Lithium battery multi-stage charging control method and system

Abstract

The invention discloses a lithium battery multi-order charging control method and system, and relates to the related field of battery management, and the method comprises the steps: carrying out the charging test of a to-be-analyzed lithium battery model, and obtaining the charging test data; configuring a charging duration prediction network according to the pre-charging current gradient sequence record data, the constant current gradient sequence record data and the charging duration record data; setting a pre-filling gradient climbing constraint step length and a constant-current gradient sinking constraint step length; constructing an unknown population A; constructing a historical population B; receiving an expected charging duration selected by a user side, performing double-population optimization, and generating a pre-charging current optimization gradient sequence and a constant current optimization gradient sequence; and executing lithium battery multi-stage charging control. The technical problem that the balance between the charging efficiency and the battery safety cannot be realized in the existing lithium battery charging control is solved, and the technical effect of realizing the balance between the charging efficiency and the battery safety is achieved.

Description

Technical Field

[0001] The present application relates to the field of battery management, and in particular to a multi-stage charging control method and system for lithium batteries. Background Art

[0002] As an important energy source for modern electronic devices, the control of the charging process of lithium batteries is crucial to ensure charging efficiency, battery safety and extend battery life. The traditional lithium battery charging method is mainly the constant current and constant voltage charging method, which is divided into three stages during the charging process: trickle charging, constant current charging and constant voltage charging. In the constant current charging stage, the battery is charged at a fixed current until the voltage reaches a preset threshold, and then enters the constant voltage charging stage. The advantages of this method are its simplicity and reliability, which can effectively reduce the charging time. However, as the charging proceeds, the acceptable charging capacity of the lithium battery will gradually decrease. Especially in the later stage of charging, if a larger charging current continues to be used, bubbles may be generated inside the battery, which may cause battery damage or reduce its service life. Therefore, how to reduce the risk of battery damage while ensuring charging efficiency has become a key issue that needs to be urgently solved in the current lithium battery charging technology.

[0003] In the current related technologies, there is a technical problem in lithium battery charging control that it is unable to achieve a balance between charging efficiency and battery safety. Summary of the invention

[0004] The present application provides a lithium battery multi-stage charging control method and system, which adopts charging tests on the lithium battery model to be analyzed, configures a charging time prediction network, sets the pre-charge gradient climbing constraint step and the constant current gradient sinking constraint step, constructs an unknown population A and a historical population B, and performs dual-population optimization and other technical means. By finely controlling the change of charging current, the optimal charging strategy that can both ensure charging efficiency and reduce the risk of battery damage is found, thereby achieving the technical effect of achieving a balance between charging efficiency and battery safety.

[0005] The present application provides a lithium battery multi-stage charging control method, comprising:

[0006] Perform a charging test on the lithium battery model to be analyzed to obtain charging test data, wherein the charging test data includes pre-charge current gradient sequence recording data, constant current gradient sequence recording data and charging time recording data, wherein the pre-charge current gradient sequence recording data is a current rising gradient, and the constant current gradient sequence recording data is a current falling gradient;

[0007] According to the pre-charge current gradient sequence record data, the constant current gradient sequence record data and the charging time record data, a charging time prediction network is configured;

[0008] Set the pre-charge gradient climbing constraint step length and constant current gradient sinking constraint step length;

[0009] Constructing an unknown population A according to the pre-charge gradient climbing constraint step length and the constant current gradient sinking constraint step length;

[0010] Constructing a historical population B according to the charging test data;

[0011] Receiving the expected charging time selected by the user end, performing dual population optimization based on the unknown population A and the historical population B through the charging time prediction network, and generating a pre-charge current optimization gradient sequence and a constant current optimization gradient sequence;

[0012] Multi-stage charging control of the lithium battery is performed according to the pre-charge current optimization gradient sequence and the constant current optimization gradient sequence.

[0013] In a possible implementation, a charging test is performed on the lithium battery model to be analyzed, charging test data is obtained, and the following processing is performed:

[0014] Configure pre-charge cut-off voltage, constant current cut-off voltage and constant voltage cut-off current;

[0015] Setting the number of pre-charge current gradients and the number of constant current gradients, wherein the number of pre-charge current gradients is ≥2, and the number of constant current gradients is ≥2;

[0016] According to the pre-charge cut-off voltage, the constant current cut-off voltage, the constant voltage cut-off current, the pre-charge current gradient number and the constant current gradient number, the lithium battery model to be analyzed is charged and tested at least 50,000 times to obtain the charging test data.

[0017] In a possible implementation, a charging duration prediction network is configured according to the pre-charge current gradient sequence recording data, the constant current gradient sequence recording data and the charging duration recording data, and the following processing is performed:

[0018] The charging duration record data includes pre-charging duration record data and constant current charging duration record data;

[0019] Constructing a first input vector according to the pre-charge current gradient sequence recording data, and constructing a second input vector according to the constant current gradient sequence recording data;

[0020] Supervise the training of a pre-charge duration prediction sub-network through the first input vector and the pre-charge duration record data;

[0021] Supervise and train a constant current charging duration prediction subnetwork through the second input vector and the constant current charging duration record data;

[0022] Construct a fully connected layer, which is used to add the preset constant voltage duration, the first output value of the pre-charge duration prediction subnetwork, and the second output value of the constant current charging duration prediction subnetwork and output them through the output layer to complete the configuration of the charging duration prediction network.

[0023] In a possible implementation, an unknown population A is constructed according to the pre-charge gradient climbing constraint step length and the constant current gradient sinking constraint step length, and the following processing is performed:

[0024] Construct an individual distance evaluation function:

[0025]

[0026]

[0027] Wherein, D1 represents the first deviation distance of the pre-charge current gradient sequence, D2 represents the second deviation distance of the constant current gradient sequence, and I 1,k1 Characterizes the kth sequence gradient current value of the first body of the pre-charge current gradient sequence, I 1,k2 Characterizing the kth sequence gradient current value of the second individual of the pre-charge current gradient sequence, U(I 1,k1 ) represents the preset cutoff voltage of the kth sequence gradient current value of the first body of the pre-charge current gradient sequence, U(I 1,k2 ) represents the preset cut-off voltage of the kth sequence gradient current value of the second individual of the pre-charge current gradient sequence, N is the number of pre-charge current gradients, M is the number of constant current gradients, I 2,j1 Characterizes the jth sequence gradient current value of the first body of the constant current gradient sequence, I 2,j2 Characterizes the jth sequence gradient current value of the second individual of the constant current gradient sequence, U(I 2,j2 ) represents the preset cutoff voltage of the jth sequence gradient current value of the second individual of the constant current gradient sequence, U(I 2,j1 ) represents a preset cutoff voltage of the jth sequence gradient current value of the first body of the constant current gradient sequence;

[0028] Among them, the first deviation distance between any two individuals of the unknown population A is greater than or equal to the first deviation distance threshold, and the second deviation distance is greater than or equal to the second deviation distance threshold.

