Battery Thermal Runaway Prediction Using Early-State Data

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods fail to predict thermal runaway in batteries effectively, leading to late detection and potential destruction of the cell and surrounding components.

Innovation Solution

A model is trained using a cross-entropy cost function to predict thermal runaway in batteries based on data from battery internal states, sensor data, and actuator signals, enabling early detection and prevention.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional detection methods based on temperature, current, and voltage measurements are used, then detection of thermal runaway is achieved, but detection occurs too late to take corrective action

Engineering Contradiction:
Improvedetection accuracyVSAvoidresponse time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies preliminary action by training a machine learning model in advance using historical battery data that includes precursor patterns of thermal runaway. The model is prepared beforehand to recognize early warning signs and predict thermal runaway events before they occur, enabling proactive rather than reactive detection.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements beforehand cushioning by creating a predictive safety buffer through the trained model. The model identifies early-stage anomalies and predicts future thermal runaway events, providing a time buffer that allows corrective actions to be taken before the actual thermal runaway occurs, cushioning against the harmful effects.

Inventive Principle:
Principle #11Beforehand cushioning (Prior cushioning)

2Reliability

If a model is trained to predict thermal runaway events, then early detection and prevention are enabled, but computational resources and training time are required

Engineering Contradiction:
Improveprediction accuracyVSAvoidmodel training complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent applies self-service by having the system train its own prediction model using its own operational data. The battery management system collects data from its own sensors and uses this data to train the machine learning model, making the system self-improving and reducing the need for external intervention in model development and maintenance.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent implements parameter changes by transforming raw battery operational parameters (temperature, current, voltage) into meaningful features that the machine learning model can process. The system adjusts and optimizes these parameters to improve prediction accuracy while managing computational complexity through feature selection and engineering.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4648171A1Prediction of thermal runaway in batteries
Publication Date: 2025.11.12 VOLVO TRUCK CORP
  • EP4648171A1 patent drawingFigure 1~2
  • EP4648171A1 patent drawingFigure 3~4
  • EP4648171A1 patent drawingFigure 5

AI summary

A computer system is disclosed for training a model to predict a thermal runaway event in a battery cell, the computer system comprising processing circuitry configured to acquire data regarding one or more operating states associated with thermal runaway of a battery cell, and train a model to predict a thermal runaway event based on the acquired data using a cross-entropy cost function. A computer system for predicting a thermal runaway event in a battery cell using a trained model is also disclosed.