Data driving fusion quantum optimization battery health state estimation method and system

Abstract

The invention provides a battery health state estimation method and system based on data-driven fusion quantum optimization, and the method comprises the steps: constructing a battery health state estimation and prediction model, introducing a dynamic decline strategy of a contraction-expansion factor into a quantum particle swarm optimization algorithm, training the model through a training set, and carrying out the prediction of the battery health state. The improved quantum particle swarm optimization algorithm is utilized to optimize the battery health state estimation and prediction model in the training process; verifying by using the test set; and acquiring a target battery data set, selecting constant-current charging time, constant-voltage charging time and internal resistance in the battery charging process as input characteristics, performing preprocessing, and then estimating and predicting the health state of the target battery by using the battery health state estimation and prediction model passing verification to obtain an estimation and prediction result. According to the method, the current SOH estimation and the future SOH prediction of the battery can be efficiently and accurately carried out at the same time.

Description

Technical Field

[0001] The present invention belongs to the field of battery state of health (SOH) estimation, and particularly relates to a method and system for estimating the battery state of health by data-driven fusion quantum optimization. Background Art

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

[0003] The state of health (SOH) of a battery, as a key indicator to measure the battery health level, can be represented by the ratio of the current capacity of the battery to the initial capacity. During the use of the battery, the SOH will continuously decrease over time and with the increase of the number of cycles. For the currently widely used lithium-ion batteries, usually when the SOH reaches 0.8, it is considered to reach the end of life (EOL). Therefore, the accurate estimation and prediction of SOH are crucial for the battery management system (BMS).

[0004] Currently, the methods for estimating and predicting SOH are mainly divided into two categories: model-based methods and data-driven methods. Model-based methods need to consider complex electrochemical processes, internal resistance growth, capacity degradation and other factors to establish a physical model to describe the battery state of health. This not only requires an in-depth understanding of the underlying physical and chemical processes of the battery to obtain accurate modeling parameters, but also faces huge computational amounts. While data-driven methods do not need to consider the complex electrochemical reaction mechanism inside the battery, and only need to establish the mapping relationship between the characteristic data of the battery (such as current, voltage, temperature, etc.) and SOH. With the development of artificial intelligence technology, data-driven methods are gradually becoming the mainstream of SOH estimation and prediction methods.

[0005] However, the currently data-driven methods still have obvious limitations. On the one hand, they usually can only achieve the estimation of the current SOH or the prediction of the future SOH, and rarely can achieve both functions in one model. On the other hand, although many studies incorporate optimization algorithms into the data-driven model, the model optimization process and the training process are often separated and cannot be carried out simultaneously. And with the increase of the training data volume, the optimal structure of the data-driven model is not immutable. Especially in the context of the gradual popularization of battery cloud data applications, it is difficult to use cloud data to optimize the structure of the data-driven model in real time, and thus it is impossible to maximize the computational accuracy and efficiency of the model.

[0006] In summary, the prior art has the problems that it is impossible to efficiently and accurately estimate the current SOH and predict the future SOH of the battery at the same time, and it is impossible to optimize the model structure in real time during the training process of the model to improve the computational accuracy and efficiency. Summary of the Invention

[0007] To solve the above problems, the present invention proposes a method and system for estimating the state of health of a battery by data-driven fusion of quantum optimization. The present invention can solve the problems in the prior art that it is impossible to simultaneously estimate the current SOH of the battery and predict the future SOH efficiently and accurately, and it is impossible to optimize the model structure in real time during the training process of the model to improve the calculation accuracy and efficiency.

