A monitoring network developed to diagnose early faults in power generators is described. Nowadays, the appropriate operation of generators resides in the information provided by the online monitoring system, which is used to implement predictive maintenance for these machines and to provide early warning of potential failures. In this paper, the configuration of the monitoring system and a detailed study of the measured variables related to their physical behaviour within the generators’ operation are presented. The sensors used and the waveform analysis procedures are explained. Additionally, the architecture that constitutes the application programs and their characteristics as a data network is also described. Finally, some results recorded by the system are shown and discussed.