This paper introduces a new control approach based on multi-layered perceptron to handle reactive power of the system with the help of a synchronous motor compensator. The neural system was trained with backpropagation with momentum. The perceptron parameters were obtained from an off-line training and then inserted in a microcontroller for the correction. The results have shown that the approach presented in this work provides accurate, simple, low-cost and fast correction than the conventional compensators available in the literature. © 2009 CRL Publishing Ltd.