Battery health state evaluation method and device based on neural network

Abstract

The invention discloses a neural network-based battery health state evaluation method and device, and the method comprises the steps: carrying out the data collection of a battery attribute parameter, and storing a data collection result as a vector; the category of the data annotation comprises three states of health, sub-health and unhealth; constructing a sample data set in combination with the vectors and the corresponding labeling results; performing sample generation by using a generative adversarial network based on rule optimization, and expanding a sample data set; using the expanded sample data set to train the constructed battery health state evaluation model; and collecting current attribute parameters of a to-be-evaluated battery, inputting the current attribute parameters into the trained battery health state evaluation model, and outputting the category of the battery health state. According to the method, the problem of insufficient battery performance data volume can be solved, and the generalization ability of the model under various battery types and use conditions is enhanced.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and apparatus for assessing battery health status based on neural networks. Background Technology

[0002] Lithium-ion batteries are an indispensable component of modern electronic devices and electric vehicles. As usage time increases, battery performance gradually degrades, affecting storage capacity and output efficiency. Therefore, accurate assessment of battery health is crucial for ensuring safe device operation and extending battery life. Traditional battery health assessment methods and devices primarily rely on periodic battery performance tests, such as capacity testing and internal resistance measurement. For example, patent document CN117970153A obtains multiple battery health assessment indicators and calculates the variation and mean of each indicator within a target range to determine the actual value of the degradation coefficient; the actual degradation coefficient is compared with a standard coefficient to comprehensively derive the battery health. Patent document CN118425819A records the results of multiple charge-discharge cycles (including charging, discharging, and resting processes) and determines candidate battery health levels based on the differences between the current and previous cycles, finally combining multiple candidate values ​​to derive the target health level. However, these methods often fail to predict the future health status of the battery or provide accurate health assessments under complex usage conditions.

[0003] In recent years, machine learning and neural network technologies have demonstrated their powerful data processing and pattern recognition capabilities in multiple fields. For example, patent document CN119001506A combines a parameter prediction model with a battery physics model to predict battery status and simulate physicochemical characteristics using historical data. The first predicted data is input into the physics model to update features, and then the parameter prediction model is adjusted to obtain the second predicted data, thereby accurately assessing battery health. Another example is patent document CN118746773A, which obtains multiple battery parameters and processes them to obtain initial battery parameters; these initial parameters are then input into a pre-trained model to obtain reconstructed battery parameters; the reconstruction error is obtained based on the initial and reconstructed battery parameters; and the battery health is then determined based on the reconstruction error. However, existing battery health assessment technologies still face problems such as insufficient data and poor model generalization ability, making them unable to effectively address accurate assessments of different battery types and complex usage environments. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method and apparatus for assessing battery health status based on neural networks. By utilizing generative adversarial networks for data expansion, the problem of insufficient battery performance data is solved, enhancing the model's generalization ability across various battery types and usage conditions. The support vector machine based on quantum state simulation utilizes the superposition and entanglement states of qubits to enhance the classifier's expressive power, improving classification accuracy and efficiency.

[0005] The present invention adopts the following technical solution.

[0006] This invention proposes a battery health status assessment method based on neural networks, comprising:

[0007] S1. Collect data on battery attribute parameters, and store the results of the data collection as a vector; the data labeling categories include three states: healthy, sub-healthy, and unhealthy; construct a sample dataset by combining the vector and the corresponding labeling results;

[0008] S2. Based on the battery degradation pattern, a generative adversarial network is used to generate samples and expand the sample dataset. When the generative adversarial network is updated, the parameters of the generator are updated by combining performance constraints to ensure that the generated samples meet the physical characteristics of the battery.

[0009] S3. Use the expanded sample dataset to train the constructed battery health status assessment model;

[0010] S4. Collect the current attribute parameters of the battery to be evaluated, input them into the trained battery health status evaluation model, and output the category of battery health status.

