Battery life prediction method and system based on adaptive generative adversarial network

Abstract

The invention discloses a battery life prediction method and system based on an adaptive generative adversarial network. The method comprises the steps of collecting battery use data through a sensor of a battery management system; performing manual labeling on the collected data, wherein labeling categories comprise'normal ', 'abnormal' and'performance reduction '; constructing a data set in combination with the data and the annotations; generating additional training data by adopting a generative adversarial network based on adaptive chaos optimization, and adding the training data into the data set to complete data expansion; reading the data set, and training a battery life prediction model by adopting an extreme learning machine algorithm based on a quantum topological phase to obtain a classification result; and inputting a real-time sample into the battery life prediction model for battery life prediction. According to the method, the accuracy and the reliability of a battery life prediction result can be improved, nonlinear characteristics of battery performance degradation can be more effectively captured, and higher efficiency and accuracy are provided when complex data are processed.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for predicting battery life based on adaptive generative adversarial networks. Background Technology

[0002] In the field of battery technology, particularly in electric vehicles and renewable energy storage systems, accurate prediction of battery life is crucial for maintenance, cost-benefit analysis, and resource optimization. Battery life is influenced by a variety of factors, including charge / discharge cycles, depth of charge, temperature, current, and voltage. The complex interactions of these factors make battery performance prediction a challenge.

[0003] Traditional battery life prediction technologies primarily rely on empirical models or basic machine learning methods. These methods often depend on large amounts of historical data and simplified mathematical models, making it difficult to capture the non-linear characteristics of battery performance degradation. Furthermore, they often exhibit poor generalization ability and insufficient prediction accuracy when faced with new battery technologies. In addition, the process of collecting and labeling battery data is cumbersome and costly, and the amount of data is often insufficient to train complex models, which limits the development of prediction models.

[0004] Existing technologies, such as patent application CN117067992A, acquire battery charging process data, input it into a curve prediction model based on generative adversarial networks to obtain a predicted charging voltage curve image, extract feature data from it, and then input it into a lifespan prediction model based on a deep learning model to output the battery life value. Another example is patent application CN116359773A, which extracts and filters features from the original dataset to construct a prediction model. It employs an LSTM-LSGAN combined prediction model, using an LSTM network for the generator and an MLP network for the discriminator, to solve the gradient vanishing problem of traditional GANs and achieve lithium-ion battery remaining life prediction. However, this technology uses fixed noise input and does not optimize the generation process for the time-series characteristics of battery degradation. Furthermore, the generation and prediction modules are separated in this technology, failing to achieve joint optimization.

[0005] Therefore, there is an urgent need for a method that can capture the nonlinear characteristics of battery performance degradation, has strong generalization ability, can be trained on limited data, and can balance real-time performance and accuracy to make more accurate and faster battery life prediction. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a battery life prediction method and system based on adaptive generative adversarial networks. The method enhances the generalization ability of the battery life prediction model and improves prediction accuracy through adaptive generative adversarial networks; it utilizes topological adaptive learning algorithms to solve the training instability problem; and it combines quantum information features with quantum topological phase classification methods to achieve more efficient and accurate classification.

[0007] The present invention adopts the following technical solution.

[0008] This invention proposes a battery lifetime prediction method based on adaptive generative adversarial networks, comprising:

[0009] S1. Collect historical battery usage data, including battery charge cycles, battery discharge depth, battery maximum temperature, battery minimum temperature, average voltage, and average current; manually annotate the battery life status of the historical battery usage data, with annotation categories including "normal," "abnormal," and "performance degradation"; construct a dataset by combining the data and the annotations;

[0010] S2. Generative adversarial network (GAN) is used to generate additional training data, which is then added to the dataset to complete data augmentation; the GAN is optimized using adaptive chaotic sequence parameters.

[0011] S3. Read the dataset, train the battery life prediction model using the Extreme Learning Machine algorithm, and obtain the classification results; the Extreme Learning Machine algorithm is optimized using the quantum state measurement algorithm; the quantum state measurement algorithm is combined with multi-scale analysis to construct adaptive quantum measurement configuration parameters, and calculate the weights of the output layer of the quantum state measurement algorithm;

[0012] S4. Input the real-time samples into the battery life prediction model to predict the battery life.

[0013] Furthermore, battery usage history data also includes charging duration, discharging duration, charging efficiency, and battery health status.

