Battery OCV Estimation via Hybrid Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing battery management systems struggle to accurately estimate Open Circuit Voltage (OCV) versus capacity curves in real-life applications, as they require slow charge or discharge cycles and long rests, making them impractical for dynamic operating conditions.
Innovation Solution
A hybrid algorithm that combines machine learning models with physical properties to estimate OCV and capacity using current, voltage, and temperature measurements during normal operation, mitigating biases and providing real-time OCV estimates aligned with physical properties of the battery.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional OCV measurement methods are used (slow charge/discharge cycles and long rests), then measurement precision is improved, but productivity deteriorates due to long measurement time
Solution Approach 1:
The patent replaces the traditional electrochemical measurement system with a machine learning-based estimation system. Instead of physically measuring OCV through slow charge/discharge cycles and long rests, the system uses a trained neural network model that predicts OCV from normal operational measurements (current, voltage, temperature), thereby achieving both high accuracy and fast measurement speed
Solution Approach 2:
The patent creates a virtual copy of the battery's electrochemical behavior through a machine learning model trained on extensive operational data. This digital twin replicates the battery's OCV characteristics without requiring actual physical measurement time, allowing instantaneous estimation while maintaining measurement precision
2Device complexity
If OCV curve is kept fixed for entire battery lifetime, then device complexity is reduced, but reliability deteriorates due to battery aging effects
Solution Approach 1:
The patent transforms the static OCV curve into a dynamic, adaptive model that automatically updates with battery aging. The machine learning model continuously learns from new operational data and adjusts its predictions to reflect changing battery characteristics, maintaining high reliability throughout the battery's lifetime without requiring manual intervention or complex reconfiguration
Solution Approach 2:
The system implements feedback mechanisms where actual battery measurements are continuously fed back into the machine learning model to refine and update the OCV estimation. This feedback loop enables the model to adapt to battery aging effects, maintaining estimation accuracy while keeping the overall system architecture relatively simple
3Measurement precision
If machine learning model with bias correction is used, then measurement precision is improved, but device complexity increases due to multiple processing algorithms
Solution Approach 1:
The patent performs bias correction and model training in advance during an offline phase. Extensive pre-processing, feature engineering, and model training are completed before deployment, so that during actual battery operation, the system only needs to execute simple inference operations. This preliminary action reduces the complexity of real-time processing while maintaining high measurement precision
Data Source
AI summary
A battery management system includes a memory, a current sensor that measures a current flow through a battery to a load, a voltage sensor that measures a voltage level between a first terminal and a second terminal of the battery that are each connected to the load, and the memory, a temperature sensor that measures a temperature level of the battery; and a controller configured to be operatively connected to the current sensor, temperature sensor, and voltage sensor. The controller is configured to receive a measurement of a first current level and a first voltage level and utilize a corrected capacity and corrected open circuit voltage estimate to output an estimated open circuit voltage of the battery as compared to an estimated capacity.