Battery remaining life prediction method and system based on decomposition integration asymmetric strategy

Through the battery residual life prediction method based on decomposition integrated asymmetric strategy, the enhanced Northern Goshawk optimization-variant modal decomposition and neural network model is used to solve the complexity problem of battery capacity sequence decomposition and prediction, and achieve more accurate Battery remaining service life forecast.

CN118033431BActive Publication Date: 2024-12-03SHANDONG UNIV
View PDF 20 Cites 0 Cited by

Patent Information

Application Number
CN202410359169.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-03-27
Publication Date
2024-12-03
Estimated Expiration
2044-03-27

AI Technical Summary

Technical Problem

The prior art has problems such as modal aliasing phenomenon in the prediction of the remaining service life of the battery, low algorithm efficiency, and the difficulty of a single algorithm to accurately describe the complex degradation process of the battery, resulting in unstable prediction and poor generalization.

Method used

Using a method based on decomposition integrated asymmetric strategy, the battery capacity sequence is decomposed through the enhanced Northern Goshawk optimization-variant modal decomposition method, combined with asymmetric dimensionality reduction and neural network models to predict the future discharge capacity of the battery, dynamic weighted summing is generated accurately Prediction of battery life.

Benefits of technology

It effectively reduces the complexity of the battery capacity sequence, retains key data information, improves model performance and efficiency, and obtains more accurate battery capacity prediction results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118033431B_ABST
    Figure CN118033431B_ABST
Patent Text Reader

Abstract

The present invention provides a method and system for predicting the remaining life of a battery based on a decomposition-integration asymmetric strategy, which obtains the discharge capacity data during the battery charge and discharge cycle, generates a battery capacity sequence, and decomposes and divides it to generate a low-frequency and high-frequency component sequence; uses an asymmetric dimensionality reduction strategy to reduce the dimension of the low-frequency and high-frequency component sequences respectively to generate low-frequency and high-frequency component sequences; uses different neural network models to predict the corresponding sequence of the battery's future discharge capacity for different component sequence historical data; calculates the weight of the low-frequency and high-frequency component historical sequence data, and combines the future discharge capacity low-frequency and high-frequency component sequences for dynamic weighted summation to obtain the future discharge capacity data sequence; calculates the remaining service life of the battery based on the future discharge capacity data sequence and the configured battery failure threshold, and obtains the prediction result. The present invention effectively describes the degradation of the battery discharge capacity and accurately predicts the remaining service life of the battery.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of battery remaining service life prediction, and in particular relates to a battery remaining service life prediction method and system based on a decomposition integration asymmetric strategy. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] With the rapid development of modern industry, energy demand is also increasing. However, given the non-renewable nature of traditional energy, the demand for efficient and clean energy is imminent. Batteries, as an electrochemical energy storage device, can store energy from clean energy in the form of electrical energy. Safety and reliability have always been key issues in the practical use of batteries. As the number of charge and discharge cycles increases, battery performance will continue to decline, leading to potential safety hazards. Therefore, accurate prediction of the remaining service life of the battery is crucial for its safe application.

[0004] During the entire life cycle of the battery, the battery storage capacity will decrease with the increase of the number of charge and discharge cycles, but due to the influence of the electrochemical reaction inside the battery, the battery capacity will suddenly and instantaneously increase during use, which is the battery capacity regeneration phenomenon. Correctly handling the battery capacity regeneration phenomenon is crucial for the prediction of the remaining service life. Empirical mode decomposition and its improved algorithm are often used to adaptively decompose the battery capacity sequence data into a series of intrinsic mode functions, but there are some problems, such as the lack of rigorous mathematical proof, the existence of mode aliasing, and low algorithm efficiency, which cannot effectively separate useful information from noise. Compared with empirical mode decomposition and its improved algorithm, the variational mode decomposition algorithm has obvious advantages in processing nonlinear and non-stationary signals. The algorithm has a solid mathematical theoretical foundation and high efficiency. In the variational mode decomposition algorithm, the penalty factor and the number of decomposition layers have a greater impact on the decomposition effect, and manual determination of these parameters will bring large errors.

[0005] Data-driven methods avoid complex chemical, physical, and mathematical modeling processes. They analyze and mine historical data through various data processing methods, and establish statistical models or machine learning models based on the data to predict battery aging trends. In recent years, deep learning has been widely used in the field of battery remaining service life prediction due to its powerful adaptive feature extraction and generalization capabilities. However, it is difficult to accurately describe the complex degradation process of the battery using only a single algorithm, and there are problems such as unstable predictions and poor generalization. Summary of the invention

[0006] In order to solve the above problems, the present invention proposes a battery remaining life prediction method and system based on a decomposition integration asymmetric strategy. The present invention can effectively describe the degradation of the battery discharge capacity and accurately predict the remaining service life of the battery.