[0014] Furthermore, in S2, training a generative adversarial network based on adaptive chaotic optimization includes:

[0015] S201. Initialize the network weights of the generator G and discriminator D, and initialize the chaotic sequence parameters θ;

[0016] S202. In each training iteration, the generator generates a physically consistent battery degradation trajectory and adaptively updates the chaotic sequence parameter θ based on the quality of the previously generated battery degradation trajectory. (0) ;

[0017] S203. The data samples generated by the generator are sent to the discriminator. The discriminator guides the generator to synthesize data that conforms to the physical laws of the battery by dynamically evaluating the authenticity, and updates the parameters of the generator G and the discriminator.

[0018] S204. Repeat the iteration until the preset stopping iteration condition is met.

[0019] Furthermore, in S201, the formula for calculating the initial chaotic sequence parameter θ is:

[0020] θ (0) =σ b ·randn(D b );

[0021] Where, θ (0) θ represents the initial chaotic sequence parameters, generated through a normal distribution. (0) The dynamic evolution characteristics simulate the randomness of battery degradation, including abrupt temperature changes and fluctuations in charging efficiency; σ b The standard deviation of a chaotic sequence is represented by randn(D). b D represents the initial value randomly drawn from a normal distribution. b The dimension representing the chaotic sequence is aligned with the dimension of the battery features;

[0022] Furthermore, in S202, the chaotic sequence parameter θ (0) The adaptive update formula is:

[0023]

[0024] Where, θ (t) and θ (t+1) Let λ represent the chaotic sequence parameters for the t-th and t+1-th iterations, respectively; b Indicates the learning rate; Indicates the discriminator loss L D The gradient with respect to θ; D(G(θ) (t) )) indicates that the generator G is in parameter θ (t) The generated data is output by discriminator D; L real Indicates the actual data label.

[0025] Further, the dataset is input into the battery life prediction model for training, which includes:

[0026] S301. Initialize the parameters of the classifier and randomly generate the weights and biases of the hidden layers of the extreme learning machine.

[0027] S302. The quantum state measurement algorithm is used on the input feature vector, and multi-scale analysis is combined to calculate the multi-scale quantum information feature matrix of each sample; combined with the multi-scale quantum feature matrix, adaptive quantum measurement configuration parameters are constructed to calculate the enhanced feature matrix;

[0028] S303. Using the enhanced feature matrix, construct the objective function of the output layer, and solve the weights of the output layer analytically using the least squares method.

[0029] S304. After determining the output weights, the battery life prediction model is trained and the remaining battery life is predicted based on the input feature vector, including the number of battery charging cycles, voltage, and temperature.

[0030] Furthermore, in S302, multi-scale analysis is performed on different usage stages of the battery, and the calculation formula for the multi-scale quantum information feature matrix of each sample is expressed as follows:

[0031]

[0032] In the formula, qF is the multi-scale quantum feature matrix; T is the size of the time window; qX(t) is the feature vector input at time t in the time window; qP is the quantum measurement configuration parameter; Qua() is the function to perform quantum measurement. Through quantum state measurement, the complex dynamic behavior inside the battery is obtained, including the performance degradation of the battery due to temperature and current fluctuations during charging and discharging.

[0033] Furthermore, in S302, an adaptive quantum measurement configuration parameter qP is constructed, and qP is automatically adjusted through adaptive mechanisms including but not limited to gradient descent methods or genetic algorithms. The implementation of Qua() is then expressed as follows:

[0034] Qua(qX,qP)=qX·(qP+ΔqP) T ·log(σ q ((qP+ΔqP)·qX+qB));

[0035] In the formula, qP T σ represents the transpose of the quantum measurement matrix; ΔqP represents the adjustment amount calculated by the adaptive mechanism, indicating the dynamic adjustment of the quantum measurement configuration parameters, which is updated based on the battery's historical health status information and the currently input feature vector qX; q qB is the nonlinear activation function used in quantum measurement; in this embodiment, it is the Sigmoid function.

[0036] The input feature vector qX is combined with the multi-scale quantum feature matrix qF to obtain the enhanced feature matrix.

[0037] This invention also proposes a battery life prediction system based on an adaptive generative adversarial network, which runs the battery life prediction method based on an adaptive generative adversarial network as described in any one of the above descriptions, including:

[0038] The data acquisition module acquires the test data from the target device;

[0039] The execution module inputs the test data into the battery life prediction model, performs data identification on the test data through the battery life prediction model, and outputs the identification results.

