Evaluation of electrical insulation using genetically evolved artificial neural nets

The degree of electrical insulation degradation depends strongly on the insulation's defects. Different types of insulation defects generate different partial discharge (PD) patterns. The correlation that exists between a defect and its pattern is identified and recognized by popular artificial neural nets (ANN) as they significantly improve the recognition of complex patterns in noisy data. An evolutionary design concept is used to realize such an ANN, to avoid the problem of stagnation, and satisfactory results are obtained.