[0008] According to some embodiments, the present invention adopts the following technical solutions:

[0009] A method for estimating the state of health of a battery by data-driven fusion of quantum optimization, comprising the following steps:

[0010] Obtain a battery data set, select the constant current charging time, constant voltage charging time and internal resistance during the battery charging process as input features, and perform preprocessing to divide the training set and the test set;

[0011] Construct a battery state of health estimation and prediction model, introduce a dynamic decreasing strategy of a contraction-expansion factor in the quantum particle swarm optimization algorithm, use the training set to train the model, and optimize the battery state of health estimation and prediction model by using the improved quantum particle swarm optimization algorithm during the training process;

[0012] Use the test set to verify the trained battery state of health estimation and prediction model;

[0013] Obtain a target battery data set, select the constant current charging time, constant voltage charging time and internal resistance during the battery charging process as input features, and perform preprocessing, and then use the verified battery state of health estimation and prediction model to estimate and predict the target battery state of health to obtain the estimation and prediction results.

[0014] As an alternative implementation, the process of preprocessing includes using a Hampel filter to process the outliers in the data of each input feature and normalizing the input feature data.

[0015] As an alternative implementation, the battery state of health estimation and prediction model includes an estimation sub-framework and a prediction sub-framework. The estimation sub-framework includes a convolutional neural network, a fully connected layer, a ReLU activation function, a fully connected layer, a ReLU activation function and a fully connected layer connected in sequence;

[0016] The prediction sub-framework includes a convolutional neural network, a gated recurrent unit, a fully connected layer, a ReLU activation function and a fully connected layer connected in sequence;

[0017] The convolutional neural network is used to extract deep hidden features for battery state of health estimation and prediction from the input features.

[0018] As a further implementation, the convolutional neural network uses convolutional kernels in the convolutional layer to extract hidden features in the input layer data layer by layer, and then preprocesses the extracted features through the pooling layer to reduce the size of the feature matrix. After another operation of the convolutional layer - pooling layer, it further extracts deep hidden features and reduces the size of the feature matrix.

[0019] As an alternative implementation, the process of introducing a dynamic decreasing strategy of the contraction - expansion factor in the quantum particle swarm optimization algorithm includes: at the initial stage of the iteration of the quantum particle swarm optimization algorithm, making the value of the contraction - expansion factor greater than the set value to accelerate the search speed and quickly approach the global extreme value; at the later stage of the iteration, making its value less than the set value to enhance the local search ability.

[0020] As a further implementation, the value of the contraction - expansion factor is updated dynamically, and the update process is as follows:

[0021]

[0022] where α initial represents the initial value of the contraction - expansion factor α, t max represents the maximum number of iterations, t represents the current number of iterations, and n is a preset value.

[0023] As an alternative implementation, the process of optimizing the battery health state estimation and prediction model using the improved quantum particle swarm optimization algorithm during the training process includes using the improved quantum particle swarm optimization algorithm during the training process to optimize the number of convolutional kernels in the convolutional neural network of the battery health state estimation and prediction model, the number of hidden layers of the gated recurrent unit, and the input batch size during the training process.

[0024] Furthermore, the input weights, output weights, and bias values of the model are optimized using the Adam algorithm.

[0025] As an alternative implementation, during the process of validating the trained battery health state estimation and prediction model using the test set, the method of shuffling the data order of the test set and randomly taking values within the range of changes in the battery health state is adopted; the mean square error, root mean square error, mean absolute error, and coefficient of determination are used as error evaluation indicators.

[0026] A data - driven fusion quantum - optimized battery health state estimation system includes:

[0027] A data processing module, configured to obtain a battery data set, select the constant - current charging time, constant - voltage charging time, and internal resistance during the battery charging process as input features, and perform preprocessing to divide the training set and the test set;

[0028] The model construction and optimization module is configured to construct a battery health state estimation and prediction model, introduce a dynamic decreasing strategy of a contraction-expansion factor into the quantum particle swarm optimization algorithm, train the model using a training set, and optimize the battery health state estimation and prediction model using the improved quantum particle swarm optimization algorithm during the training process;

[0029] The testing module is configured to verify the trained battery health state estimation and prediction model using a test set;

[0030] The estimation and prediction module is configured to obtain a target battery dataset, select the constant current charging time, constant voltage charging time, and internal resistance during the battery charging process as input features, perform preprocessing, and estimate and predict the target battery health state using the verified battery health state estimation and prediction model to obtain estimation and prediction results.