[0011] Furthermore, in S2, the generator G receives data from a standard normal distribution p. z The random noise z of (z) and the actual degradation law of the battery are converted into generated data x related to battery properties;

[0012] Battery degradation is a nonlinear mapping that includes physical degradation laws and is related to battery temperature, charging current, and depth of discharge.

[0013] Furthermore, when training the generative adversarial network, in order to ensure the consistency between the physical characteristics of the battery properties in the generator's production samples and the actual data, the convex constraints and performance constraints C of the Frank-Wolfe algorithm are jointly optimized.

[0014] Furthermore, performance constraint C includes: constraining the generator's generated data within the constraint boundaries of battery performance;

[0015] Define the loss function L of the generator. GThe generator's loss function L G As the objective function, the update direction of the generator is adjusted by gradient calculation to obtain the optimal update direction;

[0016] Based on the calculated optimal update direction, the generator's parameter θ g The update rule in the (t+1)th iteration is related to the calculated optimal update direction, step size, and the generator parameters in the tth iteration.

[0017] Furthermore, the battery health status assessment model employs a support vector machine based on quantum state simulation as the classification algorithm; the specific steps for training the support vector machine algorithm based on quantum state simulation are as follows:

[0018] S301. Each vector in the dataset is converted into a set of qubit states; based on the qubit states of each vector, the corresponding quantum state is calculated.

[0019] S302. Quantum gate operations are used to simulate nonlinear kernel functions, and quantum gate operations are used to realize quantum kernel function mapping, mapping the quantum state of each vector to a new quantum state;

[0020] S303. Train a quantum support vector machine using a new quantum state and use the loss function L. q Conduct training;

[0021] S304, Update the loss function L q Parameters in;

[0022] S305. Repeat the iteration until the preset stopping iteration condition is met, and the model training is complete.

[0023] Furthermore, in S301, the input i-th vector q xi The corresponding quantum state |ψ> is encoded into the corresponding qubit of the quantum state, and the corresponding quantum state is calculated using the qubit;

[0024] In S302, quantum gate operations are used to simulate nonlinear kernel functions, utilizing quantum gate operations.

[0025] The quantum kernel function mapping is implemented to map the quantum state |ψ> of each vector to a new quantum state |ψ′>, including mapping the nonlinear correlation between the battery's capacity decay rate and discharge rate, and providing high-dimensional correlation information to the quantum support vector machine.

[0026] Furthermore, in S303, the quantum state |ψ′> is used to train a quantum support vector machine to capture changes in battery performance; the loss function L is used. q The formula is:

[0027]

[0028] Where m is the number of training samples; ξ qj It is a slack variable; y i y k These are the labels corresponding to the j-th and k-th new quantum states, respectively; α qj It is the Lagrange multiplier corresponding to the j-th training sample; α qk It is the Lagrange multiplier corresponding to the k-th training sample; <ψ′ j |ψ′ k > is a new quantum state |ψ′ j > and |ψ′ k The inner product of > represents the kernel function.

[0029] Furthermore, in S304, the loss function L... q Lagrange multipliers α qj Update the Lagrange multiplier α. qj Update with learning rate and loss function L q Regarding α qj It is related to the gradient;

[0030] With Lagrange multipliers α qj Similarly, the loss function L q Lagrange multipliers α qk It will also be updated at the same time.

[0031] This invention also proposes a battery health status assessment device based on neural networks, comprising:

[0032] Data acquisition module 10 acquires the test data of the target device;

[0033] The execution module 20 inputs the data to be tested into the battery health status assessment model, performs data identification on the data to be tested, and outputs the identification results;

[0034] The data processing module 30 performs fault diagnosis on the target device based on the identification results.

[0035] The present invention also proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, when the computer program is loaded onto the processor, it implements a neural network-based battery health status assessment method according to any one of the above.

[0036] The present invention also proposes a computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, it implements a neural network-based battery health status assessment method according to any one of the preceding claims.

