Two large data centres in Dublin had seen malfunction of their UPS systems. Our studies were able to identify the level of harmonic distortion at each monitored location and assess any power quality events that may have been occurring. Using this data we could then propose mitigation solutions for the issues identified.
The client required Premium Power to conduct fault finding on site with portable monitoring equipment without switching of the UPS or variable frequency drives during the survey.
We conducted an extensive power quality metering assessment with a selection of advanced portable devices to identify vulnerable points on the network. Voltage total harmonic distortion (5%) and individual harmonic levels recorded at the sites did not exceed the maximum recommended limit under IEC61000-2-4.
However, it should be noted that IEC61000 doesn’t specify limits for harmonics above the order of 50. We found that content of high frequency harmonics contained in the voltage waveform was excessive and may have been impairing performance of sensitive equipment.
Sensitive devices such as P-N-P transistors use zero crossing detection for switching operation. With multiple zero crossings appearing on the voltage waveform those devices are forced to operate with higher frequency than they have been designed for, which may lead to deterioration and ultimately failure.
“Portable metering can be used to measure power issues on data centres. But the identification of the source of the faults may be difficult when test scenarios are not allowed. Having power analysis capabilities and modelling the network allows us additional tools when troubleshooting. This is particularly useful for live data centres, and advised particularly following energy efficiency initiatives such as the introduction of VSDs or VFDs.”
Recommendations were given to reduce the occurrence of the high frequency harmonics. Budget pricing for design, supply, and installation and commissioning of the required filtration devices was costed. This price was deemed by the client as a tiny fraction of the cost of potential downtime to the data centre and a budget was immediately allocated for the work.