Battery State Estimation Using Corrected SOC and Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting a battery's state of health (SOH) are inaccurate due to errors in calculating the state of charge (SOC), which can lead to permanent damage from overcharge or overdischarge, and require previous estimation tables that do not account for all battery characteristics.
Innovation Solution
A device and method that measure current, voltage, and temperature data to calculate a corrected state of charge (SOC), using machine learning techniques such as decision trees or neural networks to estimate the battery's state without relying on previous estimation tables, thereby improving accuracy and preventing error accumulation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If SOC is calculated using existing methods, then battery state prediction can be performed, but calculation errors accumulate over time due to sensor noise leading to inaccurate SOH estimation
Solution Approach 1:
The patent implements a feedback mechanism where the estimated SOH is used to correct the SOC calculation. The system continuously monitors the relationship between measured voltage/current and estimated SOH, then adjusts the SOC calculation accordingly to prevent error accumulation from sensor noise
Solution Approach 2:
The patent introduces an intermediate correction factor derived from the relationship between measured battery parameters and estimated SOH. This intermediary value serves to bridge the gap between raw sensor data and accurate SOC calculation, filtering out noise-induced errors
2Productivity
If previous estimation tables are used for SOH prediction, then estimation can be performed quickly, but the tables do not reflect all battery characteristics leading to inaccurate predictions
Solution Approach 1:
The patent transitions from static estimation tables to a dynamic estimation system that adapts to individual battery characteristics. The system continuously learns from actual battery behavior and adjusts its predictions in real-time, allowing it to capture unique battery features that fixed tables cannot represent
Solution Approach 2:
The patent changes the approach from using fixed parameter tables to dynamically adjusting estimation parameters based on real-time battery measurements and learned characteristics. This allows the system to adapt parameters like capacity fade rates and resistance changes to match the specific battery being monitored
Data Source
AI summary
An apparatus and a method for predicting a state of a battery are provided. The apparatus includes a data measuring unit that measures information about the battery and outputs first data, a data producing unit that reflects a change in available capacity of the battery based on at least a portion of the first data to calculate a corrected state of charge and processes the first data based on the corrected state of charge to generate second data, and outputs the second data, and a battery state estimating unit that estimates state information of the battery based on the second data.