Currently, an improvement in the energy performance of energy equipment and processes is required due to the high global energy consumption and polluting emissions. Centrifugal pumps have been extensively implemented in several processes for fluid transportation. Due to the above, in this research, a methodology is proposed. A backpropagation neural network and a theoretical model are used together to predict energy performance parameters such as real head, power consumption, and efficiency. The neural network is built based on different geometric and operational parameters. The results obtained show that the proposed neural network allows predicting the real head, the power consumption, and the efficiency parameters with a maximum error of 0.26%, 2.27%, and 0.15%, which is a precision similar to the one obtained in methodologies such as computational fluid mechanics. The inclusion of the theoretical model in the neuronal network makes it possible to reduce the error within the energy performance prediction of the turbomachinery. In general, an increase in the maximum error of 34.61%, 61.67%, and 126.67% has been observed for the real head, the power consumption, and the efficiency when the theoretical model is not considered. In general, the use of the proposed neural network allows predicting with high precision and rapidly energy performance parameters such as real head, power consumption, and efficiency, which results in a methodology with the potential to be used in optimization processes in centrifugal pumps.