Simulated Partial Discharge Harmonic Data generation for Neural Network Training in the Absence of real measurements: A first Approach

Partial Discharges measurements on the High Voltage equipment are assumed as one of the most critical condition assessment measurements which can propose a future lifetime of power equipment and a suitable maintenance schedule. Neural Networks are assumed as one of the most accepted Artificial Intelligence techniques for the condition assessment of High Voltage power equipment. However, the correct use of this technique demands the existence of a large number of datasets to provide valuable results. Many times the required large datasets in specific investigation areas such as the partial discharges on the high voltage electricity network do not exist for several reasons. These reasons are described in this manuscript. In the absence of these datasets, which are typically real measurements, there is a specific need to construct them by using small datasets on the field under interest. This manuscript describes such a data construction algorithm based on the authors’ knowledge and experience in the field of partial discharge measurements.