[0037] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0038] 1. This invention utilizes generative adversarial networks (GANs) for data augmentation, effectively addressing the problem of insufficient battery performance data. By generating new samples, the diversity of the training dataset is increased, thereby enhancing the model's generalization ability across various battery types and usage environments. GANs can simulate different battery usage conditions and environmental changes, generating diverse samples to help the training model better adapt to various battery states, improving its predictive ability on unknown samples, and avoiding underfitting issues caused by insufficient data.

[0039] 2. This invention combines standard normally distributed random noise with the actual degradation patterns of batteries, enabling the generator to produce data that more closely reflects the actual battery degradation process. This not only enhances the diversity and physical consistency of the generated data but also ensures that the generated samples have high realism, better simulating the degradation trends of batteries under different conditions, thereby providing reliable data support for battery health management and prediction.

[0040] 3. This invention combines the convex constraints and performance constraints of the Frank-Wolfe algorithm with the design of the generator's loss function, helping the generator to follow the physical laws of the battery during the optimization process. Convex constraints ensure that the generator parameters are always within a legal range, avoiding the generation of data that does not conform to physical limitations and improving the rationality and accuracy of the generated data. Performance constraints, by introducing a penalty term, maintain the consistency between the generated samples and the actual battery performance. The generator loss function considers both sample quality and physical constraints, driving the generator to gradually approach the real data distribution during training. In this way, the generated battery data can both "deceive" the discriminator and possess higher physical feasibility.

[0041] 4. This invention employs a support vector machine based on quantum state simulation as the classification algorithm, leveraging the superposition and entanglement properties of qubits to enhance the classifier's expressive power. Quantum state simulation-based support vector machines can handle complex, high-dimensional data, improving the efficiency and accuracy of the classifier. Compared to traditional methods, they exhibit significant advantages in feature space representation and data classification capabilities. This enables the battery health status assessment model to provide more accurate health status assessments when faced with complex battery data, improving prediction accuracy and efficiency. Attached Figure Description

[0042] Figure 1 This is a flowchart of a battery health status assessment method based on neural networks according to the present invention;

[0043] Figure 2 This is a schematic diagram of a battery health status assessment device based on a neural network according to the present invention.

[0044] Figure 3 This is a schematic diagram of the structure of an electronic device for assessing battery health status based on a neural network, according to the present invention. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0046] Example 1

[0047] This invention proposes a battery health status assessment method based on neural networks, such as... Figure 1 As shown, the steps include the following.

[0048] S1. Data is collected for various battery usage environments, and the results of the data collection are stored as vectors; the data labels include three states: healthy, sub-healthy, and unhealthy; a dataset is constructed by combining the vectors and the corresponding labels.

[0049] Specifically, the data collected in this invention comes from various battery usage environments, including charging speed, temperature, battery cycle count, etc., and the data storage format is vector.

[0050] In one embodiment, the attributes of the data include:

[0051] ax1 (battery capacity): Represents the maximum energy storage capacity of the battery.

[0052] ax2 (charging time): This indicates the time required for the battery to go from fully discharged to fully charged.

[0053] ax3 (discharge rate): indicates the battery's discharge speed.

[0054] ax4 (cycle count): The number of charge-discharge cycles of the battery.

[0055] ax5 (temperature fluctuation): The average temperature change during battery use.

[0056] ax6 (voltage fluctuation): The range of voltage fluctuation during battery use.

[0057] ax7 (Internal Resistance Change): The change in the battery's internal resistance.

[0058] ax8 (capacity decay rate): The rate at which the battery capacity decays over time.

[0059] ax9 (Battery Leakage Current): The leakage current of the battery when it is not in operation.

[0060] ax10 (charging efficiency): The charging efficiency of a battery, that is, the efficiency of converting electrical energy into chemical energy during charging.

[0061] It should be emphasized that this embodiment is only to illustrate one data format and type of the present invention. In practical applications, the data usually has more than 10 attributes, and the number of data attributes may reach dozens or even hundreds.

