Lithium battery health status prediction method and system based on neural network integration

By constructing multiple neural network models and iteratively adjusting coefficients, combining convolutional neural network, residual network and gated recursive network, the problem of low prediction accuracy of single network is solved, significantly improving the accuracy of lithium battery health status prediction, ensuring that Safety and reliability of battery systems.

CN114757340BActive Publication Date: 2024-11-08XIAN UNIV OF POSTS & TELECOMM
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202210462396.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-04-28
Publication Date
2024-11-08
Estimated Expiration
2042-04-28

AI Technical Summary

Technical Problem

In the prior art, a single neural network is difficult to capture all data characteristics during the degradation of lithium batteries, resulting in low prediction accuracy of lithium batteries' health status, affecting the safety and reliability of the battery system.

Method used

Using a neural network integration method, multiple lithium battery health status prediction network models are constructed, including convolutional neural network, residual network and gated recursive network, combined with series or parallel structures, iteratively adjusts the coefficients of each network model, and comprehensively Predict the health status of lithium batteries.

Benefits of technology

It significantly improves the accuracy of predicting the health status of lithium batteries, ensures the safety and reliability of the lithium battery system, and provides accurate evaluation and maintenance support for battery performance degradation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114757340B_ABST
    Figure CN114757340B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and system for predicting the health status of a lithium battery based on neural network integration. According to data collected during the cyclic charging and discharging process of a lithium battery and multiple sets of preset neural networks, multiple lithium battery health status prediction models are constructed. The results of multiple models are integrated, and the linear combination coefficients of each network model are iteratively adjusted to obtain an integrated lithium battery health status prediction model. According to various parameters in the actual charging process of the lithium battery, the lithium battery health status comprehensive prediction model is used to calculate the health status of the lithium battery. Compared with the traditional prediction method that only uses a single model, the present invention can significantly improve the accuracy of lithium battery health status prediction.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of lithium ion battery health status assessment, and in particular to a lithium battery health status prediction method and system based on neural network integration. Background Art

[0002] Lithium-ion batteries have become the core energy supply components of many electrical devices due to their superior performance such as light weight, high energy density, no memory effect, and low self-discharge rate. However, the charging and discharging process of lithium-ion batteries will form a solid electrolyte intermediate phase, which seriously affects its internal electrochemical reaction. This irreversible process will cause the capacity of lithium batteries to continue to decay. When lithium batteries are continuously charged and discharged, their chemical composition will tend to age, and the safety of the entire system will be affected by the degradation of battery performance. If the health state (State of Health, SOH) of lithium batteries cannot be accurately evaluated and maintenance or replacement measures cannot be taken in time, catastrophic consequences may occur, causing serious economic losses or even casualties. Therefore, when lithium batteries degrade to a certain extent, they should be retired to ensure the safety and reliability of lithium battery power supply systems. Indicators for evaluating the performance degradation of lithium batteries include battery capacity, output power, internal resistance, etc. Predicting the changing trends of these indicators can estimate the health state of lithium batteries and maintain the batteries in time, which provides strong support for the reliability of system operation and is of great significance to the safety of industrial systems.

[0003] In recent years, with the rapid improvement of computer computing power, neural network technology has developed rapidly and has achieved remarkable results in the fields of image processing and natural language processing. Therefore, applying neural network technology to the prediction of the health status of lithium-ion batteries has great practical significance and application prospects. However, the degradation process of lithium batteries is highly uncertain. Due to the limitations of its structural characteristics, a single neural network often cannot capture all data features and it is difficult to achieve higher accuracy.

[0004] In order to enable lithium-ion batteries to operate safely and efficiently on electrical equipment, a multi-network fusion lithium-ion battery health status prediction method is urgently needed to solve the problem of low accuracy of single network prediction results. Summary of the invention

[0005] In view of the problems existing in the prior art, the present invention provides a lithium battery health status prediction method and system based on neural network integration, which solves the problem of low accuracy of existing single network prediction results.

[0006] The present invention is achieved through the following technical solutions:

[0007] A method for predicting the health status of a lithium battery based on neural network integration comprises the following steps:

[0008] Step 1: construct multiple lithium battery health status prediction network models based on multiple neural networks;

[0009] Step 2: Determine the projection of the true value of the health status of the lithium battery in the projection space according to the prediction results and initial coefficients of each constructed network model;

[0010] Step 3: determine the loss between the projection and the true value of the health status of the lithium battery, and adjust the initial coefficients of each network model according to the loss to obtain the adjustment coefficients of each network model;

[0011] Step 4: Use the adjustment coefficient as the initial coefficient and repeat the above steps 3 and 4 until the projection of the true value of the health status of the lithium battery in the projection space and the loss of the given label drop to the set threshold, and then obtain the final coefficient of each network model;

[0012] Step 5: Determine the weighted prediction result based on the prediction result of each network model and the final coefficient, add the weighted prediction results of each network model, and obtain the comprehensive prediction result of the health status of the lithium battery.

