Lithium battery health state prediction method, storage medium and program product

Abstract

The invention provides a lithium battery health state prediction method, a storage medium and a program product. The method comprises a step of obtaining a preset health state prediction model, and a step of predicting the health state of a lithium battery by using the preset health state prediction model. The step of obtaining the preset health state prediction model comprises the following steps: initializing a set hyper-parameter of the CNN-Bi-LSTM neural network, and performing optimization update of set times on the set hyper-parameter by adopting an improved Schwander optimization algorithm so as to obtain an optimal set hyper-parameter in the CNN-Bi-LSTM neural network; and constructing an initial health state prediction model based on a CNN-Bi-LSTM neural network by using the optimal set hyper-parameter, and training the initial health state prediction model by using a preset training data set to obtain a preset health state prediction model.

Description

Technical Field

[0001] This invention relates to the field of lithium battery health status prediction technology, and in particular to a method, storage medium and program product for predicting the health status of lithium batteries. Background Technology

[0002] Lithium-ion batteries are a type of battery that uses lithium metal or lithium alloys as positive / negative electrode materials and a non-aqueous electrolyte solution. They are characterized by long cycle life, high power density, and low self-discharge rate, and are widely used in distributed energy storage, becoming a primary energy storage device in drones, electric vehicles, and robots. However, the continuous chemical reactions within lithium-ion batteries can lead to problems such as reduced battery performance and permanent capacity degradation. Therefore, accurate health assessment of lithium-ion batteries has become a key factor in ensuring the safe and stable operation of electric equipment such as drones.

[0003] Currently, methods for predicting the health status of lithium batteries include model estimation and data-driven methods. Model estimation methods establish physical or electrochemical models of the battery, simulate the electrochemical mechanisms between internal battery materials, obtain the relationship between the internal state and external characteristics of the battery system, and understand its aging evolution patterns, thereby predicting the health status of the lithium battery. Data-driven methods directly extract health indicators characterizing battery degradation from historical degradation data of lithium batteries and combine them with intelligent algorithms to deeply explore the battery's lifespan degradation behavior.

[0004] The accuracy of data-driven methods in predicting the health status of lithium batteries depends on the superiority of the hyperparameters in the intelligent algorithm. Therefore, how to select appropriate hyperparameters to construct an intelligent algorithm model is an important factor in improving the accuracy of lithium battery health status prediction. Summary of the Invention

[0005] This invention provides a method, storage medium, and program product for predicting the health status of lithium batteries, which are used to improve the accuracy of predicting the health status of lithium batteries.

[0006] Specifically, the present invention provides a method for predicting the health status of a lithium battery, comprising:

[0007] The steps of obtaining a preset health status prediction model, and the steps of using the preset health status prediction model to predict the health status of a lithium battery.

[0008] The step of obtaining the preset health status prediction model includes:

[0009] The hyperparameters of the CNN-Bi-LSTM neural network are initialized, and the improved Snow Goose optimization algorithm is used to optimize and update the hyperparameters a set number of times to obtain the optimal hyperparameters in the CNN-Bi-LSTM neural network; wherein each optimization update includes:

[0010] The snow goose population is constructed based on the current optimized hyperparameter settings, and the current optimization update stage is determined based on the current number of optimization updates.

[0011] If the current optimization and update type is the exploration phase, then the snow goose population will be optimized and updated using a V-formation flight mode;

[0012] If the current optimization and update type is the development stage, then the snow goose population will be optimized and updated using a straight flight mode;

[0013] Based on the optimized and updated snow goose population, the optimized and updated hyperparameters were obtained;

[0014] An initial health status prediction model based on a CNN-Bi-LSTM neural network is constructed using the optimal hyperparameter settings, and the initial health status prediction model is trained using a preset training dataset to obtain the preset health status prediction model.

[0015] The step of predicting the health status of the lithium battery using the preset health status prediction model includes:

[0016] The set health characteristic factors of the lithium battery are obtained, and the set health characteristic factors are input into the preset health state prediction model to obtain the health state of the lithium battery.