[0040] The data processing module performs fault diagnosis on the target device based on the identification results.

[0041] The present invention also proposes a terminal, including a processor and a storage medium:

[0042] The storage medium is used to store instructions;

[0043] The processor is configured to operate according to the instructions to perform the steps of any of the methods described above.

[0044] The present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described above.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0046] 1. The adaptive generative adversarial network used in this invention has a powerful sample generation capability, which can generate samples that are close to the statistical characteristics of real data. This feature provides a rich and diverse dataset for model training, which can enhance the generalization ability of the model and thus improve the accuracy and reliability of prediction.

[0047] 2. This invention utilizes a topology-adaptive learning algorithm, enabling the neural network to adjust its structure according to actual needs, thus avoiding the training instability problem common in traditional neural networks. This not only improves training efficiency but also provides a more reliable guarantee for the application of neural networks in complex tasks.

[0048] 3. By introducing adaptive quantum measurement configuration, quantum feature extraction across multiple time scales, and fusing quantum states with traditional physical models, this invention significantly enhances the intelligence, accuracy, and stability of battery life prediction models. This improvement not only enables fine-grained analysis of battery degradation processes but also addresses complex battery usage scenarios, providing battery management systems with more accurate prediction and optimization capabilities, thereby extending battery lifespan. Attached Figure Description

[0049] Figure 1This is a flowchart of a battery life prediction method based on an adaptive generative adversarial network according to the present invention.

[0050] Figure 2 This is a schematic diagram of the structure of an electronic device for predicting battery life based on an adaptive generative adversarial network according to the present invention. Detailed Implementation

[0051] 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.

[0052] Example 1

[0053] This invention proposes a battery lifetime prediction method based on adaptive generative adversarial networks, such as... Figure 1 As shown, including,

[0054] S1. Collect battery usage data through the sensors of the battery management system; manually label the collected data, with label categories including "normal", "abnormal" and "performance degradation"; construct a dataset by combining the data and the labels;

[0055] S2. Generative adversarial network based on adaptive chaos optimization is used to generate additional training data, which is then added to the dataset to complete data expansion;

[0056] S3. Read the dataset, train the battery life prediction model using the quantum topological phase-based extreme learning machine algorithm, and obtain the classification results; combine the quantum state measurement algorithm with multi-scale analysis to construct adaptive quantum measurement configuration parameters, and calculate the weights of the output layer of the quantum state measurement algorithm;

[0057] S4. Input the real-time samples into the battery life prediction model to predict the battery life.

[0058] In step S1, the data involved in this invention is acquired by the battery management system, specifically through real-time monitoring by sensors and transmitted to the storage system through the data acquisition module. The data is stored in a structured data format, specifically CSV format.

[0059] In one embodiment, the attributes of the data include:

[0060] a1: Number of battery charging cycles, integer type;

[0061] a2: Battery depth of discharge, percentage;

[0062] a3: Maximum battery temperature, in degrees Celsius;

[0063] a4: Minimum battery temperature, in degrees Celsius;

[0064] a5: Average voltage, measured in volts;

[0065] a6: Average current, in amperes;

[0066] a7: Charging duration, in hours;

[0067] a8: Discharge duration, in hours;

[0068] a9: Charging efficiency, percentage;

[0069] a 10 Battery health status, percentage.

[0070] 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.

[0071] Furthermore, the collected data is labeled. The labeling method of this invention is manual labeling. In one embodiment, the labeling categories include three types: "normal", "abnormal", and "performance degradation".

[0072] It is understandable that in the task of this invention, the acquisition, annotation, 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 employs a generative adversarial network (GAN) based on adaptive chaotic optimization for sample generation, thereby achieving data augmentation. The GAN based on adaptive chaotic optimization comprises two components: a generator (G) and a discriminator (D). The generator aims to produce new data that is almost indistinguishable from real data, while the discriminator attempts to distinguish the generated data from the real data. This invention uses a chaotic sequence in the input layer of the generator, dynamically adjusting it based on feedback during the generation process to enhance the diversity and coverage of the generated samples.