[0007] According to some embodiments, the present invention adopts the following technical solutions:

[0008] In a first aspect, a method for predicting remaining battery life based on a decomposition integrated asymmetric strategy is provided, comprising the following steps:

[0009] Obtain the discharge capacity data during the battery charge and discharge cycle, generate a battery capacity sequence, and decompose it to generate a series of intrinsic mode component sequence data;

[0010] Perform cumulative summation operations on the sequence data of each intrinsic mode component respectively, divide them according to the summation results, and generate low-frequency component sequences and high-frequency component sequences;

[0011] Using an asymmetric dimensionality reduction strategy, the low-frequency and high-frequency component sequences are respectively reduced in dimension to generate a low-frequency component sequence and a high-frequency component sequence;

[0012] Different neural network models are used to predict the corresponding sequence of battery future discharge capacity for different component sequence historical data;

[0013] Calculate the weights of the historical sequence data of low-frequency and high-frequency components, and dynamically weight the sum of the future discharge capacity low-frequency component sequence and the future discharge capacity high-frequency component sequence to obtain the future discharge capacity data sequence;

[0014] According to the future discharge capacity data series and the configured battery failure threshold, the remaining service life of the battery is calculated to obtain the prediction result.

[0015] As an optional implementation method of the first aspect, when decomposing the battery capacity sequence, an enhanced northern goshawk optimization-variational modal decomposition method is used for decomposition, and the specific process includes:

[0016] Initialize the variational mode decomposition parameter decomposition layer number K and penalty factor α and its range, initialize the northern goshawk population parameters, input battery capacity data, and use the randomly determined variational mode decomposition parameter [K, α] value to initialize the initial position of the population members in the search space;

[0017] Calculate the fitness function value, use the variational mode decomposition method to decompose the battery capacity sequence, calculate the sample entropy value and sort it, and find the optimal fitness value and its corresponding position;

[0018] According to the prey recognition operation in the optimization algorithm of the northern goshawk, the updated fitness value is calculated. If the updated position is better than the position before the update, the position of the northern goshawk is updated, otherwise it remains unchanged. When the position is updated, a nonlinear decreasing weight algorithm is introduced for improvement.

[0019] According to the chasing and escaping behavior operations in the optimization algorithm of the northern goshawk, the updated fitness value is calculated. If the updated position is better than the position before the update, the position of the northern goshawk is updated, otherwise it remains unchanged, and when the position is updated, the golden sine algorithm is used for improvement;

[0020] Determine whether the iteration termination condition is met. If not, return to the step of calculating the fitness function value. If yes, use the optimal parameters obtained by the enhanced Northern Goshawk optimization algorithm as the final parameters of the variational mode decomposition method to adaptively decompose the battery capacity sequence.

[0021] As a further limited implementation method, the specific process of the improvement using the nonlinear decreasing weight algorithm includes: the improved position update method is:

[0022]

[0023] in, is the new position of the i-th northern goshawk in the j-dimension in the first stage, x i,j is the position of the i-th northern goshawk in the j-dimension, P i,j is the optimal position of the i-th northern goshawk in the j-th dimension. Parameters r and I are random numbers used in search and update, where R is a random number in the interval [0,1], I is 1 or 2, and F i is the objective function of the i-th northern goshawk, F Pi is the objective function of the i-th northern goshawk prey location. η is the weight control factor, t is the current iteration number, and T is the maximum iteration number.

[0024] The specific process of using the golden sine algorithm for improvement includes: the improved position update method is:

[0025]

[0026] in is the new position of the i-th northern goshawk in the j-th dimension in the second stage, R is the search radius, r is a random number in the interval [0,1], and g i,j is the golden sine algorithm, g i,j =x i,j |sin(R1)|+R2sin(R2)|aP i,j -bx i,j|, where R1 is a random number in [0,2π], R2 is a random number in [0,π], a and b are the golden section coefficients, and the calculation formulas are as follows: a=π(1-2τ), b=π(1+2τ),

[0027] If the updated position is better than the previous position, then use Updates the position of the Northern Goshawk, otherwise unchanged.

[0028] As an optional implementation method of the first aspect, the specific process of generating low-frequency component sequence and high-frequency component sequence by dividing according to the summation result includes comparing the cumulative sum of the values ​​of each intrinsic modal component sequence data with the set threshold value. If the sum value is greater than the set threshold value, it is a high-frequency component, otherwise it is defined as a low-frequency component.

[0029] As an optional implementation manner of the first aspect, the specific process of reducing the dimensions of the low-frequency and high-frequency component sequences respectively by using an asymmetric dimension reduction strategy includes:

[0030] A linear dimensionality reduction method is used to reduce the dimensionality of a series of low-frequency components, and a nonlinear dimensionality reduction method is used to reduce the dimensionality of a series of high-frequency components to generate a low-frequency component sequence and a high-frequency component sequence. The specific expression is as follows:

[0031]

[0032] Among them, LDR is a linear dimensionality reduction method, NDR is a nonlinear dimensionality reduction method, and imf i , i∈[1,j] is the i-th low-frequency component sequence, imf k , k∈[j+1,n] is the kth high frequency component sequence, imf d and IMF g They are respectively the low-frequency component sequence and the high-frequency component sequence after dimensionality reduction.