[0017] Furthermore, the step of determining the current optimization update stage based on the current number of optimization updates includes:

[0018] A nonlinear transition strategy is adopted to calculate the deflection angle of the snow goose population based on the current number of optimization updates, and to determine the current optimization update stage based on the magnitude of the deflection angle.

[0019] Furthermore, after the step of optimizing and updating the snow goose population using a V-formation flight pattern, the method further includes: obtaining the optimal snow goose individual in the snow goose population, and updating the leader snow goose individual in the snow goose population based on the optimal snow goose individual.

[0020] Further, the step of obtaining the set health characteristic factors of the lithium battery includes:

[0021] Obtain multiple reference charge and discharge parameters of the lithium battery, and calculate the correlation coefficient between each reference charge and discharge parameter and the battery capacity of the lithium battery.

[0022] The types of the reference charge-discharge parameters with the largest correlation coefficient are taken as preset types, and the preset health characteristic factors are obtained according to the preset types.

[0023] Furthermore, the step of initializing the hyperparameters of the CNN-Bi-LSTM neural network includes:

[0024] Obtain the value range of the set hyperparameter, and randomly obtain the initial value of the set hyperparameter within the value range.

[0025] Further, the step of obtaining the value range of the set hyperparameter includes:

[0026] A preset set of optimal points is used to obtain the optimal set of hyperparameters, thereby obtaining the value range. Further, the step of initializing the hyperparameters of the CNN-Bi-LSTM neural network includes:

[0027] Obtain initial values ​​for the initial learning rate, maximum number of iterations, and number of hidden neurons of the CNN-Bi-LSTM neural network.

[0028] Further, the step of training the initial health status prediction model using a preset training dataset includes:

[0029] Obtain a test dataset and use the test dataset to calculate the mean absolute error, root mean square error, and mean absolute percentage error of the initial health status prediction model;

[0030] The accuracy of the initial health state prediction model is determined based on the mean absolute error, root mean square error, and mean absolute percentage error. Training of the initial health state prediction model ends when the accuracy is greater than a preset accuracy threshold.

[0031] On the other hand, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the steps of the lithium battery health status prediction method described in any of the above claims.

[0032] In another aspect, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the lithium battery health status prediction method described in any of the above claims.

[0033] In the technical solution of this invention, an improved Snow Goose optimization algorithm is used to obtain the optimal hyperparameters of the CNN-Bi-LSTM neural network, and a predictive health status model based on the CNN-Bi-LSTM neural network is constructed according to these optimal hyperparameters. Since the improved Snow Goose optimization algorithm can guarantee the accuracy, robustness, and stability of the problem solution, it can improve the accuracy and reliability of the optimal hyperparameters. Furthermore, constructing a preset health status prediction model based on these optimal hyperparameters can improve the accuracy and robustness of the model. This preset health status prediction model is then used to predict the health status of the lithium battery. Based on the above factors, compared with the prior art, the technical solution of this invention can guarantee the accuracy of the health status prediction.

[0034] The above and other objects, advantages and features of the present invention will become more apparent to those skilled in the art from the following detailed description of specific embodiments of the invention in conjunction with the accompanying drawings. Attached Figure Description

[0035] The following sections will describe some specific embodiments of the invention in detail by way of example and not limitation, with reference to the accompanying drawings. The same reference numerals in the drawings denote the same or similar parts or portions. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings:

[0036] Figure 1 This is a schematic flowchart of a method for predicting the health status of a lithium battery according to an embodiment of the present invention;

[0037] Figure 2 This is a schematic flowchart illustrating the acquisition of a preset health status prediction model based on a CNN-Bi-LSTM neural network model according to an embodiment of the present invention;

[0038] Figure 3 This is a schematic flowchart illustrating the use of an improved Snow Goose optimization algorithm to obtain the optimal hyperparameters in a CNN-Bi-LSTM neural network according to an embodiment of the present invention.

