ANN for Motor Loading Diagnosis under Voltage Variation Conditions

Globally, AC electrical systems can operate at different voltage magnitudes following the IEC 60038-2009 standard, which produces an inevitable deviation in the operation of certain loads, including electric motors. Even though standards such as NEMA MG establish that motors should operate satisfactorily at voltage variation (VV) of ±10% in relation to nominal conditions. These deviations can result in inaccurate interpretations during measurements, such as the motor output-load when considering the nameplate current, without considering that the voltage can influence this parameter. In this paper, a methodology based on artificial neural networks (ANNs) is proposed to identify the line-star permanent magnet motor (LSPMM) output load in the presence of VV with undervoltage and overvoltage conditions. The proposed methodology is constituted of two stages. The first includes measurement campaigns to obtain the PM motor response in VV conditions. Then, Spearman correlation was used to select the best variables to use in the prediction, in order to finally implement ANN to identify the LSPMM output load in each voltage condition. The results validate the proposed methodology and its application in the industry for motors operating under these conditions.