[0031] An electronic device includes a memory, a processor, and computer instructions stored on the memory and running on the processor. When the computer instructions are run by the processor, the steps in the above method are completed.

[0032] Compared with the prior art, the beneficial effects of the present invention are:

[0033] The present invention combines a convolutional neural network (CNN) with a gated recurrent unit (GRU). The constructed framework can accurately estimate the current health state of the battery and effectively predict the development trend of the future health state at the same time, providing comprehensive technical support for the management and maintenance of the battery.

[0034] The present invention uses an improved QPSO algorithm to optimize the key structural parameters of the neural network model, introduces a dynamic decreasing strategy of a contraction-expansion factor, avoids particles falling into local optimal values or crossing the optimal value, and significantly improves the SOH estimation and prediction accuracy.

[0035] The present invention seamlessly integrates the training process with the optimization process of the neural network structural parameters using the improved QPSO algorithm, enabling the neural network to dynamically adjust the structural parameters according to the improved QPSO algorithm during training, thereby accelerating convergence to the global optimal solution and improving the modeling efficiency and training efficiency.

[0036] The present invention can realize the immediate utilization of a large amount of high-quality data on the cloud platform, optimize the neural network structure within the framework in real-time and dynamically, and has great application potential in future BMSs closely integrated with the cloud, opening up a new way for the performance improvement and intelligent management of BMSs.

[0037] The present invention verifies through multiple battery datasets, indicating that the framework has good generalization and robustness in both the estimation and prediction of SOH.

[0038] To make the above objects, features, and advantages of the present invention more apparent and understandable, the following provides preferred embodiments in conjunction with the accompanying drawings and detailed descriptions are as follows. Description of the Drawings

[0039] The accompanying drawings forming a part of this specification are used to provide a further understanding of the present invention. The schematic embodiments and descriptions thereof of the present invention are used to explain the present invention and do not constitute an improper limitation of the present invention.

[0040] Figure 1 Schematic diagram of input features used in one embodiment;

[0041] Figure 2 Schematic diagram of the overall structure of the model framework in one embodiment;

[0042] Figure 3 Schematic diagram of the CNN structure used in one embodiment;

[0043] Figure 4 Schematic diagram of the GRU structure used in one embodiment;

[0044] Figure 5 Schematic diagram of the change of the loss function during the training process of the model framework in one embodiment;

[0045] Figure 6 Schematic diagram of the comparison between the SOH estimated values and the experimental values and the error schematic diagram of the CS_35, CS2_36, CS2_37, and CS2_38 batteries;

[0046] Figure 7 Schematic diagram of the comparison between the SOH predicted values and the experimental values and the error schematic diagram of the CS2_35, CS2_36, CS2_37, and CS2_38 batteries;

[0047] Figure 8 Schematic diagram of an electronic device in one embodiment. Detailed Embodiments

[0048] The present invention will be further described below in conjunction with the drawings and embodiments.

[0049] It should be noted that the following detailed descriptions are all illustrative and are intended to provide further explanations of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0050] Note that the terms used herein are for the purpose of describing particular embodiments only and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly dictates otherwise, the singular forms are also intended to include the plural forms. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they specify the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0051] In the case of no conflict, the embodiments in the present application and the features in the embodiments may be combined with each other.

[0052] Embodiment 1

[0053] A method for estimating the state of health of a battery by data-driven fusion quantum optimization includes the following steps:

[0054] Step 1, data preprocessing: Obtain a battery data set, select the constant current charging time (CCCT), constant voltage charging time (CVCT), and internal resistance (Resistance) during the battery charging process as input features, use a Hampel filter to process the outliers in the data set, and normalize the data.