[0029] In a possible implementation, the expected charging time selected by the user end is received, and dual population optimization is performed based on the unknown population A and the historical population B through the charging time prediction network to generate a pre-charge current optimization gradient sequence and a constant current optimization gradient sequence, and perform the following processing:

[0030] Through the charging time prediction network, traverse the unknown population A to obtain the charging prediction time of the unknown population;

[0031] Traversing the predicted charging time of the unknown population and the expected charging time to find the distance, and obtaining the fitness set of the unknown population;

[0032] Traversing the charging time record data and calculating the distance between the expected charging time, and obtaining a historical population fitness set;

[0033] Adding the fitness of the unknown population fitness set and the fitness of the historical population fitness set to obtain a fitness sum result;

[0034] Traversing the unknown population fitness set and the historical population fitness set, comparing them with the fitness summation result, and obtaining an individual elimination probability set;

[0035] Randomly eliminate half of the individuals according to the individual elimination probability set to generate a selected population C;

[0036] Performing individual directed expansion according to the selected population C to generate an expanded population D;

[0037] Processing the expanded population D through the charging time prediction network to obtain a global fitness minimum value;

[0038] When the global fitness minimum value is less than or equal to the fitness convergence threshold, the pre-charge current optimization gradient sequence and the constant current optimization gradient sequence are output.

[0039] In a possible implementation, the following processing is performed:

[0040] When the global fitness minimum value is greater than the fitness convergence threshold, determining whether the update times of the unknown population A meets the update times threshold;

[0041] When the update times of the unknown population A does not meet the update times threshold, the inferior individuals of the selected population C are replaced by the better individuals of the expanded population D to update the historical population B, and the unknown population A is updated by the pre-charged gradient climbing constraint step and the constant current gradient sinking constraint step, and a new cycle is executed.

[0042] In a possible implementation, the following processing is performed:

[0043] When the update times of the unknown population A meets the update times threshold, the individual corresponding to the global fitness minimum is output and set as the pre-charge current optimization gradient sequence and the constant current optimization gradient sequence.

[0044] In a possible implementation, individual directional expansion is performed according to the selected population C to generate an expanded population D, and the following processing is performed:

[0045] According to the selected population C, obtaining individuals with the minimum fitness of the selected population and a preset number of individuals with larger fitness;

[0046] According to the selected individual with the smallest fitness of the population as the target direction, the preset number of individuals with larger fitness are randomly disturbed for several times to generate the expanded population D.

[0047] The present application also provides a lithium battery multi-stage charging control system, comprising:

[0048] A charging test data acquisition module, the charging test data acquisition module is used to perform a charging test on the lithium battery model to be analyzed to obtain charging test data, wherein the charging test data includes pre-charge current gradient sequence record data, constant current gradient sequence record data and charging time record data, wherein the pre-charge current gradient sequence record data is a current rising gradient, and the constant current gradient sequence record data is a current falling gradient;

[0049] A charging time prediction network configuration module, the charging time prediction network configuration module is used to configure a charging time prediction network according to the pre-charge current gradient sequence recording data, the constant current gradient sequence recording data and the charging time recording data;

[0050] A constraint step length setting module, the constraint step length setting module is used to set the pre-charge gradient climbing constraint step length and the constant current gradient sinking constraint step length;

[0051] An unknown population A construction module, wherein the unknown population A construction module is used to construct an unknown population A according to the pre-charge gradient climbing constraint step length and the constant current gradient sinking constraint step length;

[0052] A historical population B construction module, wherein the historical population B construction module is used to construct a historical population B according to the charging test data;

[0053] A dual-population optimization module, the dual-population optimization module is used to receive the expected charging time selected by the user end, perform dual-population optimization based on the unknown population A and the historical population B through the charging time prediction network, and generate a pre-charge current optimization gradient sequence and a constant current optimization gradient sequence;

[0054] A lithium battery multi-stage charging control module, the lithium battery multi-stage charging control module is used to perform lithium battery multi-stage charging control according to the pre-charge current optimization gradient sequence and the constant current optimization gradient sequence.

[0055] A lithium battery multi-stage charging control method and system proposed in this application first performs a charging test on the lithium battery model to be analyzed to obtain charging test data, wherein the charging test data includes pre-charge current gradient sequence recording data, constant current gradient sequence recording data and charging time recording data, wherein the pre-charge current gradient sequence recording data is a current rising gradient, and the constant current gradient sequence recording data is a current falling gradient, and then according to the pre-charge current gradient sequence recording data, the constant current gradient sequence recording data and the charging time recording data, a charging time prediction network is configured, and then the pre-charge gradient climbing network is set. According to the pre-charge gradient climbing constraint step and the constant current gradient sinking constraint step, an unknown population A is constructed, and then according to the charging test data, a historical population B is constructed. Then, the expected charging time selected by the user is received, and a dual-population optimization is performed based on the unknown population A and the historical population B through the charging time prediction network to generate a pre-charge current optimization gradient sequence and a constant current optimization gradient sequence. Finally, multi-stage charging control of the lithium battery is performed according to the pre-charge current optimization gradient sequence and the constant current optimization gradient sequence, thereby achieving a technical effect of balancing the charging efficiency and battery safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the accompanying drawings of the embodiment of the present invention will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the system according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0057] Figure 1 A schematic diagram of a flow chart of a multi-stage charging control method for a lithium battery provided in an embodiment of the present application;

[0058] Figure 2 A schematic diagram of the structure of a lithium battery multi-stage charging control system provided in an embodiment of the present application.

