Harnessing AI for Safer, Faster Thermal Runaway Testing

­Energy storage batteries are at the heart of the renewable energy transition, powering residential, commercial and grid-scale applications. However, they also carry a critical risk: thermal runaway. This phenomenon is a chain reaction within cells that can escalate into a fire or even an explosion. As global deployment of energy storage systems accelerates, the industry faces mounting pressure to ensure not only performance but also safety, making the need for rigorous, reliable and efficient safety testing greater than ever. SGS is addressing this demand by leveraging AI and automation to enhance accuracy, efficiency and safety in laboratory testing. In collaboration with Chongqing Energy College (CEC), the company has pioneered an AI-powered thermal runaway testing solution.

Causes of Thermal Runaway

Thermal runaway occurs when a battery cell experiences an uncontrollable rise in temperature and pressure. This dangerous condition may be triggered by internal defects, overcharging or exposure to external heat sources. Once initiated, chemical reactions within the cell generate heat at a rate faster than it can be dissipated, creating a self-reinforcing cycle that may ultimately lead to venting, smoke, fire or explosion.

In battery energy storage systems (BESS), the risks are compounded. A single cell entering thermal runaway can quickly propagate to adjacent cells and modules, amplifying the scale of the hazard. This cascading effect is particularly concerning for large-scale BESS installations in urban, industrial and residential settings, where safety is paramount and the consequences of failure can be severe. To address these challenges, fire propagation testing has become essential, a requirement now reflected in international safety standards such as ANSI/CAN/UL 9540A:2025.

Traditional Testing in the Industry

Traditionally, thermal runaway tests have been both labor-intensive and time-consuming, requiring extensive human oversight at nearly every stage:

• Pre-test preparation: multiple instruments, such as heating power supplies, data acquisition units, gas analyzers and cameras, had to be manually initialized, often consuming several hours before the actual experiment could begin

• Manual monitoring: technicians were required to observe data streams in real-time, make occasional adjustments to heating parameters and simultaneously record experimental phenomena such as venting, smoke or visible flames

• High safety risks: personnel had to remain in close proximity to test setups, exposing themselves to potential hazards, including explosions, toxic fumes and sudden ignition events

• Cooling phase supervision: even after thermal runaway had occurred, samples required hours of cooling. Staff had to be physically present throughout this stage to watch for possible re-ignition and ensure safe handling

Altogether, a single thermal runaway test cycle could take two to three days to complete, placing heavy strain on laboratory technicians, reducing operational efficiency and severely limiting throughput.

Implementing AI and Automation

To overcome the inefficiencies and safety concerns of traditional methods, SGS has integrated deep learning-based computer vision and automated system control into its UL 9540A thermal runaway test platform. This innovative approach streamlines operations, enhances accuracy and reduces the need for direct human intervention in hazardous environments.

The system architecture includes:

• Centralized software (STAS-PLUS): coordinates equipment initialization, test sequencing and real-time data acquisition within a single, unified interface

• Sensors and data loggers: continuously capture a range of parameters, including temperature, voltage, gas emissions and combustion dynamics, ensuring precision and comprehensive datasets

• AI-driven computer vision: trained on thousands of images using advanced models like YOLO, ResNet, VGG and TensorFlow, the system can reliably identify smoke, fire and venting events with an accuracy exceeding 85%

• Automated safety protocols: when abnormal phenomena are detected, the platform immediately initiates recording, activates alarms and secures data storage, minimizing reliance on manual intervention while improving operator safety

Click image to enlarge

Figure 2: Centralized software (STAS-PLUS) coordinates equipment initialization, test sequencing and real-time data acquisition within a single, unified interface

Benefits of AI-Powered Testing

1. Dramatic reduction in test duration:

·      Data processing time has been shortened from two to three days to less than half a day, significantly accelerating testing turnaround

·       Overall testing efficiency has increased by approximately 50%, allowing laboratories to handle more test cycles within the same timeframe

2. Higher data reliability:

·       Automated initialization and real-time data synchronization eliminate misalignment issues that were common in manual testing workflows

·       AI-based recognition minimizes errors in detecting smoke and fire events, greatly reducing the risk of mistakes in data processing and interpretation

3. Enhanced safety:

·       Remote monitoring combined with automated controls reduces the need for personnel to remain in hazardous test environments

·       AI-driven systems ensure reliable and repeatable test outcomes without requiring constant human supervision, lowering occupational risks

4. Improved customer value:

·       Faster access to accredited results shortens time-to-market for battery manufacturers and energy storage system developers

·       Reduced reliance on manual operations lowers overall testing costs, delivering measurable efficacy gains

·       Transparent, high-precision reporting strengthens client confidence and supports compliance with global market access requirements

Click image to enlarge

Figure 3: The control software for UL 9540A thermal runaway test

Chongqing Renewable & Advanced Energy Laboratory

This breakthrough solution is now fully operational at SGS’s Chongqing Renewable & Advanced Energy Laboratory, the first ISO/IEC 17025-accredited laboratory for UL 9540A testing in China. At this facility, AI not only automates thermal runaway assessments, but it also integrates seamlessly into SGS’s broader TeREES L2 digital framework, which includes automated testing of EV chargers and power conversion systems (PCSs).

Together, these capabilities enable fully digitalized laboratory operations, delivering transparency, intelligent scheduling and advanced analytics across the entire testing lifecycle.

Statistics: AI vs. Traditional Testing

Click image to enlarge

Meeting Future Demands

As the global energy transition accelerates, demand for safe, reliable and high-performing BESS will only intensify. With regulatory standards becoming increasingly stringent, AI-powered testing is no longer a luxury; it is rapidly becoming a necessity to safeguard industry safety, efficiency and long-term competitiveness.

Conclusion

Thermal runaway remains one of the most persistent challenges in battery safety. SGS’s AI-powered automation system is revolutionizing the way the industry conducts testing, delivering faster, safer and more accurate results that directly benefit manufacturers while supporting the broader adoption of renewable energy worldwide.

SGS