Battery Soaking Prediction for Early Health Screening
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current battery testing methods are resource-intensive and time-consuming, particularly when evaluating battery health through pack soaking behavior, which requires completing the entire soaking process to assess battery suitability for vehicle installation.
Innovation Solution
A computer-implemented method using machine learning to predict battery health by analyzing initial battery measurement data during a shorter soaking duration, computing features such as individual cell voltage drop rates, and employing a Bayesian neural network for time-series prediction to forecast battery state after completion of the soaking process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the entire soaking process is completed to assess battery suitability, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The patent applies partial action by performing battery health prediction during only a portion of the soaking process (first duration) rather than waiting for the complete soaking period. The machine learning model computes features from measurement data collected during this partial duration and predicts the battery state after full soaking, thereby reducing the actual measurement time while maintaining assessment accuracy.
Solution Approach 2:
The patent implements preliminary action by using machine learning to predict the final battery health state before the soaking process completes. The system computes features from early measurement data and forecasts the outcome, enabling early determination of battery suitability without waiting for the entire soaking process to finish.
2Reliability
If the entire soaking process is completed to assess battery suitability, then reliability is improved, but productivity worsens
Solution Approach 1:
The patent reduces the measurement duration from the complete soaking process to a partial duration, collecting measurement data only during this shortened period. The machine learning model then predicts the battery state as if the full soaking had occurred, thereby improving productivity while maintaining reliability through accurate prediction.
Solution Approach 2:
The patent creates a computational model that replicates the outcome of the complete soaking process. The machine learning model is trained to copy the results that would be obtained from full soaking measurements, enabling rapid prediction without actually performing the time-consuming complete soaking test.
3Loss of time
If measurement data is collected for a shorter duration, then loss of time is reduced, but measurement precision worsens
Solution Approach 1:
The patent replaces the mechanical approach of collecting data throughout the entire soaking process with a computational approach. A machine learning model processes measurement data from a short duration and predicts the final battery state, substituting physical time consumption with computational prediction to maintain accuracy while reducing time.
Solution Approach 2:
The patent changes the parameters of the measurement process by collecting data at higher frequency or selecting specific critical parameters during the shortened duration. The machine learning model is trained to extract meaningful features from this condensed data set, maintaining prediction accuracy despite reduced measurement time.
Data Source
AI summary
A computer-implemented method for predicting a quality of a battery includes receiving a first battery measurement data for a first duration of soaking the battery, the first duration shorter than or equal to the soaking. The method further includes computing a plurality of features based on the first battery measurement data. The method further includes predicting, based on the plurality of features, a state of the battery after completion of the soaking. The method further includes outputting suitability of the quality of the battery based on the state of the battery as predicted.