A method and system for estimating the state of charge of a battery based on an improved neural network model.

By combining the improved 2-GLSTM model and the CSA algorithm, the problems of insufficient feature extraction and hyperparameter dependence in the state of charge estimation of lithium-ion batteries are solved, achieving high-precision and stable SOC estimation, which is suitable for new energy vehicles and energy storage systems.

CN122307408APending Publication Date: 2026-06-30ANHUI UNIVERSITY OF TECHNOLOGY
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
ANHUI UNIVERSITY OF TECHNOLOGY
Filing Date
2026-03-30
Publication Date
2026-06-30

AI Technical Summary

Technical Problem

Existing methods for estimating the state of charge of lithium-ion batteries suffer from insufficient feature extraction, reliance on human experience in setting hyperparameters, and poor adaptability to complex operating conditions, resulting in insufficient estimation accuracy and stability.

Method used

A dual-dimensional feature extraction method based on an improved 2D grid long short-term memory network (2-GLSTM) model is used, and adaptive hyperparameter optimization is performed by combining the Cuckoo Search algorithm (CSA). Through collaborative feature extraction of the time and depth dimensions, high-precision and high-stability SOC estimation is achieved.

Benefits of technology

It improves the accuracy and stability of lithium-ion battery state of charge estimation, enabling high-precision SOC estimation under different dynamic operating conditions, and is applicable to fields such as new energy vehicles and energy storage systems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122307408A_ABST
    Figure CN122307408A_ABST
Patent Text Reader

Abstract

This application relates to the field of new energy vehicle battery management technology, specifically a battery state-of-charge (SOC) estimation method and system based on an improved neural network model. The method involves standardizing and preprocessing time-series data of current, voltage, and temperature during battery charging and discharging to construct a sample set suitable for the model. Subsequently, a 2-GLSTM model is constructed, capturing both the time-series dynamic features and cross-layer coupling features of the battery through a collaborative feature extraction mechanism of time and depth dimensions. Comparative auto-analysis (CSA) is introduced to adaptively optimize hyperparameters such as the number of hidden layer neurons and the learning rate of the 2-GLSTM model globally. Model training is then completed based on the optimal hyperparameters, and an early stopping mechanism is used to avoid overfitting. Finally, real-time data from the battery under test is input into the trained model, directly outputting an accurate SOC estimate. This method solves the problems of insufficient feature extraction, reliance on manual hyperparameter tuning, and susceptibility to overfitting inherent in traditional SOC estimation methods.
Need to check novelty before this filing date? Find Prior Art