Battery State of Charge Estimation via Regime-Specific Regression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating the state of charge (SoC) of batteries, particularly in electric or hybrid vehicles, face challenges due to the complexity of battery behavior influenced by various parameters, temperature, and aging, leading to imprecise measurements and laborious calibration procedures, which limit their applicability in real-time applications.
Innovation Solution
A method and system that acquire time series of voltage and other physical parameter measurements to determine the operating regime of the battery, selecting a nonlinear regression model from a predefined set to estimate SoC, allowing for real-time estimation using embedded sensors and data fusion techniques, including kernel regression and support vector machines.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If direct calculation methods using voltage charts are used to estimate SoC, then the estimation can be performed in real-time, but the precision is reduced due to the non-one-to-one relationship between voltage and SoC for all battery types
Solution Approach 1:
The patent segments the continuous voltage-SoC relationship into discrete intervals or ranges. By dividing the voltage range into multiple segments and assigning different functional relationships or calibration parameters to each segment, the system can achieve both real-time calculation capability and improved precision. This segmentation allows the model to capture the non-linear and non-one-to-one characteristics of battery voltage-SoC relationships without requiring complex continuous models.
Solution Approach 2:
The patent changes parameters such as temperature, aging state, and charge/discharge rate to create multiple sets of voltage-SoC relationships. By adapting the estimation model to these varying parameters, the system maintains real-time performance while improving precision across different operating conditions. The model dynamically adjusts parameters based on measured conditions to select or weight appropriate voltage-SoC relationships.
2Measurement precision
If no-load voltage measurement is used to estimate SoC, then the precision is improved, but the applicability is limited to periods when the battery is not being used
Solution Approach 1:
The patent performs preliminary calibration or characterization of the battery's voltage-SoC relationship during periods when the battery is at rest or not in use. This preliminary action captures the no-load voltage characteristics and stores them as reference data or calibration parameters. During active operation, the system applies these pre-acquired parameters to estimate SoC, combining the precision of no-load measurements with the versatility of real-time operation.
Solution Approach 2:
The patent introduces an intermediary model or algorithm that bridges the gap between no-load voltage measurements and real-time operation conditions. This intermediary component processes the no-load voltage data and adapts it to account for load effects, temperature variations, and dynamic conditions, enabling the system to maintain high precision during battery operation without requiring direct no-load measurements at all times.
3Measurement precision
If battery impedance is used to estimate SoC, then the precision can be improved under certain conditions, but the measurement is strongly influenced by temperature and current conditions
Solution Approach 1:
The patent implements feedback mechanisms that continuously monitor temperature and current conditions during impedance measurements. Based on this feedback, the system adjusts the measurement protocol, selects appropriate reference values, or applies correction factors to compensate for environmental influences. This feedback loop enables the system to maintain high precision across varying conditions without requiring complex manual control of measurement conditions.
Solution Approach 2:
The patent develops a universal impedance-based SoC estimation model that functions across a wide range of temperature and current conditions. By creating a multi-functional model that accounts for various operating conditions through parameter adaptation or selection, the system achieves high precision without requiring separate measurement procedures for each condition, thereby reducing operational complexity.
4Measurement precision
If calibration procedures are performed to improve SoC estimation accuracy, then the precision is improved, but additional costs and time are required
Solution Approach 1:
The patent applies partial calibration procedures that focus on the most critical parameters or operating ranges rather than performing exhaustive calibration across all conditions. By identifying and calibrating only the most influential parameters or the most frequently encountered operating conditions, the system achieves sufficient precision with reduced calibration time and resources. This selective approach balances accuracy requirements with practical constraints.
Data Source
Figure 1~4
Figure 2
Figure 3A~3B
AI summary
The invention relates to a method for estimating the charge state of a battery (BATT) comprising the following steps: a) acquiring a time series of voltage measurements (Uτ) across the terminals of said battery, as well as at least one other time series of measurements of another physical parameter of said battery or its environment; b) as a function of said measurements, determining an operating regime of said battery; c) as a function of said operating regime, choosing a regression model from among a predefined set of such models; and d) estimating the state of charge (SoC) of said battery by applying said regression model to said time series of voltage measurements and said or at least one other said time series of measurements. The invention also relates to a device for estimating the charge state of a battery for implementing one such method. The invention also relates to a device for learning regression models of the charge state of a battery, suitable for implementing such a method. The invention also relates to a system for estimating the charge state of a battery comprising such a device for estimating the charge state and such a device for learning regression models.