[0055] Specifically, it includes:

[0056] (a) Obtain the data set: In this embodiment, the CS2 series battery data set from the University of Maryland (CALCE) is used. Randomly select the 35th battery data set for the training and accuracy verification of the SOH estimation and prediction framework, and the 36th, 37th, and 38th battery data sets for verifying the generalization ability of the framework. The cathodes of these batteries are all lithium cobalt oxide (LiCoO 2 ), and the initial capacity is 1100 mAh.

[0057] (b) Feature selection: Select the constant current charging time (CCCT), constant voltage charging time (CVCT), and internal resistance (Resistance) during the battery charging process as input features, as shown in Figure 1 (a) and (b) respectively.

[0058] (c) Outlier processing: Use a Hampel filter to process the outliers in each input feature. The Hampel filter is based on the median (m) and the median absolute deviation (MAD). By calculating the median and MAD, it is determined whether a data point is an outlier. If it is an outlier, it is replaced with the median of the data set. The median and MAD are calculated according to formulas (1) and (2):

[0059]

[0060] MAD = m(|X i-m(X)|)(2)

[0061] where X i represents each data point in the dataset, m represents the median, and MAD is the median of the absolute differences between each data point X i and the median of the dataset X, and N is the number of data points in the dataset.

[0062] If the data point X i is identified as an outlier (i.e., its absolute deviation divided by MAD exceeds the threshold), it is replaced with the median of the dataset X; otherwise, the original value of the data point is retained, as shown in formula (3):

[0063]

[0064] where Threshold represents the set threshold.

[0065] (d) Normalization: The selected input feature data is normalized using the normalization formula to improve the training efficiency and accuracy of the SOH estimation and prediction framework. The normalization processing formula is as shown in formula (4).

[0066]

[0067] where x * represents the data value after filtering, x * max represents the maximum value of the data after filtering, x * min represents the minimum value of the data after filtering, and x represents the data value after normalization.

[0068] Step 2: Construct the SOH estimation and prediction model framework: Combine the convolutional neural network (CNN) with the gated recurrent unit (GRU), and use the improved quantum particle swarm optimization (QPSO) algorithm to optimize the neural network structure inside the framework. The improved QPSO algorithm introduces a dynamic decreasing strategy for the contraction-expansion factor, which can simultaneously optimize the number of convolutional kernels (CNN_Kernal) of the CNN, the number of hidden layers (Hidden_Layer) of the GRU layer, and the input batch size (Batch_Size) during the training process, so as to determine the optimal neural network structure.

[0069] The overall structure of the SOH estimation and prediction model framework constructed by the present invention is as Figure 2 shown, and specifically includes the following process:

[0070] (a) Invoke CNN: It includes an input layer, two convolutional layers, and two pooling layers, and the structure is as Figure 3As shown in the figure. The basic principle of its operation is as follows: First, the convolutional kernels in the convolutional layer are used to extract the hidden features in the input layer data layer by layer. Then, the features extracted are preprocessed by the pooling layer to reduce the size of the feature matrix. After that, another convolutional layer-pooling layer operation is performed to further extract the deep hidden features and reduce the size of the feature matrix. Finally, the feature matrix is output for the simultaneous estimation and prediction of the SOH.

[0071] (b) Build the SOH estimation sub-framework and the SOH prediction sub-framework:

[0072] SOH estimation sub-framework: CNN - fully connected layer (FC) - ReLU activation function - fully connected layer (FC) - ReLU activation function.

[0073] SOH prediction sub-framework: CNN - gated recurrent unit (GRU) - fully connected layer (FC) - ReLU activation function - fully connected layer (FC).

[0074] Among them, the SOH prediction sub-framework needs to call the GRU, including the reset gate, update gate, candidate hidden state, and final hidden state structures, and the structure is as Figure 4 shown.

