improve electrical isolation. n6 .T4 I 11. EXPERIENCES INSTALLING SENSORS INSIDE A POWER TRANSFORMER There is not much knowledge about neither the internal behaviour of real transformers, nor its internal instrumentation. To overcome this lack of experience, internal sensors were installed inside a real 75 kVA - 15 kVl380 V transformer of the Narcea coal power plant, and several tests were carried out. The most important objectives of this research were, the study of possible location of vibration and temperature sensors, the analysis of possible materials for fixing the sensors to the internal parts of the transformer, the installation of sensors inside the transformer and the evaluation of sensor performance and sensibility at different Figure 1: 1" experience sensors location To acquire the vibration signals a HP 3565 with 4 simultaneous fast channels controlled by a HP330 workstation, and a HP spectrum analyser with one input channel were used. Vibrations were sampled at around 1 kHz frequency. For temperature signals a Datalogger with 23 slow channels controlled fiom a conventional PC was used. Temperatures were sampled every 5 seconds. During this first experience the transformer operated unloaded and at 30% of its rated load. Temperature signals were clean enough from noise (probably due to average and filter effect of the Datalogger) and non significative differences (less than 1 "C) were detected between the 0-7803-394M)/97/S10.00 zyxwvutsrq 0 1997 IEEE. TC3-11.1