Battery State of Health Determination via Data Segmentation

The patent applies preliminary action by using historical battery data to train machine learning models in advance, enabling the system to predict future battery degradation trends before failures occur. The model learns patterns from past performance and proactively identifies potential issues, transforming passive monitoring into active prediction.

The system implements feedback by continuously comparing real-time battery measurements against the machine learning model's predictions. When deviations are detected, the system adjusts its monitoring parameters and alerts operators, creating a closed-loop system that adapts to changing battery conditions and improves prediction accuracy over time.

The patent applies segmentation by dividing the battery data into discrete time intervals and operational modes (such as charging, discharging, resting). The machine learning model processes data segment by segment rather than handling the entire continuous stream at once, reducing computational complexity while maintaining real-time monitoring capability.

The system extracts only the most critical features and parameters from the raw battery data (such as voltage, current, temperature, and their rates of change), discarding redundant information. This feature extraction approach reduces the amount of data that needs to be processed in real-time while preserving the essential information needed for accurate degradation prediction.

The patent uses machine learning models that create a virtual copy or representation of the battery's operational characteristics. Instead of performing complex physical measurements or detailed simulations for every state of health assessment, the system uses the trained model to quickly predict SoH based on input data, achieving high precision with reduced processing time.

The system dynamically adjusts the level of modeling detail based on operational conditions. During normal operation, the system uses simplified models with fewer parameters for quick assessments. When higher precision is required (such as during critical states or initial calibration), the system switches to more detailed models, optimizing the balance between accuracy and processing time.