[0062] Furthermore, the collected data is labeled. The labeling method of this invention is manual labeling. In one embodiment, the labeling categories include three states: "healthy", "sub-healthy" and "unhealthy".

[0063] Furthermore, in step S2, a rule-based generative adversarial network algorithm is used to generate samples, thereby expanding the dataset.

[0064] It is understandable that in the task of this invention, the acquisition, labeling, and preprocessing of training data are time-consuming and labor-intensive, and insufficient training samples can easily lead to poor model generalization ability and affect model accuracy. This invention proposes a rule-optimized generative adversarial network (GAN) algorithm for sample generation, thereby achieving data augmentation. The rule-optimized GAN consists of a generator (G) and a discriminator (D). The generator is responsible for generating sufficiently realistic data, while the discriminator attempts to distinguish between generated and real data. In traditional GANs, the generator and discriminator are often trained using gradient descent, which may lead to training instability or convergence to local optima in some cases. This invention improves upon the Frank-Wolfe algorithm, which is based on a linear programming optimization strategy and updates the direction of the solution by solving a linear subproblem in each iteration.

[0065] The training process for the generative adversarial network algorithm is as follows:

[0066] S201. Initialize the parameters of generator G and discriminator D. In this embodiment, the initialization method adopts a standard normal distribution, which can be expressed as:

[0067]

[0068] In the formula, θ g θ is the parameter of the generator. d These are the parameters of the discriminator; This indicates a normal distribution with a mean of 0 and a covariance matrix equal to the identity matrix.

[0069] S202, Generator G receives data from a standard normal distribution p. zThe random noise z of (z) and the actual degradation law of the battery are converted into generated data x related to battery properties, including battery voltage, temperature fluctuation, internal resistance change and charging efficiency. The generation process can be expressed as:

[0070] x = G(z, C(t); θ g );

[0071] In the formula, x represents the generated data sample, which includes the battery's multidimensional attributes (capacity, voltage, internal resistance, etc.); θ g These are the parameters of the generator; C(t) represents the battery degradation law; This represents random noise sampled from a standard normal distribution, which serves as the input to the generator and implicitly encodes battery operating conditions. For example, different dimensions of the noise vector can correspond to attributes such as the number of cycles and temperature range. The generator maps these into multidimensional battery data.

[0072] Specifically, the battery degradation law C(t) is a nonlinear mapping that incorporates the physical degradation law, and the specific calculation formula is as follows:

[0073]

[0074] Where C0 is the initial battery capacity; T(t) is the battery temperature at time t; I(t) is the battery charging current at time t; D(t) is the battery depth of discharge at time t; I max and T max These are the maximum values ​​of battery charging current and temperature, respectively; α, β, and γ are the battery charging current factor, battery temperature factor, and battery discharge depth factor, respectively, which are set according to actual conditions.

[0075] By combining generative adversarial networks with the battery degradation process, more realistic and physically consistent battery degradation data can be generated, which is of great significance for battery health management and prediction.

[0076] S203, The discriminator evaluates the generated data x and the real data x. real It provides a true / false judgment, guiding the generator to adjust the generated data to simulate the battery's health state. The discriminator's loss function L... D The calculation method can be expressed as:

[0077]

[0078] In the formula, x real Represents a real data sample; p data Represents the distribution of real data; θ d These are the parameters of the discriminator; P represents the expectation operation; data Represents the true data distribution; pz This represents a standard normal distribution.

[0079] Furthermore, D(x) can be expressed as the output of the sigmoid function, that is:

[0080]

[0081] In the formula, f(x; θ) d ) is based on input x and parameter θ d A linear combination of can be expressed as:

[0082]

[0083] In the formula, θ di This indicates that the discriminator parameters correspond to feature x. i The weighting, for example, assigning a high weight to capacity decay rate (a sensitive indicator), and assigning a dynamic weight to temperature fluctuations (adjusted according to operating conditions), b d It is a bias term.

[0084] S204. The generator's parameter updates are not based on gradient descent, but rather on finding the optimal update direction by solving a linear optimization problem.