[0013] Preferably, the multiple lithium battery health status prediction network models described in step 1 include a convolutional neural network model, a residual network model, a gated recursive network model, and a combined network model formed by connecting the convolutional neural network model and the residual network model in series or in parallel with the gated recursive network model, respectively.

[0014] Preferably, the initial coefficients of each network model are obtained according to the prediction results of each network model and the loss of a given label.

[0015] Preferably, the expression of the initial coefficient of the network model in step 2 is as follows:

[0016]

[0017] in, represents the initial coefficient of the i-th neural network model; k is the initial setting coefficient.

[0018] Preferably, the calculation expression of the projection of the true value of the health status of the lithium battery in the projection space in step 2 is as follows:

[0019]

[0020]

[0021] in, represents the coefficient of the kth adjustment of the i-th network model, is the actual degradation state of a lithium battery pack, for The projection in space S.

[0022] Preferably, the expression of the adjustment coefficient in step 3 is as follows:

[0023] ω i (k+1) =ω i (k) +ηZ i

[0024] Among them, ω i (k) Represents the coefficient of the last iteration of the i-th network model, ω i (k+1) represents the adjusted coefficient of the i-th network model, η represents the learning rate, Z i Represents the loss between the current projection value and the optimal value.

[0025] Preferably, in step 5, the prediction result of each network model is multiplied by the final coefficient to obtain a weighted prediction result; the expression of the comprehensive prediction model of the health status of the lithium battery is as follows:

[0026]

[0027]

[0028] in, Indicates setting threshold; represents the final coefficient of the i-th group of network models; f represents the comprehensive prediction model of lithium battery health status.

[0029] Preferably, the method further includes step 6, inputting the voltage, current and temperature of the battery during a single charging process into a comprehensive prediction model for the health status of the lithium battery, which outputs a comprehensive prediction result for the health status of the lithium battery.

[0030] A system for predicting the health status of lithium batteries based on neural network integration, comprising:

[0031] Network model building module, used to build multiple lithium battery health status prediction network models;

[0032] A projection module is used to determine the projection of the true value of the health status of the lithium battery in the projection space according to the prediction results and initial coefficients of each constructed network model;

[0033] The coefficient adjustment module is used to determine the loss between the projection and the true value of the health status of the lithium battery, adjust the initial coefficients of each network model according to the loss, and output the adjustment coefficients of each network model;

[0034] The coefficient optimization module is used to iteratively calculate the adjustment coefficient as the initial coefficient until the projection of the true value of the health status of the lithium battery in the projection space and the loss of the given label drop to a set threshold, thereby obtaining the final coefficient of each network model;

[0035] The prediction module is used to determine the weighted prediction result according to the prediction result of each network model and the final coefficient, and add the weighted prediction results of each network model to obtain the comprehensive prediction result of the health status of the lithium battery.

[0036] Compared with the prior art, the present invention has the following beneficial technical effects:

[0037] The present invention provides a lithium battery health status prediction method based on neural network integration. According to the data collected during the lithium battery cyclic charge and discharge process and multiple preset neural network groups, multiple lithium battery health status prediction network models are constructed; the results of the multiple network models are integrated, and the coefficients of each network model are iteratively adjusted to obtain an integrated lithium battery health status prediction model; according to various parameters in the actual charging process of the lithium battery, the lithium battery health status comprehensive prediction model is used to calculate the health status of the lithium battery; compared with the traditional prediction method that only uses a single model, the present invention can significantly improve the accuracy of lithium battery health status prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a flow chart of the lithium battery health status prediction method based on neural network integration of the present invention;

[0039] Figure 2 It is a schematic diagram of the structure of the high-dimensional space of the present invention. DETAILED DESCRIPTION

[0040] The present invention will be further described in detail below in conjunction with the accompanying drawings, which are intended to explain the present invention rather than to limit it.

[0041] See also Figure 1 , a lithium battery health status prediction method based on neural network integration, comprising the following steps:

[0042] Step 1: construct multiple lithium battery health status prediction network models based on the data collected during the lithium battery charge and discharge cycle and multiple preset neural networks;

[0043] The above collected data are the voltage, current, temperature and charge state data of this model of lithium battery during the cyclic charge and discharge process.

[0044] The above-mentioned preset multiple groups of neural networks include a convolutional neural network model (CNN) for extracting local features, a residual network model (ResNet) for extracting timing features, a gated recurrent unit (GRU) for extracting timing features, and a combined network model that extracts local features and timing features by combining the above three network models in series and parallel.

