Battery health state prediction method, device, equipment, medium and program product

Abstract

The invention relates to a battery health state prediction method and device, equipment, a medium and a program product. The method comprises the following steps: acquiring power data of a to-be-tested battery under a plurality of loop turns, and inputting the power data into a target health state prediction network to obtain a health state value of the to-be-tested battery. Wherein the target health state prediction network is obtained by training an initial health state prediction network according to a training data set; the training data set is used for training the network prediction precision of the initial health state prediction network in the training process. The method can improve the battery health state detection efficiency and reduce the detection cost.

Description

Technical Field

[0001] The present application relates to the field of new energy technology, and in particular to a battery health status prediction method, device, equipment, medium and program product. Background Art

[0002] With the rapid development of new energy technologies, batteries are increasingly used in various fields, such as electric vehicles and energy storage. Based on this, it is particularly important to evaluate the health status of batteries.

[0003] In the related art, when the health status of a battery is detected, the battery is usually cyclically charged and discharged in a constant voltage and current limiting manner until the health status of the battery is reduced to a preset threshold value to obtain the health status of the battery.

[0004] However, the battery health status detection method in the related art has the problems of low detection efficiency and high detection cost. Summary of the invention

[0005] Based on this, it is necessary to provide a battery health status prediction method, device, equipment, medium and program product that can improve the battery health status detection efficiency and reduce the detection cost in response to the above technical problems.

[0006] In a first aspect, the present application provides a method for predicting a battery health state, the method comprising:

[0007] Obtaining power data of the battery under test under multiple cycle numbers;

[0008] The power data is input into the target health status prediction network to obtain the health status value of the battery to be tested; the target health status prediction network is obtained by training the initial health status prediction network according to the training data set; the training data set is used to train the network prediction accuracy of the initial health status prediction network during the training process.

[0009] In one embodiment, obtaining power data of a battery to be tested under multiple cycle numbers includes:

[0010] Obtain the discharge time, discharge capacity and charging temperature change value of the battery to be tested at each cycle number;

[0011] The discharge time, discharge capacity and charging temperature change values ​​at each cycle number are processed by dimensionality upgrading to obtain the power data of the battery to be tested at each cycle number.

[0012] In one embodiment, the training process of the target health status prediction network includes:

[0013] Obtain a battery data set; the battery data set includes actual power data and actual health status values ​​of multiple batteries at each cycle number;

[0014] Divide the battery data set into a training data set and a validation data set;

[0015] Based on the training data set and the validation data set, the initial health state prediction network is trained to obtain the target health state prediction network.

[0016] In one embodiment, based on the training data set and the validation data set, the initial health state prediction network is trained to obtain the target health state prediction network, including:

[0017] The initial health status prediction network is trained according to the training data set to obtain a candidate health status prediction network;

[0018] Perform relevance verification on the candidate health status prediction network according to the verification data set to obtain a relevance verification result;

[0019] When the correlation verification result is verified to be passed, the candidate health state prediction network is determined as the target health state prediction network.

[0020] In one embodiment, the training data set includes training power data and training health state values; the initial health state prediction network is trained according to the training data set to obtain a candidate health state prediction network, including:

[0021] Input the training power data into the initial health status prediction network to obtain the predicted health status value of the training data set;

[0022] Obtain error data between the predicted health status value and the training health status value of the training data set;

[0023] If the error data is less than a preset error threshold, it is determined that the initial health state prediction network training is completed, and a candidate health state prediction network is obtained.

[0024] In one embodiment, the verification data set includes verification power data and verification health status value; the candidate health status prediction network is verified for relevance according to the verification data set to obtain a relevance verification result, including:

[0025] Inputting the verification power data into the candidate health status prediction network to obtain the predicted health status value of the verification data set;

[0026] Obtaining a correlation coefficient between the predicted health status value and the verified health status value;

[0027] If the correlation coefficient is greater than the preset correlation threshold, the correlation verification result is determined to be verified passed;

[0028] If the correlation coefficient is not greater than the preset correlation threshold, the correlation verification result is determined to be verification failure.

[0029] In one embodiment, the method further comprises:

[0030] When the correlation verification result is verification failure, the battery data set is re-divided to obtain a training data set and a verification data set, and the candidate health state prediction network is trained according to the divided training data set until the correlation verification result of the candidate health state prediction network is verification passed, thereby obtaining the target health state prediction network.

[0031] In a second aspect, the present application also provides a battery health status prediction device, comprising:

[0032] A data acquisition module is used to acquire power data of the battery to be tested under multiple cycle numbers;

[0033] The state prediction module is used to input power data into the target health state prediction network to obtain the health state value of the battery to be tested; the target health state prediction network is obtained by training the initial health state prediction network according to the training data set; the training data set is used to train the prediction accuracy of the initial health state prediction network during the training process.

[0034] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method in any one of the embodiments of the first aspect are implemented.

[0035] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method in any one of the embodiments of the first aspect above.

[0036] In a fifth aspect, the present application further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the method in any one of the embodiments of the first aspect above.