[0085] For the generator optimization problem, in order to ensure the consistency between the physical characteristics of battery properties (such as voltage, battery temperature, etc.) and actual data, the convex constraints of the Frank-Wolfe algorithm and the performance constraint C are combined for optimization; the formula for the performance constraint C is as follows:

[0086]

[0087] Among them, v i The constraint boundary represents the battery performance, z is random noise, G(z,C(t); θ g () represents the generated data of the generator;

[0088] Define the loss function L of the generator. G The generator's loss function L G The generator's update direction is adjusted through gradient calculation as the objective function; the generator's loss function L... G The formula is as follows:

[0089]

[0090] Wherein, λ is the penalty coefficient, used to control the deviation between the generated data samples and the physical limitations of the battery; p represents the expectation operation. z This represents a standard normal distribution.

[0091] The generator's loss function comprehensively considers the quality and physical constraints of the generated samples, ensuring that the generated data not only "fools" the discriminator but also meets the physical characteristics of the battery. The loss function captures subtle changes in battery performance, improving the model's accuracy in classifying battery health status. Its optimization helps generate more realistic and practical battery data, thereby improving the reliability and effectiveness of battery health assessment.

[0092] By linearizing the objective function, the optimal update direction d is obtained. (t) :

[0093]

[0094] in, The loss function L of the generator G Relative to the generator's parameter θ g The gradient of ; d represents the update direction of the search; γ is the adjustment compensation parameter.

[0095] Based on the calculated optimal update direction d (t) The generator's parameter θ g The update rule in the (t+1)th iteration is:

[0096]

[0097] Where, α t It's the step length.

[0098] By introducing convex constraints and penalty terms, the generator can automatically comply with the boundary requirements of battery performance (such as voltage fluctuation range) during the update process, thereby achieving a more stable and effective training process. Ultimately, the generated data better meets the needs of the battery health assessment model, providing a reliable data foundation for subsequent battery health status assessment.

[0099] S205. Based on the comparison results between the generated samples and the actual samples output by the generator, adjust the parameters of the discriminator to more accurately evaluate the authenticity of the samples. The parameter update method of the discriminator can be expressed as:

[0100]

[0101] In the formula, These are the updated discriminator parameters; γ pb It is the learning rate; The discriminator loss L D Regarding θ d The gradient of γ. In this embodiment, γ pb Set to 0.01.

[0102] S206. Repeat the above steps until a preset stopping iteration condition is met, indicating that the model training is complete. In this embodiment, the preset stopping iteration condition is reaching a preset maximum number of iterations. Preferably, the preset maximum number of iterations is set to 1000.

[0103] Furthermore, S3, using the dataset, the battery health status assessment model is trained; the battery health status assessment model uses a support vector machine based on quantum state simulation as the classification algorithm;

[0104] S301. Each vector in the dataset is converted into a set of quantum bit states. For example, the "charging time" of a battery can be converted into a value in the range of 0 to 1 through a normalization function, and then this value is mapped to the corresponding quantum bit through a quantum gate. Through quantum state encoding, various characteristics of the battery (such as charging time, temperature fluctuation, battery internal resistance change, etc.) can be mapped to different quantum states.

[0105] The input i-th vector q xi Encoding the quantum state into the corresponding qubit, the computation process of the quantum state |ψ> is expressed as follows:

[0106]

[0107] in, θ represents the tensor product from the 1st to the nth qubit; qi and φ qi These are the amplitude angle and phase angle of the qubit corresponding to the i-th vector, respectively; |0> and |1> are the two ground states of the qubit;

[0108] Furthermore, the amplitude angle θ qi and phase angle φ qi Based on the input feature vector q x The calculated result can be expressed as:

[0109] θ qi =π·norm(q) xi );

[0110] φ qi =2π·norm(q) xi );

[0111] In the formula, norm(q) xi ) indicates that feature q xi A function normalized to the interval [0,1].