[0073] Specifically, the training process of the generative adversarial network based on adaptive chaotic optimization is as follows:

[0074] S201. Initialize the network weights of the generator G and discriminator D, and initialize the chaotic sequence parameters θ to provide randomness to the generator for the first batch of inputs, expressed as:

[0075] θ (0) =σ b ·randn(D b );

[0076] Where, θ (0) The initial chaotic sequence parameters are represented by σ. Because battery life data (such as charge cycles, discharge depth, and temperature fluctuations) is highly nonlinear and dynamically time-series, the random noise input of traditional generative adversarial networks (GANs) is insufficient to capture complex patterns. Therefore, the initial parameters of the chaotic sequence are generated using a normal distribution, and their dynamic evolution characteristics can simulate the randomness in the battery degradation process (such as sudden temperature changes and charging efficiency fluctuations). b The standard deviation of a chaotic sequence is represented by randn(D). b D represents the initial value randomly drawn from a normal distribution. b The dimension representing the chaotic sequence is aligned with the dimension of the battery features.

[0077] S202. In each training iteration, the generator can generate physically consistent battery degradation trajectories (such as high discharge depth accompanied by higher temperature rise), improving the prediction model's generalization ability to unseen operating conditions. The generator adaptively updates the chaotic sequence parameter θ based on the quality of the previously generated battery degradation trajectory. (0) , represented as:

[0078]

[0079] In the formula, θ (t) and θ (t+1) Let λ represent the chaotic sequence parameters for the t-th and t+1-th iterations, respectively; b Indicates the learning rate; Indicates the discriminator loss L D The gradient with respect to θ; D(G(θ) (t) )) indicates that the generator G is in parameter θ (t) The generated data is output by discriminator D; L real This represents the actual data label. Preferably, λ b Set to 0.01.

[0080] Furthermore, the gradient term This includes adversarial loss and structural risk minimization loss. The structural risk minimization loss constrains model complexity and prevents overfitting to noisy data (such as single abnormal charging current). The calculation method can be expressed as:

[0081]

[0082] In the formula, x i Represents a real data sample, y j Represents the sample label, β b The weights represent the loss minimizing structural risk, where N is the number of samples. Preferably, β b Set it to 0.3.

[0083] S203. The data samples generated by the generator are sent to the discriminator. The discriminator guides the generator to synthesize data that conforms to the physical laws of the battery (such as the negative correlation between charging duration and current) by dynamically evaluating the authenticity, and updates the parameters of the generator G and the discriminator.

[0084] The update method for generator G can be represented as:

[0085]

[0086] In the formula, and Let α represent the parameters of the generator at step t and step t+1, respectively; b This represents the learning rate of the generator; Indicates gradient operation; Let α represent the loss function of the generator. Preferably, α b Set to 0.05.

[0087] Furthermore, The calculation method can be expressed as:

[0088]

[0089] In the formula, This represents the gradient of the generated data with respect to the generator parameters after passing through the discriminator. Furthermore, the discriminator evaluates the difference between these samples and the real samples, calculates the loss, and feeds the result back to the generator, guiding it to further optimize its data generation strategy. The update method of the discriminator D can be represented as:

[0090]

[0091] In the formula, and These represent the parameters of the discriminator at step t and step t+1, respectively; γ b x represents the learning rate of the discriminator. real Represents real data; This represents gradient operation.

[0092] Furthermore, The calculation method can be expressed as:

[0093]

[0094] and, The calculation method can be expressed as:

[0095]

[0096] S204. Repeat the iteration until a preset stopping iteration condition is met, indicating that 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.

[0097] By using a generative adversarial network based on adaptive chaotic optimization to generate additional training data to expand the dataset, the problems of tedious battery data labeling and insufficient data volume are solved, providing sufficient and diverse data for model training. The generated samples have statistical characteristics close to those of real data, which can enhance the model's generalization ability, allowing the model to learn more data features and patterns, and improve its adaptability to different operating conditions. The generator generates data that conforms to the physical laws of batteries, enabling the model to more accurately capture the characteristics of battery performance degradation, thereby improving the accuracy of battery life prediction.

