Battery SOC Estimation With AI-Calibrated Coulomb Counting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Extended Kalman Filter (EKF) based battery state of charge (SOC) estimation methods are computationally intensive and require accurate initial state knowledge, making them inefficient for high-voltage battery packs with many cells, and they struggle with initial SOC uncertainty and parameter dependency.
Innovation Solution
A system combining coulomb counting (CC) estimation with periodic artificial intelligence (AI) calibration, using an AI estimator like LSTM, to periodically recalibrate SOC estimates, reducing computational load while maintaining accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Extended Kalman Filter (EKF) is used for SOC estimation, then measurement precision is improved, but use of energy increases due to computational intensity
Solution Approach 1:
The patent implements periodic calibration using AI-based estimation at specific intervals (e.g., when SOC reaches certain thresholds or after specific time periods) rather than continuous EKF computation. This periodic intervention corrects cumulative errors in coulomb counting while avoiding continuous high computational overhead, thereby reducing energy consumption while maintaining accuracy.
Solution Approach 2:
The patent segments the SOC estimation process into two distinct phases: (1) continuous low-computational coulomb counting for tracking charge/discharge, and (2) periodic AI-based calibration for error correction. This segmentation allows the system to use computational resources efficiently by applying intensive processing only when necessary, rather than continuously.
2Measurement precision
If EKF is used for SOC estimation, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The system segments functionality between a simple coulomb counter (continuous operation) and an AI calibration module (periodic operation). The coulomb counter provides continuous tracking with minimal complexity, while the AI module handles complex computations only periodically, reducing overall system complexity compared to continuous EKF implementation.
Solution Approach 2:
The coulomb counter continuously tracks SOC without requiring complex processing, serving itself for routine monitoring. The AI calibration module periodically intervenes to correct errors, allowing the simple coulomb counter to maintain operation without complex control logic, thereby reducing device complexity.
3Use of energy by moving object
If coulomb counting is used for SOC tracking, then use of energy is reduced, but reliability deteriorates due to de-calibration over time
Solution Approach 1:
The AI-based estimator periodically provides feedback to correct cumulative errors in the coulomb counting process. By comparing AI-estimated SOC with coulomb counting results and applying corrections when discrepancies exceed thresholds, the system maintains reliability while keeping computational energy consumption low during intermediate periods.
Solution Approach 2:
Periodic AI calibration is implemented to reset and correct cumulative errors in coulomb counting before they compromise reliability. This periodic intervention ensures the coulomb counter remains accurate over extended operation periods without requiring continuous high-energy computation.
4Measurement precision
If AI estimation is performed continuously, then measurement precision is improved, but productivity decreases due to computational demands
Solution Approach 1:
AI-based estimation is performed periodically rather than continuously, intervening only when calibration is needed (e.g., at SOC thresholds or after specific time intervals). This periodic execution maintains measurement precision by correcting errors when necessary while maximizing processing efficiency by avoiding unnecessary continuous AI computations.
Solution Approach 2:
The system applies AI calibration partially - only when and where needed - rather than continuously across all operations. By triggering AI calibration only when error thresholds are exceeded or at scheduled intervals, the system achieves sufficient precision without the excessive computational burden of continuous AI processing, thereby improving productivity.
Data Source
AI summary
A battery management system for tracking a state of charge of a battery including multiple cells including a coulomb counter (CC) estimator and an artificial intelligence (AI) estimator. The CC estimator determines a charge difference over time for tracking changes of a state of charge of each cell and updates a corresponding state of charge value. The AI estimator converts a set of sample values of a cell into an estimated state of charge for calibrating the cell. A controller may periodically invoke the AI estimator to calibrate the state of charge of each cell and to update a corresponding state of charge value. Processing by the AI estimator may be minimized by being used only to calibrate each cell before combined error of the CC estimator reaches a predetermined threshold. AI based calibration may use Long Short-Term Memories and may further incorporate voltage mean and voltage variance over time.