[0039] Figure 4 This is a schematic diagram of a computer program product according to an embodiment of the present invention;

[0040] Figure 5 This is a schematic diagram of a computer-readable storage medium according to an embodiment of the present invention. Detailed Implementation

[0041] The following reference Figures 1 to 5This invention describes a method, storage medium, and program product for predicting the health status of a lithium battery according to embodiments of the present invention. In this description, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first" and "second" may explicitly or implicitly include at least one of that feature, that is, include one or more of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified. When a feature "includes or contains" one or more of the features it encompasses, unless otherwise specifically described, this indicates that other features are not excluded and may be further included.

[0042] In the description of this embodiment, the terms "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0043] Please see Figure 1 , Figure 1 The illustration shows an embodiment of the present invention: a method for predicting the health status of a lithium battery. This method employs a preset health status prediction model based on a CNN-Bi-LSTM neural network to predict the health status of the lithium battery according to its health characteristic factors, thereby achieving the goal of accurate prediction results.

[0044] Specifically, the health status prediction method in this embodiment includes the following steps:

[0045] Step S101: Obtain a preset health status prediction model based on CNN-Bi-LSTM neural network;

[0046] Step S102: Predict the health status of the lithium battery using a preset health status prediction model.

[0047] In step S101, the CNN-Bi-LSTM neural network is a combination of a convolutional neural network and a bidirectional long short-term memory neural network, and the method for obtaining a preset health status prediction model based on the CNN-Bi-LSTM neural network model is as follows: Figure 2 As shown, it includes the following steps:

[0048] Step S111: Use the improved Snow Goose optimization algorithm to obtain the optimal hyperparameters of the CNN-Bi-LSTM neural network;

[0049] Step S112: Construct an initial health status prediction model based on a CNN-Bi-LSTM neural network using the obtained optimal hyperparameter settings;

[0050] Step S113: Obtain a preset training dataset and use the preset training dataset to train the initial health status prediction model to obtain the preset health status prediction model.

[0051] In step S111 above, the optimal hyperparameters of the CNN-Bi-LSTM neural network obtained using the improved Xueyan optimization algorithm include the optimal initial learning rate, optimal number of iterations, and optimal number of hidden neurons. Furthermore, in this embodiment, the method for obtaining the optimal hyperparameters of the CNN-Bi-LSTM neural network using the improved Xueyan optimization algorithm is as follows: Figure 3 As shown, it includes the following steps:

[0052] Step S201: Initialize the snow goose population within a preset spatial range;

[0053] Step S202: The improved snow goose optimization algorithm is used to optimize and update the snow goose population a set number of times, and the optimal hyperparameters of the CNN-Bi-LSTM neural network are obtained based on the best snow goose individual in the optimized and updated snow goose population.

[0054] In step S201 above, taking a single snow goose from the snow goose population as an example, let the position of this snow goose be P. The position of this snow goose can be randomly generated using the following formula:

[0055] P = lb + rand * (ub - lb)

[0056] Where ub and lb represent the upper and lower bounds of the preset spatial range, respectively. The upper bound includes the preset upper limit of the initial learning rate, the preset upper limit of the number of iterations, and the preset upper limit of the number of hidden neurons. The lower bound includes the preset lower limit of the initial learning rate, the preset lower limit of the number of iterations, and the preset lower limit of the number of hidden neurons. rand is a random number in the range [0, 1], and the characteristics of each individual snow goose in the snow goose population include the initial learning rate, the number of iterations, and the number of hidden neurons.

[0057] In the process of optimizing and updating the snow goose population using the improved snow goose optimization algorithm in step S202 above, the current optimization and update stage of the snow goose population can be determined based on the current number of optimization and updates. If it is the exploration stage, the snow goose population is optimized and updated using a V-shaped flight pattern. If it is the development stage, the snow goose population is optimized and updated using a straight flight pattern.

[0058] Specifically, taking the t-th optimization update of the snow goose population as an example, we first obtain the deflection angle of this optimization update, and let this deflection angle be θ. t ,but

[0059]

[0060] Where M is the set number of times the snow goose population is optimized and updated, and if 0 ≤ θ t If π < θ, then the current optimization and update phase is the exploration phase; if π ≤ θ t If the value is less than 2π, then the current optimization and update is in the development phase.