[0112] S302. A nonlinear kernel function is simulated using quantum gate operations, utilizing the quantum gate operation U(θ). qk ,φ qk ,λ qkThis involves implementing quantum kernel function mapping. For example, there might be a non-linear relationship between a battery's "capacity decay rate" and "discharge rate." The quantum kernel function is calculated through the inner product between quantum states. This non-linear mapping helps quantum support vector machines better identify the battery's health status and improve classification accuracy. Mapping the input state |ψ> to a new quantum state |ψ′> is represented as:

[0113] |ψ′>=U(θ qk ,φ qk ,λ qk )|ψ>;

[0114] Among them, U(θ qk ,φ qk ,λ qk It depends on the parameter θ qk φ qk and λ qk quantum gate operations, θ qk φ represents the rotation angle about the y-axis. qk λ represents the first rotation angle around the z-axis. qk This represents the second rotation angle around the z-axis, which enables effective mapping of battery data, allowing the model to capture more complex data relationships.

[0115] Furthermore, the quantum gate U(θ) qk ,φ qk ,λ qk This can be achieved by combining a revolving door and a controlled door, and can be represented as:

[0116] U(θ qk ,φ qk ,λ qk ) = R z (φ qk )R y (θ qk )R z (λ qk );

[0117] In the formula, R z (φ) and R y (θ) represents rotations about the z-axis and y-axis, respectively, and is used to implement state rotations on the Bloch sphere.

[0118] S303. Train a quantum support vector machine using the new quantum state |ψ′>.

[0119] Because changes in battery performance are typically gradual, loss functions, by capturing these subtle changes, help improve the accuracy of classifying battery health status. Using the loss function L... q Conduct training:

[0120]

[0121] Where m is the number of training samples; ξ qj It is a slack variable; y j It is the label of the j-th sample; y j y k These are the labels corresponding to the j-th and k-th new quantum states, respectively; α qj It is the Lagrange multiplier corresponding to the j-th training sample; α qk It is the Lagrange multiplier corresponding to the k-th training sample; α qj It is a Lagrange multiplier; <ψ′ j |ψ′ k > is a new quantum state |ψ′ j > and |ψ′ k The inner product of > represents the kernel function.

[0122] Furthermore, the inner product <ψ′ j |ψ′ k The calculation method for > can be expressed as:

[0123]

[0124] S304. Utilize the properties of quantum states to perform parameter optimization and update parameter α. qj The process is represented as:

[0125]

[0126] in, It is the updated Lagrange multiplier; η q It is the learning rate; The loss function L q Regarding α qj The gradient.

[0127] With Lagrange multipliers α qj The update steps are the same, and the Lagrange multipliers α are updated simultaneously. qk .

[0128] Furthermore, the loss function L q Regarding α qj The gradient can be calculated as follows:

[0129]

[0130] S305. Repeat the iteration until a preset stopping iteration condition is met, and the model training is complete. In one embodiment, the preset stopping iteration condition is reaching a preset maximum number of iterations. Preferably, the preset maximum number of iterations is set to 1000.

[0131] Quantum computing utilizes the superposition and entanglement of qubits to process information. This allows quantum models to map data more effectively in high-dimensional spaces, capturing nonlinear features and correlations between high-dimensional features that traditional computational models cannot reveal. For example, in battery health assessment, the relationships between parameters such as temperature, capacity, and internal resistance can be highly complex and nonlinear. Quantum state simulation, through quantum computing, can effectively capture these complex physical relationships and improve the expressive power of the model. Furthermore, for battery health assessment, traditional methods may require substantial computational resources to handle a large number of features and complex nonlinear relationships, while quantum computing, through the processing of quantum states, can significantly reduce computational complexity and provide accurate assessments in a shorter time.

[0132] Further, in step S4, the new sample is processed using the battery health status assessment model to complete the battery health status assessment. In this embodiment, the classification categories include three states: "healthy," "sub-healthy," and "unhealthy."

