“The reliability of a data center electrical network is critically important for its smooth and continuous operation. While the data center is operational, identifying the foreseen risks and finding suitable solutions with the help of real-time operational data will enhance reliability. Electrical elements as individuals may perform excellently but when it’s operational as an integrated system with all its operational philosophy implemented may not provide the expected outcome. Therefore, modelling of data center electrical networks in simulation software’s, inputting real-time information from the site and performing no. of simulations by subjecting range of different possible operating conditions will provide a conclusive solution to improve the reliability.”
Voltage Study For Large Data Centre
A major data centre client came to us with concerns over the reliability of their electrical system. They had noticed some unusual operational issues that were cause for concern.
Large Data Centre
Premium Power Role:
Power System Consultants
Between July and August 2020 a number of emergency generators started up to support the loads, despite the fact that certain voltage deviations were within the allowable operating range, short in duration, and grid voltage was restored to the supply threshold outlined by the TSO in a timely manner. So the backup system was doing its job, but not in the way it was intended. This made things difficult because there was not an obvious error in the system that one could just identify and fix, and it also raised concerns that maybe the issue was due to the design of the system.
Having previously conducted a number of studies when this data centre was being built, (short circuit study, protection coordination study and arc flash studies) we had detailed knowledge of the site and a model of the network built that we could begin working off. When these models are developed at the design stage they are not always 100% perfect and can present deviations from the actual behaviour on site, especially when it comes to voltage response. After much deliberation our power quality team developed a strategy that required the gathering of load data from the site in real time and inputting this into the model. The model would then be subjected to a range of different over and undervoltage conditions, switching scenarios etc, the aim being to simulate the behaviour and static response of the data centre to identify what was causing the issues. The results of the simulations were then compared with the values seen on site and the model was adjusted to reflect the real behaviour of the system. This is extremely difficult to do and not common practice, but we were able to achieve an error margin of approximately 0.25%.
They had noticed some unusual operational issues that were cause for concern, specifically with their backup generators transfer sequence in the event of utility parameter variations. They also observed that the predefined equipment set thresholds were also contributing to these undesirable operational issues at the site. Unresolved, this had the potential to impact the reliability of the system and increase downtime which for this particular client would result in losses of millions. Being such a big energy user, it also raised concerns with TSO because the fluctuations they were experiencing were having a knock-on effect on the wider grid.
Following extensive simulations (a total of 102 simulation scenarios were developed to include various utility voltage scenarios, site transformer tap positions, and loading conditions) and examination of the data centre’s different types of equipment, sequence of operations, a number of adjustments to the electrical network were proposed to the client. These changes included the replacement of some malfunctioning equipment, adjustments to defined equipment thresholds, as well as modifications to the sequence of operations of the network. Inadequate settings defined within the generator start-up sequence were identified to be the issue causing the unnecessary changeover from the utility supply to the backup generator supply.
During the analysis, other sequences of operation issues were also identified in the network. Adjustment to the equipment settings allowed another contingency protocol to be included which provided an additional reserve supply to the downstream loads if any of the emergency generators failed to come online.
The proposed changes not only improved the reliability of the network and eliminated the generator start-up issue, but the corrective protocols implemented can now be replicated in the client's other data centres and be included in any future designs. With a properly functioning system, the client can ensure they provide a reliable service to their customers, they have saved costs on their energy bills, and the utility provider can ensure an uninterrupted supply to the rest of the grid.