[0061] This embodiment employs a V-shaped flight pattern to update the snow goose population during the exploration phase, including the following methods:

[0062] First, calculate the fitness value of each individual snow goose in the snow goose population, and sort the snow goose individuals in the population according to the fitness value from largest to smallest. Then, the top 20% of the snow goose individuals in the sorted population are regarded as young and middle-aged snow goose individuals, the bottom 20% of the snow goose individuals are regarded as old snow goose individuals, and the other snow goose individuals are regarded as juvenile snow goose individuals.

[0063] Then, optimization and updates were carried out on young and middle-aged snow geese, elderly snow geese, and juvenile snow geese in the snow goose population during the exploratory phase.

[0064] In this embodiment, during the t-th optimization update process, when optimizing and updating the i-th snow goose individual, if the i-th snow goose individual is a young adult snow goose individual, then the update formula for the snow goose individual is:

[0065]

[0066] Where b is the first weighting coefficient, the value of which is obtained empirically; It represents the position of the i-th snow goose individual in the snow goose population before the t-th iteration update. It represents the position of the i-th individual snow goose in the snow goose population after the t-th iteration update. Let the flight speed of the i-th snow goose be optimized and updated in the t-th iteration, and:

[0067] b = 4·rand(0,1)-2

[0068]

[0069] In the above formula, rand(0, 1) is a random function, representing the generation of a random value within the range of 0 to 1; a i Let be the flight acceleration of the i-th snow goose in the snow goose population, and c be a preset weighting factor. The acceleration is calculated using the following formulas:

[0070]

[0071] If the i-th snow goose is an older snow goose, then the formula for updating the position of this snow goose is:

[0072]

[0073] Where d is the second weighting coefficient, the value of which is obtained empirically, and:

[0074] d = 3·rand(0,1) - 1.5

[0075] If the i-th snow goose is a juvenile snow goose, then the formula for updating the position of this snow goose is:

[0076]

[0077] In the above formula, This indicates the central individual position of the snow goose population in the t-th optimization update. This represents the position of the optimal individual snow goose. This represents the position of the snow goose with the lowest fitness value, and 'e' is the third weighting coefficient, whose value is obtained empirically.

[0078]

[0079] e = 2·rand(0,1)-1

[0080] During the development phase, methods for optimizing and updating the snow goose population using a straight-line flight pattern include:

[0081] First, a random number r is generated within the range (0,1). If the random number r > 0.5, then the individual snow geese in the snow goose population exhibit "collective guidance" during the development phase. At this time, the snow goose population will follow experienced and energetic snow geese to find the best destination. Therefore, the position of the i-th snow goose in the t-th optimization update is calculated using the following formula:

[0082]

[0083] If the random number r ≤ 0.5, then individual snow geese in the snow goose population exhibit "random movement" during the development phase. If the snow goose population is trapped in a local solution, individual snow geese will exhibit random behavior similar to Brownian motion. Therefore, the position of the i-th snow goose in the t-th optimization update is calculated using the following formula:

[0084]

[0085] r represents a random number in the range (0,1). During the development phase, individual snow geese in a population exhibit two behaviors: "collective guidance" and "random movement." "Collective guidance" refers to the snow goose flock following experienced and energetic geese to find the optimal destination when r > 0.5. "Random behavior" describes the random behavior of individual snow geese, similar to Brownian motion, when the flock is trapped in a local solution and r ≤ 0.5. α in the formula represents the differential operator, which, combined with the Brownian equation, represents the random differential of Brownian motion. The expression represents element-wise multiplication. Brownian() represents the Brownian motion function, which is characterized by the strong randomness of particle motion and its independence from external driving and control.

[0086] In step S112 above, the number of hidden neurons in the CNN-Bi-LSTM neural network can be set to the optimal number of hidden neurons, thereby obtaining the initial health status prediction model based on the CNN-Bi-LSTM neural network.

[0087] In step S113 above, the acquired preset training dataset includes multiple health feature samples of lithium batteries and the health status corresponding to each health feature sample. The initial learning rate of the preset health status prediction model is set to the optimal initial learning rate, and the preset health status prediction model is optimized and updated with the optimal number of iterations to obtain the preset health status prediction model.