[0033] As a further limited implementation method, the linear dimensionality reduction method uses principal component analysis. The newly constructed cumulative contribution rate formula is used in the principal component analysis, and the specific expression is as follows:

[0034]

[0035] Among them, Me is the median and μ is the cumulative contribution rate.

[0036] In this embodiment, the nonlinear dimensionality reduction method uses an autoencoder. The autoencoder uses a newly constructed activation function, and the specific expression is as follows:

[0037]

[0038] Among them, x is the input, f(x) is the output, λ is the weight coefficient, RReLU is the random rectified linear unit, Sigmoid is the Sigmoid function, and the constructed activation function is more suitable for processing nonlinear data.

[0039] As an optional implementation manner of the first aspect, the specific process of using the first neural network model to predict the low-frequency component sequence of the future discharge capacity of the battery includes: using the pre-trained convolutional neural network-bidirectional long short-term memory model as the first neural network model; the specific process of using the second neural network model to predict the high-frequency component sequence of the future discharge capacity of the battery includes: using the pre-trained time convolution network-attention mechanism-deep neural network model as the second neural network model; the specific expression is:

[0040]

[0041] Among them, IMF t d ,t∈[1,l] is the historical low-frequency component value, l is the length of the historical sequence data, is the future low-frequency component value, imf t g ,t∈[1,l] is the historical high-frequency component value, is the future high-frequency component value.

[0042] As an optional implementation method of the first aspect, the process of calculating the weight of the historical sequence data of low-frequency and high-frequency components includes:

[0043]

[0044]

[0045] Among them, α l is the weight coefficient of the low-frequency component sequence, β l is the weight coefficient of the high-frequency component sequence, ε is a positive decimal to avoid the denominator being zero, and p and q are preset exponents.

[0046] As an optional implementation of the first aspect, the specific process of calculating the remaining service life of the battery according to the future discharge capacity data sequence and the configured battery failure threshold includes:

[0047] The remaining service life of the battery is the number of charge and discharge cycles from the current moment to the predefined failure threshold of the battery, denoted by RUL = T EOL -T i , where RUL is the remaining battery life, T EOL is the number of charge and discharge cycles when the battery capacity reaches the failure threshold, T i The number of cycles for the current battery capacity.

[0048] In a second aspect, a battery remaining life prediction system based on a decomposition integration asymmetric strategy is provided, comprising:

[0049] A data decomposition module is configured to obtain discharge capacity data during a battery charge and discharge cycle, generate a battery capacity sequence, and decompose the battery capacity sequence to generate a series of intrinsic mode component sequence data;

[0050] The data partitioning module is configured to perform cumulative summation operations on each intrinsic mode component sequence data, and partition according to the summation results to generate a low-frequency component sequence and a high-frequency component sequence;

[0051] A data dimension reduction module is configured to use an asymmetric dimension reduction strategy to reduce the dimensions of the low-frequency component sequence and the high-frequency component sequence respectively to generate a low-frequency component sequence and a high-frequency component sequence;

[0052] A low-frequency component prediction module is configured to predict the low-frequency component sequence of the future discharge capacity of the battery using the first neural network model based on the historical data of the low-frequency component sequence;

[0053] A high-frequency component prediction module is configured to predict the high-frequency component sequence of the future discharge capacity of the battery using a second neural network model based on the historical data of the high-frequency component sequence;

[0054] A dynamic weighting module is configured to calculate the weight of the historical sequence data of low-frequency and high-frequency components, and dynamically weight the sum of the future discharge capacity low-frequency component sequence and the future discharge capacity high-frequency component sequence to obtain the future discharge capacity data sequence;

[0055] The calculation prediction module is configured to calculate the remaining service life of the battery according to the future discharge capacity data sequence and the configured battery failure threshold to obtain a prediction result.

[0056] According to a third aspect, an electronic device is provided, comprising a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein the steps in the above method are completed when the computer instructions are executed by the processor.

[0057] Compared with the prior art, the present invention has the following beneficial effects:

[0058] (1) The present invention adaptively decomposes the battery capacity sequence through the enhanced Northern Goshawk optimization-variational modal decomposition method, which can effectively decompose any battery capacity sequence data and reduce the complexity of the battery capacity sequence.

[0059] (2) The present invention performs cumulative summation operations on each intrinsic mode component sequence data, sets a rated threshold, divides according to the summation result, generates a series of low-frequency component sequences and high-frequency component sequences, and can effectively classify the decomposed sequences.

[0060] (3) The present invention adopts an asymmetric dimensionality reduction strategy and uses different dimensionality reduction methods to reduce the dimensionality according to the characteristics of different decomposition sequences. This can retain key data information, reduce the complexity of calculation, and improve the performance and efficiency of the model.

