Lithium ion battery state-of-charge prediction method based on EKF combined with CNN-BiLSTM-Attention

By combining EKF with CNN-BiLSTM-Attention, a lithium-ion battery state-of-charge prediction model is constructed, which solves the problems of insufficient accuracy and poor robustness of lithium-ion batteries under complex operating conditions. It achieves adaptation to the strong nonlinearity and time-varying characteristics of the battery and avoids prediction drift, thereby improving prediction accuracy and interpretability.

CN122109842APending Publication Date: 2026-05-29HEFEI UNIV OF TECH
1 Cites 2 Cited by

Patent Information

Application Number
CN202610195487.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-02-11
Publication Date
2026-05-29

AI Technical Summary

Technical Problem

Existing methods for estimating the state of charge of lithium-ion batteries are insufficient in accuracy and robustness under complex operating conditions. They cannot adapt to the strong nonlinearity and time-varying characteristics of batteries, lack physical prior constraints, are prone to prediction drift, and are sensitive to noise and sensor errors.

Method used

We employ a method based on EKF combined with CNN-BiLSTM-Attention. By constructing a second-order Thevenin equivalent circuit model, we identify key parameters online using recursive least squares method, extract features by combining extended Kalman filter and convolutional neural network, and introduce attention mechanism and bidirectional long short-term memory network to capture temporal dynamic dependence, thereby achieving accurate prediction of battery state of charge.

Benefits of technology

It effectively reduces the impact of data noise and sensor errors, improves the accuracy and robustness of battery state of charge prediction, adapts to the strong nonlinearity and time-varying characteristics of batteries, avoids prediction drift in multiple operating conditions and boundary intervals, and enhances the interpretability of the model.

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

Abstract

The invention provides a lithium ion battery charge state prediction method based on EKF combined with CNN-BiLSTM-Attention, and belongs to the field of electrochemical energy storage. The prediction comprises the steps of constructing an equivalent circuit model and identifying key parameters of the model on line; and obtaining a pseudo battery state-of-charge estimation value through extended Kalman filtering. The method comprises the following steps: extracting a complex feature vector by using a convolutional neural network layer, obtaining a weighted feature vector through an attention mechanism layer, obtaining a BiLSTM feature vector through a bidirectional long-short-term memory network layer, and finally outputting a final battery state-of-charge prediction value SOCpred. According to the method, the physical model and the data driving model are cooperatively adapted to the characteristics of the battery, the SOC prediction precision is effectively improved, the influence of data noise and sensor errors can be greatly reduced, the method is adapted to industrial and commercial energy storage power stations, electric vehicles and other scenes depending on the lithium ion battery, and the accuracy and robustness of SOC prediction are guaranteed.
Need to check novelty before this filing date? Find Prior Art