Battery State Estimation via Segmented Adaptive Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing battery state estimation techniques face inefficiencies in adaptive learning and large errors in estimating State of Health (SOH), particularly due to reliance on impedance and lack of specific equations for SOH estimation.
Innovation Solution
A battery internal state estimating apparatus and method that employs a simulation model with a constant phase element (CPE) and internal resistance, using an extended Kalman filter for adaptive learning based on discharge current and electrolyte concentration, to efficiently determine SOH.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If parameters are adaptively learned in parallel using Kalman filter, then SOC estimation is achieved, but learning efficiency is low
Solution Approach 1:
The patent segments the adaptive learning process by dividing parameters into two groups: parameters learned during charging periods and parameters learned during discharging periods. This segmentation allows the system to focus learning on relevant parameters during each period, improving learning efficiency and reducing computational time compared to parallel learning of all parameters simultaneously.
Solution Approach 2:
The patent implements dynamic parameter selection where the set of parameters to be learned changes based on the battery's operating state (charging or discharging). The system dynamically switches between different parameter subsets, adapting the learning process to current conditions, which enhances efficiency by avoiding unnecessary computations on irrelevant parameters.
2Measurement precision
If SOH is estimated solely based on impedance, then estimation is simplified, but error is large
Solution Approach 1:
The patent merges multiple estimation approaches by combining impedance-based SOH estimation with simulation model-based estimation. The system integrates results from both methods, using the simulation model (which incorporates electrochemical reactions and mass transport) to correct and refine the impedance-based estimates, thereby improving accuracy while maintaining practical complexity through modular integration.
Solution Approach 2:
The patent introduces a simulation model as an intermediary between impedance measurements and final SOH estimation. The simulation model acts as a mediator that processes impedance data along with other battery state information to produce more accurate SOH estimates, bridging the gap between simple measurement and complex electrochemical reality.
3Measurement precision
If no specific equation for estimating SOH is used, then method is flexible, but SOH cannot be estimated
Solution Approach 1:
The patent employs parameter changes by using different equations and models appropriate for different battery states and operating conditions. The system selects and switches between various estimation equations based on parameters such as charge/discharge state, temperature, and current magnitude, enabling accurate SOH estimation across diverse conditions while managing complexity through conditional selection.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
To learn a parameter of a simulation model of a battery efficiently. A battery internal state estimating apparatus, estimating an internal state of a battery based on a simulation model of the battery, includes a storing section (RAM10c) that stores a plurality of parameters of the simulation model, a detecting section (1/F10d) that detects a discharge current flowing from the battery to a load, a selecting section (CPU 10a) that selects a parameter to be subjected to adaptive learning based on a value of the discharge current detected by the detecting section, and an adaptive learning section (CPU 10a) that performs adapting learning on a parameter selected by the selecting section.