[0061] (4) The present invention uses the historical sequence data of low-frequency and high-frequency components to calculate their respective weights, and combines the future discharge capacity low-frequency component sequence with the future discharge capacity high-frequency component sequence to dynamically weight and sum to obtain the future discharge capacity data sequence. Based on data changes, the weights are adjusted in real time to obtain more accurate battery capacity prediction results.

[0062] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0064] Figure 1 is a flow chart of a method for predicting remaining battery life based on a decomposition-integration asymmetric strategy in an embodiment;

[0065] Figure 2 is a flowchart of the enhanced northern goshawk optimization-variational modal decomposition in step S2 of an embodiment;

[0066] Figure 3 This is a flow chart of dynamic weighted summation in step S7 in an embodiment. DETAILED DESCRIPTION

[0067] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0068] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0069] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0070] In the absence of conflict, the embodiments in this application and the features in the embodiments may be combined with each other.

[0071] Embodiment 1

[0072] like Figure 1 As shown, a method for predicting the remaining battery life based on a decomposition integration asymmetric strategy specifically includes the following steps:

[0073] Step S1, measuring the discharge capacity data of the battery during the charge and discharge cycle to generate a battery capacity sequence;

[0074] Step S2, based on the adaptive decomposition method, the battery capacity sequence data is decomposed to generate a series of intrinsic mode component sequence data; the adaptive decomposition method means that any battery capacity sequence data can be effectively decomposed without adjusting parameters, thereby reducing the complexity of the battery capacity sequence. The enhanced Northern Goshawk Optimization (NGO)-Variational Mode Decomposition (VMD) method is used to effectively decompose any battery capacity sequence data to reduce the complexity of the battery capacity sequence, such as Figure 2 As shown, the specific steps are as follows:

[0075] S21: Initialize the VMD parameter decomposition layer number (K) and penalty factor (α) and their range, initialize the northern goshawk population parameters, input battery capacity data, and use the randomly determined VMD parameter [K, α] values ​​to initialize the initial positions of the population members in the search space.

[0076] S22: Calculate the fitness function value, apply VMD to decompose the battery capacity sequence, calculate the sample entropy value and sort it, and find the optimal fitness value and its corresponding position.

[0077] S23: For the prey identification operation in the original NGO algorithm, in order to balance the global search and local search of the algorithm and reasonably allocate the search methods in the early and late stages of the algorithm search, the original position update formula is improved. The improved formula is as follows:

[0078]

[0079] in, is the new position of the i-th northern goshawk in the j-dimension in the first stage, x i,j is the position of the i-th northern goshawk in the j-dimension, P i,j is the optimal position of the i-th northern goshawk in the j-th dimension. Parameters r and I are random numbers used in search and update, where R is a random number in the interval [0,1], I is 1 or 2, and F iis the objective function of the i-th northern goshawk, is the objective function of the i-th northern goshawk prey position. η is the weight control factor, t is the current iteration number, and T is the maximum iteration number. If the updated position is better than the position before the update, the above formula is used to update the position of the northern goshawk, otherwise it remains unchanged.

[0080] S24: For the chasing and escaping behavior operations in the original NGO algorithm, its original position update formula may fall into the local optimum. Therefore, in order to further improve the global search ability of the algorithm, the original position update formula is improved, and the improved formula is as follows:

[0081]

[0082] in is the new position of the i-th northern goshawk in the j-th dimension in the second stage, R is the search radius, r is a random number in the interval [0,1], and g i,j It is the golden sine algorithm, and the specific formula is as follows:

[0083] g i,j =x i,j |sin(R1)|+R2sin(R2)|aP i,j -bx i,j |

[0084] Where R1 is a random number in [0,2π], R2 is a random number in [0,π], a and b are the golden section coefficients, and the calculation formulas are as follows: a=π(1-2τ), b=π(1+2τ), If the updated position is better than the position before the update, then use Updates the position of the Northern Goshawk, otherwise unchanged.

[0085] S25: Determine the algorithm termination condition. If the iteration termination condition is not met, return to step S22. If the termination condition is met, output the optimal parameters calculated by the enhanced NGO as the final parameters of the VMD algorithm to adaptively decompose the battery capacity sequence.

[0086] Step S3, perform cumulative summation operations on each intrinsic mode component sequence data, record the summation result as Q, and set the threshold T to 0.01. If Q is greater than T, the component is defined as a high-frequency component, otherwise it is defined as a low-frequency component, and a series of low-frequency component sequences and high-frequency component sequences are generated;

[0087] Step S4, based on an asymmetric dimensionality reduction strategy including a linear dimensionality reduction method and a nonlinear dimensionality reduction method, respectively, a series of low-frequency and high-frequency component sequences are reduced in dimension to generate a low-frequency component sequence and a high-frequency component sequence. The linear dimensionality reduction method includes but is not limited to the principal component analysis (PCA) method, and the nonlinear dimensionality reduction method includes but is not limited to the auto-encoder (AE).

