High voltage (HV) power apparatus are fundamental parts in the industrial process. It is also well Known that electrical insulation systems are the parts that present more problems through the life cycle of an HV apparatus. The dielectric properties of materials are altered during operation time and for this reason they should be monitored. Among the main anomalies that can affect the insulation systems are the Partial Discharges (PD). They are result of a partial failure of the dielectric rigidity of the insulating material submitted to an intense electric field.Until recently, probably due to the inexistence of measurement mechanisms that made possible the storage of data, very a little could be done to accompany the actual condition of insulation systems. With the coming of digital systems, this limitation was overcome.However, the obtention of real data and their subsequent evaluation are mechanisms that still will demand a good number of years of experience to perfect . This fact occurs, mainly, due to the difficulty in obtaining those data .As a way to the problem of lack of real data, special cells development with the purpose of simulating the principal types of defects that occur in HV insulation equipment . In that way, it was possible to do both the simulation of the defects and the measurement and acquisition, through the electromagnetic measurement method, of an expressive number of data. Those data were used to adjust a system of pattern recognition of PD, using artificial neural networks, where it is also made use of the fractal theory, as a mechanism of data compression, by measures of fractal dimension and lacunarity. That stage, denomited laboratory, was of fundamental importance to the obtention of experimental . Knowledge very useful to the development of real operating systems .Another source of information that was also available refers to the levels of gases found in oil samples from the interior of power transformers. This analysis, known as analysis of dissolved gases or chromatography, is based on the existence of certain types of gases as an indication of the type of anomaly on the electrical insulation system. Although some mehods exist, capable of accomplishing such diagnosis, the thesis made use of a specialist`s information. Those statements were used then in a hierarchical neuro-fuzzy model, designed method, with the type of anomaly existent in the insulation system. After that process, a real sample of operating HV power apparatus was monitored using the two analysis techniques, the chromatography and the electromagnetic method of PD measurement.Therefore these two techniques, are used to estimate the PD activity in the insulation systems and, consequently, their level of degradation . In resume the thesis presents a diagnosis system that makes use of internal data from the electrical insulation system with base in the PD occurrence. That system indicates, at the end, a concept of its current operative state, being able to be used in a more appropriate planning in order to prevent inopportune losses. The evaluation of the model, as much of simulating as in field, present high indexes and a visual confirmation intern of a real equipment, that proves the efficiency and the importance of system.