Ship power lithium battery state of health prediction method based on hybrid deep model

By using hybrid deep models and cross-domain knowledge transfer technology, the problems of accuracy and generalization ability in predicting the health status of lithium-ion batteries in marine power systems were solved, achieving high-precision and robust health status management that can adapt to changing operating conditions and battery material variations.

CN121008189BActive Publication Date: 2026-06-19WEIHAI OCEAN VOCATIONAL COLLEGE
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
WEIHAI OCEAN VOCATIONAL COLLEGE
Filing Date
2025-09-09
Publication Date
2026-06-19

AI Technical Summary

Technical Problem

Existing methods for predicting the health status of lithium-ion batteries struggle to balance high accuracy and strong generalization in marine power systems, especially with a significant decline in prediction performance under different battery materials and complex application scenarios.

Method used

A hybrid deep model is adopted, combining attention and gating mechanisms. A universal combination of health features is selected through clustering ensemble method, and cross-domain knowledge transfer is introduced to construct a multi-scale feature encoder, which is combined with physical mechanism-assisted branches for prediction.

Benefits of technology

It achieves high-precision prediction of the health status of marine power lithium batteries throughout their entire life cycle, possesses good interpretability and robustness, adapts to different electrochemical material systems and variable operating conditions, reduces data acquisition costs, and supports real-time and accurate health status management.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121008189B_ABST
    Figure CN121008189B_ABST
Patent Text Reader

Abstract

This invention discloses a method for predicting the health status of marine power lithium batteries based on a hybrid deep model, specifically relating to the field of lithium-ion battery health status prediction technology. The method involves acquiring historical monitoring datasets of marine power lithium batteries during operation and extracting a set of health features. Using a clustering ensemble method, the optimal combination of health features with universal applicability is selected under unlabeled conditions. Based on this feature combination, a hybrid deep prediction model is constructed, transferring source domain knowledge from different electrochemical material systems and varying operating conditions to the target domain, forming an adaptive prediction model. This model is then used to predict the health status of the target marine power lithium battery, outputting the current health status value and future health trend. This invention can achieve high-precision, highly generalizable battery health status prediction in diverse operating environments, and has the advantages of low data requirements, strong physical interpretability, and good cross-domain adaptability.
Need to check novelty before this filing date? Find Prior Art