[0037] The above-mentioned battery health state prediction method, device, equipment, medium and program product obtain the power data of the battery to be tested under multiple cycle numbers, and input the power data into the target health state prediction network to obtain the health state value of the battery to be tested. Among them, the target health state prediction network is obtained by training the initial health state prediction network according to the training data set; the training data set is used to train the network prediction accuracy of the initial health state prediction network during the training process. In this method, on the basis of obtaining the power data under multiple cycle numbers, the target health state prediction network is used to predict it to obtain the predicted health state value. Such a prediction process avoids obtaining the power data of the battery to be tested in all cycle numbers, which greatly reduces the waste of power resources. In addition, the method of predicting by the target health state prediction network shortens the detection time, avoids the interference of human testing factors, and improves the prediction speed of the health state. Furthermore, the present application uses the training data set to train the network prediction accuracy of the initial health state prediction network during the training process, which ensures the accuracy of the health state value of the battery to be tested output by the target health state prediction network. In summary, by adopting the battery health status prediction method of the embodiment of the present application, while saving battery testing resources, the battery health status value can be obtained quickly and accurately, thereby improving battery testing efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0039] FIG1 is a diagram showing an application environment of a method for predicting a battery health state in one embodiment;

[0040] FIG2 is a schematic diagram of a flow chart of a method for predicting a battery health state in one embodiment;

[0041] FIG3 is a schematic diagram of a flow chart of a power data acquisition step in one embodiment;

[0042] FIG4 is a schematic diagram of a flow chart of a network training step in one embodiment;

[0043] FIG5 is a schematic diagram of a flow chart of a network training step in another embodiment;

[0044] FIG6 is a schematic diagram of a flow chart of a network training step in another embodiment;

[0045] FIG7 is a schematic diagram of model training parameters in one embodiment;

[0046] FIG8 is a schematic diagram of a flow chart of a network verification step in one embodiment;

[0047] FIG9 is a schematic flow chart of a method for predicting a battery health state in another embodiment;

[0048] FIG10 is a schematic diagram of verifying network training results in one embodiment;

[0049] FIG11 is a schematic diagram of verifying network training results in another embodiment;

[0050] FIG12 is a schematic diagram of verifying network training results in another embodiment;

[0051] FIG13 is a schematic diagram of verifying network training results in another embodiment;

[0052] FIG14 is a schematic diagram of verifying network training results in another embodiment;

[0053] FIG15 is a structural block diagram of a battery health status prediction device in one embodiment;

[0054] FIG. 16 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0056] The battery health status prediction method provided in the embodiment of the present application can be applied in the application environment shown in Figure 1. The computer device 102 is connected to the data acquisition device 104, and the data acquisition device 104 can collect the power data of the battery and transmit it to the computer device 102. The computer device 102 processes the power data of the battery to obtain the health status value of the battery.

[0057] With the rapid development of new energy technologies, batteries are increasingly used in various fields, such as electric vehicles, energy storage, etc. In this context, battery developers need to consider many factors when designing battery products. Among them, the battery health status value, that is, the life cycle, is one of the most critical indicators of battery design projects.

[0058] Taking energy storage batteries as an example, during the design and mass production of energy storage batteries, in order to ensure that the project requirements are met and the qualified standards are achieved, a portion of the batteries to be tested are usually sampled from the batch-generated energy storage batteries, and the batteries to be tested are subjected to long-term and large-scale testing, namely, cycle testing, to obtain the life of the sampled batteries, and use this to evaluate the life cycle of the entire batch of batteries.

[0059] However, in actual testing, given the characteristics of high energy, long test time and long cycle life of energy storage battery cells, a complete test cycle of charging and discharging energy storage batteries at 25℃ 0.5P often takes 3 to 4 years, which will consume a lot of manpower and material resources; and, in the case of a long test cycle, there is a probability of unstable testing, which will also have a great impact on the accuracy of the test results. Obviously, such a testing process not only slows down the R&D efficiency of new products, but also causes a huge waste of electricity resources.

[0060] In summary, how to reduce the test time and power resources required for battery life detection while ensuring the test accuracy is a technical problem that needs to be solved in this field. To this end, the embodiment of the present application proposes to predict the power data of the battery to be tested under multiple cycle numbers through the target health state prediction network to obtain the health state value of the battery to be tested. On the one hand, the prediction method avoids obtaining the power data of the battery to be tested in all cycle numbers, which greatly reduces the waste of power resources. On the other hand, the prediction method through the target health state prediction network shortens the detection time, avoids the interference of factors of human testing, and improves the prediction speed of the health state. Further, the present application uses a training data set to train the network prediction accuracy of the initial health state prediction network during the training process, and uses a verification data set to verify the correlation of the initial health state prediction network during the training process. The target health state prediction network obtained in this way has the characteristics of high prediction accuracy and strong correlation between the prediction result and the real result, which further ensures the accuracy of the health state value of the battery to be tested output by the target health state prediction network. In summary, the battery health state prediction method of the embodiment of the present application can save battery testing resources while obtaining the health state value of the battery quickly and accurately, thereby improving the battery testing efficiency.

[0061] In an exemplary embodiment, as shown in FIG2 , a method for predicting the health status of a battery is provided, which is described by taking the method applied to a computer device as an example, and includes the following steps:

[0062] S201, obtaining power data of a battery to be tested under multiple cycle numbers.

[0063] The battery to be tested refers to a secondary battery that can be charged and discharged for multiple cycles, and may be a nickel-cadmium battery, a nickel-hydrogen battery, a lithium iron phosphate battery, a polymer lithium-ion battery, or the like.

[0064] The power data of the battery to be tested may be charging data of the battery to be tested under multiple cycles, and / or discharging data, such as charging time, charging temperature difference, temperature value under a preset charging time period, discharging time, discharging temperature difference, temperature value under a preset discharging time period.

[0065] In practical applications, the power data of the battery to be tested is obtained during the constant power charge and discharge cycle test of the battery to be tested, including the power data of the battery to be tested under a fixed number of cycles during the test process. For example, the power data of the battery to be tested under the first 50 cycles, or the power data of the battery to be tested under the 10th to 60th cycles.

[0066] It should be noted that, considering that the greater the number of cycles, the more power resources are required to test the battery to be tested, the multiple cycle numbers in this embodiment can be multiple cycle numbers preset in the test process of the battery to be tested to save power resources to the greatest extent.