[0098] Furthermore, the dataset is input into the battery life prediction model for training. The battery life prediction model employs an extreme learning machine classification algorithm based on quantum topological phases, and uses novel quantum information features through the measurement of quantum states, including the following steps:

[0099] S301. Initialize the classifier parameters, and randomly generate the weights and biases of the hidden layers of the extreme learning machine; this can be expressed as:

[0100] qW=σ q ·randn(qN,qD);

[0101] qb=σ q rand(qN);

[0102] In the formula, qW is the hidden layer weight matrix; qN is the number of hidden layer nodes; qD is the dimension of the input feature; σ q q is the initial standard deviation; randn(qN,qD) generates normally distributed random numbers of shape qN×qD; rand(qN) generates uniformly distributed random numbers of shape qN; qb is the hidden layer bias vector.

[0103] S302. A quantum state measurement algorithm is used on the input feature vector to calculate the quantum information feature matrix for each sample. To improve adaptability to different battery usage stages (such as the initial charging stage, the final charging stage, and the cycling process), multi-scale analysis is introduced, extracting quantum information through a sliding time window. The multi-scale quantum information feature matrix is ​​represented as follows:

[0104]

[0105] In the formula, qF is the multi-scale quantum feature matrix. Multi-scale quantum information features extracted through quantum measurement algorithms (such as "maximum battery temperature" and "battery health state") can provide the model with more fine-grained information; T is the size of the time window; qX(t) is the feature vector input at time t within the time window; qP is the quantum measurement configuration parameter; Qua() is the function to perform quantum measurement. Through quantum state measurement, the complex dynamic behavior inside the battery can be revealed, especially the performance degradation caused by temperature and current fluctuations during charging and discharging. In this way, not only can the quantum features of the current state be captured, but also the quantum information of historical states can be integrated to obtain a multi-scale quantum feature matrix.

[0106] Specifically, the adaptive quantum measurement configuration parameter qP is constructed and automatically adjusted through adaptive mechanisms, including but not limited to gradient descent methods or genetic algorithms. The implementation of Qua() is then expressed as follows:

[0107] Qua(qX,qP)=qX·(qP+ΔqP) T ·log(σ q ((qP+ΔqP)·qX+qB));

[0108] In the formula, qP T σ represents the transpose of the quantum measurement matrix; ΔqP represents the adjustment amount calculated by the adaptive mechanism, indicating the dynamic adjustment of the quantum measurement configuration parameters, which is updated based on the battery's historical health status information and the currently input feature vector qX; q qB is the nonlinear activation function used in quantum measurement; in this embodiment, it is the Sigmoid function. qB is the bias term used in quantum measurement.

[0109] The input feature vector qX is combined with the multi-scale quantum information feature matrix qF to obtain the enhanced feature matrix, which is used to improve classification accuracy. This can be represented as:

[0110] qX enhanced =[qX qF];

[0111] In the formula, qX enhanced The enhanced feature matrix is ​​formed by horizontally splicing the original feature vector qX and the multi-scale quantum feature matrix qF.

[0112] Quantum computing possesses stronger pattern recognition capabilities in high-dimensional spaces, enabling it to capture nonlinear relationships that are difficult for traditional models to handle. Through quantum state mapping and feature extraction, models can discover complex nonlinear characteristics hidden in the battery performance degradation process. By combining battery behavior at different time scales, quantum models can capture the nonlinear characteristics of the degradation process at different scales. For example, a battery may exhibit a gradually accelerating degradation trend over a long period, while experiencing drastic nonlinear fluctuations in the short term; quantum models can flexibly handle these complex dynamics. Adaptive quantum measurement can adjust the measurement strategy in real time according to the actual operating state of the battery, thereby more accurately capturing the nonlinear changes in the battery degradation process. By dynamically optimizing the measurement method, the model can more effectively and accurately identify the state of the battery at different degradation stages. This improvement not only enables fine-grained analysis of the battery degradation process but also addresses complex battery usage scenarios, providing battery management systems with more accurate prediction and optimization capabilities, thereby improving battery lifespan.

[0113] S303. Using the enhanced feature matrix, construct the objective function of the output layer, and analytically solve for the weights of the output layer using the least squares method, which can be expressed as:

[0114] qH=tanh(qW·qX enhanced +qb);

[0115] qβ=(qH T ·qH+qI / γ q ) -1 ·qH T ·qY;

[0116] In the formula, qH is the hidden layer output matrix; tanh() is the Tanh activation function; qβ is the output layer weight; qY is the label vector; qI is the identity matrix; γ q To prevent overfitting, the regularization parameter can be increased appropriately for batteries with poor health conditions. This reduces the model's sensitivity to outlier data points, thereby improving the model's generalization ability and avoiding performance degradation caused by imbalanced historical data. In this embodiment, γ... q Set it to 3.