[0045] The combined network models include a combined model of a convolutional neural network model and a gated recurrent network model in series (CNN-GRU), a combined model of a convolutional neural network model and a gated recurrent network model in parallel (CNN-GRU parallel), a network model of a residual network and a gated recurrent network in series (ResNet-GRU), and a network model of a residual network and a gated recurrent network in parallel (ResNet-GRU parallel).

[0046] Next, the constructed convolutional neural network model, residual network model, gated recursive model and combined network model are integrated to adjust the coefficients of each model, and the prediction result of each network model is multiplied by the corresponding coefficient to obtain the weighted prediction result. All weighted prediction results are added to obtain the comprehensive lithium battery health status prediction model. The specific method is as follows:

[0047] Step 2: Calculate the loss of the prediction results of each network model and the given label to obtain the initial coefficients of each network model.

[0048] Initialize the initial adjustment coefficient of each network model. The specific method is as follows:

[0049] See also Figure 2 , let space H represent a high-dimensional space, which can represent any degradation state of lithium batteries, where is the actual degradation state of a lithium battery pack. The space S represents a low-dimensional space. The prediction results of each network model are the basis of the space. for The projection in space S.

[0050] Initially, the real value of the lithium battery health status The projection in space S is set to the origin, and the loss between the true value and the projection is Define d as the inner product of this difference and the basis:

[0051]

[0052] in, represents the prediction result of the i-th neural network; di Represents the prediction result and true value of the i-th network The inner product of ; N represents the number of preset neural network models.

[0053] The initial coefficients of each network model can be defined as:

[0054]

[0055] in, represents the initial coefficient of the i-th neural network model; k is the initial setting coefficient, and its value is 2.

[0056] Step 3: Multiply the prediction result of each network model by the corresponding initial coefficient to obtain the weighted prediction result of the network model. Add all weighted prediction results to obtain the projection of the true value of the health status of the lithium battery in the projection space. The calculation expression of the projection is as follows:

[0057]

[0058]

[0059] in, Represents the coefficient of the kth adjustment of the ith network model.

[0060] Step 4: Calculate the loss between the projection and the true value of the health status of the lithium battery, adjust the adjustment coefficient of each network model according to the loss, and obtain a new adjustment coefficient of each network model. The specific method is as follows:

[0061] If the vector in space S With the true value The error is minimal and the following conditions should be met:

[0062]

[0063] But at this time Not really The projection in space S, so the above formula does not satisfy it, let:

[0064]

[0065] Among them, Z i Represents the loss between the current projection value and the optimal value. The adjustment coefficient is calculated by the following formula to obtain the new projection value:

[0066] ω i (k+1) =ω i (k) +ηZ i

[0067] Among them, ωi (k) Represents the coefficient of the last iteration of the i-th network model, ω i (k+1) represents the adjusted coefficient of the ith network model, η represents the learning rate, and its value is 0.002.

[0068] Step 5: Use the new adjustment coefficient as the initial coefficient and repeat the above steps 3 and 4 until the projection of the true value of the lithium battery health status in the projection space and the loss of the given label drop to the set threshold, and then obtain the final coefficient of each network model.

[0069] Step 6: Multiply the prediction result of each network model by the final coefficient to obtain a weighted output result, and add all weighted output results to obtain a comprehensive prediction model for the health status of lithium batteries. The expression of the comprehensive prediction model for the health status of lithium batteries is as follows:

[0070]

[0071]

[0072] Among them, SSE represents a method of calculating loss; Indicates setting threshold; represents the final coefficient of the i-th group of network models; f represents the comprehensive prediction model of lithium battery health status.

[0073] Step 7: Collect the voltage, current, and temperature data of the lithium battery during a single charge, input them into each network model, multiply the output of each network model by the corresponding final coefficient to obtain a weighted output result, and add all the weighted output results to obtain a comprehensive prediction result of the lithium battery health status.

[0074] The lithium battery health status prediction method based on neural network integration constructs multiple lithium battery health status prediction models based on data collected during the lithium battery cyclic charge and discharge process and multiple preset neural network groups; the results of multiple models are integrated, and the linear combination coefficients of each network model are iteratively adjusted to obtain the integrated lithium battery health status prediction model; according to various parameters in the actual charging process of the lithium battery, the lithium battery health status comprehensive prediction model is used to calculate the health status of the lithium battery; compared with the traditional prediction method that only uses a single model, the present invention can significantly improve the prediction accuracy of the lithium battery health status.

