Investigating Unexpected UPS Behavior During Data Center Commissioning
Even in the world’s most advanced data centers, the smallest distortion can make the biggest difference. During the commissioning of a new hyperscale data center in France, an unexpected UPS behavior revealed just how crucial continuous waveform recording is in ensuring seamless power-up operations.
In this newly built hyperscale data center commissioning, everything appeared stable – until the moment the second MV transformer was energized. Without any loss of power, several UPS systems in another building suddenly switched to battery mode, triggering multiple alarms across the facility.
The power supply was intact. The control systems showed no fault.
Yet something had clearly disturbed the network.
To understand the event, the engineering team from ANALYSEO, Elspec's agent in France, used Elspec's PureBB plug & play analyzers with continuous waveform recording across the site to capture and record the data across both MV and LV levels.
What were the hidden interactions that affected stability during commissioning?
Here’s a clue:
Read the full case to see what caused the disturbance and how Elspec’s advanced Analyzers helped the team uncover it.
Reviewing the data from the Analyzers on the Building #1 at the same time as the main inrush current spike (17:07:02), revealed voltage drops and THDv spikes. The maximum THDv% value recorded on Building #1 low voltage installation was out of tolerance rang of the UPS systems which were been operating under an “Eco/high-efficiency mode” which is more efficient but more sensitive to volage distortions. From this data it appears that the disturbances on Building #2 propagated through the shared HV substation to Building #1, where UPS systems tolerate THDv up to 5%, by generating voltage sags and temporary THDv% spikes due to current non-linearity. Once this limit was exceeded, the UPSs automatically switched to battery mode, interpreting the event as a power anomaly.
Continuous PQ recording provided full visibility into the true transient behavior of the electrical system – capturing and recording every waveform, every cycleand every disturbance in real time and enables access to historical data.
This allowed engineers to:
This level of visibility exposed subtle interactions between transformers, UPS systems and the HV network, insights that standard PQ meters or transient recorders could not provide due to their limited recording duration, sampling rate or trigger dependency.
Solutions
Based on the power quality findings, the engineering team evaluated several technical options to mitigate transformer energization effects and prevent UPS disturbances. This comprehensive analysis enabled the team to propose effective solutions balancing reliability, efficiency, and cost, ensuring stable operation during future load transfers.
Results
Power quality measurement was key to diagnosing complex energization interactions in the data center’s MV/LV network. Continuous waveform recording provided complete visibility into the real electrical behavior – beyond what event-based tools could show. The analysis demonstrated that energy efficiency (PUE) must be balanced with operational reliability, especially in hyperscale environments. PQ insights enabled data-driven engineering decisions, ensuring stable operation and alignment with design and uptime targets.
Conclusions
This case highlights the essential role of power quality measurement during commissioning. Power quality measurement was key to diagnosing complex energization interactions in the data center’s MV/LV network. Continuous waveform recording provided complete visibility into the real electrical behavior – beyond what event-based tools could show. PQ insights enabled data-driven engineering decisions, ensuring stable operation and alignment with design and uptime targets. Through detailed PQ investigation, the engineering team gained a clear understanding of MV substation dynamics and implemented corrective measures that secured reliable, disturbance-free operation for this data center.