When it comes to assessing the asset health of generators, removing cross coupled signals from partial discharge data is essential for accurate insulation defect diagnosis. In the case of hydro generators, an increase in PD activity, especially in the phase winding, can be indicative of insulation degradation in the stator. Hence, effective noise removal software is imperative for establishing reliable monitoring results.

Denoising is the process of isolating and eliminating unwanted noise signals from the generator's data stream. At Monitra, our Kronos Ultimate Software, complete with ML-42 machine learning algorithm and synchronous acquisition, is able to differentiate between noise signals and PD pulses coming from the hydro generator under test. Combined with Monitra's patented monitoring solutions and long term trending of data, this enables us to obtain cleaner, more accurate data, which is essential for effective condition monitoring.

Some of the advantages of the Kronos Ultimate Software are:

• Enhanced Precision: Our ML-42 machine learning algorithm can discern even subtle deviations in generator performance, improving early fault detection.
• Reduced False Alarms: By eliminating noise interference, a clear PRPD pattern can be recognised in the software and false alarms are minimised, allowing for more focused and efficient maintenance efforts.

Below you can see some examples of data before and after the decoupling noise removal process.

The PRPD patterns of the generator show that slot discharges are present on all three phases. The magnitude of the PD signals put the generator in the "Unreliable" condition category.

In this case, it is advised that the generator undergoes an inspection on the next available outage, looking for signs of slot damage in particular and that the generator is considered for remedial action.

Watch our video here.