[0067] For example, the battery to be tested can be connected to multiple different types of sensors, including but not limited to temperature sensors, cycle counting sensors, and timing sensors, and the computer device can be communicatively connected to each sensor. In this way, the computer device can interact with each sensor to collect the charge and discharge data of the battery to be tested under multiple cycle numbers in real time or periodically, construct the charge and discharge data curve of the battery to be tested under different cycle numbers, and then determine the power data of the battery to be tested under multiple cycle numbers from each charge and discharge data curve.

[0068] Exemplarily, the power data of the battery to be tested can be pre-stored in the memory space, and the computer device extracts the power data of the battery to be tested under multiple cycle numbers from the content space according to the specified starting cycle number and ending cycle number. For example, the power data under 1 to 100 cycle numbers are pre-stored in the memory space of the computer device. If the starting cycle number is 1 and the ending cycle number is 50, the power data under 1 to 50 cycle numbers are extracted from the memory space.

[0069] S202, inputting the power data into the target health status prediction network to obtain the health status value of the battery to be tested; the target health status prediction network is obtained by training the initial health status prediction network according to the training data set; the training data set is used to train the network prediction accuracy of the initial health status prediction network during the training process.

[0070] Among them, the state of health (SOH) value refers to a quantitative indicator that describes battery performance and life. It can refer to the total number of cycles corresponding to the battery capacity dropping from 100% to the preset capacity threshold during multiple cycles of charge and discharge testing, that is, the total number of cycles of the battery. The target health state prediction network refers to a deep learning neural network that characterizes the logical relationship between the battery's power data and the overall predicted health state value, such as the back propagation (BP) neural network, the recurrent neural network (RNN), and so on.

[0071] In the embodiment of the present application, the target health status prediction network is a network whose network prediction accuracy training of the initial health status prediction network has been completed and verified.

[0072] In one scenario, the prediction accuracy and verification index can be two quantitative index values ​​that characterize the degree of closeness between the output results of the target health status prediction network and the actual results, such as accuracy and correlation, to ensure the effectiveness of the prediction results of the target health status prediction network from the two dimensions of accuracy and correlation.

[0073] Optionally, the network prediction accuracy of the initial health status prediction network is trained according to the training data set, and then the trained initial health status prediction network is verified for relevance using the verification data set, and the initial health status prediction network that passes the relevance verification is determined as the target health status prediction network.

[0074] In another scenario, the prediction accuracy and the verification index can be the same quantitative index that characterizes the degree of proximity between the output result of the target health state prediction network and the actual result. In this case, the training process of the target health state prediction network is: the network prediction accuracy of the initial health state prediction network is trained according to the training data set, and then the verification data set is input into the trained initial health state prediction network to obtain the prediction accuracy of the verification data set, and then when the prediction accuracy of the verification data set is greater than the preset accuracy threshold, it is determined that the training of the initial health state prediction network is completed, and the target health state prediction network is obtained.

[0075] After acquiring the power data of the battery to be tested, a pre-trained target health state prediction model is called to predict the health state value of the power data, and the prediction result of the target health state prediction model is used as the health state value of the battery to be tested.

[0076] The above-mentioned battery health state prediction method obtains power data of the battery to be tested under multiple cycle numbers, and inputs the power data into the target health state prediction network to obtain the health state value of the battery to be tested. Among them, the target health state prediction network is obtained by training the initial health state prediction network according to the training data set; the training data set is used to train the network prediction accuracy of the initial health state prediction network during the training process. In this method, on the basis of obtaining power data under multiple cycle numbers, the target health state prediction network is used to predict it to obtain the predicted health state value. Such a prediction process avoids obtaining the power data of the battery to be tested in all cycle numbers, which greatly reduces the waste of power resources. In addition, the method of predicting by the target health state prediction network shortens the detection time, avoids the interference of factors of human testing, and improves the prediction speed of the health state. Further, the present application uses the training data set to train the network prediction accuracy of the initial health state prediction network during the training process, and ensures the accuracy of the health state value of the battery to be tested output by the target health state prediction network. In summary, the battery health state prediction method of the embodiment of the present application can save battery testing resources while also obtaining the health state value of the battery quickly and accurately, thereby improving the efficiency of battery testing.

[0077] Next, another possible implementation method of the power data in the above embodiment is described.

[0078] In an exemplary embodiment, as shown in FIG3 , obtaining power data of a battery to be tested under multiple cycle numbers includes:

[0079] S301, obtaining the discharge time, discharge capacity and charging temperature change value of the battery to be tested at each cycle number.

[0080] It should be noted that a complete cycle includes a set of charging process and discharging process. In the embodiment of the present application, for any cycle of the battery to be tested, the discharge time and discharge capacity corresponding to the discharge process under the cycle are obtained, and the charging temperature change value corresponding to the charging process under the cycle is obtained.

[0081] Among them, the discharge time refers to the time required for the tested battery to release all the electricity under this number of cycles, the discharge capacity refers to the total amount of electricity released by the tested battery under this number of cycles, and the charging temperature change value refers to the temperature rise value corresponding to the full charge of the tested battery during the charging process under this number of cycles.

[0082] In practical applications, the battery to be tested can be cycled and charged and discharged in accordance with the test method of "stationary-constant power charging-stationary-constant power discharging". During the test, the discharge time, discharge capacity and temperature value of the battery to be tested at each cycle number are obtained at a sampling frequency of 30 s / point, and the discharge time-discharge capacity curve and discharge time-discharge temperature curve at each cycle number are constructed. Then, the discharge time and discharge capacity of the battery to be tested at the same cycle number are obtained from the discharge time-discharge capacity curve corresponding to the same cycle number, and the charging temperature change value of the battery to be tested at the same cycle number is obtained from the discharge time-discharge temperature curve corresponding to the same cycle number.

