Batteries have been present in the data center since their introduction, connected to the UPS system to ensure smooth operations without service outages. In this article we deep dive into the battery world to identify preventive measures and considerations, with a specific focus on battery monitoring and predictive data analytics.

Data center battery systems are designed to provide instantaneous power continuity to support power drawn down from the grid. Batteries are usually designed to take up short-term power needs. Beyond that time and when longer support is needed, a generator or other means of power generation such as fuel cells start and continue to power the site.

Recently, new opportunities have been generated by the possibility of using back up batteries to provide grid services and, therefore, become a revenue source for data center operators.

Battery technology options

For decades lead batteries have been basically the only option available in the market. Their design has evolved through the years, improving energy density and reducing service needs.

Over the past 10 years, lithium solutions have been developed as an alternative to lead. Lithium offers several advantages, like lower weight and, in some designs, better cycle life. But it also comes with its drawbacks, mainly linked to the inherent instability of lithium which needs proper control and management to remain within safe operational limits.

For this reason, all lithium-based batteries include a Battery Management System (BMS) which is designed to protect the battery, provide alarms, and disconnect it from the load in case the controlled parameters get outside the preset safety window.

In any case, due to the instability of lithium – which can sometimes cause fires that cannot be easily controlled – lithium batteries are not recommended to be installed on data center premises, but in separate dedicated rooms/containers.

Battery system design criteria

When selecting, designing and sizing a battery system several factors should be taken in account. As with any other component or device, batteries may fail. Therefore, the system must be fault tolerant by design.

Commonly this is achieved by selecting multiple battery strings in n+1 configuration in conjunction with n+1 UPS design, specifically in the case of UPSs connected to common battery systems.

That said, redundancy comes with its issues in terms of space, weight and cost linked to the additional devices. The level of redundancy should therefore be selected in accordance with the level of availability requested by the system.

Battery performances are also degrading during the operational life. This should also be considered during the design phase in order to get the right back up time at the end of the battery life. Unfortunately, according to studies examining data center outages, it looks that in several cases these recommendations are not always fully implemented.

Battery monitoring

As with any other component, batteries need to be serviced regularly by qualified staff. However, onsite servicing is expensive and testing procedures may clash with the end user’s reluctance to take the risk associated with performing actual battery discharge.

The difficulties associated with battery servicing have been identified as one of the major causes behind outages. To overcome this problem, battery monitoring systems have been developed for all battery technologies.

These devices are live connected to battery cell terminals, to constantly check functional parameters like voltage, temperature, and internal impedance.

Cell balancing is also an important feature. Although this kind of system is meant as an optional feature for lead technology (i.e., sending alerts and alarms in case of malfunctioning), for lithium batteries it is a fundamental and mandatory part which is necessary to keep the battery within the safety window’s functional parameters.

Whenever the parameters exceed the designated settings, the lithium battery is disconnected from critical load and an alarm is raised. These devices, while not replacing onsite servicing completely, are providing essential information about battery operational and functional parameters.

It’s worth noting that when alarm thresholds are reached, it means that batteries are working in critical conditions and therefore can’t provide full performance, which can in turn lead to a critical load power outage.

Battery data analysis

Fiamm has been collecting and analyzing data from a large number of installed batteries for more than 15 years. In the beginning, the goal was to detect any battery issues and alert the customer in order to provide speedy battery replacement.

After years of collecting and analyzing data, Fiamm engineers developed an advanced algorithm to predict a battery failure based on the information detected by the Battery Management System (i.e. voltage, temperature and impedance.)

At first, data analysis highlighted that in many cases batteries are not operated as they should be. Temperature is the main factor affecting battery life and, in several installations, it was much higher than necessary.

This resulted in shorter-than-expected operational life, putting the critical load at risk. In order to improve the situation and reduce overall costs, temperature resistant battery installation is recommended in rooms where cooling implementation is difficult or expensive.

Another critical element raised by preliminary data analysis was the mismatch between the charging voltage and the battery temperature, which can often lead to battery deterioration. In this case the problem can be solved by activating the battery voltage compensation in the UPS charger and placing the battery temperature sensor in the right place.

In addition to these initial findings, big data elaboration was launched, crossing data collected from BMS with information from faulty battery laboratory tear down analysis. The elaboration target was matching laboratory findings on faulty batteries with data coming from their operational life.

This resulted in reliable methodologies allowing prediction and identification of potential problems in operational batteries based on parameters available in BMS (voltage, temperature, impedance.) With this information data center operators have full control of their batteries and can plan their replacement in the right moment, before they reach a critical point – but not too early, therefore disposing of a still valuable asset.

In order to carry out this analysis the BMS should be either live connected to Fiamm or, when this is not possible, data should be downloaded regularly and sent to Fiamm R&D engineers. In this way, the risk of outage due to battery aging can be reduced to nearly zero.

Fiamm Energy Technology recommendations

BMSs are widely implemented within the data center industry, but most customers are looking at them just in case of any alert or alarm. Our experience shows that data collected and stored in a BMS has huge value in optimizing battery life and predicting potential system failure.

Fiamm Energy Technology recommends making full use of this value by properly analyzing the data available in order to take any necessary action.