[0059] Explanation of the reference numerals: charging test data acquisition module 10, charging time prediction network configuration module 20, constraint step size setting module 30, unknown population A construction module 40, historical population B construction module 50, dual population optimization module 60, lithium battery multi-stage charging control module 70. DETAILED DESCRIPTION

[0060] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0061] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0062] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0063] The present application embodiment provides a lithium battery multi-stage charging control method, such as Figure 1 As shown, the method includes:

[0064] Step S100, a charging test is performed on the lithium battery model to be analyzed to obtain charging test data, wherein the charging test data includes pre-charge current gradient sequence recording data, constant current current gradient sequence recording data and charging time recording data, wherein the pre-charge current gradient sequence recording data is a current rising gradient, and the constant current current gradient sequence recording data is a current falling gradient. Specifically, the charging test data refers to the current change sequence and charging time recorded during the charging process. Select the lithium battery model to be analyzed, prepare the charging test equipment, set the charging test parameters, such as the starting charging current, the ending charging current, the charging voltage, etc., perform the charging test, obtain the pre-charge current gradient sequence recording data, the pre-charge current gradient sequence recording data indicates the process in which the current gradually increases from a smaller value to a preset value at the beginning of charging, which is characterized by a current rising gradient. Continue to perform the charging test, obtain the constant current gradient sequence recording data, the constant current gradient sequence recording data indicates the process in which the current gradually decreases under a constant voltage at a certain stage of charging, which is characterized by a current falling gradient. Record the charging time of the entire charging process.

[0065] In a possible implementation, a charging test is performed on the lithium battery model to be analyzed to obtain charging test data, and step S100 further includes step S110, configuring the pre-charge cut-off voltage, constant current cut-off voltage and constant voltage cut-off current. Specifically, according to the specifications, material properties and safety requirements of the lithium battery to be analyzed, the pre-charge cut-off voltage is determined, and the pre-charge cut-off voltage is in the pre-charge current stage. When the voltage reaches this value, the pre-charge current stage ends and enters the next charging stage. Similarly, the constant current cut-off voltage is determined, that is, in the constant current charging stage, when the voltage reaches this value, the constant current charging stage ends. Set the constant voltage cut-off current, the constant voltage cut-off current is in the constant voltage charging stage, when the current drops below this value, it is considered that the battery is full and the charging process ends. Step S120, set the pre-charge current gradient number and the constant current current gradient number, wherein the pre-charge current gradient number ≥ 2, and the constant current current gradient number ≥ 2. Specifically, according to the charging characteristics and test requirements of the lithium battery to be analyzed, the number of pre-charge current gradients is set. The number of pre-charge current gradients is the number of different levels of current change in the pre-charge current stage. This number is at least 2, that is, at least two different current values ​​are used in the pre-charge stage. Similarly, the number of constant current current gradients is set. The number of constant current current gradients is the number of different levels of current change in the constant current charging stage. This number is also at least 2 to meet the multi-gradient current test in the constant current charging stage. Step S130, according to the pre-charge cut-off voltage, the constant current cut-off voltage, the constant voltage cut-off current, the number of pre-charge current gradients and the number of constant current gradients, the lithium battery model to be analyzed is charged and tested at least 50,000 times to obtain the charging test data. Specifically, a charging test device is used to design a charging test plan according to the set pre-charge cut-off voltage, constant current cut-off voltage, constant voltage cut-off current, and the number of gradients of pre-charge current and constant current. According to the test plan, the lithium battery to be analyzed is charged for at least 50,000 times to cover all possible charging conditions, and the charging current gradient sequence, charging time and other data of each test are recorded. This implementation method comprehensively understands the behavior of the lithium battery to be analyzed under different charging conditions by setting multiple current gradients, and obtains accurate charging characteristic data. By configuring the cut-off voltage and cut-off current, it ensures that the test is carried out under the premise of battery safety, avoiding battery damage or safety problems caused by overcharging or over-discharging. Through at least 50,000 charging tests, it is ensured that the obtained charging test data is sufficiently universal and can reflect the average charging behavior of the lithium battery to be analyzed under different conditions. Therefore, this implementation method achieves the technical effect of obtaining more accurate and comprehensive charging test data.

[0066] Step S200, configuring a charging time prediction network according to the pre-charge current gradient sequence recording data, the constant current gradient sequence recording data and the charging time recording data. The charging time prediction network is a model constructed based on a neural network, which is used to predict the charging time of the lithium battery to be analyzed according to a given current gradient sequence. Specifically, the charging test data is preprocessed, such as normalization, denoising, etc., and the preprocessed charging test data is used to train the neural network to predict the charging time under different charging current gradients.

[0067] In a possible implementation, step S200 further includes step S210, and the charging duration record data includes pre-charging duration record data and constant current charging duration record data. Specifically, the charging duration of the pre-charging stage and the charging duration of the constant current stage are separated from the charging test data, and the pre-charging duration and the constant current charging duration are recorded to form pre-charging duration record data and constant current charging duration record data. Among them, the pre-charging duration record data is a record of the time taken by the lithium battery to be analyzed from the start of charging to reaching the pre-charging cut-off voltage during the pre-charging current stage; the constant current charging duration record data is a record of the time taken by the lithium battery to be analyzed from the end of pre-charging to reaching the constant current cut-off voltage during the constant current charging stage. Step S220, construct a first input vector based on the pre-charging current gradient sequence record data, and construct a second input vector based on the constant current current gradient sequence record data. Specifically, the pre-charge current gradient sequence record data is encoded in a certain order or structure to form a first input vector for input into the pre-charge duration prediction subnetwork; the constant current current gradient sequence record data is also encoded according to the same rule or structure to form a second input vector for input into the constant current charging duration prediction subnetwork. Step S230, supervise the training of the pre-charge duration prediction subnetwork through the first input vector and the pre-charge duration record data. Specifically, the first input vector is used as the input of the pre-charge duration prediction subnetwork, and the corresponding pre-charge duration record data is used as the target output, and the pre-charge duration prediction subnetwork is trained using a supervised learning algorithm (such as a back propagation algorithm) so that the pre-charge duration prediction subnetwork can accurately predict the pre-charge duration. Step S240, supervise the training of the constant current charging duration prediction subnetwork through the second input vector and the constant current charging duration record data. Specifically, the second input vector is used as the input of the constant current charging time prediction subnetwork, the corresponding constant current charging time record data is used as the target output, and the constant current charging time prediction subnetwork is trained using the same supervised learning algorithm, so that the constant current charging time prediction subnetwork can accurately predict the constant current charging time. Step S250, construct a fully connected layer, the fully connected layer is used to add the preset constant voltage time, the first output value of the pre-charge time prediction subnetwork, the second output value of the constant current charging time prediction subnetwork and output through the output layer to complete the configuration of the charging time prediction network. Specifically, create a fully connected layer, the fully connected layer accepts the output of the pre-charge time prediction subnetwork and the constant current charging time prediction subnetwork as input, in the fully connected layer, the preset constant voltage time is added to the pre-charge time prediction value and the constant current charging time prediction value, and the summed result is output through the output layer as the final charging time prediction value. Wherein, the preset constant voltage time is the preset charging time in the constant voltage charging stage determined based on the characteristics and charging requirements of the lithium battery to be analyzed.This implementation method reduces the complexity of the problem by decomposing the charging time prediction into two subtasks: pre-charging time prediction and constant current charging time prediction. This allows each sub-network to focus on learning the charging time rules of a specific stage, achieving the technical effect of improving prediction accuracy.

