Battery SOH Prediction Using 2D CNN With Fewer Charge Cycles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing battery management systems struggle to accurately predict the state of health (SOH) of battery cells using a small number of charging/discharging cycles, which is crucial for timely replacement and ensuring safety in devices like electric vehicles.
Innovation Solution
A battery management apparatus utilizing a two-dimensional convolutional neural network (2D CNN) trained on extensive charging/discharging data from a second battery cell to predict the SOH of a first battery cell based on a smaller number of cycles, with mechanisms to handle incomplete data and adjust the cycle number for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If a small number of charging/discharging cycles are used for SOH prediction, then the time and data requirements are reduced, but the prediction accuracy deteriorates
Solution Approach 1:
The system performs preliminary training of the 2D CNN model using extensive charging/discharging data from a second battery cell before predicting SOH for the first battery cell. This preliminary action enables the model to make accurate predictions with minimal input data by leveraging the pre-learned patterns and relationships from the training phase.
Solution Approach 2:
The patent uses a second battery cell to generate training data that copies and replicates the charging/discharging characteristics of the first battery cell. The 2D CNN model learns from this copied data pattern, allowing accurate SOH prediction for the first battery cell without requiring extensive cycling data from it directly.
2Measurement precision
If extensive charging/discharging data is collected for training, then the prediction accuracy is improved, but the data processing complexity and time increase
Solution Approach 1:
The patent extracts only the essential features and patterns from extensive charging/discharging data using the 2D CNN model during training. By extracting and retaining only the critical information needed for accurate SOH prediction, the system avoids processing unnecessary data complexity while maintaining high prediction accuracy.
Solution Approach 2:
The system transforms the extensive raw charging/discharging data into compressed feature representations through the 2D CNN training process. This parameter transformation reduces the complexity of data processing by converting large volumes of raw data into condensed, meaningful features that capture the essential degradation patterns.
3Productivity
If the number of charging/discharging cycles is reduced for prediction, then the efficiency is improved, but the reliability of the prediction deteriorates
Solution Approach 1:
The system incorporates feedback mechanisms where the 2D CNN model continuously refines its predictions based on the charging/discharging data patterns it has learned during training. This feedback loop enables the model to maintain high reliability with fewer input cycles by adjusting and optimizing its internal parameters based on the learned relationships between charging/discharging patterns and SOH degradation.
Data Source
AI summary
According to an embodiment disclosed herein, a battery management apparatus includes a memory and a controller, in which the controller may be configured to obtain first charging/discharging data by performing a charging/discharging cycle on a first battery cell a first number of times, input input data comprising the first charging/discharging data to a two-dimensional (2D) convolutional neural network (CNN) trained based on second charging/discharging data obtained by performing the charging/discharging cycle on a second battery cell a second number of times, and predict a state of health (SOH) of the first battery cell, based on result data output through the 2D CNN in response to the input data.