[0088] The specific expression is as follows:

[0089]

[0090] Among them, LDR is a linear dimensionality reduction method, NDR is a nonlinear dimensionality reduction method, and imf i , i∈[1,j] is the i-th low-frequency component sequence, imf k , k∈[j+1,n] is the kth high frequency component sequence, imf d and IMF g They are respectively the low-frequency component sequence and the high-frequency component sequence after dimensionality reduction.

[0091] In this embodiment, the linear dimensionality reduction method uses principal component analysis. The newly constructed cumulative contribution rate formula is used in PCA, and the specific expression is as follows:

[0092]

[0093] Among them, Me is the median and μ is the cumulative contribution rate.

[0094] In this embodiment, the nonlinear dimensionality reduction method uses an automatic encoder. The newly constructed activation function is used in AE, and the specific expression is as follows:

[0095]

[0096] Among them, x is the input, f(x) is the output, λ is the weight coefficient, RReLU is the random rectified linear unit, Sigmoid is the Sigmoid function, and the constructed activation function is more suitable for processing nonlinear data.

[0097] Step S5, establish a Convolutional Neural Network (CNN)-Bidirectional Long Short-Term Memory (BiLSTM) model, the input of the CNN-BiLSTM model is the historical sequence data of the low-frequency component, and the output is the future low-frequency component value; establish a Temporal Convolutional Network (TCN)-Attention Mechanism (AM)-Deep Neural Network (DNN) model, the input of the TCN-AM-DNN model is the historical sequence data of the high-frequency component, and the output is the future high-frequency component value; the specific expression is as follows:

[0098]

[0099] Among them, IMF t d ,t∈[1,l] is the historical low-frequency component value, l is the length of the historical sequence data, is the future low-frequency component value, imf t g ,t∈[1,l] is the historical high-frequency component value, is the future high-frequency component value.

[0100] Step S6, using the historical sequence data of low-frequency and high-frequency components, calculate the respective weights, and dynamically weight the sum of the future discharge capacity low-frequency component sequence and the future discharge capacity high-frequency component sequence to obtain the future discharge capacity data sequence, such as Figure 3 As shown, the specific steps are as follows:

[0101] S61: Based on the historical sequence data of low-frequency and high-frequency components, calculate the respective weights. The calculation formula is as follows:

[0102]

[0103]

[0104] Among them, α l is the weight coefficient of the low-frequency component sequence, β l is the weight coefficient of the high frequency component sequence. ε is a positive decimal to avoid the denominator being zero. p is set to 2 and q is set to 0.5.

[0105] S62: Based on the obtained weights of the low-frequency and high-frequency components, the future discharge capacity low-frequency component sequence and the future discharge capacity high-frequency component sequence are dynamically weighted and summed to calculate the future battery capacity prediction result. The calculation formula is as follows:

[0106]

[0107] Among them, α l and β l The value of changes dynamically with the data length l, and b is the bias coefficient.

[0108] Step S7, based on the battery failure threshold, deduce the remaining service life of the battery, the specific steps are as follows:

[0109] S71: The battery capacity failure threshold is when the battery discharge capacity drops to 80% of its rated capacity;

[0110] S72: The remaining service life of the battery is the number of charge and discharge cycles from the current time to the predefined failure threshold of the battery, denoted by RUL=T EOL -T i , where RUL is the remaining battery life, T EOL is the number of charge and discharge cycles when the battery capacity reaches the failure threshold, T i The number of cycles for the current battery capacity.

[0111] The specific numerical values ​​in the above embodiments can be adaptively adjusted in other embodiments according to specific application scenarios and predicted needs. These are easily conceivable by those skilled in the art and should fall within the scope of protection of the present invention.

[0112] Embodiment 2

[0113] A battery remaining life prediction system based on a decomposition integration asymmetric strategy, comprising:

[0114] A data decomposition module is configured to obtain discharge capacity data during a battery charge and discharge cycle, generate a battery capacity sequence, and decompose the battery capacity sequence to generate a series of intrinsic mode component sequence data;

[0115] The data partitioning module is configured to perform cumulative summation operations on each intrinsic mode component sequence data, and partition according to the summation results to generate a low-frequency component sequence and a high-frequency component sequence;

[0116] A data dimension reduction module is configured to use an asymmetric dimension reduction strategy to reduce the dimensions of the low-frequency component sequence and the high-frequency component sequence respectively to generate a low-frequency component sequence and a high-frequency component sequence;

[0117] A low-frequency component prediction module is configured to predict the low-frequency component sequence of the future discharge capacity of the battery using the first neural network model based on the historical data of the low-frequency component sequence;

[0118] A high-frequency component prediction module is configured to predict the high-frequency component sequence of the future discharge capacity of the battery using a second neural network model based on the historical data of the high-frequency component sequence;

[0119] A dynamic weighting module is configured to calculate the weight of the historical sequence data of low-frequency and high-frequency components, and dynamically weight the sum of the future discharge capacity low-frequency component sequence and the future discharge capacity high-frequency component sequence to obtain the future discharge capacity data sequence;

[0120] The calculation prediction module is configured to calculate the remaining service life of the battery according to the future discharge capacity data sequence and the configured battery failure threshold to obtain a prediction result.

