Cross-type battery recession prediction method based on transfer learning

By constructing a general mathematical model and designing a new loss function, combined with the LSTM-seq2seq model of global attention mechanism, the problem of insufficient interpretability of the decline prediction method of span type batteries in the existing technology is solved, and accurate decay trajectory prediction of different types of batteries is achieved.

CN119763721APending Publication Date: 2025-04-04BEIJING JIAOTONG UNIV
0 Cites 0 Cited by

Patent Information

Application Number
CN202411543994.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-31
Publication Date
2025-04-04

AI Technical Summary

Technical Problem

The existing cross-type battery decline prediction method based on transfer learning lacks analysis of the intrinsic differences and connections of different battery cycle data, resulting in insufficient interpretability. Especially when there are large differences in capacity, life and charge and discharge curves, the prediction accuracy decreases and even causes negative migration.

Method used

A general mathematical model is constructed to describe the voltage-capacity (V-Q) curve relationship of different materials, reveal the commonality between different types of battery data, design a new loss function for parameter optimization in deep learning, and establish a LSTM-seq2seq target domain V-Q curve reconstruction model based on the global attention mechanism to reduce the negative impact of domain differences on transfer learning.

Benefits of technology

The interpretability and rationality of the transfer learning model are enhanced through stable mathematical mapping relationships. The reconstructed discharge curve pattern and trend are closer to the discharge curve of the source domain battery, effectively improving the accuracy of prediction of the decline trajectory of the cross-type battery.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses a cross-type battery recession prediction method based on transfer learning. The method comprises the following steps: 1, analyzing the relationship between discharge V-Q curves of batteries made of different materials, and constructing a general mathematical model for describing the potential relationship between the discharge V-Q curves of the batteries made of different materials; 2, designing a new loss function based on a mathematical model between V-Q curves of different types of batteries; 3, establishing an LSTM-seq2seq model based on a global attention mechanism, and optimizing model parameters by using a new loss function to reconstruct a V-Q curve of a battery in a target domain; and step 4, applying the reconstructed V-Q data of the battery, taking a Siamese-CNN model as a pre-training model, and proposing a migration-based recession trajectory prediction method suitable for different types of batteries. According to the invention, degradation trajectory prediction of different types of batteries can be realized based on a small number of cycles before the batteries are used in groups.
Need to check novelty before this filing date? Find Prior Art