AI driven data analytics– a solution for battery energy storage fires?

The global utility scale battery energy storage sector is having its moment.

Over the last few years, the trajectory of growth globally within this sector has been phenomenal with the latest Battery Energy Storage report by the IEA (International Energy Agency) showing over 40 GW of capacity was added in 2023, double the previous year’s increase1. Over the same time however, according to the EPRI (Electric Power Research Institute) BESS (Battery Energy Storage System) Failure Incident Database2, there have been several notable utility scale battery energy storage fires which involved thermal runaway of lithium-ion chemistry batteries. Use of data science-based techniques, such as advanced AI (Artificial Intelligence) cloud-based analytics, to perform predictive monitoring of BESS, has been suggested as a solution to prevent these types of fires occurring, justifying more competitive BESS insurance offerings.


Utility scale battery energy storage systems are data rich, considering multiple parameters such as temperature, voltage and current are monitored and analysed in real time by the battery management system (BMS), the brain of the battery energy storage system. Advanced cloud-based analytics, including AI, that can be processed offsite, offers a way to significantly enhance the BMS capability to perform predictive monitoring. This allows adverse trends in the operational performance of the BESS to be identified at a very early stage. Analytics also offers other advantages including intelligently managing state of charge (SOC), optimising balancing, enhancing trading strategies and improving maintenance strategies of the BESS.


It is worth reflecting that the battery energy storage sector is following a similar path to the wind industry in terms of adopting advanced, real-time data analytics to predict operational failures at a very early stage before they occur. Key lessons from the wind sector that asset owners and developers within the BESS sector need to consider, to assess the capability of BESS AI data analytics solutions include:

  1. Building databases of a sufficient size which reflect normal and abnormal operational BESS behaviour.
  2. Having an awareness of data variability when building predictive behaviour models, as they may not be universally applicable i.e. not all BESS behave the same, as there is variability from different manufacturers at cell, module and unit level.
  3. Setting appropriate alarm thresholds based on real world BESS behaviour for highlighting possible adverse trends.
  4. Being aware of conscious bias in the predictive model for a particular BESS system which can result in false positive or false negative outcomes that can have an adverse influence on operational decision making, impacting revenue generation.

At the moment, there are several new commercial solutions and products on offer to support advanced analytics capabilities for battery energy storage systems. However, there is limited operational experience to provide reassurance that these AI solutions are a panacea for preventing catastrophic BESS failure events such as thermal runaway. Owners and operators who are considering their deployment, need to identify where it provides an operational and commercial benefit.


The adoption of AI based data analytics for battery energy storage still needs to undergo a significant learning curve. At present, commercial solutions and products available for BESS AI data analytics are not sufficiently mature enough, given the pace of innovation taking place in the battery sector and the diversity/performance in battery performance from different manufacturers and integrators. Robust criteria needs to be agreed within the BESS sector to evaluate the suitability and viability of these commercial offerings to generate consistent and reliable operational outcomes for BESS assets. AI based data analytics can become another defensive layer for reliable operation of BESS assets. It should not be considered as a replacement to existing robust barriers in place such as spacing, battery cooling and an effective emergency response plan supported by the local fire services.

Tariq Dawood
Technical Account Manager, Renewables 
 

1 IEA (2024), Batteries and Secure Energy Transitions, IEA, Paris https://www.iea.org/reports/batteries-and-secure-energy-transitions, Licence: CC BY 4.0

2 BESS Failure Incident Database. (2024, May 20). EPRI, Program 94. Retrieved 15:49, May 30, 2024 from https://storagewiki.epri.com/index.php?title=BESS_Failure_Incident_Database&oldid=5032