[0068] Step S300, set the pre-charge gradient climbing constraint step length and the constant current gradient sinking constraint step length. Specifically, according to the charging characteristics and safety requirements of the lithium battery to be analyzed, the pre-charge gradient climbing constraint step length is set, that is, the maximum step length allowed when the current rises during the pre-charge current stage, which is used to ensure the safety and efficiency of charging. Similarly, according to the charging characteristics and safety requirements of the lithium battery to be analyzed, the constant current gradient sinking constraint step length is set, that is, the maximum step length allowed when the current drops during the constant current stage, which is also used to ensure the safety and efficiency of charging.

[0069] Step S400, construct an unknown population A according to the pre-charge gradient climbing constraint step and the constant current gradient sinking constraint step. Specifically, initialize the unknown population A, which is a set of multiple possible charging current gradient sequences, and each individual in the unknown population A represents a possible charging current gradient sequence, which is used to optimize the search process. According to the set pre-charge gradient climbing constraint step and constant current gradient sinking constraint step, constrain the individuals in the unknown population A to meet the charging requirements.

[0070] In a possible implementation, step S400 further includes step S410, constructing an individual distance evaluation function:

[0071]

[0072] Wherein, D1 represents the first deviation distance of the pre-charge current gradient sequence, D2 represents the second deviation distance of the constant current gradient sequence, and I 1,k1 Characterizes the kth sequence gradient current value of the first body of the pre-charge current gradient sequence, I 1,k2 Characterizing the kth sequence gradient current value of the second individual of the pre-charge current gradient sequence, U(I 1,k1 ) represents the preset cutoff voltage of the kth sequence gradient current value of the first body of the pre-charge current gradient sequence, U(I 1,k2 ) represents the preset cut-off voltage of the kth sequence gradient current value of the second individual of the pre-charge current gradient sequence, N is the number of pre-charge current gradients, M is the number of constant current gradients, I 2,j1 Characterizes the jth sequence gradient current value of the first body of the constant current gradient sequence, I 2,j2 Characterizes the jth sequence gradient current value of the second individual of the constant current gradient sequence, U(I 2,j2) represents the preset cutoff voltage of the jth sequence gradient current value of the second individual of the constant current gradient sequence, U(I 2,j1 ) characterizes the preset cutoff voltage of the jth sequence gradient current value of the first individual in the constant current gradient sequence. Specifically, the individual distance evaluation function is used to calculate the deviation distance between any two individuals in the unknown population A, and the individual distance evaluation function consists of two parts: the first deviation distance of the pre-charge current gradient sequence and the second deviation distance of the constant current gradient sequence. The first deviation distance measures the current gradient difference between any two individuals in the pre-charge current gradient sequence, and is calculated by comparing the sequence gradient current value of each individual and its corresponding preset cutoff voltage. If the sequence gradient current value or cutoff voltage of the two individuals is different, the first deviation distance is larger. The second deviation distance performs a similar calculation for the constant current gradient sequence, and measures the difference between any two individuals in the constant current stage current gradient sequence. The individual distance evaluation function quantifies the similarity or difference between individuals in the unknown population A, and is used for screening and optimization based on these differences. Step S420, wherein the first deviation distance of any two individuals of the unknown population A is greater than or equal to the first deviation distance threshold, and the second deviation distance is greater than or equal to the second deviation distance threshold. Specifically, the first deviation distance threshold and the second deviation distance threshold are used to evaluate the differences between any two individuals in the unknown population A in the pre-charge current gradient sequence and the constant current gradient sequence. The specific screening conditions are: the first deviation distance between any two individuals in the unknown population A must be greater than or equal to the first deviation distance threshold, and the second deviation distance between the two individuals must also be greater than or equal to the second deviation distance threshold. Such a setting ensures that the individuals in the unknown population A have sufficient differences in both the pre-charge current gradient sequence and the constant current gradient sequence, which helps to explore a wider solution space during the optimization process and avoid falling into the local optimal solution. This implementation method constructs an unknown population A that is both diverse and efficient by constructing an individual distance evaluation function and setting threshold conditions, achieving the technical effect of making the optimization process more efficient and effective.

[0073] Step S500, constructing a historical population B according to the charging test data. Specifically, extracting the historical charging current gradient sequence in the charging test data, constructing the historical population B according to the historical data, the historical population B is a set including multiple historical charging current gradient sequences, each individual in the population represents a historical charging current gradient sequence, which is used for comparison and optimization with the unknown population A.

[0074] Step S600, receiving the expected charging time selected by the user end, and performing dual population optimization based on the unknown population A and the historical population B through the charging time prediction network to generate a pre-charge current optimization gradient sequence and a constant current optimization gradient sequence. Specifically, the expected charging time input by the user end is received, and the charging time prediction network is used to predict the charging time of individuals in the unknown population A and the historical population B. According to the prediction results and the expected charging time, a dual population optimization algorithm (such as a genetic algorithm, a particle swarm algorithm, etc.) is used to perform an optimization search, and after multiple iterations, a pre-charge current optimization gradient sequence and a constant current optimization gradient sequence that meet the expected charging time are generated.