[0083] S302, performing dimension-upgrading processing on the discharge time, discharge capacity and charging temperature change value at each cycle number to obtain power data of the battery to be tested at each cycle number.

[0084] Dimensionality increase refers to converting data with nonlinear relationships into data with linear relationships, which can better explore the potential characteristics of the data and expand the distinction between different groups of data.

[0085] The discharge time, discharge capacity and charging temperature change values ​​corresponding to each cycle of the battery to be tested are upgraded to continuous and differentiable four-dimensional data or high-dimensional data higher than four dimensions. In the embodiment of the present application, the method of data dimensionality upgrade is not limited.

[0086] For example, the discharge time, discharge capacity and charging temperature change values ​​at each cycle number are input into the data processing model, and the three-dimensional input data is mapped to a higher-dimensional space through the data processing model to obtain the power data at each cycle number.

[0087] In the embodiment of the present application, the discharge time, discharge capacity and charging temperature change value of the battery to be tested under each cycle number are processed by dimension upgrade to obtain power data, which is equivalent to taking into account the risk of data missing during the sampling process, and ensuring the integrity of the power data through data dimension upgrade to avoid the problem that the initial health state prediction network identifies the missing data as the power data characteristic value of the battery during training, resulting in low training accuracy of the health state prediction network. In addition, the power data includes the time and capacity during the discharge process, as well as the temperature change value during the charging process, which can more comprehensively characterize the power state of the charging process and the discharge process under each cycle number to be tested.

[0088] In the process of predicting the battery health status, the battery power data is predicted by the pre-trained target health status prediction network to obtain the battery health status value. It can be seen that the network quality of the target health status prediction network is crucial to the accuracy of the health status value result of the battery to be tested. Based on this, the training process of the target health status prediction network is described below through an embodiment.

[0089] In an exemplary embodiment, as shown in FIG4 , the training process of the target health status prediction network includes:

[0090] S401, obtaining a battery data set; the battery data set includes actual power data and actual health status values ​​of multiple batteries at each cycle number.

[0091] The battery data set includes actual power data and actual health status values ​​of multiple batteries under multiple cycle numbers. The method for obtaining the actual power data under each cycle number is the same as the method for obtaining the power data of the battery to be tested in the aforementioned embodiment. For details, please refer to the description of the embodiment shown in FIG3 above, which will not be repeated here.

[0092] The actual health status value refers to the total number of cycles corresponding to the battery capacity dropping from 100% to the preset capacity threshold during multiple cycles of charge and discharge testing. This means that the actual number of test cycles of multiple batteries in the battery data set needs to meet the battery capacity change conditions, and the actual number of cycles is much larger than the number of cycles corresponding to the power data of the battery to be tested.

[0093] Optionally, the preset capacity threshold is set based on an empirical value. In this case, the battery capacity corresponding to each cycle number of the battery is recorded, and the number of the cycle number corresponding to the first time the battery capacity is equal to or less than the preset capacity threshold is determined as the actual health status value of the battery.

[0094] Optionally, the preset capacity threshold is set according to a ratio of the initial battery capacity of the battery, such as 70% of the initial capacity. In this case, the battery capacity corresponding to each cycle number of the battery is recorded, and the number of the cycle number corresponding to the first time the battery capacity is equal to or less than 70% of the initial capacity is determined as the actual health status value of the battery.

[0095] In one embodiment, Charge_ori represents a battery data set, Charge_exe represents actual power data, and Charge_SOH represents an actual health status value. The relationship between the three is: Charge_ori includes Charge_exe and Charge_SOH, and the battery corresponding to Charge_exe and the battery corresponding to Charge_SOH are the same battery.

[0096] S402, dividing the battery data set to obtain a training data set and a verification data set.

[0097] The battery data set is divided according to the preset division ratio to obtain a training data set and a verification data set.

[0098] Continuing to take Charge_ori as an example, the battery data set is divided according to a 6:4 ratio, 60% of the battery data set is determined as the training data set (Charge_ori_F), and the battery data set excluding the training data set, that is, 40% of the battery data set is determined as the verification data set (Charge_ori_S).

[0099] S403: Based on the training data set and the verification data set, the initial health status prediction network is trained to obtain a target health status prediction network.

[0100] It should be noted that the parameters of the initial health status prediction network include parameters and hyperparameters. Parameters are variables learned by the network during the training process, such as weights, biases, etc.; hyperparameters are data that control the network training process, such as the number of iterations, the number of network layers, the learning rate, etc.

[0101] In this case, the training of the initial health status prediction network can be divided into two stages: in the first training stage, the hyperparameters of the initial health status prediction network are pre-set and frozen, the training data set is input into the initial health status prediction network, and the parameters in the initial health status prediction network are trained; in the second training stage, the parameters in the trained initial health status prediction network are frozen, and the hyperparameters in the trained initial health status prediction network are fine-tuned according to the verification data set to obtain the target health status prediction network.

[0102] In an embodiment of the present application, the initial health status prediction network is trained in stages based on a training data set and a verification data set, so that the target health status prediction network obtained by training can better learn the logical relationship between the power data of multiple batteries and the battery health status values, thereby improving the effectiveness of the target health status prediction network.

[0103] In an exemplary embodiment, as shown in FIG5 , the initial health state prediction network is trained based on the training data set and the validation data set to obtain the target health state prediction network, including:

[0104] S501, training an initial health status prediction network according to a training data set to obtain a candidate health status prediction network.

