Partner POV | Reimagining UPS uptime at data centers: How condition-based analytics can power the AI revolution

This article was written by Krzysztof Supierz, an IT/OT Platform Leader, at Schneider Electric.

The rise of AI is creating big challenges for data center managers. The need for their services, which was already rocketing, is now increasing even more quickly. According to McKinsey, global demand for data center capacity could more than triple by 2030 – with around three quarters of this (70 per cent) to support advanced AI. In response, the industry is expanding rapidly. Existing facilities are growing fast, and new ones are being built on a scale never seen before. But this isn't just a question of capacity – it's also about complexity. As well as being large, AI workloads are unpredictable, putting additional pressure on both computing power and energy.

Source: McKinsey

The indispensable role of UPSs – and the challenge of UPS maintenance plans

As the demands on AI-ready data centers intensify, their electrical distribution systems are becoming larger and more intricate. In turn, this means that the Uninterruptible Power Supply (UPS), a key data center asset, is becoming an ever more critical component of their infrastructure. A UPS is a device that keeps electrical equipment running when the power goes out, providing a backup electrical power source to prevent interruptions. It works by using a battery that kicks in during an outage.

UPSs have long played an important role in this sector. But as data centers scale up, it will be increasingly common for them to incorporate considerable numbers of these units into their power systems. So operating and maintaining a large installed base of UPSs is set to be a growing challenge for data center managers, who will need to weigh up considerations such as efficiency, uptime and sustainability as they seek to meet soaring demand. New 3-Phase UPS models, such as Schneider Electric's Galaxy VXL, offer the modular design, scalability and power density needed to support modern uptime expectations.

But even with the best equipment, there are still substantial challenges. Over time, key components in each UPS will need replacing as they reach the end of their design life. And these parts will age differently in different conditions, making it difficult to keep track of what's needed in terms of replacement and maintenance – and when. Considering the sheer numbers of UPSs needed for AI-ready data centers, and the current shortage of electrical engineering skills, it's clear that the need to shift the strategy to a proactive asset management approach – including effective service plans for this infrastructure – could quickly become overwhelming for data center operators.

The path towards predictive maintenance through condition-based analytics

To address these complexities, data center operators will need to consider new ways of managing their UPSs – and this is where Schneider Electric's digital service plans, such as EcoCare membership for 3-Phase UPS, can really make a difference. Our natively connected UPS models include network management cards that continuously take readings of a wide range of variables such as wear, aging and temperature. This data is uploaded via the Schneider Electric gateway to our EcoStruxure IoT platform, where it can be monitored remotely by the experts in our Connected Services Hub.

This team uses AI-powered analytics to provide accurate models of the health of the UPS and each component within it – distilling the information into the Asset Health Index. This allows our experts to monitor the health status of batteries, power modules, fans and other critical components, developing insights into the overall life expectancy of the UPS. Based on their analysis, the team provides actionable recommendations to help reduce the risk of downtime and optimize asset lifespans – helping facility operators avoid unnecessary spending by replacing parts only when required. The reliable, detailed picture provided by our analytics helps data center managers optimize the total cost of ownership over the lifecycle of the UPS.

At the same time, data center managers can also optimize the maintenance of their UPSs – shifting to smarter, proactive electrical asset management by basing interventions on the condition of equipment rather than time-based schedules. Through assessing critical data points such as wear, aging and temperature readings from the UPS – and at a component level – our experts can recommend when preventive maintenance should be carried out. To further support this approach, we combine health insights with data on servicing history to create an Asset Maintenance Index score and recommendations for when the best time is for the next maintenance intervention. This allows businesses to potentially extend the period between their maintenance visits from one to up to two years – helping reduce on-site interventions by up to 50% through a condition-based maintenance approach. As a consequence, unnecessary site interruptions can be avoided – a must for data center facilities – and the risk of carrying out maintenance too soon, with all the associated costs, can be mitigated. Just as importantly, a condition-based approach can also help reduce the risk of performing maintenance too late, helping to avoid downtime.

Because engineers know exactly what to focus on, the amount of downtime can potentially be reduced, too. By adding greater clarity to maintenance work, our condition-based analytics also reduce the risk of human error causing a serious outage. And by supporting the optimal functioning of a UPS, they give engineers more freedom to focus on higher-value tasks.

Proactive electrical asset management for data center facilities

Moving to condition-based UPS maintenance, then, can be the start of a journey towards managing electrical assets more proactively, which offers many benefits. And the advantages of this approach will grow in future as the technology supporting it develops. At Schneider Electric, with more than 300 in-house data scientists we're constantly improving our insights, creating greater capacity to detect anomalies and enhance performance. As we incorporate more UPS data into our predictive analytics models, we'll help customers get more out of their equipment – planning and managing their infrastructure more strategically, and further reducing the need for intervention.

This will become increasingly important as the demands on UPSs continue to escalate. New models pack ever-greater power density into smaller volumes, but this efficiency leaves little margin for error. For data centers seeking to make the most of their real estate, an ongoing and comprehensive understanding of how their UPSs are functioning – and how they can help ensure the devices keep running smoothly – will be essential. Those who can draw on this will not only ensure more reliable services now, but also position themselves favorably for future growth.