Evaluation of Pattern Recognition Algorithms for Applications on Power Factor Compensation

• Autor(a) principal: Moreira, Alexandre C.
• Data de Publicação: 2018
• Outros Autores: Paredes, Helmo K. M. [UNESP], de Souza, Wesley A., Nardelli, Pedro H. J., Marafão, Fernando P. [UNESP], da Silva, Luiz C. P.
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://dx.doi.org/10.1007/s40313-017-0352-9; http://hdl.handle.net/11449/179511
• topic: Active compensators; Passive compensators; Pattern recognition; Power factor; Reactive and harmonic compensation

This paper assesses different applied pattern recognition algorithms to decide the most appropriate power factor compensator for a particular point of common coupling. Power factor, current unbalance factor, total demand distortion, voltage harmonic distortion and reactive power daily variation, as well as human expertise, are the key parameters used to set each recognition algorithm. These algorithms are then trained with a series of both simulation and experimental data. Numerical results consistently indicate the decision-tree algorithm with depth 20 as the best classifier for power factor improvement in terms of all metrics considered in this work.