Neural Network Battery Model for Accurate State Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current battery management systems face challenges in accurately estimating the state of charge (SoC) and state of health (SoH) of batteries, which limits their efficient operation and expands their available uses.
Innovation Solution
A method and apparatus that utilize a neural network-based battery model to calculate internal battery information, including ion concentration, current density distribution, and overpotential, using sensed physical quantities like current, voltage, and temperature, to determine the state of a battery, such as SoC, by training on previous potential information from a reference battery.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional battery management systems are used for state estimation, then the system structure is simple, but the estimation accuracy of SoC and SoH is insufficient
Solution Approach 1:
The patent introduces a neural network model as an intermediary between the battery and the management system. This neural network acts as a mediator that processes complex electrochemical relationships internally, providing accurate SoC and SoH estimates without requiring the external management system to implement complex estimation algorithms directly.
Solution Approach 2:
The patent creates a virtual copy of the battery's internal electrochemical state through the neural network model. Instead of directly measuring difficult-to-obtain parameters like ion concentration and overpotential, the system uses the neural network to generate accurate copies of these internal states based on easily measurable external parameters, achieving high estimation accuracy without complex physical sensing.
2Measurement precision
If complex electrochemical models are used to calculate internal battery information, then the estimation accuracy improves, but the calculation time increases
Solution Approach 1:
The neural network model is trained in advance using comprehensive electrochemical data and models. This preliminary training phase allows the network to learn complex relationships offline, so that during actual battery operation, the system can quickly infer internal states without performing time-consuming real-time calculations of ion concentration distributions and overpotential fields.
Solution Approach 2:
The patent replaces complex mechanical/electrochemical calculation systems with an information processing system (neural network). Instead of solving differential equations and performing iterative calculations in real-time, the system uses the trained neural network to rapidly estimate internal battery states, substituting computational mechanics with intelligent information processing that achieves both accuracy and speed.
Data Source
AI summary
According to one aspect, a method to estimate a state of a battery includes receiving physical quantity information about a sensed physical quantity of a battery, obtaining estimated information of the battery from a battery model based on the received physical quantity information. The battery model includes a training model configured to determine internal battery information comprising potential information of an internal material of the battery based on the physical quantity information and a mathematical function. The method further includes calculating an ion concentration in the battery using the mathematical function based on the internal battery information to determine the estimated information of the battery.