[0075] In a possible implementation, step S600 further includes step S610, traversing the unknown population A through the charging time prediction network to obtain the unknown population charging prediction time; traversing the unknown population charging prediction time and the expected charging time to find the distance, and obtaining the unknown population fitness set. Specifically, traverse each individual in the unknown population A, use the charging time prediction network to predict its charging time, compare each charging prediction time with the expected charging time, calculate the distance between them (for example, using absolute error, square error, etc.), and use these distances as the fitness of each individual in the unknown population A to form an unknown population fitness set. Step S620, traverse the charging time record data and the expected charging time to find the distance, and obtain the historical population fitness set. Specifically, traverse each record in the charging time record data (representing the historical population), compare the actual charging time in each record with the expected charging time, calculate the distance between them, and use these distances as the fitness of each individual in the historical population to form a historical population fitness set. Step S630, add the fitness of the unknown population fitness set and the fitness of the historical population fitness set to obtain the fitness sum result; traverse the unknown population fitness set and the historical population fitness set, compare with the fitness sum result, and obtain the individual elimination probability set. Specifically, add the fitness of the unknown population fitness set and the historical population fitness set to obtain the total fitness sum result. Traverse the unknown population fitness set and the historical population fitness set, compare the fitness of each individual with the total fitness sum result, and obtain the individual elimination probability. These elimination probabilities constitute the individual elimination probability set, which is used for individual elimination operations. Step S640, randomly eliminate half of the individuals according to the individual elimination probability set to generate a selected population C; perform individual directional expansion according to the selected population C to generate an expanded population D. Specifically, according to the individual elimination probability set, randomly eliminate half of the individuals in the unknown population A and the historical population B to generate a selected population C. Perform individual directional expansion on the selected population C, that is, generate new individuals according to a certain strategy (such as crossover and mutation operations in genetic algorithms) to form an expanded population D. Step S650, process the expanded population D through the charging time prediction network to obtain the global minimum fitness. Specifically, use the charging time prediction network to predict the charging time of each individual in the expanded population D, calculate the distance between each predicted charging time and the expected charging time, obtain the fitness of each individual in the expanded population D, and find the global minimum fitness from these fitness, that is, the fitness of the optimal individual. Step S660, when the global minimum fitness is less than or equal to the fitness convergence threshold, output the pre-charge current optimization gradient sequence and the constant current optimization gradient sequence.Specifically, if the global minimum fitness value is less than or equal to the fitness convergence threshold, it means that an optimal solution that meets the requirements is found, and the pre-charge current optimization gradient sequence and constant current optimization gradient sequence corresponding to the global minimum fitness value are output. This implementation method updates the population by eliminating individuals with poor fitness and expanding new individuals, improves the diversity of the knot, and achieves the technical effect of facilitating the rapid search for the global optimal solution.

[0076] In a possible implementation, step S600 further includes step S670, when the global fitness minimum value is greater than the fitness convergence threshold, it is determined whether the update times of the unknown population A meets the update times threshold. Specifically, when the global fitness minimum value is still greater than the fitness convergence threshold, it is checked whether the update times of the unknown population A has reached the preset update times threshold. The update times threshold is a control parameter used to limit the maximum update times of the unknown population A to avoid the algorithm from falling into an infinite loop or over-calculation. Step S680, when the update times of the unknown population A does not meet the update times threshold, the inferior individuals of the selected population C are replaced by the superior individuals of the expanded population D, the historical population B is updated, the unknown population A is updated by the pre-charged gradient climbing constraint step and the constant current gradient sinking constraint step, and a new cycle is executed. Specifically, if the update times of the unknown population A has not reached the update times threshold, the update process is entered. First, the historical population B is updated by replacing the inferior individuals in the selected population C with the better individuals in the expanded population D, that is, the selected population C is improved by using the excellent individuals explored in the expanded population D to improve its overall fitness level. Then, the unknown population A is updated according to the pre-charged gradient climbing constraint step and the constant current gradient sinking constraint step. After completing the above update, return to step S650 and start a new round of cycles to continue evaluating the fitness of the new population and try to find a better solution. This implementation method increases the judgment of the update times of the unknown population A and the update steps of the population and historical data, so that the algorithm can adjust the strategy more flexibly during the optimization process, thereby achieving the technical effect of improving the efficiency and accuracy of finding the optimal solution.

[0077] In a possible implementation, step S600 further includes step S690, when the unknown population A is updated the number of times to meet the update number threshold, the global fitness minimum corresponding individual is output, set as the pre-charge current optimization gradient sequence and the constant current current optimization gradient sequence. Specifically, when the number of updates of the unknown population A meets (i.e. reaches or exceeds) the update number threshold, it means that the algorithm has been iterated enough times, but the global fitness minimum is still greater than the fitness convergence threshold. At this time, the algorithm cannot continue to find the current gradient sequence that meets the expected charging time through the current optimization strategy. In this case, the individual corresponding to the current global fitness minimum is output, that is, the "optimal" pre-charge current optimization gradient sequence and constant current optimization gradient sequence currently found are output. Although these sequences do not reach the desired fitness convergence threshold, they are the best solutions that the algorithm can find under given conditions. This implementation method sets the update number threshold so that the algorithm can stop trying after reaching a certain number of iterations, avoiding falling into an infinite loop, and achieving the technical effect of ensuring that the algorithm can give a usable result within a limited time while avoiding the dilemma of falling into an infinite loop.

[0078] In a possible implementation, individual directional expansion is performed according to the selected population C to generate an expanded population D, and step S640 further includes step S641, according to the selected population C, obtaining the individual with the minimum fitness of the selected population and a preset number of individuals with larger fitness. Specifically, the individual with the minimum fitness is found from the selected population C. This individual represents the solution with the best performance in the current population, and is therefore selected as the target direction to guide subsequent individual expansion. In addition to the individual with the minimum fitness, a certain number of individuals with larger fitness are selected from the selected population C. Although these individuals are not as good as the individuals with the minimum fitness, their existence can increase the diversity of the population and help explore a wider solution space in the subsequent expansion process. Step S642, according to the individual with the minimum fitness of the selected population as the target direction, the preset number of individuals with larger fitness are randomly perturbed several times to generate the expanded population D. Specifically, the individual with the minimum fitness selected in step S641 is taken as the target direction, that is, the subsequent expansion process attempts to explore in this direction in order to find a better solution. The individual with the larger fitness selected in step S641 is randomly perturbed. The random perturbation can be achieved by adding noise, performing a small-scale mutation operation, or generating new individuals through a certain strategy. The purpose of this process is to introduce some new elements while maintaining a certain similarity, in the hope of generating individuals with better performance. Repeat the random perturbation operation several times to generate a sufficient number of new individuals to form an expanded population D. The expanded population D introduces new possibilities through random perturbations, providing more options for the optimization process. This implementation method ensures that the individuals in the expanded population D are consistent in the optimization direction by selecting the individual with the minimum fitness of the population as the target direction, and achieves the technical effect of generating an expanded population D with excellent performance and diversity.