[0117] Furthermore, the inverse matrix is ​​calculated (qH) T ·qH+qI / γ q ) -1 The implementation can be represented as:

[0118] (qH T ·qH+qI / γ q ) -1 =qI-qH T ·(qH·qHT +γ q ·qI) -1 ·qH;

[0119] In the formula, qI is the identity matrix, and γ q It is the regularization coefficient.

[0120] S304. After determining the output weights, the battery life prediction model is trained and can predict the remaining life of the battery based on the input features (such as the number of times the battery is charged, voltage, temperature, etc.).

[0121] Training a battery lifetime prediction model using an extreme learning machine algorithm based on quantum topological phases yields numerous significant benefits. Quantum feature extraction, previously unreported in the field of battery prediction, is a crucial innovation. Quantum information features extracted through quantum state measurement algorithms, such as the battery's maximum temperature and health state, provide the model with fine-grained information of a new dimension—something difficult to obtain using traditional methods. These unique quantum features are integrated into the enhanced feature matrix, greatly enriching the model's input and enabling it to capture more complex features during battery performance degradation. This significantly improves the accuracy and reliability of battery lifetime prediction, opening a new direction in the development of battery prediction technology.

[0122] Finally, in step S4, the real-time samples are input into the battery life prediction model to predict battery life. In one embodiment, the input data of the real-time samples is input into a classifier model for classification. In this embodiment, the classification categories include three types: "normal," "abnormal," and "performance degradation."

[0123] Example 2

[0124] This invention also proposes a battery life prediction system based on an adaptive generative adversarial network, comprising:

[0125] The data acquisition module acquires the test data from the target device;

[0126] The execution module inputs the test data into the battery life prediction model, performs data identification on the test data through the battery life prediction model, and outputs the identification results.

[0127] The data processing module performs fault diagnosis on the target device based on the identification results;

[0128] The battery life prediction system provided in this embodiment of the invention has the same technical features as the battery life prediction model construction method provided in the above embodiment, so it can also solve the same technical problems and achieve the same technical effects.

[0129] Example 3

[0130] 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... Figure 1 The steps of the method shown.

[0131] Example 4

[0132] This invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the above-described actions. Figure 1 The steps of the method shown.

[0133] This invention also provides a schematic diagram of the structure of an electronic device, such as... Figure 2 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... Figure 1 The method shown.

[0134] exist Figure 2 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.

[0135] 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 2 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.

[0136] 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. Figure 1 The method shown.

[0137] The present invention provides a computer program product for a battery life prediction method based on an adaptive generative adversarial network (including a method and system for constructing a battery life prediction model, and a battery life prediction method and system). 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.

[0138] 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.

[0139] 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.

[0140] 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.

[0141] 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 life prediction method based on adaptive generative adversarial networks, characterized in that, include, S1. Collect historical battery usage data, including battery charge cycles, battery discharge depth, battery maximum temperature, battery minimum temperature, average voltage, and average current; manually annotate the battery life status of the historical battery usage data, with annotation categories including "normal," "abnormal," and "performance degradation"; construct a dataset by combining the data and the annotations; S2. Generative adversarial networks are used to generate additional training data, which is then added to the dataset to complete data augmentation. The generative adversarial network is optimized using adaptive chaotic sequence parameters; S3. Read the dataset, train the battery life prediction model using the Extreme Learning Machine algorithm, and obtain the classification results; the Extreme Learning Machine algorithm is optimized using the quantum state measurement algorithm; the quantum state measurement algorithm is combined with multi-scale analysis to construct adaptive quantum measurement configuration parameters, and calculate the weights of the output layer of the quantum state measurement algorithm; S4. Input the real-time samples into the battery life prediction model to predict the battery life.

2. The battery life prediction method based on adaptive generative adversarial networks according to claim 1, characterized in that: Battery usage history data also includes charging duration, discharging duration, charging efficiency, and battery health status.

3. The battery life prediction method based on adaptive generative adversarial networks according to claim 1, characterized in that: In S2, training a generative adversarial network based on adaptive chaotic optimization includes: S201. Initialize the network weights of the generator G and discriminator D, and initialize the chaotic sequence parameters θ; S202. In each training iteration, the generator generates a physically consistent battery degradation trajectory and adaptively updates the chaotic sequence parameter θ based on the quality of the previously generated battery degradation trajectory. (0) ; S203. The data samples generated by the generator are sent to the discriminator. The discriminator guides the generator to synthesize data that conforms to the physical laws of the battery by dynamically evaluating the authenticity, and updates the parameters of the generator G and the discriminator. S204. Repeat the iteration until the preset stopping iteration condition is met.