[0133] Battery health assessment involves numerous and complex parameters. Using quantum support vector machines (SVMs) enables more accurate classification in high-dimensional space, exhibiting stronger generalization ability than traditional methods, especially in various battery types and complex usage environments. The combination of quantum states and SVMs effectively addresses problems in traditional battery health assessment methods, such as insufficient data, poor model generalization, and high computational resource requirements. Quantum states provide the ability to map high-dimensional data and capture nonlinear features, while SVMs utilize quantum kernel functions to further improve classification performance and efficiency. This combination provides a more accurate and faster battery health assessment when dealing with complex battery data, significantly improving the performance of the battery management system.

[0134] Example 2

[0135] This invention also proposes a battery health status assessment device based on neural networks, such as... Figure 2 As shown, it includes:

[0136] Data acquisition module 10 acquires the test data of the target device;

[0137] The execution module 20 inputs the data to be tested into the battery health status assessment model, performs data identification on the data to be tested, and outputs the identification results;

[0138] The data processing module 30 performs fault diagnosis on the target device based on the identification results.

[0139] The battery health status assessment device provided in this embodiment of the invention has the same technical features as the battery health status assessment model construction method provided in the above embodiment, so it can also solve the same technical problems and achieve the same technical effects.

[0140] Example 3

[0141] This invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the above-described actions. Figures 1 to 2 The steps of the method shown.

[0142] Example 4

[0143] This invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described... Figures 1 to 2 The steps of the method shown. (As...) Figure 3 The diagram shows the structure of the electronic device, which includes a processor 71 and a memory 70. The memory 70 stores computer-executable instructions that can be executed by the processor 71. The processor 71 executes the computer-executable instructions to implement the above-mentioned... Figures 1 to 2 The method shown.

[0144] exist Figure 3 In the illustrated embodiment, the electronic device further includes a bus 72 and a communication interface 73, wherein the processor 71, the communication interface 73, and the memory 70 are connected via the bus 72.

[0145] The memory 70 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 73 (which can be wired or wireless), such as the Internet, wide area network, local area network, or metropolitan area network. The bus 72 may be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, or an AMBA (Advanced Microcontroller Bus Architecture) bus. AMBA defines three types of buses: APB (Advanced Peripheral Bus), AHB (Advanced High-performance Bus), and AXI (Advanced eXtensible Interface). The bus 72 can be divided into address bus, data bus, and control bus. For ease of representation, Figure 3 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0146] The processor 71 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 71 or by instructions in software form. The processor 71 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in the memory. The processor 71 reads the information in the memory and, in conjunction with its hardware, completes the aforementioned task. Figures 1 to 2 Any of the methods shown.

[0147] The present invention provides a computer program product of a battery health status assessment method based on a neural network (including a method and apparatus for constructing a battery health status assessment model, and a battery health status assessment method and apparatus). The product includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the preceding method embodiments. For specific implementation details, please refer to the method embodiments, which will not be repeated here.

[0148] Those skilled in the art will readily understand that, for the sake of convenience and brevity, the specific working process of the system described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here. Furthermore, in the description of the embodiments of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0149] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0150] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0151] Finally, it should be noted that the above embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A battery health status assessment method based on neural networks, characterized in that, include: S1. Collect data on battery attribute parameters, and store the results of the data collection as a vector; The data annotation categories include three states: healthy, sub-healthy, and unhealthy; a sample dataset is constructed by combining the vectors and the corresponding annotation results. S2. Based on the battery degradation pattern, a generative adversarial network is used to generate samples and expand the sample dataset. When the generative adversarial network is updated, the parameters of the generator are updated by combining performance constraints to ensure that the generated samples meet the physical characteristics of the battery. S3. Use the expanded sample dataset to train the constructed battery health status assessment model; S4. Collect the current attribute parameters of the battery to be evaluated, input them into the trained battery health status evaluation model, and output the category of battery health status.