[0121] According to implementation requirements, the computer instructions in the present invention can be loaded / stored on a computer or other programmable device to execute a series of operation steps. The functions specified by a process / multiple processes / a step / multiple steps in the flowchart described in the present invention can be synthesized into a new process / step to be executed on a computer or other programmable device to achieve the purpose of the present invention.

[0122] 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.

[0123] 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.

[0124] 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 including an instruction device, which implements the process Figure 1A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0125] 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 in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0126] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made by those skilled in the art within the spirit and principle of the present invention without creative labor shall be included in the protection scope of the present invention.

Claims

1. A method for predicting remaining battery life based on a decomposition-integration asymmetric strategy, characterized in that: The following steps are involved: Obtain the discharge capacity data during the battery charge and discharge cycle, generate a battery capacity sequence, and decompose it to generate a series of intrinsic mode component sequence data; Perform cumulative summation operations on the sequence data of each intrinsic mode component respectively, divide them according to the summation results, and generate low-frequency component sequences and high-frequency component sequences; Using an asymmetric dimensionality reduction strategy, the low-frequency and high-frequency component sequences are respectively reduced in dimension to generate a low-frequency component sequence and a high-frequency component sequence; Different neural network models are used to predict the corresponding sequence of battery future discharge capacity for different component sequence historical data; Calculate the weights of the historical sequence data of low-frequency and high-frequency components, and dynamically weight the sum of the future discharge capacity low-frequency component sequence and the future discharge capacity high-frequency component sequence to obtain the future discharge capacity data sequence; According to the future discharge capacity data sequence and the configured battery failure threshold, the remaining service life of the battery is calculated to obtain the prediction result; When decomposing the battery capacity sequence, the enhanced Northern Goshawk optimization-variational mode decomposition method is used for decomposition. The specific process includes: Initialize the variational mode decomposition parameter decomposition layer number K and penalty factor α and its range, initialize the northern goshawk population parameters, input battery capacity data, and use the randomly determined variational mode decomposition parameter [K, α] value to initialize the initial position of the population members in the search space; Calculate the fitness function value, use the variational mode decomposition method to decompose the battery capacity sequence, calculate the sample entropy value and sort it, and find the optimal fitness value and its corresponding position; According to the prey recognition operation in the optimization algorithm of the northern goshawk, the updated fitness value is calculated. If the updated position is better than the position before the update, the position of the northern goshawk is updated, otherwise it remains unchanged. When the position is updated, a nonlinear decreasing weight algorithm is introduced for improvement. According to the chasing and escaping behavior operations in the optimization algorithm of the northern goshawk, the updated fitness value is calculated. If the updated position is better than the position before the update, the position of the northern goshawk is updated, otherwise it remains unchanged, and when the position is updated, the golden sine algorithm is used for improvement; Determine whether the iteration termination condition is met, if not, return to the step of calculating the fitness function value, if yes, use the optimal parameters obtained by the enhanced Northern Goshawk optimization algorithm as the final parameters of the variational mode decomposition method to adaptively decompose the battery capacity sequence; The specific process of improving by using the nonlinear decreasing weight algorithm includes: the improved position update method is: in, is the new position of the i-th northern goshawk in the j-dimension in the first stage, x i,j is the position of the i-th northern goshawk in the j-dimension, P i,j is the optimal position of the i-th northern goshawk in the j-th dimension. Parameters r and I are random numbers used in search and update, where r is a random number in the interval [0,1], I is 1 or 2, and F i is the objective function of the i-th northern goshawk, is the objective function of the i-th northern goshawk prey position; η is the weight control factor, t is the current iteration number, and T is the maximum iteration number; if the updated position is better than the position before the update, the above formula is used to update the position of the northern goshawk, otherwise it remains unchanged; The specific process of using the golden sine algorithm for improvement includes: the improved position update method is: in is the new position of the i-th northern goshawk in the j-th dimension in the second stage, R is the search radius, g i,j is the golden sine algorithm, g i,j =x i,j |sin(R1)|+R2sin(R2)|aP i,j -bx i,j |, where R1 is a random number in [0,2π], R2 is a random number in [0,π], a and b are the golden section coefficients, and the calculation formulas are as follows: a=π(1-2τ), b=π(1+2τ), If the updated position is better than the previous position, then use Update the position of the northern goshawk, otherwise it remains unchanged; The specific process of dividing the low-frequency component sequence and the high-frequency component sequence according to the summation result includes comparing the cumulative summation value of each intrinsic mode component sequence data with the set threshold value. If the summation value is greater than the set threshold value, it is a high-frequency component, otherwise it is defined as a low-frequency component. The specific process of reducing the dimensions of the low-frequency and high-frequency component sequences respectively by using an asymmetric dimension reduction strategy includes: A linear dimensionality reduction method is used to reduce the dimensionality of a series of low-frequency components, and a nonlinear dimensionality reduction method is used to reduce the dimensionality of a series of high-frequency components to generate a low-frequency component sequence and a high-frequency component sequence. The specific expressions are as follows: Among them, LDR is a linear dimensionality reduction method, NDR is a nonlinear dimensionality reduction method, and imf i , i∈[1,j] is the i-th low-frequency component sequence, imf k , k∈[j+1,n] is the kth high frequency component sequence, imf d and IMF g They are the low-frequency component sequence and high-frequency component sequence after dimension reduction respectively; The linear dimensionality reduction method uses principal component analysis, and the newly constructed cumulative contribution rate formula is used in the principal component analysis. The specific expression is as follows: Among them, Me is the median, μ is the cumulative contribution rate; The nonlinear dimensionality reduction method uses an automatic encoder, and the newly constructed activation function is used in the automatic encoder. The specific expression is as follows: Among them, x is the input, f(x) is the output, λ is the weight coefficient, RReLU is the random rectified linear unit, Sigmoid is the Sigmoid function, and the constructed activation function is more suitable for processing nonlinear data; The specific process of using the first neural network model to predict the low-frequency component sequence of the future discharge capacity of the battery includes: using the pre-trained convolutional neural network-bidirectional long short-term memory model as the first neural network model; the specific process of using the second neural network model to predict the high-frequency component sequence of the future discharge capacity of the battery includes: using the pre-trained temporal convolutional network-attention mechanism-deep neural network model as the second neural network model; the specific expression is: Among them, IMF t d ,t∈[1,l] is the historical low-frequency component value, l is the length of the historical sequence data, is the future low-frequency component value, imf t g ,t∈[1,l] is the historical high-frequency component value, is the future high-frequency component value.

