AI-Based Battery SoH Testing from Partial Discharge Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for accurately and efficiently evaluating the state of health (SoH) of batteries, particularly in electric vehicles and energy storage systems, are inadequate, necessitating faster and more precise evaluation techniques.
Innovation Solution
An apparatus and method utilizing a discharging unit, data collection, preprocessing, training with an AI learning model, and prediction unit to quickly and accurately determine SoH by collecting and processing data from a test battery, employing techniques like Gramian angular field and recurrent neural networks to enhance precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional SoH evaluation methods are used, then measurement precision can be achieved, but testing time and efficiency are excessive
Solution Approach 1:
The system performs preliminary actions by pre-processing battery data (normalization, feature extraction) and pre-training the AI model offline. During actual SoH testing, only inference is needed, which significantly reduces testing time while maintaining measurement precision through the trained model's accurate predictions.
Solution Approach 2:
The patent replaces traditional mechanical/electrical measurement systems with an AI-based prediction system. Instead of using complex electrical testing equipment to directly measure battery properties, the system uses a trained neural network that predicts SoH from simplified input data, thereby reducing testing time while maintaining or improving measurement precision.
2Measurement precision
If comprehensive battery testing is performed, then measurement precision improves, but device complexity and testing procedure complexity increase
Solution Approach 1:
The system extracts only the essential features from comprehensive battery testing data. Instead of using all available data and test parameters, the pre-processing unit selects and extracts the most relevant features (voltage, current, temperature, time) and presents them to the AI model in a simplified format, reducing device complexity while maintaining measurement precision.
Solution Approach 2:
The patent creates a virtual copy of the complex battery testing system in the form of an AI model. The physical testing apparatus collects basic data, and the AI model (trained on comprehensive datasets) replicates the complex evaluation logic, thereby simplifying the physical device while maintaining high measurement precision through the intelligent prediction algorithm.
3Productivity
If fast SoH testing is implemented, then productivity improves, but measurement precision may deteriorate
Solution Approach 1:
The system incorporates feedback mechanisms where the AI model continuously refines its predictions based on the battery's response during testing. The pre-processing unit analyzes the battery's voltage, current, and temperature changes in real-time and feeds this information back to the model, allowing the system to achieve both fast testing and high measurement precision through iterative prediction and adjustment.
Data Source
AI summary
There is provided a discharging unit configured to discharge a test battery as a test target in a preset environment, a data collection unit configured to select a preset number of points while the test battery is being discharged by the discharging unit and collect input data during the discharging process or within the selected points, a sensor unit configured to sense a voltage of the test battery, a preprocessing unit configured to preprocess the data collected by the data collection unit, a training unit configured to train an artificial intelligence learning model by performing training on a preset architecture using the data subjected to the preprocessing unit as an input value and a maximum battery available capacity as an output value, and a prediction unit configured to test an SoH of the test battery using the input value subjected to the preprocessing unit into the artificial intelligence learning model.