[0079] Step S700, perform multi-stage charging control of the lithium battery according to the pre-charge current optimization gradient sequence and the constant current optimization gradient sequence. Specifically, the multi-stage charging control is to perform staged charging control on the lithium battery according to the optimized current gradient sequence to achieve a more efficient and safer charging process. Specifically, the generated pre-charge current optimization gradient sequence and constant current optimization gradient sequence are applied to the charging process of the lithium battery. In the early stage of charging, the charging current is gradually increased according to the pre-charge current optimization gradient sequence until the preset current value is reached, and then the constant current charging stage is entered to keep the current constant, and the charging current is gradually reduced according to the constant current optimization gradient sequence. The charging process is monitored to keep the charging current and voltage within the allowable range, and necessary adjustments are made according to the actual situation. When the preset charging termination condition is reached, the charging is stopped. The embodiment of the present application adopts technical means such as performing charging tests on the lithium battery model to be analyzed, configuring a charging time prediction network, setting a pre-charge gradient climbing constraint step and a constant current gradient sinking constraint step, constructing an unknown population A and a historical population B, and performing dual-population optimization. It achieves the optimal charging strategy that can both ensure charging efficiency and reduce the risk of battery damage by precisely controlling the change of the charging current, thereby achieving the technical effect of achieving a balance between charging efficiency and battery safety.

[0080] In the above, refer to Figure 1 A multi-stage charging control method for a lithium battery according to an embodiment of the present invention is described in detail. Figure 2 A lithium battery multi-stage charging control system according to an embodiment of the present invention is described.

[0081] A lithium battery multi-stage charging control system according to an embodiment of the present invention is used to solve the technical problem that the existing lithium battery charging control cannot achieve a balance between charging efficiency and battery safety, and achieve the technical effect of achieving a balance between charging efficiency and battery safety. A lithium battery multi-stage charging control system includes: a charging test data acquisition module 10, a charging time prediction network configuration module 20, a constraint step setting module 30, an unknown population A construction module 40, a historical population B construction module 50, a dual population optimization module 60, and a lithium battery multi-stage charging control module 70.

[0082] The charging test data acquisition module 10 is used to perform a charging test on the lithium battery model to be analyzed to obtain charging test data, wherein the charging test data includes pre-charge current gradient sequence record data, constant current gradient sequence record data and charging time record data, wherein the pre-charge current gradient sequence record data is a current rising gradient, and the constant current gradient sequence record data is a current falling gradient;

[0083] The charging time prediction network configuration module 20 is used to configure a charging time prediction network according to the pre-charging current gradient sequence recording data, the constant current gradient sequence recording data and the charging time recording data;

[0084] The constraint step length setting module 30 is used to set the pre-charge gradient climbing constraint step length and the constant current gradient sinking constraint step length;

[0085] The unknown population A construction module 40 is used to construct the unknown population A according to the pre-charge gradient climbing constraint step length and the constant current gradient sinking constraint step length;

[0086] The historical population B construction module 50 is used to construct the historical population B according to the charging test data;

[0087] The dual population optimization module 60 is used to receive the expected charging time selected by the user end, and perform dual population optimization based on the unknown population A and the historical population B through the charging time prediction network to generate a pre-charge current optimization gradient sequence and a constant current optimization gradient sequence;

[0088] The lithium battery multi-stage charging control module 70 is used to perform lithium battery multi-stage charging control according to the pre-charge current optimization gradient sequence and the constant current optimization gradient sequence.

[0089] The specific configuration of the charging test data acquisition module 10 will be described in detail below. As described above, the charging test is performed on the lithium battery model to be analyzed to obtain the charging test data. The charging test data acquisition module 10 may further include: a configuration unit for configuring the pre-charge cut-off voltage, the constant current cut-off voltage and the constant voltage cut-off current; a current gradient quantity setting unit for setting the pre-charge current gradient quantity and the constant current gradient quantity, wherein the pre-charge current gradient quantity is ≥2, and the constant current current gradient quantity is ≥2; a charging test unit for performing a charging test on the lithium battery model to be analyzed at least 50,000 times according to the pre-charge cut-off voltage, the constant current cut-off voltage, the constant voltage cut-off current, the pre-charge current gradient quantity and the constant current gradient quantity to obtain the charging test data.

[0090] The specific configuration of the charging time prediction network configuration module 20 will be described in detail below. As described above, according to the pre-charge current gradient sequence recording data, the constant current current gradient sequence recording data and the charging time recording data, the charging time prediction network is configured, and the charging time prediction network configuration module 20 may further include: a charging time recording data construction unit for the charging time recording data including the pre-charge time recording data and the constant current charging time recording data; an input vector construction unit for constructing a first input vector according to the pre-charge current gradient sequence recording data, and a second input vector according to the constant current gradient sequence recording data; a pre-charge time prediction unit for constructing a charging time prediction network; a charging ... The subnetwork training unit is used to supervise the training of the pre-charge time prediction subnetwork through the first input vector and the pre-charge time record data; the constant current charging time prediction subnetwork training unit is used to supervise the training of the constant current charging time prediction subnetwork through the second input vector and the constant current charging time record data; the fully connected layer construction unit is used to construct a fully connected layer, and the fully connected layer is used to add the preset constant voltage time, the first output value of the pre-charge time prediction subnetwork, and the second output value of the constant current charging time prediction subnetwork and output through the output layer to complete the configuration of the charging time prediction network.

