A method for liquid metal battery curve reconstruction and SOH estimation based on fragmented data

CN122154492BActive Publication Date: 2026-07-03HEBEI UNIV OF TECH

Patent Information

Authority / Receiving Office: CN · China
Patent Type: Patents(China)
Current Assignee / Owner: HEBEI UNIV OF TECH
Filing Date: 2026-05-08
Publication Date: 2026-07-03

AI Technical Summary

Technical Problem

Existing technologies cannot effectively address the health status estimation requirements of liquid metal batteries under high-temperature operating conditions with fragmented and sparse data and frequent capacity regeneration scenarios. Traditional methods are insufficient to meet the requirements for high-precision and robust SOH estimation.

Method used

A curve reconstruction model based on weak physical information neural network is adopted, which combines one-dimensional convolutional neural network, bidirectional long short-term memory network and output layer, embeds physical prior constraints for data reconstruction, and performs SOH estimation through self-feedback online sequential extreme learning machine to achieve adaptive updating of the model.

Benefits of technology

It achieves high-fidelity voltage-capacity curve reconstruction and SOH estimation of liquid metal batteries throughout their entire life cycle under fragmented data conditions, improving the accuracy and robustness of health state estimation, reducing errors, and meeting the real-time monitoring needs of engineering projects.

✦ Generated by Eureka AI based on patent content.

Smart Images

Figure CN122154492B_ABST

Patent Text Reader

Abstract

This invention discloses a method for curve reconstruction and SOH estimation of liquid metal batteries based on fragmented data. First, fragmented short-term discharge voltage-capacity data segments generated by random grid scheduling are collected. Next, a curve reconstruction model is constructed, consisting of a one-dimensional convolutional neural network layer, a bidirectional long short-term memory network layer, and a weak physical information constraint layer, to recover the complete discharge curve from the sparse fragments. Then, multidimensional health features are extracted from the reconstructed curve, enhanced by classical correlation analysis, and input into an online sequential extreme learning machine model with a self-feedback pseudo-label mechanism. This model generates self-feedback pseudo-labels through exponential moving average adaptive bias correction and updates weights in real time using recursive least squares, thereby achieving adaptive tracking of capacity regeneration phenomena and accurate SOH estimation throughout the entire life cycle of the liquid metal battery. This invention effectively solves the estimation problem under extreme data shortage conditions, significantly improving the accuracy and robustness of SOH estimation.

Need to check novelty before this filing date? Find Prior Art