[0075] A system for predicting the health status of lithium batteries based on neural network integration, comprising:

[0076] Network model building module, used to build multiple lithium battery health status prediction network models;

[0077] A projection module is used to determine the projection of the true value of the health status of the lithium battery in the projection space according to the prediction results and initial coefficients of each constructed network model;

[0078] The coefficient adjustment module is used to determine the loss between the projection and the true value of the health status of the lithium battery, adjust the initial coefficients of each network model according to the loss, and output the adjustment coefficients of each network model;

[0079] The coefficient optimization module is used to iteratively calculate the adjustment coefficient as the initial coefficient until the projection of the true value of the health status of the lithium battery in the projection space and the loss of the given label drop to a set threshold, thereby obtaining the final coefficient of each network model;

[0080] The prediction module is used to determine the weighted prediction result according to the prediction result of each network model and the final coefficient, and add the weighted prediction results of each network model to obtain a comprehensive prediction model for the health status of the lithium battery.

[0081] In another embodiment of the present invention, a computer device is provided, the computer device including a processor and a memory, the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, which are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions in a computer storage medium to implement the corresponding method flow or corresponding function; the processor described in the embodiment of the present invention can be used for the operation of the lithium battery health status prediction method based on neural network integration.

[0082] In another embodiment of the present invention, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It can be understood that the computer-readable storage medium here can include both built-in storage media in the computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the lithium battery health status prediction method based on neural network integration in the above embodiment.

[0083] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0084] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0085] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0086] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for predicting the health status of a lithium battery based on neural network integration, characterized in that: The following steps are involved: Step 1: construct multiple lithium battery health status prediction network models based on multiple neural networks; The multiple lithium battery health status prediction network models include a convolutional neural network model, a residual network model, a gated recursive network model, and a combined network model formed by connecting the convolutional neural network model and the residual network model in series or in parallel with the gated recursive network model; Step 2: Determine the projection of the true value of the health status of the lithium battery in the projection space according to the prediction results and initial coefficients of each constructed network model; The expression of the initial coefficient of the network model is as follows: in, represents the initial coefficient of the i-th neural network model; k is the initial setting coefficient; Step 3: determine the loss between the projection and the true value of the health status of the lithium battery, and adjust the initial coefficients of each network model according to the loss to obtain the adjustment coefficients of each network model; The expression of the adjustment coefficient is as follows: oh i (k+1) =ω i (k) +ηZ i Among them, ω i (k) Represents the coefficient of the last iteration of the i-th network model, ω i (k+1) represents the adjusted coefficient of the i-th network model, η represents the learning rate, Z i Represents the loss between the current projection value and the optimal value; Step 4: Use the adjustment coefficient as the initial coefficient and repeat the above steps 3 and 4 until the projection of the true value of the health status of the lithium battery in the projection space and the loss of the given label drop to the set threshold, and then obtain the final coefficient of each network model; Step 5: Determine the weighted prediction result according to the prediction result of each network model and the final coefficient, add the weighted prediction results of each network model, and obtain the comprehensive prediction result of the health status of the lithium battery; The prediction result of each network model is multiplied by the final coefficient to obtain the weighted prediction result; the expression of the comprehensive prediction model of lithium battery health status is as follows: in, Indicates setting threshold; represents the final coefficient of the i-th group of network models; f represents the comprehensive prediction model of lithium battery health status; The voltage, current and temperature of the battery during a single charge process are input into a comprehensive prediction model of the health status of the lithium battery, which outputs a comprehensive prediction result of the health status of the lithium battery.

2. The method for predicting the health status of a lithium battery based on neural network integration according to claim 1, characterized in that: According to the prediction results of each network model and the loss of the given label, the initial coefficients of each network model are obtained.

3. The method for predicting the health status of a lithium battery based on neural network integration according to claim 1, characterized in that: The calculation expression of the projection of the true value of the health status of the lithium battery in the projection space in step 2 is as follows: in, represents the coefficient of the kth adjustment of the i-th network model, is the actual degradation state of a lithium battery pack, for The projection in space S.

4. A system for predicting the health status of a lithium battery based on neural network integration according to any one of claims 1 to 3, characterized in that: include, Network model building module, used to build multiple lithium battery health status prediction network models; A projection module is used to determine the projection of the true value of the health status of the lithium battery in the projection space according to the prediction results and initial coefficients of each constructed network model; The coefficient adjustment module is used to determine the loss between the projection and the true value of the health status of the lithium battery, adjust the initial coefficients of each network model according to the loss, and output the adjustment coefficients of each network model; The coefficient optimization module is used to iteratively calculate the adjustment coefficient as the initial coefficient until the projection of the true value of the health status of the lithium battery in the projection space and the loss of the given label drop to a set threshold, thereby obtaining the final coefficient of each network model; The prediction module is used to determine the weighted prediction result according to the prediction result of each network model and the final coefficient, and add the weighted prediction results of each network model to obtain the comprehensive prediction result of the health status of the lithium battery.

Citation Information

Cited By

  • Lithium battery coating temperature control parameter optimization method and system based on deep learning model

    CN122885625A