Electrical Power Equipment Maintenance and Testing

Comparison of Artificial Neural Network and Multiple Regression for Partial Discharge Sources Recognition

“…The multilayer perceptron neural network (MLPN) using the back propagation (BP) is the most widely applied for PD classification because of its capability to classify complex PD fingerprints [1,3]. The basic structure of an MLPN consists of the input layer, hidden layer and the output layer.…”

“…1(b). Nine layers were created similar to the literature [3]. The inception voltage for void 5 was 3.4kV and all measurements were taken at approximately 4.4kV.…”

“…If PD, such as internal, is not detected at the initial stage, these discharges can bridge the insulation system with complete breakdown of the insulation and serious financial implications [2]. Therefore, PD measurement and recognition has been the focus of condition monitoring engineers [3].…”

“…These statistical fingerprints are extracted from the φ-q-n (phase-amplitude-number) patterns representing different PD sources. For φ-q-n evaluation, statistical fingerprints have been widely applied because of their capability for well-defined pattern quantification [1,3]. In order to simplify the φ-q-n analysis, statistical fingerprints are usually extracted from 2D plots derived from the φ-q-n patterns.…”

Separation of sources in radiofrequency measurements of partial discharges using time–power ratio maps

“…Knowing the condition of the insulation is key for the reliability of power systems, Gill (2008). Partial discharges (PD) are low-energy ionizations in sites where highly divergent fields are present, Boggs (1990) so they represent a measurable manifestation of electrical stress and a symptom of other problems in the electrical asset.…”