First AI enabled battery test system for thermal runaway

Researchers in China have combined several AI models to create the world’s first automated system that tests for thermal runaway of battery energy storage systems (BESS). The system was developed by Swiss test house SGS with the Chongqing Energy College (CEC) to address fire safety concerns around the rapid global growth of BESS systems in commercial, industrial, and residential deployments.

China has seen a huge growth in solar power production in recent years, which requires BESS battery systems to smooth out the load. The global market for these battery storage systems is expanding rapidly. Valued at US$468.2 billion in 2024, it is projected to reach $1,200 billion by 2031, a CAGR growth of 15.2%, according to SGS. This is driving the need for more test and validation capacity.

Thermal runaway can cause uncontrollable increases in temperature and pressure within a battery cell leading to fires or explosions. SGS’s thermal runaway testing solution aligns with ANSI/CAN/UL 9540A:2025 standard to assess thermal runaway fire propagation in BESS and provide essential data on potential risks during thermal runaway events.

The system combines Yolo, TensorFlow, ResNet, and VGG AI models to customize the data processing of temperature, voltage, gas emissions, and combustion collected during tests. The researchers developed algorithms that automatically adjust the test parameters, synchronously collecting data in real-time. The Yolo and Resnet computer vision models are used to detect and log any smoke and fire occurrences, minimizing manual intervention.

The test system, now in operation at SGS’s Chongqing Renewable and Advanced Energy Laboratory, has centralized software controls all testing equipment, automates initialization, parameter adjustments, and data collection. This frees up test engineers from manual operations and reduces exposure to hazardous conditions, enhancing the safety of the lab.

“Thermal runaway incidents have become critical challenges in the industry, and we are delighted to launch our fully automated testing solution which addresses the need for internationally accredited safety testing, faster certification turnaround, and improved test transparency,” said Walter Zheng, Connectivity & Products at SGS. “With automated data processing, our customers benefit from faster access to the market, reduced costs, and higher testing precision to ensure reduced risk and greater competitiveness.”

SGS’s Chongqing lab obtained the ISO/IEC 17025 accreditation in 2022 to become the first lab in China accredited to offer testing services according to UL9540A standards.