[0091] The specific configuration of the unknown population A construction module 40 will be described in detail below. As described above, according to the pre-charge gradient climbing constraint step length and the constant current gradient sinking constraint step length, the unknown population A is constructed, and the unknown population A construction module 40 may further include: an individual distance evaluation function construction unit for constructing an individual distance evaluation function:

[0092]

[0093] Wherein, D1 represents the first deviation distance of the pre-charge current gradient sequence, D2 represents the second deviation distance of the constant current gradient sequence, and I 1,k1 Characterizes the kth sequence gradient current value of the first body of the pre-charge current gradient sequence, I 1,k2 Characterizing the kth sequence gradient current value of the second individual of the pre-charge current gradient sequence, U(I 1,k1 ) represents the preset cutoff voltage of the kth sequence gradient current value of the first body of the pre-charge current gradient sequence, U(I 1,k2 ) represents the preset cut-off voltage of the kth sequence gradient current value of the second individual of the pre-charge current gradient sequence, N is the number of pre-charge current gradients, M is the number of constant current gradients, I 2,j1 Characterizes the jth sequence gradient current value of the first body of the constant current gradient sequence, I 2,j2 Characterizes the jth sequence gradient current value of the second individual of the constant current gradient sequence, U(I 2,j2) represents the preset cutoff voltage of the jth sequence gradient current value of the second individual of the constant current gradient sequence, U(I 2,j1 ) characterizes a preset cutoff voltage of the j-th sequence gradient current value of the first individual of the constant current gradient sequence, wherein the first deviation distance between any two individuals of the unknown population A is greater than or equal to the first deviation distance threshold, and the second deviation distance is greater than or equal to the second deviation distance threshold.

[0094] The specific configuration of the dual population optimization module 60 will be described in detail below. As described above, the expected charging time selected by the user end is received, and dual population optimization is performed based on the unknown population A and the historical population B through the charging time prediction network to generate a pre-charge current optimization gradient sequence and a constant current optimization gradient sequence. The dual population optimization module 60 may further include: an unknown population charging prediction time acquisition unit is used to traverse the unknown population A through the charging time prediction network to obtain the unknown population charging prediction time; an unknown population fitness set acquisition unit is used to traverse the unknown population charging prediction time and calculate the distance between the expected charging time and obtain the unknown population fitness set; a historical population fitness set acquisition unit is used to traverse the charging time record data and calculate the distance between the expected charging time and obtain the historical population fitness set; a fitness summing unit is used to sum the fitness of the unknown population The fitness of the set is compared with the fitness of the historical population fitness set to obtain the fitness sum result; the individual elimination probability set acquisition unit is used to traverse the unknown population fitness set and the historical population fitness set, and compare with the fitness sum result to obtain the individual elimination probability set; the selected population C generation unit is used to randomly eliminate half of the individuals according to the individual elimination probability set to generate the selected population C; the expanded population D generation unit is used to perform individual directional expansion according to the selected population C to generate the expanded population D; the global fitness minimum value acquisition unit is used to process the expanded population D through the charging time prediction network to obtain the global fitness minimum value; the output unit is used to output the pre-charge current optimization gradient sequence and the constant current optimization gradient sequence when the global fitness minimum value is less than or equal to the fitness convergence threshold.

[0095] Among them, the dual-population optimization module 60 can further include: a judgment unit is used to judge whether the number of updates of the unknown population A meets the update number threshold when the global fitness minimum value is greater than the fitness convergence threshold; a loop execution unit is used to replace the inferior individuals of the selected population C with the better individuals of the expanded population D when the number of updates of the unknown population A does not meet the update number threshold, update the historical population B, update the unknown population A through the pre-charge gradient climbing constraint step and the constant current gradient sinking constraint step, and execute a new cycle.

[0096] Among them, the dual-population optimization module 60 can further include: an individual output unit is used to output the individual corresponding to the global fitness minimum when the update times of the unknown population A meets the update times threshold, and set it as the pre-charge current optimization gradient sequence and the constant current optimization gradient sequence.

[0097] Among them, individual directional expansion is performed according to the selected population C to generate an expanded population D, and the expanded population D generation unit may further include: an individual acquisition subunit is used to obtain the individual with the smallest fitness of the selected population and a preset number of individuals with larger fitness according to the selected population C; a random perturbation subunit is used to randomly perturb the preset number of individuals with larger fitness for several times according to the individual with the smallest fitness of the selected population as the target direction to generate the expanded population D.

[0098] A lithium battery multi-stage charging control system provided in an embodiment of the present invention can execute a lithium battery multi-stage charging control method provided in any embodiment of the present invention, and has corresponding functional modules and beneficial effects of the execution method.

[0099] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0100] The above specific implementation manner does not constitute a limitation to the protection scope of the present application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be performed in an order different from that in the embodiment and can still achieve the desired results. In addition, the process depicted in the accompanying drawings does not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A multi-stage charging control method for a lithium battery, characterized in that: include: Perform a charging test on the lithium battery model to be analyzed to obtain charging test data, wherein the charging test data includes pre-charge current gradient sequence recording data, constant current gradient sequence recording data and charging time recording data, wherein the pre-charge current gradient sequence recording data is a current rising gradient, and the constant current gradient sequence recording data is a current falling gradient; According to the pre-charge current gradient sequence record data, the constant current gradient sequence record data and the charging time record data, a charging time prediction network is configured; Set the pre-charge gradient climbing constraint step length and constant current gradient sinking constraint step length; Constructing an unknown population A according to the pre-charge gradient climbing constraint step length and the constant current gradient sinking constraint step length; Constructing a historical population B according to the charging test data; Receiving the expected charging time selected by the user end, performing dual population optimization based on the unknown population A and the historical population B through the charging time prediction network, and generating a pre-charge current optimization gradient sequence and a constant current optimization gradient sequence; Multi-stage charging control of the lithium battery is performed according to the pre-charge current optimization gradient sequence and the constant current optimization gradient sequence.

2. A lithium battery multi-stage charging control method as claimed in claim 1, characterized in that: Perform a charging test on the lithium battery model to be analyzed and obtain charging test data, including: Configure pre-charge cut-off voltage, constant current cut-off voltage and constant voltage cut-off current; Setting the number of pre-charge current gradients and the number of constant current gradients, wherein the number of pre-charge current gradients is ≥2, and the number of constant current gradients is ≥2; According to the pre-charge cut-off voltage, the constant current cut-off voltage, the constant voltage cut-off current, the pre-charge current gradient number and the constant current gradient number, the lithium battery model to be analyzed is charged and tested at least 50,000 times to obtain the charging test data.

