Battery Diagnosis via Full-Cell Profile Correction Under High Stimulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing battery diagnosis methods using high electric stimulation suffer from reduced diagnostic accuracy due to overpotential noise, leading to longer diagnosis times and inconsistent results compared to actual charge/discharge performance.
Innovation Solution
A battery diagnosis apparatus and method that applies high electric stimulation to obtain charge/discharge information and uses a machine learning-based factor correction model to remove overpotential noise, generating an estimated full-cell profile for accurate diagnosis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If high electric stimulation is applied to the battery to shorten diagnosis time, then diagnosis speed is improved, but measurement precision deteriorates due to overpotential noise
Solution Approach 1:
The patent segments the full-cell profile into multiple components: observed profile from high-rate measurement, overpotential profile (noise), and corrected profile (useful information). By separating and independently analyzing these components, the system extracts accurate charge/discharge performance data while eliminating overpotential interference, thus achieving both speed and precision.
Solution Approach 2:
The patent introduces a machine learning model as an intermediary that processes the relationship between high-rate and low-rate full-cell profiles. This intermediary learns the transformation pattern and enables accurate prediction of low-rate performance from high-rate measurements, bridging the gap between speed and accuracy requirements.
2Measurement precision
If low electric stimulation is applied to the battery to improve diagnosis accuracy, then measurement precision is improved, but productivity deteriorates due to longer diagnosis time
Solution Approach 1:
The patent performs preliminary action by conducting high-rate charge/discharge measurements first to obtain the observed full-cell profile quickly. Then, using the pre-trained machine learning model, it directly predicts the corrected full-cell profile without performing time-consuming low-rate measurements, thus achieving fast diagnosis with accurate results.
Solution Approach 2:
The patent changes the stimulation rate parameter from low-rate (accurate but slow) to high-rate (fast but noisy), and uses the machine learning model to compensate for the parameter change effect. This allows the system to operate at high stimulation rates while maintaining accuracy equivalent to low-rate measurements.
3Productivity
If high electric stimulation is applied to the battery, then diagnosis speed is improved, but reliability deteriorates due to polarization phenomenon and overpotential
Solution Approach 1:
The patent replaces the direct physical measurement approach (low-rate charge/discharge) with a computational approach (machine learning-based profile correction). Instead of physically performing slow measurements, the system uses algorithms to predict accurate profiles from fast measurements, substituting mechanical measurement time with computational processing.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Provided is a battery diagnosis apparatus and a battery diagnosis method. The battery diagnosis apparatus includes a data obtaining unit configured to obtain a first target full-cell profile representing a correspondence between a capacity factor and a voltage of a target cell while a first electric stimulation is being applied to the target cell, and a control circuit configured to generate an estimated full-cell profile based on the first target full-cell profile and an overpotential profile. The control circuit determines a first performance factor group as a primary estimation result for charge/discharge performance of the target cell by applying a cell diagnosis logic to the estimated full-cell profile. The control circuit determines a second performance factor group as a secondary estimation result for the charge/discharge performance of the target cell by applying a factor correction model to the first performance factor group.