[0105] The training data set is input into the initial health state prediction network, and the initial health state prediction network is instructed to learn the logical relationship between the power data and the health state value in the initial data set until the training termination condition is met, thereby obtaining a candidate health state prediction network. The training termination condition may be the number of iterations, the prediction error, etc., which is not limited in the embodiment of the present application.

[0106] S502, performing relevance verification on the candidate health status prediction network according to the verification data set to obtain a relevance verification result.

[0107] In practical applications, the performance of the candidate health status prediction network on the training data set meets the prediction requirements and can be used to predict the health status of the battery to be tested. However, considering that the training data set may have too many negative samples, invalid samples, and relatively one-sided characteristics, resulting in the problem of deviation in the prediction logic of the candidate health status prediction network, the embodiment of the present application further uses verification data different from the training data set to verify the relevance of the candidate health status prediction network.

[0108] It should be noted that the validation data set includes the actual power data and actual health status values ​​of multiple batteries at each cycle number. Correspondingly, the correlation verification result obtained by performing correlation verification on the candidate health status prediction network based on the validation data set refers to the degree of closeness between the prediction results of each battery in the validation data set and the actual verification results. The larger the correlation verification result, the better the performance of the candidate health status prediction network.

[0109] S503: When the correlation verification result is verified as passed, the candidate health status prediction network is determined as the target health status prediction network.

[0110] If the correlation verification of the candidate health state prediction network passes, it means that the prediction accuracy of the candidate health state prediction network on the training data set and the verification data set meets the requirements, then the candidate health state prediction network is determined as the target health state prediction network.

[0111] In an embodiment of the present application, an initial health status prediction network is trained according to a training data set to obtain a candidate health status prediction network, so that the candidate health status prediction network learns the logical relationship between power data and health status values. The candidate health status prediction network is then verified according to a verification data set, the logical relationship of the candidate health status prediction network is verified, and the verified candidate health status prediction network is determined as a target health status prediction network to improve the accuracy of the target health status prediction network.

[0112] In an exemplary embodiment, the battery health status prediction method further includes:

[0113] When the correlation verification result is verification failure, the battery data set is re-divided to obtain a training data set and a verification data set, and the candidate health state prediction network is trained according to the divided training data set until the correlation verification result of the candidate health state prediction network is verification passed, thereby obtaining the target health state prediction network.

[0114] If the correlation verification of the candidate health status prediction network fails, it means that the candidate health status prediction network meets the requirements in the training data set, but the prediction accuracy on the verification data set does not meet the requirements. At this time, the correlation verification result is determined to be verification failure.

[0115] Furthermore, when the correlation verification result is verification failure, the battery data set can be re-divided according to a preset ratio (or randomly) to obtain a training data set and a verification data set, and the candidate health status prediction network is trained according to the latest divided training data set, and the correlation of the trained candidate health status prediction network is verified according to the latest divided verification data set. The above steps are repeated until the correlation verification result of the candidate health status prediction network is verification passed, and the target health status prediction network is obtained.

[0116] In an embodiment of the present application, when the correlation verification result is failure to pass, the battery data set is re-divided and the candidate health status prediction network is trained until the correlation verification of the candidate health status prediction network passes. This is equivalent to using different training data sets and verification data sets each time to train and verify the network that failed the previous verification, so as to continuously improve the network accuracy and obtain a more accurate and effective target health status prediction network.

[0117] In an exemplary embodiment, the training data set includes training power data and training health status values; an implementable method of the aforementioned step S501 "training the initial health status prediction network according to the training data set to obtain a candidate health status prediction network" is described, as shown in FIG6, including:

[0118] S601, inputting training power data into an initial health status prediction network to obtain a predicted health status value of the training data set.

[0119] The training power data and the training health state value are input into the candidate health state prediction network, and the initial health state prediction network is instructed to train with the training health state value as the training target, and the network training setting parameters are shown in Figure 7. It should be noted that the various parameter values ​​shown in Figure 7 are schematic values. In actual applications, other values ​​can be flexibly set according to their own needs.

[0120] After the initial health state prediction network training is completed, the output result obtained by the initial health state prediction network for predicting each battery in the training data set is obtained, that is, the predicted health state value of the training data set.

[0121] S602, obtaining error data between the predicted health status value and the training health status value of the training data set.

[0122] The predicted health state value refers to the output result obtained by the initial health state prediction network for each battery in the training data set. The training data set includes multiple batteries, and the predicted health state value of the corresponding training data set also includes the prediction results of the initial health state prediction network for multiple batteries.

[0123] Furthermore, the absolute value of the error between the predicted health state value and the training health state value of each battery in the training data set is calculated, that is, the error data of each battery. In other words, the error data in this embodiment refers to the distance between the predicted health state value and the training health state value.

[0124] S603: If the error data is less than a preset error threshold, it is determined that the initial health status prediction network training is completed, and a candidate health status prediction network is obtained.

[0125] The error data is negatively correlated with the prediction accuracy of the initial health state prediction network. The smaller the error data, the closer the predicted health state value is to the training health state value, and the higher the prediction accuracy of the initial health state prediction network; the larger the error data, the farther the predicted health state value is from the training health state value, and the lower the prediction accuracy of the initial health state prediction network.

[0126] When the error data of each battery in the training data set is less than the preset error threshold, the trained initial health state prediction network is determined as a candidate health state prediction network, otherwise, the initial health state prediction network is continuously instructed to perform more rounds of training.

[0127] The preset error threshold is a preset minimum error value. Taking the preset error threshold of 0.001 as an example, if the error data of each battery in the training data set are all less than 0.001, it is determined that the initial health status prediction network training is completed, and a candidate health status prediction network is obtained, which can be represented by NET_F.

