If you have already understood the previous article, “Thermal Runaway and Internal Short Circuits in Lithium-Ion Batteries,” you should be able to infer how our company addresses the thermal runaway problem.
Our solution framework is as follows:
In the process described above, training a highly accurate RNN network is the core issue. This includes the neural network input data structure, training data, output, etc.
In our model, the data inputs include the voltage and temperature of each individual cell. The neural network output is the severity/probability of an internal short circuit for each cell.
Internal Short-Circuit Severity：
We implement a dedicated algorithm to assess the severity level of the internal short circuit.
- Early-stage internal short: Issue warnings and recommend replacing the problematic cell.
- Mid-stage internal short: Recommend immediate replacement or isolation of the entire PACK.
If a thermal-runaway suppression system is available, it can be activated. - Late-stage internal short: Immediately activate the thermal-runaway suppression system.
Thermal-Runaway Suppression System:
Based on the application scenario of the host system, our company provides or recommends an appropriate thermal-runaway suppression system design.
Why Is Xun’An Intelligent the First to Achieve This Technology?
Large companies typically invest resources into solving thermal runaway at the root, such as developing all-solid-state or semi-solid-state batteries, improving manufacturing processes, or optimizing cell structural design. However, these approaches are either extremely difficult to realize or provide limited real-world effectiveness.
Xunan Intelligent takes a different view:
We acknowledge that liquid lithium-ion batteries will inevitably experience a certain rate of internal short circuits throughout their life cycle. As long as these internal shorts can be detected in the early or mid stages, there is sufficient time to prevent thermal runaway. Even if detected in the late stage, there are still tens of seconds to over a minute available to take action.
Another reason—often overlooked—is that few expect algorithms to play such a decisive role. You may think of our method as similar to facial-recognition systems, but applied to lithium batteries.
Our breakthrough comes from the integration of multiple disciplines, including:
- lithium-ion cell internal mechanisms
- physics-based computation
- advanced software algorithms
- hardware-level data acquisition
- software engineering
- simulation technologies
- reliability engineering