[0088] In step S102 above, a set health characteristic factor of the lithium battery can be collected and input into a preset health state prediction model to obtain the health state of the lithium battery.

[0089] In this embodiment, the health status of the lithium battery is determined based on its capacity. It is assumed that the maximum usable capacity of the lithium battery in its current state is Q. c The rated capacity of the lithium battery is Q. R The health status score of the lithium battery is then...

[0090]

[0091] As can be seen from the above, this embodiment uses an improved Snow Goose optimization algorithm to obtain the optimal hyperparameters of the CNN-Bi-LSTM neural network, and constructs a predictive health status model based on the CNN-Bi-LSTM neural network according to the optimal hyperparameters. Since the improved Snow Goose optimization algorithm can ensure the accuracy, robustness and stability of the problem solution, it can improve the accuracy and reliability of the optimal hyperparameters. Furthermore, constructing a preset health status prediction model based on the optimal hyperparameters can improve the accuracy and robustness of the model. Thus, using the preset health status prediction model to predict the health status of the lithium battery can ensure the accuracy of the health status prediction.

[0092] In some embodiments of the present invention, the method for determining the current optimization update stage based on the current number of optimization updates during each optimization update process using the improved Xueyan optimization algorithm includes:

[0093] A nonlinear transition strategy is adopted to calculate the deflection angle of the snow goose population based on the current number of optimization updates, and to determine the current optimization update stage based on the magnitude of the deflection angle.

[0094] For example, if the current number of optimization updates is t, and a nonlinear transition strategy is used, the deflection angle of the snow goose population can be obtained as follows:

[0095]

[0096] This embodiment allows for an exploratory phase update of the snow goose population during the early stages of optimization and updating. Since the update function grows faster during the exploratory phase, this accelerates the exploration of the snow goose population. In the later stages of optimization and updating, a development phase update is performed on the snow goose population. Since the update function grows slower during the development phase, this facilitates more refined exploration during the development phase.

[0097] In some embodiments of the present invention, during each optimization and update of the set hyperparameters using the improved snow goose optimization algorithm, after each optimization and update of the snow goose population during the exploration phase, the position of the leading snow goose individual in the snow goose population is also updated.

[0098] Specifically, after optimizing and updating the snow goose population during the exploratory phase, the fitness of each individual snow goose in the population is calculated, and the central position of the top 20% of snow goose individuals with the highest fitness is obtained. Let this central position be... but

[0099]

[0100] Where U = 0.2·N indicates that only the top 20% of snow goose individuals in the population compete for leadership; then, the positions of the top 20% of snow goose individuals in the population are updated again using the following formula:

[0101]

[0102] In the above formula, k is the fourth weighting coefficient, and

[0103] k = rand(0, 1) - 0.5

[0104] Then the leaders in the snow goose population are updated, including:

[0105] like Then As the position of the i-th individual snow goose in the snow goose population after the t-th iteration update;

[0106] like Then This represents the position of the i-th individual snow goose in the snow goose population after the t-th iteration update.

[0107] In this embodiment, after optimizing and updating individual snow geese in the snow goose population during the development phase, the leading snow goose in the population is also updated to ensure the superiority of the leading snow goose in the population, thereby improving the reliability and accuracy of the optimization and updating of the snow goose population.

[0108] In some embodiments of the present invention, the method for obtaining the set health characteristic factor of the lithium battery in step S102 includes: first, obtaining a variety of reference charge and discharge parameters of the lithium battery, and calculating the correlation coefficient between each reference charge and discharge parameter and the battery capacity respectively; then, taking the type of a set number of reference charge and discharge parameters with the largest correlation coefficient with the battery capacity as a preset type, and obtaining the set health characteristic factor of the lithium battery according to the preset type.

[0109] In this embodiment, the charging and discharging times of the constant current charging stage, constant voltage charging stage, and constant current discharging stage were obtained at the 40th, 80th, 120th, and 160th cycles. As the number of cycles increased, the charging time of the battery in the constant current charging stage gradually decreased from 3097s to 1635s, the charging time in the constant voltage charging stage gradually increased from 6918s to 8949s, and the discharging time in the constant current discharging stage gradually decreased from 3185s to 2346s, thus obtaining various reference charging parameters for the lithium battery.