2. The method for predicting the remaining battery life based on the decomposition integration asymmetric strategy as claimed in claim 1, characterized in that: The process of calculating the weight of the low-frequency and high-frequency component historical series data includes: Among them, α l is the weight coefficient of the low-frequency component sequence, β l is the weight coefficient of the high-frequency component sequence, ε is a positive decimal to avoid the denominator being zero, and p and q are preset exponents.

3. The method for predicting the remaining battery life based on the decomposition integration asymmetric strategy as claimed in claim 1, characterized in that: The specific process of calculating the remaining service life of the battery based on the future discharge capacity data series and the configured battery failure threshold includes: The remaining service life of the battery is the number of charge and discharge cycles from the current moment to the predefined failure threshold of the battery, denoted by RUL = T EOL -T i , where RUL is the remaining battery life, T EOL is the number of charge and discharge cycles when the battery capacity reaches the failure threshold, T i The number of cycles for the current battery capacity.

4. A battery remaining life prediction system based on a decomposition integration asymmetric strategy, characterized in that: include: A data decomposition module is configured to obtain discharge capacity data during a battery charge and discharge cycle, generate a battery capacity sequence, and decompose the battery capacity sequence to generate a series of intrinsic mode component sequence data; The data partitioning module is configured to perform cumulative summation operations on each intrinsic mode component sequence data, and partition according to the summation results to generate a low-frequency component sequence and a high-frequency component sequence; A data dimension reduction module is configured to use an asymmetric dimension reduction strategy to reduce the dimensions of the low-frequency component sequence and the high-frequency component sequence respectively to generate a low-frequency component sequence and a high-frequency component sequence; A low-frequency component prediction module is configured to predict the low-frequency component sequence of the future discharge capacity of the battery using the first neural network model based on the historical data of the low-frequency component sequence; A high-frequency component prediction module is configured to predict the high-frequency component sequence of the future discharge capacity of the battery using a second neural network model based on the historical data of the high-frequency component sequence; A dynamic weighting module is configured to calculate the weight of the historical sequence data of low-frequency and high-frequency components, and dynamically weight the sum of the future discharge capacity low-frequency component sequence and the future discharge capacity high-frequency component sequence to obtain the future discharge capacity data sequence; A calculation prediction module is configured to calculate the remaining service life of the battery according to the future discharge capacity data sequence and the configured battery failure threshold to obtain a prediction result; When decomposing the battery capacity sequence, the enhanced Northern Goshawk optimization-variational mode decomposition method is used for decomposition. The specific process includes: Initialize the variational mode decomposition parameter decomposition layer number K and penalty factor α and its range, initialize the northern goshawk population parameters, input battery capacity data, and use the randomly determined variational mode decomposition parameter [K, α] value to initialize the initial position of the population members in the search space; Calculate the fitness function value, use the variational mode decomposition method to decompose the battery capacity sequence, calculate the sample entropy value and sort it, and find the optimal fitness value and its corresponding position; According to the prey recognition operation in the optimization algorithm of the northern goshawk, the updated fitness value is calculated. If the updated position is better than the position before the update, the position of the northern goshawk is updated, otherwise it remains unchanged. When the position is updated, a nonlinear decreasing weight algorithm is introduced for improvement. According to the chasing and escaping behavior operations in the optimization algorithm of the northern goshawk, the updated fitness value is calculated. If the updated position is better than the position before the update, the position of the northern goshawk is updated, otherwise it remains unchanged, and when the position is updated, the golden sine algorithm is used for improvement; Determine whether the iteration termination condition is met, if not, return to the step of calculating the fitness function value, if yes, use the optimal parameters obtained by the enhanced Northern Goshawk optimization algorithm as the final parameters of the variational mode decomposition method to adaptively