Determining the Responsibility Sharing of Harmonic Distortion: An Approach Based on Decision Trees and Neural-Fuzzy Systems

Nonlinear devices and the high penetration of inverter-based distributed generations have contributed to increase harmonic distortions in distribution systems, affecting the quality of power delivered to consumers. In this sense, determining the responsibility sharing of harmonic distortions allows the application of effective mitigating actions. Thus, the present paper aims to determine the harmonic contributions at points of common coupling (PCC). From voltages and currents measured at the PCC, a feature extraction stage was employed, where root mean square, crest factor, form factor and total harmonic distortion were calculated. These features were used as inputs to decision trees responsible to identify the contribution side (none, utility-side, consumer-side or both sides). Next, adaptive neural-fuzzy inference systems were used to estimate the harmonic contribution, if necessary, for each side. The decision trees were able to reach more than 99% of accuracy, while the neural-fuzzy systems obtained mean square errors between 1.1e-2 and 3.0e-9.