[0110] During the constant current charging phase, as the number of cycles increases, the constant current charging time decreases, and the battery capacity corresponding to the number of cycles also decreases. During the constant voltage charging phase, as the number of cycles increases, the constant voltage charging time increases, showing the opposite trend to the decrease in battery capacity. During the constant current discharging phase, as the number of cycles increases, the time interval between equal voltage drops decreases, showing the same trend as the gradual decrease in battery capacity with discharge cycles.

[0111] First, the battery is charged in constant current charging mode at a current of 1.5A until the battery voltage reaches the charging cutoff voltage of 4.2V. Then, it is charged in constant voltage charging mode at a voltage of 4.2V until the battery current drops to the cutoff current of 20mA. The battery discharge process is as follows: the battery is discharged in constant current discharging mode at a current of 2A until the battery voltage drops to 2.7V. When the battery capacity decays to below 80% of the rated capacity, the battery cycle life ends and the experiment is stopped.

[0112] Taking one type of reference charging parameter as an example, suppose we obtain m such reference charging parameters, where the value of the i-th reference charging parameter is x. j The average value of m reference charging parameters is The lithium battery capacity is predicted as y based on the i-th reference charging parameter. i The average capacity of m lithium batteries is Let rel be the correlation coefficient between the reference charging parameter and the lithium battery capacity, then

[0113]

[0114] The larger the absolute value of the correlation coefficient rel between the reference charging parameters and the battery capacity, the stronger the correlation between the reference charging parameters and the health status of the lithium battery.

[0115] For example, in this embodiment, health characteristic factors are extracted from relevant data of the lithium battery during the constant current charging, constant voltage charging, and constant current discharging stages to predict the health status of the lithium battery. This allows for the acquisition of the time variations of the constant current charging, constant voltage charging, and constant current discharging stages under different cycle conditions. Furthermore, using the method of this embodiment, the constant current charging time, constant voltage charging time, and constant voltage drop time are identified as the reference charging parameters with the highest correlation coefficient with the lithium battery health status. In other words, the types of health characteristic factors are set to include constant current charging time, constant voltage charging time, and constant voltage drop time.

[0116] In this embodiment, the data type of the set health feature factor is determined according to the correlation with the health status of the lithium battery. This ensures that the correlation between the set health feature factor and the health status of the lithium battery is the strongest. Therefore, using the preset health feature factor to predict the health status of the lithium battery can improve the accuracy of the prediction results.

[0117] In this embodiment, a set health feature factor is obtained based on a preset data type with the highest correlation coefficient with the health status of the lithium battery, and the health status of the lithium battery is predicted based on the set health feature factor. This not only ensures the accuracy of the prediction of the health status of the lithium battery, but also reduces the amount of data processing of the preset health status prediction model and improves the efficiency of predicting the health status of the lithium battery.

[0118] In some embodiments of the present invention, the method for initializing the snow goose population in step S201 includes: obtaining a range of values ​​for a set hyperparameter, and randomly obtaining an initial value for the set hyperparameter within the range of values.

[0119] Furthermore, in some embodiments of the present invention, the method for obtaining the value range of the set hyperparameter includes: obtaining the set of optimal points of the set hyperparameter using a preset optimal point set strategy, and determining the value range of each set hyperparameter based on the optimal point set.

[0120] Specifically, assume that the minimum value of the j-th set hyperparameter is lb. j The maximum value is ub j Furthermore, the normalized optimal point coordinates of the i-th snow goose individual in the snow goose population on the j-th set hyperparameter dimension are g. i，j ,but

[0121]

[0122] Where, p j It is the j-th prime number whose set hyperparameter dimension is greater than its dimension, and mod() is the modulo function, i.e. express The decimal part.