[0075] In the built SOH estimation sub-framework and SOH prediction sub-framework, the CNN and GRU are combined so that the constructed framework can accurately estimate the current health state of the battery and effectively predict the development trend of the future health state at the same time, providing comprehensive technical support for the management and maintenance of the battery.

[0076] (c) Introduce an optimization algorithm: Use the improved QPSO algorithm to simultaneously optimize the number of convolutional kernels (CNN_Kernal) of the CNN, the number of hidden layers (Hidden_Layer) of the GRU layer, and the input batch size (Batch_Size) during the training process. At the same time, the Adam algorithm is used to optimize the input weights, output weights, and bias values in the network.

[0077] Among them, the particle position update of the traditional QPSO algorithm is shown in formula (5):

[0078]

[0079] where x i represents the position of the i-th particle, P i is used for the update of the position of the i-th particle, M best is the average optimal position of the particles, α is the contraction-expansion factor, and u is a uniform distribution on (0, 1).

[0080] A dynamic decreasing strategy of the contraction-expansion factor is introduced to improve the traditional QPSO algorithm. At the initial stage of algorithm iteration, the contraction-expansion factor is set to a relatively large value to accelerate the search speed and quickly approach the global extreme value. At the later stage of iteration, its value is set to a relatively small value to enhance the local search ability.

[0081] In this embodiment, the contraction-expansion factor is dynamically updated by formula (6) to prevent particles from falling into local optima or crossing the optimal value:

[0082]

[0083] where α initial represents the initial value of the contraction-expansion factor α, t max represents the maximum number of iterations, t represents the current iteration number, and n is an empirical value, which is taken as 0.5 in this embodiment. According to different values of α initial and n, different decay strategies can be achieved. By introducing the dynamic decreasing strategy of the contraction-expansion factor, particles are prevented from falling into local optimal values or crossing the optimal value, significantly improving the SOH estimation and prediction accuracy.

[0084] Step 3: Framework training and verification: The battery dataset is divided into a training set and a test set according to a certain ratio. The SOH estimation and prediction framework is trained and optimized using the training set, and the estimation performance and prediction performance of the framework are verified using the test set. The generalization of the framework is verified using other battery datasets not used for training.

[0085] Specifically, it includes:

[0086] (a) Dataset division and framework training:

[0087] In this embodiment, the dataset of the CS2_35 battery is divided into a training set and a test set according to a ratio of 7:3, and the number of training times is set to 200 times to obtain the change of the loss function.

[0088] Of course, in other embodiments, the above division ratio, number of training times, etc. can be adjusted or changed according to specific situations.

[0089] Figure 5 is the change trend of the loss function during the framework training process. The abscissa is the iteration number, and the ordinate is the mean squared error loss in the loss function. It can be seen that when the improved QPSO algorithm is used to optimize the framework, the mean squared error loss in the loss function converges very quickly during the training process, and the time required to train the framework is less.

[0090] Apply the improved QPSO algorithm to the SOH estimation sub-framework and prediction sub-framework, enabling the neural network to dynamically adjust its structural parameters according to the QPSO algorithm during training, thereby accelerating convergence to the global optimal solution and improving the modeling and training efficiency.

[0091] (b) Framework accuracy verification:

[0092] Use the divided test set to verify the estimation performance and prediction performance of the framework to test its accuracy.

[0093] For the verification of the accuracy of the SOH estimation results, a method of shuffling the order and randomly taking values within the SOH change range is adopted, which is different from the traditional method of taking values in order. Figure 6 In (a) and (b), the comparison and error between the SOH estimated value and the experimental value of the CS2_35 battery dataset are shown.

[0094] For the verification of the accuracy of the SOH prediction results, since the CS2_35 dataset has been divided into a training set and a test set in a ratio of 7:3, only the SOH of the last 30% of the CS2_35 battery needs to be predicted. Figure 7 In (a), the comparison and error between the SOH predicted value and the experimental value of the CS2_35 battery dataset are shown.