[0128] In an embodiment of the present application, based on a preset error threshold, the error data of each battery in the training data set is evaluated separately, which is equivalent to quantitatively evaluating the prediction accuracy of the initial health status prediction network for each battery in the training data set, and when the error data is less than the preset error threshold, it is determined that the prediction accuracy verification of the initial health status prediction network has passed, and a candidate health status prediction network is obtained to ensure that the prediction accuracy of the candidate health status prediction network in the training data set meets the requirements.

[0129] In an exemplary embodiment, the verification data set includes verification power data and verification health status value; an implementable method of the aforementioned step S502 "performing correlation verification on the candidate health status prediction network according to the verification data set to obtain a correlation verification result" is described, as shown in FIG8, including:

[0130] S801, inputting the verification power data into the candidate health status prediction network to obtain the predicted health status value of the verification data set.

[0131] The verification power data in the verification data set is input into the candidate health state prediction network, and the candidate health state prediction network predicts the input to obtain the predicted health state value of the verification data set.

[0132] It should be noted that the validation data set includes multiple batteries, and the predicted health status values ​​of the corresponding validation data set also include the prediction results of the candidate health status prediction network for the multiple batteries.

[0133] S802, obtaining a correlation coefficient between the predicted health status value and the verified health status value.

[0134] The predicted health status value is subtracted from the verified health status value to calculate the correlation coefficient between the two.

[0135] The correlation coefficient is R 2 For example, the corresponding calculation formula is:

[0136]

[0137] In the above formula, represents the verification health status value of the i-th battery in the verification data set, represents the predicted health status value of the candidate health status prediction model for the i-th battery in the validation data set, Represents the mean of the validation health status values ​​of each battery in the validation dataset.

[0138] It is emphasized again that the error data corresponding to the training data set includes multiple, which are the absolute values ​​of the errors of each battery in the training data set; and the correlation coefficient corresponding to the validation data set is one, which refers to a comprehensive coefficient value corresponding to each battery in the validation data set.

[0139] S803: If the correlation coefficient is greater than a preset correlation threshold, the correlation verification result is determined to be verified passed.

[0140] The correlation coefficient is positively correlated with the prediction accuracy of the candidate health status prediction network. The larger the correlation coefficient, the higher the prediction accuracy of the candidate health status prediction network on the validation data set; the smaller the correlation coefficient, the lower the prediction accuracy of the candidate health status prediction network on the validation data set.

[0141] Based on this, the correlation coefficient is compared with the preset correlation threshold. If the correlation coefficient is greater than the preset correlation threshold, it indicates that the candidate health status prediction network has a high prediction accuracy on the verification data set, and the correlation verification result is determined to be verified passed.

[0142] S804: If the correlation coefficient is not greater than the preset correlation threshold, determine that the correlation verification result is verification failure.

[0143] If the correlation coefficient is greater than the preset correlation threshold, it indicates that the prediction accuracy of the candidate health status prediction network on the verification data set is low, and the correlation verification result is determined to be verification failure.

[0144] In an embodiment of the present application, the prediction accuracy of the candidate health status prediction network on the verification data set is quantitatively evaluated by comparing the preset correlation threshold and the correlation coefficient, and when the correlation coefficient is less than the preset correlation threshold, it is determined that the prediction accuracy verification of the candidate health status prediction network has passed, and the target health status prediction network is obtained, so as to further ensure that the prediction accuracy of the target health status prediction network on both the training data set and the verification data set meets the requirements, thereby improving the effectiveness of the target health status prediction network.

[0145] In a specific embodiment, as shown in FIG9 , a method for predicting a battery health state is provided, including:

[0146] S901, obtaining discharge time, discharge capacity and charging temperature change values ​​of multiple batteries in the first 50 cycles.

[0147] S902, performing dimension-upgrading processing on the discharge time, discharge capacity and charging temperature change value of each battery in the first 50 cycles to obtain the power data of each battery in the first 50 cycles.

[0148] S903: Summarize the power data and actual health status value of each battery to obtain a battery data set.

[0149] S904, dividing the battery data set to obtain a training data set and a verification data set.

[0150] S905, inputting the training power data in the training data set into the initial health state prediction network to obtain the predicted health state value of the training data set.

[0151] S906, obtaining error data between the predicted health status value of the training data set and the training health status value in the training data set.

[0152] S907, determining whether the error data is less than a preset error threshold.

[0153] S908, if the error data is greater than the preset error threshold, it is determined that the initial health status prediction network has not been trained, and the process returns to step S905.

[0154] S909: If the error data is less than a preset error threshold, it is determined that the initial health status prediction network training is completed, and a candidate health status prediction network is obtained.

[0155] S910, inputting the verification power data into the candidate health status prediction network to obtain the predicted health status value of the verification data set.

[0156] S911, obtaining a correlation coefficient between the predicted health status value and the verified health status value.

[0157] S912, determining whether the correlation coefficient is greater than a preset correlation threshold.

[0158] S913, if the correlation coefficient is not greater than the preset correlation threshold, the correlation verification result is determined to be verification failure, and the process returns to step S904.

[0159] S914: If the correlation coefficient is greater than a preset correlation threshold, the correlation verification result is determined to be verified passed, and the candidate health status prediction network is determined as the target health status prediction network.

[0160] S915, inputting the power data of the battery to be tested under the first 50 cycles into the target health state prediction network to obtain the health state value of the battery to be tested.