[0123] Then, the normalized optimal point coordinates are linearly mapped to the actual value range of each production parameter, resulting in the dimension of the j-th set hyperparameter for the i-th snow goose individual in the snow goose population:

[0124] P i，j =lb j +g i，j ·(ub j -lb j )

[0125] In this embodiment, the optimal initialization strategy is used to obtain the initial values ​​of the set hyperparameters, which can improve the superiority of each set hyperparameter. Then, based on the initial values, the set hyperparameters are optimized and updated, which can further ensure the quality of each optimal set hyperparameter and improve the accuracy and reliability of the preset health state prediction model.

[0126] In some embodiments of the present invention, during the training of the initial health state prediction model using a preset training dataset, the process further includes the following after each training iteration:

[0127] Obtain the test dataset and use it to calculate the mean absolute error, root mean square error, and mean absolute percentage error of the initial health status prediction model.

[0128] The accuracy of the initial health status prediction model is determined based on its mean absolute error, root mean square error, and mean absolute percentage error. If the accuracy is greater than the preset accuracy threshold, the training of the initial health status prediction model is terminated and the preset health status prediction model is obtained.

[0129] In this embodiment, let the mean absolute error of the initial health status prediction model be MAE, the root mean square error be RMSE, and the mean absolute percentage error be MAPE.

[0130]

[0131] Where M is the number of test datasets, y′ i This represents the true value of the i-th health status in the test dataset. This represents the predicted health status value of the i-th unit in the test dataset, as predicted by the initial health status prediction model.

[0132] In this embodiment, the accuracy of the initial health state prediction model is obtained by using mean absolute error, root mean square error, and mean absolute percentage error. This ensures the reliability of the evaluation of the initial health state prediction model. Furthermore, the health state of the lithium battery can be predicted based on the trained preset health state prediction model, thereby improving the accuracy and reliability of the prediction results.

[0133] The flowcharts provided in this embodiment are not intended to indicate that the operations of the method will be performed in any particular order, or that all operations of the method are included in every case. Furthermore, the methods described above may include additional operations. Within the scope of the technical concept provided by the methods in this embodiment, additional variations can be made to the methods described above.

[0134] It should be understood that in some embodiments, the components may be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods may be implemented using software or firmware stored in memory and executed by a suitable instruction execution system.

[0135] This embodiment also provides a computer program product 10 and a computer-readable storage medium 20. Figure 4 This is a schematic diagram of a computer program product 10 according to an embodiment of the present invention. Figure 5 This is a schematic diagram of a computer-readable storage medium 20 according to an embodiment of the present invention. The computer program product 10 includes a computer program 11, which, when executed by a processor 32, implements the steps of any of the above-described lithium battery health state prediction methods. The computer-readable storage medium 20 stores the computer program 11 thereon, which, when executed by the processor 32, implements the steps of the lithium battery health state prediction method of any of the above-described embodiments.

[0136] The computer program 11 used to perform the operations of this invention may be assembly instructions, Instruction Set Architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, integrated circuit configuration data, or source code or object code written in any combination of one or more programming languages ​​and procedural programming languages. The computer program 11 may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter case, the remote computer may be connected to the user's computer via any type of network, including a Local Area Network (LAN) or Wide Area Network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, to perform aspects of this invention, electronic circuits, including, for example, programmable logic circuits, Field-Programmable Gate Arrays (FPGAs), or Programmable Logic Arrays (PLAs), may execute computer-readable program instructions using status information from computer-readable program instructions to personalize the electronic circuits.

[0137] For the purposes of this embodiment, computer program product 10 is a related product that includes computer program 11.

[0138] For the purposes of this embodiment, the computer-readable storage medium 20 is a tangible device capable of holding and storing a computer program 11. It can be any device capable of containing, storing, communicating, propagating, or transmitting the program 11 for use by or in conjunction with an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable storage medium 20 include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable optical disc read-only memory (CD-ROM), digital versatile disc (DVD), memory stick, floppy disk, mechanical encoding device, and any suitable combination thereof.

[0139] Therefore, those skilled in the art should recognize that although numerous exemplary embodiments of the present invention have been shown and described in detail herein, many other variations or modifications conforming to the principles of the present invention can be directly determined or derived from the disclosure of the present invention without departing from the spirit and scope of the invention. Thus, the scope of the present invention should be understood and construed as covering all such other variations or modifications.