[0095] (c) Framework generalization verification:

[0096] To test the generalization of the SOH estimation and prediction framework obtained by training, in this embodiment, three battery datasets, CS2_36, CS2_37, and CS2_38, are used for accuracy verification. The mean square error (MSE), root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R 2 ) are used as error evaluation indicators, and the calculation formulas are shown in (7)(8)(9)(10).

[0097]

[0098] The comparison and error between the SOH estimated values and the experimental values of the CS2_36, CS2_37, and CS2_38 datasets are shown in Figure 6 (c)-(d), (e)-(f), (g)-(h) as shown. The comparison and error between the SOH predicted values and the experimental values of the CS2_36, CS2_37, and CS2_38 datasets are shown in Figure 7 (b), (c), (d) as shown. It can be seen that the gaps between the SOH estimation results and prediction results of each dataset and the experimental values are relatively small.

[0099] Table 1 and Table 2 are the evaluation results of SOH estimation accuracy and SOH prediction accuracy respectively. As can be seen from Table 1, when the framework is used to estimate SOH, the MSE is less than 1%, and the RMSE and MAE are about 3%, indicating that the error between the estimated value and the experimental value is small. As can be seen from Table 2, when the framework is used to predict SOH, the MSE is less than 1%, and the RMSE and MAE are about 4%, and the coefficient of determination R 2 is about 95% on average, indicating a relatively high prediction accuracy.

[0100] The above results all prove the accuracy of the method provided in this embodiment in SOH estimation and prediction. Therefore, for the battery dataset not used for training, the solution provided in this embodiment can still maintain a relatively high accuracy in SOH estimation and prediction, showing good generalization and robustness.

[0101] Table 1 Evaluation of SOH Estimation Results

[0102]

[0103] Table 2 Evaluation of SOH Prediction Results

[0104]

[0105] Under the background of the gradual popularization of current battery cloud data applications, this embodiment can use the massive high-quality data on the cloud platform to optimize the structure of the data-driven model in real time and dynamically, maximizing the computational accuracy and efficiency of the model.

[0106] Through the above specific implementation manners, the present invention realizes a method for estimating the health state of a battery by data-driven fusion quantum optimization, effectively solves the problems in the prior art, and has significant technical advantages and application prospects.

[0107] Embodiment 2

[0108] A system for estimating the health state of a battery by data-driven fusion quantum optimization, comprising:

[0109] A data processing module, configured to obtain a battery dataset, select the constant current charging time, constant voltage charging time and internal resistance during the battery charging process as input features, and perform preprocessing to divide the training set and the test set;

[0110] The execution process of this module can specifically refer to the process of step 1 in Embodiment 1, which will not be elaborated here.

[0111] A model building and optimization module is configured to build a battery health state estimation and prediction model, introduce a dynamic reduction strategy of a shrinkage-expansion factor in a quantum particle swarm optimization algorithm, train the model using a training set, and optimize the battery health state estimation and prediction model using an improved quantum particle swarm optimization algorithm during the training process;

[0112] The execution process of this module can be specifically referred to the process of step 2 of the first embodiment, which will not be repeated here.

[0113] A testing module, configured to use a test set to verify the trained battery health state estimation and prediction model;

[0114] The execution process of this module can be specifically referred to the process of step 3 of the first embodiment, which will not be repeated here.

[0115] The estimation and prediction module is configured to obtain a target battery data set, select constant current charging time, constant voltage charging time and internal resistance during the battery charging process as input features, and perform preprocessing, and use a verified battery health state estimation and prediction model to estimate and predict the target battery health state to obtain an estimation and prediction result.

[0116] It can be understood that the above-mentioned units / modules can be separately or completely combined into one or several other units / modules, or one (some) of the units can be further divided into multiple functionally smaller units, which can achieve the same operation without affecting the realization of the technical effects of the embodiments of the present application.

