Receding Horizon Regression for Battery Impedance Estimation
Solution Overview
Problem
Hybrid electric vehicles face challenges in accurately estimating battery parameters, such as state of charge, power fade, and capacity fade, which affects the efficient operation and longevity of the battery pack, especially when the engine needs to start quickly and smoothly in response to changing power demands.
Innovation Solution
A battery management system using generalized linear regression analysis and a Randles Circuit Model to estimate battery model parameters, incorporating a receding horizon approach to minimize noise sensitivity and improve predictive accuracy, allowing for real-time control and optimization of battery power capability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional battery parameter estimation methods are used, then the system is simpler to implement, but the measurement precision of battery parameters deteriorates
Solution Approach 1:
The patent introduces an equivalent circuit model as an intermediary representation of the battery system, which transforms the complex electrochemical battery into a manageable electrical circuit model with identifiable parameters. This intermediary model enables accurate parameter estimation through regression analysis while keeping the control system implementation feasible.
Solution Approach 2:
The patent replaces traditional mechanical/mathematical filtering methods (like Kalman filters) with statistical regression analysis based on equivalent circuit models. This substitution provides more accurate parameter estimation by directly modeling the electrical characteristics of the battery rather than using indirect estimation techniques.
2Measurement precision
If real-time battery parameter estimation is performed continuously, then the measurement precision improves, but the use of energy increases
Solution Approach 1:
The patent performs regression analysis periodically at specific time intervals or under specific operating conditions rather than continuously. The system collects voltage and current data over defined time windows and performs parameter estimation at discrete intervals, reducing computational energy consumption while maintaining accurate real-time parameter tracking.
Solution Approach 2:
The patent performs parameter estimation only when sufficient data is available and when it provides meaningful updates, rather than continuously. The system uses sliding time windows and evaluates data quality metrics to determine when estimation should be performed, avoiding redundant calculations that would waste energy.
3Adaptability or versatility
If a fixed time window for data collection is used, then the device complexity is reduced, but the adaptability to changing operating conditions deteriorates
Solution Approach 1:
The patent implements a sliding time window approach where the data collection window dynamically adjusts its position and potentially its size based on operating conditions. The window slides through the data stream, continuously updating the most recent data set used for regression analysis, allowing the system to adapt to changing battery characteristics while maintaining a manageable data processing complexity.
4Measurement precision
If noise filtering is applied to improve parameter estimation, then the measurement precision improves, but the loss of information increases
Solution Approach 1:
The patent extracts and separates the noise components from the signal through regression analysis rather than applying blanket filtering. By using the equivalent circuit model to identify and isolate parameter-related signals from noise, the system achieves accurate parameter estimation while preserving the essential information in the raw voltage and current measurements.
Data Source
AI summary
A hybrid vehicle includes a battery pack having one or more cells and a controller configured to continuously update battery model parameter values based on generalized linear regression analysis. The analysis updates the parameters using a dataset of independent variable vectors and a dependent variable vector constructed from a series of input currents to the battery pack and corresponding voltage responses of the battery pack that fall within a moving window.