Local Model Network for Battery State of Charge Observer
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing battery management systems face challenges in creating a control observer for estimating the state of charge (SoC) of batteries, as each battery type requires a unique model, which is difficult to parameterize and unsuitable for real-time applications due to high computational costs, and the state of health of the battery is not effectively accounted for in existing methods.
Innovation Solution
A method using a local model network with linear, time-invariant models, determined via data-based modeling from optimized experimental design, allows for the creation of a control observer that can estimate SoC efficiently across different battery types and account for the state of health, reducing computational costs and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a nonlinear battery model is used to create a control observer for estimating SoC, then the estimation accuracy is improved, but the computational cost becomes very high making real-time application difficult
Solution Approach 1:
The patent divides the complex nonlinear battery model into multiple local linear models, each valid in specific operating ranges. This segmentation allows the system to use simple linear models for real-time computation while maintaining overall accuracy through the collection of local models that cover the full operating range.
Solution Approach 2:
The patent transforms the nonlinear model parameters into linear model parameters through local linearization around operating points. By changing the parameter representation from global nonlinear to local linear parameters, the computational complexity is reduced while preserving estimation accuracy within each local region.
2Measurement precision
If an extended Kalman filter is used for estimating nonlinear state, then the SoC can be estimated, but the linearization in each time step results in high to very high computational cost
Solution Approach 1:
Instead of continuously linearizing the nonlinear model at each time step as in extended Kalman filter, the patent pre-segments the operating range into multiple local regions and creates a separate linear model for each. This eliminates repeated linearization computations during real-time operation.
Solution Approach 2:
The patent performs the model linearization and parameter identification in advance during an offline phase, creating a library of local linear models before real-time operation. This preliminary action removes the computational burden of linearization from the real-time estimation process.
3Ease of manufacture
If a simple equivalent electric circuit model is used, then the model is easier to parameterize and compute, but it is only trustworthy within a specific parameter range such as temperature range
Solution Approach 1:
The patent segments the limited validity range of simple equivalent circuit models into multiple local models, each valid in its specific operating range. By combining these segmented local models, the overall system achieves both the simplicity of local linear models and the extended validity across the full operating range.
Solution Approach 2:
The patent creates a universal modeling framework that can handle different battery types and operating conditions. The collection of local linear models serves multiple functions: maintaining simplicity for easy parameterization while extending applicability across the entire operating range through the ensemble of local models.
4Measurement precision
If electro-chemical models are used to model nonlinear battery behavior, then the model accuracy is improved, but the models are difficult to parameterize and complex
Solution Approach 1:
The patent segments the complex electro-chemical modeling task into multiple simple linear models operating in different regions. Each local model captures the battery behavior in its specific operating range with simple linear relationships, avoiding the need for complex electro-chemical equations while maintaining accuracy through the collection of segmented models.
Data Source
AI summary
In order to create a control observer for any battery type in a structured and at least partially automated manner, first, a nonlinear model of the battery, in form of a local model network including a number of local, linear, time-invariant, and dynamic models, which each have validity in specific ranges of the input variables, is determined from the measuring data of a previously ascertained, optimized experimental design via a data-based modeling method. For each local model (LMi) of the model network determined in this manner, a local observer is then determined. The control observer (13) for estimating the SoC then results from a linear combination of the local observers.