Unlocking the power of AI

However, as more energy storage assets come online, owners and managers are facing an emerging set of common challenges that must be addressed, such as issue identification and prioritisation, maintenance planning, data management, etc. These challenges hamper profitability, increase downtime, and stymie the deployment of new assets.

To mitigate these issues, asset owners and managers find an increased necessity to address them through asset performance management software. This article looks into the challenges and how asset performance management software helps asset owners and managers overcome them, helping spend less time managing data and more time acting on it.

Data overload causes ineffective issue identification and prioritisation

With hundreds of millions of battery cells, the average 1 GW battery-based energy storage system produces 100 times the data points of a conventional 1 GW power generation plant. With data coming in every second, knowing where to look for signs of trouble is effectively impossible. Even if an asset manager did know where to look, close monitoring of a single asset would not be cost-effective because that would re-quire the work of multiple full-time analysts.

Instead, the status quo has moved on to waiting for SCADA systems to surface alerts, which are lacking in a few key ways:

• The SCADA systems are only triggered when some battery component is malfunctioning, at which point the system could already be experiencing costly loss.
• The SCADA systems offer very little detail into the shape of that loss, and virtually no window into potential future capacity loss from the issue.
• The timing of SCADA alerts makes it difficult for asset managers to diagnose problems as they arise, often finding out precious minutes or hours later, when irreparable damage has already been done.

All of this puts asset managers on their backfoot, monitoring ad-hoc and reactively troubleshooting. Operations and maintenance (O&M) technicians face the same challenge, running from one issue to the next with virtually no time or ability to proactively check the health of their systems.

Making energy storage system maintenance proactive rather than reactive

Asset managers face three top maintenance planning difficulties:

1. Anticipating where and when maintenance will be necessary.
2. Communicating maintenance priorities in time to prevent downtime or costly asset damage.
3. Tracking asset performance before and after performed maintenance to quantify impact.

Relying on SCADA alerts for system maintenance planning forces a reactive stance. Trying to deduce battery component failures before they happen, such as when they are showing above average temperatures for their operating conditions or when their temperatures are rising at an unusual rate, would be time consuming and inaccurate for an analyst looking at SCADA data alone.

Deducing battery component failures needs to happen proactively rather than reactively. Having a granular system performance view without the need for visual rotating inspections allows teams to prioritise maintenance tasks and ultimately prevent costly downtime. Granular data also offers the needed verification to ensure issues are properly resolved after maintenance tasks are complete.

Predictive maintenance capability to overcome challenges

Solution 1: How to leverage predictive maintenance for issue identification and prioritisation

Nispera leverages artificial intelligence (AI) to learn what normal and anomalous battery cell behaviour looks like across a vast range of operating conditions by studying huge amounts of linked SCADA data. It learns when cells are simply running hot because of extenuating circumstances and when their behaviour is a sign of impending failure.

Nispera deploys AI on energy storage systems without adding any hardware to predict what maximum cell temperatures should be under current operating conditions (e.g. level of charge and discharge, cooling system temperatures) and issues an alarm if measured temperatures exceed that value by a certain threshold or trend. These alarms come an average of three days before a battery outage actually occurs, giving technicians critical insights and enabling them to investigate and resolve issues before the SCADA system triggers an alert.

The software can process far more data, faster and more accurately than analysts, and it works around the clock, every day of the year. The software is manufacturer agnostic, integrating data from any battery original equipment manufacturer (OEM) on the same platform, and scalable enough to monitor a full energy storage portfolio. This results in a round-the-clock sentinel monitoring energy storage system from different manufacturers at various locations, delivering actionable alerts behind a single pane of glass.

Solution 2: How to leverage predictive maintenance capability to switch from reactive to proactive asset management

Nispera’s predictive maintenance tool anticipates issues at a level of granularity that makes proactive communication with on-site teams easy. Instead of asking O&M teams to perform rotating visual inspections of every chiller/HVAC, the software gives asset managers and on-site teams the same view of components that need attention.

Technicians can get to work before downtime occurs, diagnosing and resolving the problem, which could be anything from rack failure to chiller malfunction to unusual environmental factors. When the work is done, Nispera’s data collection makes it easy to ensure the issue is resolved and the system is performing as expected.

Data management

The asset management challenges presented by the ever-growing volume of dispersed data in renewable and storage assets are substantial. Traditional data management strategies simply cannot keep pace with the data intensity required to get optimal performance out of renewable and storage assets (Figure 2).

Even more, fragmentation of data caused by siloed systems is leading asset management teams to spend most of their time manually collecting and harmonising data across assets. This leaves little time for analysing, sharing, and acting upon that data. In recent years, asset performance management teams have often spent almost half of their time on data collection. It is highly time-consuming and error-prone work for dedicated resources to manually collect, clean, prepare, and harmonise data across assets just so it can be examined in one place. Fluence tends to see almost one-third of teams’ time spent analysing that data, looking for patterns and anomalies that can be acted on to improve asset performance. That leaves very little time for sharing data analysis findings to mobilise action and even less to act on the issues that have been discovered. This is nowhere near enough to make meaningful improvements to asset performance, leaving assets performing suboptimally.

Enjoyed what you've read so far? Read the full article and the rest of the Spring issue of Energy Global by following the link or below, or why not register today for free!

For more news and technical articles from the global renewable industry, read the latest issue of Energy Global magazine.

Energy Global's Spring 2024 issue

The Spring 2024 issue of Energy Global starts with a guest comment from Field on how battery storage sites can serve as a viable solution to curtailed energy, before moving on to a regional report from Théodore Reed-Martin, Editorial Assistant, Energy Global, looking at the state of renewables in Europe. This issue also hosts an array of technical articles on electrical infrastructure, turbine and blade monitoring, battery storage technology, coatings, and more.

Read the article online at: https://www.energyglobal.com/special-reports/01052024/unlocking-the-power-of-ai/