Claims

1. A method of predicting a state of health of a lithium battery, characterized in that, The method comprises the following steps: obtaining a preset health state prediction model based on a CNN-Bi-LSTM neural network, and predicting the health state of a lithium battery by using the preset health state prediction model; wherein the step of obtaining the preset health state prediction model comprises: initializing the set hyperparameters of the CNN-Bi-LSTM neural network, and performing optimization update on the set hyperparameters for a set number of times by using an improved snow goose optimization algorithm to obtain the optimal set hyperparameters in the CNN-Bi-LSTM neural network; wherein each time the optimization update comprises: constructing a snow goose population according to the currently optimized set hyperparameters, and determining the stage of the current optimization update according to the number of times of the current optimization update; if the type of the current optimization update is an exploration stage, then performing optimization update on the snow goose population by using a chevron formation flight mode; if the type of the current optimization update is a development stage, then performing optimization update on the snow goose population by using a straight-line flight mode; obtaining the set hyperparameters after optimization update according to the snow goose population after optimization update; constructing an initial health state prediction model based on the CNN-Bi-LSTM neural network by using the optimal set hyperparameters, and training the initial health state prediction model by using a preset training data set to obtain the preset health state prediction model; the step of predicting the health state of the lithium battery by using the preset health state prediction model comprises: obtaining the set health feature factors of the lithium battery, and inputting the set health feature factors into the preset health state prediction model to obtain the health state of the lithium battery.

2. The state of health prediction method of a lithium battery according to claim 1, characterized in that, The method comprises the following steps: the step of determining the stage of the current optimization update according to the number of times of the current optimization update comprises: using a nonlinear transition strategy to calculate the deflection angle of the snow goose population according to the number of times of the current optimization update, and determining the stage of the current optimization update according to the size of the deflection angle.

3. The state of health prediction method of a lithium battery according to claim 1, characterized by, The method comprises the following steps: after the step of performing optimization update on the snow goose population by using a chevron formation flight mode, the method further comprises: obtaining the optimal snow goose individual in the snow goose population, and updating the leader snow goose individual in the snow goose population according to the optimal snow goose individual.

4. The state of health prediction method of a lithium battery according to claim 1, characterized by, The method comprises the following steps: the step of obtaining the set health feature factors of the lithium battery comprises: obtaining a plurality of reference charge and discharge parameters of the lithium battery, and calculating the correlation coefficient between each reference charge and discharge parameter and the battery capacity of the lithium battery respectively; taking the type of a set number of reference charge and discharge parameters with the largest correlation coefficient as a preset type, and obtaining the set health feature factors according to the preset type.

5. The state of health prediction method of a lithium battery according to claim 1, characterized in that, The method comprises the following steps: the step of initializing the set hyperparameters of the CNN-Bi-LSTM neural network comprises: obtaining the value range of the set hyperparameters, and randomly obtaining the initial value of the set hyperparameters within the value range.

6. The state of health prediction method of a lithium battery according to claim 5, characterized in that, The method comprises the following steps: the step of obtaining the value range of the set hyperparameters comprises: obtaining a set of good points of the set hyperparameters by using a preset set of good points to obtain the value range.

7. The state of health prediction method of a lithium battery according to claim 1, characterized by, The method comprises the following steps: The step of setting hyperparameters of the initialized CNN-Bi-LSTM neural network comprises: obtaining initial values of an initial learning rate, a maximum number of iterations and a number of hidden neurons of the CNN-Bi-LSTM neural network.

8. The state of health prediction method of a lithium battery according to claim 1, characterized by, comprise: The step of training the initial health state prediction model using the preset training data set comprises: obtaining a test data set, and calculating the average absolute error, the root mean square error and the average absolute percentage error of the initial health state prediction model using the test data set; determining the accuracy of the initial health state prediction model according to the average absolute error, the root mean square error and the average absolute percentage error, and ending the training of the initial health state prediction model if the accuracy is greater than a preset accuracy threshold.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the lithium battery health state prediction method of any one of claims 1 to 8.

10. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the lithium battery health state prediction method of any one of claims 1 to 8.

Images