4. The battery life prediction method based on adaptive generative adversarial networks according to claim 3, characterized in that: In S201, the formula for initializing the chaotic sequence parameter θ is: i (0) =s b ·randn(D b ); Where, θ (0) θ represents the initial chaotic sequence parameters, generated through a normal distribution. (0) The dynamic evolution characteristics simulate the randomness of battery degradation, including abrupt temperature changes and fluctuations in charging efficiency; σ b The standard deviation of a chaotic sequence is represented by randn(D). b D represents the initial value randomly drawn from a normal distribution. b The dimension representing the chaotic sequence is aligned with the dimension of the battery features.

5. The battery life prediction method based on adaptive generative adversarial networks according to claim 3, characterized in that: In S202, the chaotic sequence parameter θ (0) The adaptive update formula is: Where, θ (t) and θ (t+1) Let λ represent the chaotic sequence parameters for the t-th and t+1-th iterations, respectively; b Indicates the learning rate; Indicates the discriminator loss L D The gradient with respect to θ; D(G(θ) (t) )) indicates that the generator G is in parameter θ (t) The generated data is output by discriminator D; L real Indicates the actual data label.

6. The battery life prediction method based on adaptive generative adversarial networks according to claim 1, characterized in that: The process of inputting the dataset into the battery life prediction model for training includes: S301. Initialize the parameters of the classifier and randomly generate the weights and biases of the hidden layers of the extreme learning machine. S302. The quantum state measurement algorithm is used on the input feature vector, and multi-scale analysis is combined to calculate the multi-scale quantum information feature matrix of each sample; combined with the multi-scale quantum feature matrix, adaptive quantum measurement configuration parameters are constructed to calculate the enhanced feature matrix; S303. Using the enhanced feature matrix, construct the objective function of the output layer, and solve the weights of the output layer analytically using the least squares method. S304. After determining the output weights, the battery life prediction model is trained and the remaining battery life is predicted based on the input feature vector, including the number of battery charging cycles, voltage, and temperature.

7. The battery life prediction method based on adaptive generative adversarial networks according to claim 6, characterized in that: In S302, multi-scale analysis is performed on the battery at different usage stages, and the calculation formula for the multi-scale quantum information feature matrix of each sample is expressed as follows: In the formula, qF is the multi-scale quantum feature matrix; T is the size of the time window; qX(t) is the feature vector input at time t in the time window; qP is the quantum measurement configuration parameter; Qua() is the function to perform quantum measurement. Through quantum state measurement, the complex dynamic behavior inside the battery is obtained, including the performance degradation of the battery due to temperature and current fluctuations during charging and discharging.

8. The battery life prediction method based on adaptive generative adversarial networks according to claim 7, characterized in that: In S302, an adaptive quantum measurement configuration parameter qP is constructed, and qP is automatically adjusted through adaptive mechanisms including but not limited to gradient descent methods or genetic algorithms. The implementation of Qua() is then expressed as follows: Qua(qX,qP)=qX·(qP+ΔqP) T ·log(σ q ((qP+ΔqP)·qX+qB)); In the formula, qP T σ represents the transpose of the quantum measurement matrix; ΔqP represents the adjustment amount calculated by the adaptive mechanism, indicating the dynamic adjustment of the quantum measurement configuration parameters, which is updated based on the battery's historical health status information and the currently input feature vector qX; q qB is the nonlinear activation function used in quantum measurement; in this embodiment, it is the Sigmoid function. The input feature vector qX is combined with the multi-scale quantum feature matrix qF to obtain the enhanced feature matrix.

9. A battery life prediction system based on adaptive generative adversarial networks, comprising a data acquisition module, an execution module, and a data processing module, characterized in that, include: The data acquisition module acquires the test data from the target device; The execution module inputs the test data into the battery life prediction model, performs data identification on the test data through the battery life prediction model, and outputs the identification results. The data processing module performs fault diagnosis on the target device based on the identification results.

10. A terminal, comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1-8.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1-8.

Images