3. A lithium battery multi-stage charging control method as claimed in claim 2, characterized in that: According to the pre-charge current gradient sequence record data, the constant current gradient sequence record data and the charging time record data, a charging time prediction network is configured, including: The charging duration record data includes pre-charging duration record data and constant current charging duration record data; Constructing a first input vector according to the pre-charge current gradient sequence recording data, and constructing a second input vector according to the constant current gradient sequence recording data; Supervise the training of a pre-charge duration prediction sub-network through the first input vector and the pre-charge duration record data; Supervise and train a constant current charging duration prediction subnetwork through the second input vector and the constant current charging duration record data; Construct a fully connected layer, which is used to add the preset constant voltage duration, the first output value of the pre-charge duration prediction subnetwork, and the second output value of the constant current charging duration prediction subnetwork and output them through the output layer to complete the configuration of the charging duration prediction network.

4. A lithium battery multi-stage charging control method as claimed in claim 1, characterized in that: According to the pre-charge gradient climbing constraint step length and the constant current gradient sinking constraint step length, an unknown population A is constructed, including: Construct an individual distance evaluation function: Wherein, D1 represents the first deviation distance of the pre-charge current gradient sequence, D2 represents the second deviation distance of the constant current gradient sequence, and I 1,k1 Characterizes the kth sequence gradient current value of the first body of the pre-charge current gradient sequence, I 1,k2 Characterizing the kth sequence gradient current value of the second individual of the pre-charge current gradient sequence, U(I 1,k1 ) represents the preset cutoff voltage of the kth sequence gradient current value of the first body of the pre-charge current gradient sequence, U(I 1,k2 ) represents the preset cut-off voltage of the kth sequence gradient current value of the second individual of the pre-charge current gradient sequence, N is the number of pre-charge current gradients, M is the number of constant current gradients, I 2,j1 Characterizes the jth sequence gradient current value of the first body of the constant current gradient sequence, I 2,j2 Characterizes the jth sequence gradient current value of the second individual of the constant current gradient sequence, U(I 2,j2 ) represents the preset cutoff voltage of the jth sequence gradient current value of the second individual of the constant current gradient sequence, U(I 2,j1 ) represents a preset cutoff voltage of the jth sequence gradient current value of the first body of the constant current gradient sequence; Among them, the first deviation distance between any two individuals of the unknown population A is greater than or equal to the first deviation distance threshold, and the second deviation distance is greater than or equal to the second deviation distance threshold.

5. A lithium battery multi-stage charging control method as claimed in claim 1, characterized in that: The expected charging time selected by the user end is received, and dual population optimization is performed based on the unknown population A and the historical population B through the charging time prediction network to generate a pre-charge current optimization gradient sequence and a constant current optimization gradient sequence, including: Through the charging time prediction network, traverse the unknown population A to obtain the charging prediction time of the unknown population; Traversing the predicted charging time of the unknown population and the expected charging time to find the distance, and obtaining the fitness set of the unknown population; Traversing the charging time record data and calculating the distance between the expected charging time, and obtaining a historical population fitness set; Adding the fitness of the unknown population fitness set and the fitness of the historical population fitness set to obtain a fitness sum result; Traversing the unknown population fitness set and the historical population fitness set, comparing them with the fitness summation result, and obtaining an individual elimination probability set; Randomly eliminate half of the individuals according to the individual elimination probability set to generate a selected population C; Performing individual directed expansion according to the selected population C to generate an expanded population D; Processing the expanded population D through the charging time prediction network to obtain a global fitness minimum value; When the global fitness minimum value is less than or equal to the fitness convergence threshold, the pre-charge current optimization gradient sequence and the constant current optimization gradient sequence are output.

6. A lithium battery multi-stage charging control method as claimed in claim 5, characterized in that: Also includes: When the global fitness minimum value is greater than the fitness convergence threshold, determining whether the update times of the unknown population A meets the update times threshold; When the update times of the unknown population A does not meet the update times threshold, the inferior individuals of the selected population C are replaced by the better individuals of the expanded population D to update the historical population B, and the unknown population A is updated by the pre-charged gradient climbing constraint step and the constant current gradient sinking constraint step, and a new cycle is executed.

7. A lithium battery multi-stage charging control method as claimed in claim 6, characterized in that: Also includes: When the update times of the unknown population A meets the update times threshold, the individual corresponding to the global fitness minimum is output and set as the pre-charge current optimization gradient sequence and the constant current optimization gradient sequence.

8. A lithium battery multi-stage charging control method as claimed in claim 5, characterized in that: Performing individual directional expansion according to the selected population C to generate an expanded population D includes: According to the selected population C, obtaining individuals with the minimum fitness of the selected population and a preset number of individuals with larger fitness; According to the selected individual with the smallest fitness of the population as the target direction, the preset number of individuals with larger fitness are randomly disturbed for several times to generate the expanded population D.

9. A lithium battery multi-stage charging control system, characterized in that: The system is used to implement a lithium battery multi-stage charging control method according to any one of claims 1 to 8, and the system comprises: A charging test data acquisition module, the charging test data acquisition module is used to perform a charging test on the lithium battery model to be analyzed to obtain charging test data, wherein the charging test data includes pre-charge current gradient sequence record data, constant current gradient sequence record data and charging time record data, wherein the pre-charge current gradient sequence record data is a current rising gradient, and the constant current gradient sequence record data is a current falling gradient; A charging time prediction network configuration module, the charging time prediction network configuration module is used to configure a charging time prediction network according to the pre-charge current gradient sequence recording data, the constant current gradient sequence recording data and the charging time recording data; A constraint step length setting module, the constraint step length setting module is used to set the pre-charge gradient climbing constraint step length and the constant current gradient sinking constraint step length; An unknown population A construction module, wherein the unknown population A construction module is used to construct an unknown population A according to the pre-charge gradient climbing constraint step length and the constant current gradient sinking constraint step length; A historical population B construction module, wherein the historical population B construction module is used to construct a historical population B according to the charging test data; A dual-population optimization module, the dual-population optimization module is used to receive the expected charging time selected by the user end, perform dual-population optimization based on the unknown population A and the historical population B through the charging time prediction network, and generate a pre-charge current optimization gradient sequence and a constant current optimization gradient sequence; A lithium battery multi-stage charging control module, the lithium battery multi-stage charging control module is used to perform lithium battery multi-stage charging control according to the pre-charge current optimization gradient sequence and the constant current optimization gradient sequence.

Images