Battery SOH Estimation Using Attention-Guided Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for accurately estimating the state of health (SOH) of lithium-ion batteries are limited by their reliance on non-physicochemical models, which lack accuracy and computational efficiency, making it difficult to represent various aging states and internal dynamics effectively.
Innovation Solution
A method and apparatus using a pre-trained artificial neural network to estimate SOH by generating and processing input data from battery parameters, incorporating an attention map and score to determine the reliability of health state estimation, based on a pseudo-two-dimensional (P2D) model and inverted bottleneck network (IBN) structure for efficient computation and accurate aging prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If non-physicochemical models are used for SOH estimation, then computational efficiency is improved, but accuracy and ability to represent aging states deteriorates
Solution Approach 1:
The patent segments the complex SOH estimation problem into two distinct components: (1) a P2D electrochemical model that accurately represents battery physics and aging mechanisms, and (2) an IBN neural network that efficiently processes measurements. This segmentation allows each component to specialize - the P2D model provides accurate physics-based predictions while the IBN delivers computational efficiency - resolving the contradiction between accuracy and computational efficiency.
Solution Approach 2:
The patent introduces an inverted bottleneck network as an intermediary between the complex P2D electrochemical model and the practical SOH estimation task. The IBN acts as a computationally efficient proxy that learns to replicate the P2D model's accurate aging predictions without requiring the full computational burden of solving the partial differential equations in real-time, thus maintaining accuracy while improving productivity.
2Measurement precision
If complex electrochemical models are used, then accuracy in representing aging states is improved, but computational burden increases
Solution Approach 1:
The patent creates a simplified copy of the complex P2D electrochemical model using an inverted bottleneck network. This neural network copy captures the essential aging dynamics and electrochemical behaviors of the original P2D model but can be evaluated much more quickly, reducing computational burden while preserving accuracy in representing aging states.
Solution Approach 2:
The patent performs preliminary training of the IBN offline using data from the accurate P2D model. This preliminary action allows the system to pre-learn the complex electrochemical relationships during a computationally intensive phase, so that during actual SOH estimation, the pre-trained IBN can provide accurate aging state representations with minimal computational burden.
Data Source
AI summary
A method of estimating a state of health of a battery is performed by at least one computing device, and includes preparing a pre-trained artificial neural network, generating input data by measuring at least one parameter of the battery, inputting the input data into the pre-trained artificial neural network to obtain a health state estimation value of the battery and an attention map, calculating an attention score based on the attention map, and determining whether to trust the health state estimation value based on the attention score.