Battery SOH Estimation Using Adaptive Voltage-Current Parameters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for estimating battery state of health (SOH) are inaccurate and require discontinuing battery use, are environment-dependent, and suffer from sensor error accumulation, limiting real-time accuracy and adaptability.
Innovation Solution
A method using adaptive filters to estimate the G and H parameters from real-time voltage and current measurements, allowing for accurate SOH estimation without the need for complex models or additional operating conditions, utilizing recursive least squares (RLS) for parameter updates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional methods (measuring released electric charge, OCV-SOC relation, impedance tracking) are used to estimate battery SOH, then battery state can be estimated, but the methods require discontinuing battery use, matching certain conditions, or suffer from sensor error accumulation
Solution Approach 1:
The patent replaces conventional mechanical/electrical measurement methods (charge measurement, impedance tracking) with a data-driven machine learning approach. The system uses neural networks to process voltage and current data, substituting physical measurement systems with computational models that can operate continuously without disrupting battery usage.
Solution Approach 2:
The system enables the battery management system to continuously monitor and estimate its own state of health using embedded sensors and processing. The BMS self-diagnoses by processing its own operational data (voltage, current measurements) through the neural network model, eliminating the need for external testing equipment or discontinuing operation.
2Measurement precision
If conventional SOH estimation methods are used, then battery state can be estimated, but sensor errors accumulate leading to unreliable accuracy
Solution Approach 1:
The system implements continuous feedback by constantly comparing predicted voltage (from neural network) with actual measured voltage. The loss function calculates the difference between these values and uses backpropagation to update model parameters, creating a closed-loop system that continuously corrects estimation errors and prevents error accumulation.
Solution Approach 2:
The system dynamically adjusts model parameters (weights and biases in the neural network) based on real-time data. By changing these parameters through gradient descent optimization, the system adapts to battery aging and operational conditions, maintaining precision without the error accumulation that plagues static measurement methods.
3Measurement precision
If complex battery models are used for SOH estimation, then accuracy may improve, but hardware complexity and memory requirements increase
Solution Approach 1:
The patent applies partial action by using a simplified neural network architecture that processes only essential features (voltage and current data) rather than implementing a complete physics-based battery model. This partial approach achieves sufficient accuracy for SOH estimation while dramatically reducing computational complexity and hardware requirements.
Solution Approach 2:
The system extracts only the critical information needed for SOH estimation from the battery's operational data. Instead of modeling all battery physics and chemistry, the neural network extracts key patterns from voltage-current relationships, eliminating unnecessary complexity while retaining essential estimation capability.
Data Source
AI summary
Provided is a method of estimating the state of health of the battery according to various embodiments. The method of estimating the state of health of the battery comprises: measuring a voltage and current of a battery in use to periodically generate a voltage value and a current value; using an adaptive filter to periodically update a G parameter value and an H parameter value in real time from the voltage value and the current value, said parameters indicating the present state of the battery; and using an initial value and a final value of the G parameter that is preset and a present value of the G parameter to estimate the state of health of the battery in real time. The G parameter is a parameter that represents the sensitivity of the voltage to changes in the current of the battery, and the H parameter is a parameter that represents an effective potential determined by the local equilibrium potential distribution and resistance distribution inside the battery.