2. The battery health status assessment method based on neural networks according to claim 1, characterized in that: In S2, the generator G receives data from the standard normal distribution p. z The random noise z of (z) and the actual degradation law of the battery are converted into generated data x related to battery properties; Battery degradation is a nonlinear mapping that includes physical degradation laws and is related to battery temperature, charging current, and depth of discharge.

3. The battery health status assessment method based on neural networks according to claim 1, characterized in that: In S2, when training the generative adversarial network, in order to ensure the consistency between the physical characteristics of the battery properties in the generator's produced samples and the actual data, the convex constraints and performance constraints C of the Frank-Wolfe algorithm are jointly optimized.

4. The battery health status assessment method based on neural networks according to claim 3, characterized in that: In S2, the performance constraint C includes: constraining the generator's generated data within the constraint boundaries of battery performance; Define the loss function L of the generator. G The generator's loss function L G As the objective function, the update direction of the generator is adjusted by gradient calculation to obtain the optimal update direction; Based on the calculated optimal update direction, the generator's parameter θ g The update rule in the (t+1)th iteration is related to the calculated optimal update direction, step size, and the generator parameters in the tth iteration.

5. The battery health status assessment method based on neural networks according to claim 1, characterized in that: In S3, the battery health status assessment model uses a support vector machine based on quantum state simulation as the classification algorithm; the specific steps for training the support vector machine algorithm based on quantum state simulation are as follows: S301. Each vector in the dataset is converted into a set of qubit states; based on the qubit states of each vector, the corresponding quantum state is calculated. S302. Quantum gate operations are used to simulate nonlinear kernel functions, and quantum gate operations are used to realize quantum kernel function mapping, mapping the quantum state of each vector to a new quantum state; S303. Train a quantum support vector machine using a new quantum state and use the loss function L. q Conduct training; S304, Update the loss function L q Parameters in; S305. Repeat the iteration until the preset stopping iteration condition is met, and the model training is complete.

6. The battery health status assessment method based on neural networks according to claim 5, characterized in that: In S301, the i-th input vector q xi The corresponding quantum state |ψ> is encoded into the corresponding qubit of the quantum state, and the corresponding quantum state is calculated using the qubit; In S302, quantum gate operations are used to simulate nonlinear kernel functions. Quantum gate operations are used to realize quantum kernel function mapping, mapping the quantum state |ψ> of each vector to a new quantum state |ψ′>. This includes mapping the nonlinear correlation between the battery's capacity decay rate and discharge rate, and providing high-dimensional correlation information to the quantum support vector machine.

7. The battery health status assessment method based on neural networks according to claim 6, characterized in that: In S303, the quantum state |ψ′> is used to train a quantum support vector machine to capture changes in battery performance; the loss function L is used. q The formula is: Where m is the number of training samples; ξ qj It is a slack variable; y j y k These are the labels corresponding to the j-th and k-th new quantum states, respectively; α qj It is the Lagrange multiplier corresponding to the j-th training sample; α qk It is the Lagrange multiplier corresponding to the k-th training sample; <ψ′ j |ψ′ k > is a new quantum state |ψ′ j > and |ψ′ k The inner product of > represents the kernel function.

8. The battery health status assessment method based on neural networks according to claim 6, characterized in that: In S304, the loss function L q Lagrange multipliers α qj Update the Lagrange multiplier α. qj Update with learning rate and loss function L q Regarding α qj It is related to the gradient; With Lagrange multipliers α qj Similarly, the loss function L q Lagrange multipliers α qk It will also be updated at the same time.

9. A battery health status assessment device based on neural networks, characterized in that, include: Data acquisition module 10 acquires the test data of the target device; The execution module 20 inputs the data to be tested into the battery health status assessment model, performs data recognition on the data to be tested, and outputs the recognition results; The data processing module 30 performs fault diagnosis on the target device based on the identification results.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is loaded into the processor, it implements a neural network-based battery health status assessment method according to any one of claims 1-8.

11. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements a battery health status assessment method based on any one of claims 1-8.

Images