Traction Battery Capacity Estimation Using Low-Uncertainty SOC Data

The patent segments the battery data points into multiple groups based on their associated SOC uncertainty levels. Instead of treating all data points uniformly, the system divides them into segments (e.g., high uncertainty, medium uncertainty, low uncertainty groups) and selectively uses segments with lower uncertainty for capacity estimation. This segmentation allows the system to utilize sufficient data points while maintaining estimation accuracy by excluding problematic high-uncertainty data.

The patent applies local quality by assigning different weights or selection criteria to different data points based on their local uncertainty characteristics. Data points with low SOC measurement uncertainty are selected or weighted more heavily, while data points with high uncertainty are excluded or weighted less. This local differentiation in data quality assessment enables precise capacity estimation by focusing on reliable data regions.

The patent implements a dynamic data selection approach where the set of usable battery data points changes based on real-time or historical SOC uncertainty assessments. Rather than using a fixed time window or all available data uniformly, the system dynamically identifies and selects data points from various time periods that meet uncertainty criteria. This dynamic selection maintains comprehensive time period coverage while ensuring reliability by adapting to varying measurement conditions.

The patent changes the selection parameter from simple time-based inclusion to uncertainty-based filtering. By introducing SOC uncertainty as a critical selection parameter, the system transforms how data from different time periods is evaluated. Data points are included or excluded based on their uncertainty parameter values rather than solely on their temporal position, enabling reliable capacity assessment across comprehensive time periods.

The patent applies preliminary action by pre-assessing and storing SOC uncertainty values for different battery data points before the capacity estimation process. This preliminary uncertainty characterization allows the estimation algorithm to quickly filter and select appropriate data points without performing complex real-time uncertainty analysis during the estimation itself. The preprocessing step maintains fast processing speed while enabling accurate uncertainty-based data selection.

The patent incorporates feedback mechanisms where SOC uncertainty measurements from previous operations inform the selection of data points for subsequent capacity estimations. The system uses feedback from SOC measurement quality assessments to adjust which data points are included in the estimation, creating a closed-loop process that continuously improves accuracy while maintaining efficient processing through learned patterns in uncertainty behavior.