[0117] The above modules of the system are divided based on logical functions. In practical applications, the function of one module can also be implemented by multiple modules, or the functions of multiple modules can be implemented by one module.

[0118] Similarly, in other embodiments of the present application, the system may also include other units / modules. In actual applications, these functions may also be implemented with the assistance of other units, and may be implemented by collaboration of multiple units.

[0119] According to another embodiment of the present application, the system described in this embodiment can be constructed and the method of Embodiment 1 can be implemented by running a computer program (including program code) that can execute the respective steps involved in the corresponding method described in Embodiment 1 on a general computing device such as a computer including processing elements and storage elements such as a Central Processing Unit (CPU), a Random Access Memory (RAM), and a Read-Only Memory (ROM). The computer program can be recorded on, for example, a computer-readable recording medium, loaded into the above computing device through the computer-readable recording medium, and run therein.

[0120] Embodiment 3

[0121] This implementation provides an electronic device, such as Figure 8 As shown, the electronic device includes a processor 1001, a communication interface 1002, and a computer-readable storage medium 1003. Among them, the processor 1001, the communication interface 1002, and the computer-readable storage medium 1003 can be connected through a bus or other means.

[0122] Among them, the communication interface 1002 is used to receive and send data. The computer-readable storage medium 1003 can be stored in the memory of the electronic device. The computer-readable storage medium 1003 is used to store a computer program, and the computer program includes program instructions. The processor 1001 is used to execute the program instructions stored in the computer-readable storage medium 1003.

[0123] The processor 1001 (or CPU (Central Processing Unit, central processor)) is the computing core and control core of the electronic device, which is adapted to implement one or more instructions, and is specifically adapted to load and execute one or more instructions to implement the corresponding method flow or corresponding function.

[0124] The processor 1001 is configured to execute the following process:

[0125] Obtain a battery dataset, select the constant current charging time, constant voltage charging time, and internal resistance during the battery charging process as input features, and perform preprocessing to divide the training set and the test set;

[0126] Construct a battery health state estimation and prediction model, introduce a dynamic decreasing strategy of a contraction-expansion factor in the quantum particle swarm optimization algorithm, use the training set to train the model, and optimize the battery health state estimation and prediction model using the improved quantum particle swarm optimization algorithm during the training process;

[0127] Verify the trained battery health state estimation and prediction model using a test set;

[0128] Obtain a target battery dataset, select the constant current charging time, constant voltage charging time, and internal resistance during the battery charging process as input features, and perform preprocessing. Then, use the verified battery health state estimation and prediction model to estimate and predict the target battery health state to obtain the estimation and prediction results.

[0129] Or the process of steps 1-3 in Embodiment 1, which will not be elaborated here.

[0130] Those skilled in the art should understand that the embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) containing computer-usable program code.

[0131] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the specified functions in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0132] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device implements the specified functions in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0133] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide for implementing the specified functions in Figure 1 one process or multiple processes and / or blocksFigure 1 Steps of functions specified in one or more boxes.

[0134] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made by those skilled in the art without creative efforts within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A data-driven quantum-optimized battery health state estimation method, characterized in that: The following steps are involved: Obtain a battery data set, select constant current charging time, constant voltage charging time and internal resistance during battery charging as input features, perform preprocessing, and divide the data into training and test sets; Constructing a battery health state estimation and prediction model, introducing a dynamic reduction strategy of the contraction-expansion factor in the quantum particle swarm optimization algorithm, training the model using a training set, and optimizing the battery health state estimation and prediction model using an improved quantum particle swarm optimization algorithm during the training process; Use the test set to verify the trained battery health status estimation and prediction model; The target battery data set is obtained, and the constant current charging time, constant voltage charging time and internal resistance in the battery charging process are selected as input features and preprocessed. Then, the health status of the target battery is estimated and predicted using the verified battery health status estimation and prediction model to obtain the estimation and prediction results.