decompose the battery capacity sequence; The specific process of improving by using the nonlinear decreasing weight algorithm includes: the improved position update method is: in, is the new position of the i-th northern goshawk in the j-dimension in the first stage, x i,j is the position of the i-th northern goshawk in the j-dimension, P i,j is the optimal position of the i-th northern goshawk in the j-th dimension. Parameters r and I are random numbers used in search and update, where r is a random number in the interval [0,1], I is 1 or 2, and F i is the objective function of the i-th northern goshawk, is the objective function of the i-th northern goshawk prey position; η is the weight control factor, t is the current iteration number, and T is the maximum iteration number; if the updated position is better than the position before the update, the above formula is used to update the position of the northern goshawk, otherwise it remains unchanged; The specific process of using the golden sine algorithm for improvement includes: the improved position update method is: in is the new position of the i-th northern goshawk in the j-th dimension in the second stage, R is the search radius, g i,j is the golden sine algorithm, g i,j =x i,j |sin(R1)|+R2sin(R2)|aP i,j -bx i,j |, where R1 is a random number in [0,2π], R2 is a random number in [0,π], a and b are the golden section coefficients, and the calculation formulas are as follows: a=π(1-2τ), b=π(1+2τ), If the updated position is better than the previous position, then use Update the position of the northern goshawk, otherwise it remains unchanged; The specific process of dividing the low-frequency component sequence and the high-frequency component sequence according to the summation result includes comparing the cumulative summation value of each intrinsic mode component sequence data with the set threshold value. If the summation value is greater than the set threshold value, it is a high-frequency component, otherwise it is defined as a low-frequency component. The specific process of reducing the dimensions of the low-frequency and high-frequency component sequences respectively by using an asymmetric dimension reduction strategy includes: A linear dimensionality reduction method is used to reduce the dimensionality of a series of low-frequency components, and a nonlinear dimensionality reduction method is used to reduce the dimensionality of a series of high-frequency components to generate a low-frequency component sequence and a high-frequency component sequence. The specific expressions are as follows: Among them, LDR is a linear dimensionality reduction method, NDR is a nonlinear dimensionality reduction method, and imf i , i∈[1,j] is the i-th low-frequency component sequence, imf k , k∈[j+1,n] is the kth high frequency component sequence, imf d and IMF g They are the low-frequency component sequence and high-frequency component sequence after dimension reduction respectively; The linear dimensionality reduction method uses principal component analysis, and the newly constructed cumulative contribution rate formula is used in the principal component analysis. The specific expression is as follows: Among them, Me is the median, μ is the cumulative contribution rate; The nonlinear dimensionality reduction method uses an autoencoder, and the newly constructed activation function is used in the autoencoder. The specific expression is as follows: Among them, x is the input, f(x) is the output, λ is the weight coefficient, RReLU is the random rectified linear unit, Sigmoid is the Sigmoid function, and the constructed activation function is more suitable for processing nonlinear data; The specific process of using the first neural network model to predict the low-frequency component sequence of the future discharge capacity of the battery includes: using the pre-trained convolutional neural network-bidirectional long short-term memory model as the first neural network model; the specific process of using the second neural network model to predict the high-frequency component sequence of the future discharge capacity of the battery includes: using the pre-trained temporal convolutional network-attention mechanism-deep neural network model as the second neural network model; the specific expression is: Among them, IMF t d ,t∈[1,l] is the historical low-frequency component value, l is the length of the historical sequence data, is the future low-frequency component value, imf t g ,t∈[1,l] is the historical high-frequency component value, is the future high-frequency component value.

Citation Information

Patent Citations

  • Lithium ion battery remaining life prediction method based on wolf pack optimization LSTM network

    CN109993270A

  • Lithium ion battery remaining service life prediction method based on long and short term memory LSTM and particle filter PF

    CN111103544A

  • Health state and residual life prediction method of multi-type lithium ion battery pack management system

    CN111707956A

  • Lithium battery residual life prediction method based on variational mode decomposition and integrated depth model

    CN112348185A

  • A lithium ion battery remaining useful life prediction method and device

    CN112966432A