[0161] In the embodiment of the present application, the method of predicting by the target health state prediction network saves the detection time cost, avoids the interference of human testing factors, and improves the prediction speed of the health state. In addition, on the basis of obtaining the power data under the first 50 cycles, the target health state prediction network is used to predict it to obtain the predicted health state value. Such a prediction process avoids obtaining the power data of the battery to be tested in all cycles, which greatly reduces the waste of power resources. Further, the network prediction accuracy of the initial health state prediction network during the training process is trained using a training data set, and the correlation of the initial health state prediction network during the training process is verified using a verification data set. The target health state prediction network obtained in this way has the characteristics of high prediction accuracy and strong correlation between the prediction results and the real results, which further ensures the accuracy of the health state value of the battery to be tested output by the target health state prediction network. In summary, the battery health state prediction method of the embodiment of the present application can save battery testing resources while obtaining the health state value of the battery quickly and accurately, thereby improving the efficiency of battery testing.

[0162] To further characterize the prediction effect of the battery health state prediction method provided in the embodiment of the present application, in one embodiment, taking a lithium iron phosphate battery, a target health state prediction network being a BP neural network, and a health state value being a number of cycles as an example, the battery health state prediction method is described:

[0163] (1) Build a database.

[0164] With the voltage interval set at 2.5-3.65 V, the battery was subjected to a cycle test of static-constant power charging-static-constant power discharging, with a sampling point of 30 s / point. The complete charging data of the battery for each cycle at 70%-100% SOH was extracted, and the charging curve corresponding to the battery at 70%-100% SOH was plotted; the complete discharge data of the battery for each cycle at 100%-70% SOH was extracted, and the discharge curve corresponding to the battery at 100%-70% SOH was plotted.

[0165] (2) Data processing and definition.

[0166] After arranging the data sets of constant power charge and discharge tests under the same conditions of 100%~70% SOH of multiple lithium iron phosphate energy storage batteries, it is named Charge_ori, and the data set consisting of the cycle life of each group of batteries is defined as Charge_SOH_ori. Next, the first 50 weeks of charge and discharge data of the full EOL charge and discharge curve data of multiple lithium iron phosphate energy storage batteries are extracted as a data set, which is defined as Charge_exe. Since different charging curves correspond to different lifespans, the set of lifespan data cycles corresponding to each battery in the data set is defined as Charge_SOH.

[0167] (3) Construct a target health status prediction network.

[0168] In order to avoid the risk of missing numbers and resulting in low network accuracy to the greatest extent, the dataset Charge_ori is upgraded to continuous and differentiable high-dimensional data, and the dataset Charge_ori is updated to avoid the network identifying missing data as feature values ​​during training, which may cause network errors or greatly reduce the calculation time.

[0169] According to the preset ratio, Charge_ori is randomly divided into two data sets, named Charge_ori_F and Charge_ori_S, which account for 60% and 40% of the whole data set respectively.

[0170] Correspondingly, the dataset Charge_SOH will be split into Charge_SOH_F and Charge_SOH_S.

[0171] Charge_ori_F is used as data input and Charge_SOH_F as data target to train the BP neural network. The error is set to 0.001. After the training is completed, the neural network training is completed and the network is named NET_F.

[0172] The Charge_ori_S dataset is put into NET_F, and the number of cycles of the corresponding data is calculated. The set of values ​​is named Charge_SOH_M.

[0173] Subtract Charge_SOH_M from Charge_SOH_S to solve for R 2 , if R 2 If R is greater than 0.99, the training is completed and the BP neural network is obtained. 2 If it is less than 0.99, return to the data partitioning step of Charge_ori and retrain the network until the network R 2 When it is greater than 0.99, the BP neural network is obtained.

[0174] During the training of the BP neural network based on the training data set (60% of the data set), a schematic diagram of the Pearson correlation index value (R) between the predicted number of cycles of the power data of each battery and the actual number of cycles of each battery is shown in Figure 10. In Figure 10, the correlation coefficient between the predicted number of cycles output by the BP neural network and the actual number of cycles is 0.98031.

[0175] In addition, the embodiment of the present application can also verify the BP neural network based on the verification data set (40% of the data set). In the process of verifying the BP neural network based on the verification data set, a schematic diagram of the Pearson correlation index value (R) between the predicted number of cycles of the power data of each battery and the actual number of cycles of each battery by the BP neural network is shown in Figure 11. In Figure 11, the correlation coefficient between the predicted number of cycles output by the BP neural network and the actual number of cycles is 0.97804.

[0176] Next, after the BP neural network training and verification are completed, a batch of test data sets are used to test the prediction accuracy of the BP neural network. A schematic diagram of the Pearson correlation index value (R) between the predicted number of cycles of the power data of each battery by the BP neural network and the actual number of cycles of each battery is shown in Figure 12. In Figure 12, the correlation coefficient between the predicted number of cycles output by the BP neural network and the actual number of cycles is 0.97425.

[0177] Finally, the training data set, validation data set and test data set are taken as the overall data set and input into the trained BP neural network. A schematic diagram of the Pearson correlation index value (R) between the predicted number of cycles of the power data of each battery and the actual number of cycles of each battery by the BP neural network is shown in FIG13 . In FIG13 , the correlation coefficient between the predicted number of cycles and the actual number of cycles of the BP neural network for the overall data set is 0.97899.

[0178] (4) Predict the number of cycles of the battery to be tested.

[0179] The charge and discharge data of the battery to be tested under the first 50 cycles are input into the BP neural network to obtain the cycle number of the battery to be tested.

[0180] FIG14 is a schematic diagram showing the comparison between the predicted values ​​and the true values ​​of the BP neural network for 10 batteries to be tested. In FIG14 , the correlation coefficient (R 2 ) is 0.99898. It can be seen that the BP neural network trained by the present application has a high prediction accuracy for the battery to be tested, and can be widely used in the field of battery health status value prediction to save a lot of electricity and detection time costs. While quickly detecting the life span, it can also improve the accuracy, pave the way for subsequent batch processing of battery data, and greatly improve the detection efficiency.