2. The method for estimating battery health status by data-driven fusion quantum optimization as claimed in claim 1, characterized in that: The preprocessing process includes using a Hampel filter to process abnormal values ​​of data in each input feature, and normalizing the input feature data.

3. The method for estimating battery health status by data-driven fusion quantum optimization as claimed in claim 1, characterized in that: The battery health state estimation and prediction model includes an estimation subframe and a prediction subframe, wherein the estimation subframe includes a convolutional neural network, a fully connected layer, a ReLU activation function, a fully connected layer, a ReLU activation function and a fully connected layer connected in sequence; The prediction subframework includes a convolutional neural network, a gated recurrent unit, a fully connected layer, a ReLU activation function and a fully connected layer connected in sequence; The convolutional neural network is used to extract deep hidden features from input features for battery health state estimation and prediction.

4. The method for estimating battery health status by data-driven fusion quantum optimization as claimed in claim 3, characterized in that: The convolutional neural network uses the convolution kernel in the convolution layer to extract hidden features in the input layer data layer by layer, and then preprocesses the extracted features through the pooling layer to reduce the size of the feature matrix, and then goes through a convolution layer-pooling layer operation again to further extract deep hidden features and reduce the size of the feature matrix.

5. The method for estimating battery health status by data-driven fusion quantum optimization as claimed in claim 1, characterized in that: The process of introducing a dynamic decreasing strategy of the shrinkage-expansion factor in the quantum particle swarm optimization algorithm includes: in the early iteration of the quantum particle swarm optimization algorithm, the shrinkage-expansion factor is made greater than the set value to speed up the search speed and quickly approach the global extreme value; in the later iteration, the value is made less than the set value to enhance the local search capability.

6. A data-driven, quantum-optimized battery health state estimation method as claimed in claim 5, characterized in that: The value of the shrink-expansion factor is dynamically updated, and the updating process is: where α initial represents the initial value of the contraction-expansion factor α, t max represents the maximum number of iterations, t represents the current number of iterations, and n represents the preset value.

7. A data-driven quantum-optimized battery health state estimation method as claimed in claim 1, characterized in that The process of optimizing the battery health state estimation and prediction model using the improved quantum particle swarm optimization algorithm during the training process includes optimizing the number of convolution kernels in the convolutional neural network of the battery health state estimation and prediction model, the number of hidden layers of the gated recurrent unit, and the input batch size during the training process using the improved quantum particle swarm optimization algorithm during the training process; Or / and, the input weight, output weight and bias value of the model are optimized using the Adam algorithm.

8. The method for estimating battery health status by data-driven fusion quantum optimization as claimed in claim 1, characterized in that: In the process of using the test set to verify the trained battery health state estimation and prediction model, the data order of the test set is disrupted and randomly values ​​are taken within the range of battery health state. The mean square error, root mean square error, mean absolute error and determination coefficient are used as error evaluation indicators.

9. A data-driven quantum-optimized battery health status estimation system, characterized in that: include: The data processing module is configured to obtain a battery data set, select a constant current charging time, a constant voltage charging time and an internal resistance in a battery charging process as input features, perform preprocessing, and divide the data into a training set and a test set; A model building and optimization module is configured to build a battery health state estimation and prediction model, introduce a dynamic reduction strategy of a shrinkage-expansion factor in a quantum particle swarm optimization algorithm, train the model using a training set, and optimize the battery health state estimation and prediction model using an improved quantum particle swarm optimization algorithm during the training process; A testing module, configured to use a test set to verify the trained battery health state estimation and prediction model; The estimation and prediction module is configured to obtain a target battery data set, select constant current charging time, constant voltage charging time and internal resistance during the battery charging process as input features, and perform preprocessing, and use a verified battery health state estimation and prediction model to estimate and predict the target battery health state to obtain an estimation and prediction result.

10. An electronic device, characterized in that: The method comprises a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the steps in the method according to any one of claims 1 to 8 are completed.

Images