[0181] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0182] Based on the same inventive concept, the embodiment of the present application also provides a battery health state prediction device for implementing the battery health state prediction method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more battery health state prediction device embodiments provided below can refer to the limitations of the battery health state prediction method above, and will not be repeated here.

[0183] In an exemplary embodiment, as shown in FIG15 , a battery health status prediction device is provided, including: a data acquisition module 1501 and a status prediction module 1502, wherein:

[0184] The data acquisition module 1501 is used to acquire the power data of the battery to be tested under multiple cycle numbers;

[0185] The state prediction module 1502 is used to input the power data into the target health state prediction network to obtain the health state value of the battery to be tested; the target health state prediction network is obtained by training the initial health state prediction network according to the training data set; the training data set is used to train the prediction accuracy of the initial health state prediction network during the training process.

[0186] In an exemplary embodiment, the data acquisition module 1501 further includes: a parameter acquisition unit and a data dimension upgrading unit, wherein:

[0187] A parameter acquisition unit, used to obtain the discharge time, discharge capacity and charging temperature change value of the battery to be tested at each cycle number;

[0188] The data dimension upgrading unit is used to perform dimension upgrading on the discharge time, discharge capacity and charging temperature change value under each cycle number to obtain the power data of the battery to be tested under each cycle number.

[0189] In an exemplary embodiment, the battery health status prediction device further includes:

[0190] A data set acquisition module is used to acquire a battery data set; the battery data set includes actual power data and actual health status values ​​of multiple batteries at each cycle number;

[0191] A data set partitioning module is used to partition the battery data set to obtain a training data set and a verification data set;

[0192] The network training module is used to train the initial health status prediction network based on the training data set and the verification data set to obtain the target health status prediction network.

[0193] In an exemplary embodiment, the network training module includes: a parameter training unit, a network verification unit and a network determination unit, wherein:

[0194] A parameter training unit, used for training the initial health status prediction network according to the training data set to obtain a candidate health status prediction network;

[0195] A network verification unit, used to perform relevance verification on the candidate health status prediction network according to the verification data set to obtain a relevance verification result;

[0196] The network determination unit is used to determine the candidate health status prediction network as the target health status prediction network when the correlation verification result is verification passed.

[0197] In an exemplary embodiment, the parameter training unit is also used to input the training power data into the initial health status prediction network to obtain the predicted health status value of the training data set; obtain the error data between the predicted health status value of the training data set and the training health status value; when the error data is less than a preset error threshold, determine that the training of the initial health status prediction network is completed, and obtain a candidate health status prediction network.

[0198] In an exemplary embodiment, the network verification unit is also used to input the verification power data into the candidate health status prediction network to obtain the predicted health status value of the verification data set; obtain the correlation coefficient between the predicted health status value and the verified health status value; if the correlation coefficient is greater than a preset correlation threshold, determine that the correlation verification result is verified passed; if the correlation coefficient is not greater than the preset correlation threshold, determine that the correlation verification result is verified failed.

[0199] In an exemplary embodiment, the network determination unit is also used to, when the correlation verification result is verification failure, re-divide the battery data set to obtain a training data set and a verification data set, and train the candidate health status prediction network according to the divided training data set until the correlation verification result of the candidate health status prediction network is verification passed, thereby obtaining the target health status prediction network.

[0200] Each module in the above-mentioned battery health status prediction device can be implemented in whole or in part by software, hardware and their combination. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the corresponding operations of each of the above modules.

[0201] In an exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be shown in FIG16. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. The processor, the memory, and the input / output interface are connected via a system bus, and the communication interface, the display unit, and the input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be implemented through WIFI, a mobile cellular network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, a battery health state prediction method is implemented. The display unit of the computer device is used to form a visually visible picture, which may be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0202] Those skilled in the art will understand that the structure shown in FIG. 16 is merely a block diagram of a partial structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0203] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0204] In an exemplary embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0205] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0206] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0207] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0208] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0209] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for predicting battery health status, characterized in that: The method comprises: Obtaining power data of the battery under test under multiple cycle numbers; The power data is input into a target health status prediction network to obtain the health status value of the battery to be tested; the target health status prediction network is obtained by training an initial health status prediction network according to a training data set; the training data set is used to train the network prediction accuracy of the initial health status prediction network during the training process.

2. The method according to claim 1, characterized in that The step of obtaining power data of the battery to be tested under multiple cycle numbers includes: Obtaining the discharge time, discharge capacity and charging temperature change value of the battery to be tested under each of the cycle numbers; The discharge time, the discharge capacity and the charging temperature change value at each of the cycle numbers are subjected to dimensionality upgrading processing to obtain power data of the battery to be tested at each of the cycle numbers.

3. The method according to claim 1 or 2, characterized in that: The training process of the target health status prediction network includes: Acquire a battery data set; the battery data set includes actual power data and actual health status values ​​of multiple batteries at each of the cycle numbers; Dividing the battery data set to obtain the training data set and the verification data set; Based on the training data set and the verification data set, an initial health state prediction network is trained to obtain the target health state prediction network.

4. The method according to claim 3, characterized in that The step of training the initial health status prediction network based on the training data set and the verification data set to obtain the target health status prediction network includes: Training the initial health status prediction network according to the training data set to obtain a candidate health status prediction network; Performing relevance verification on the candidate health status prediction network according to the verification data set to obtain a relevance verification result; When the correlation verification result is verified to be passed, the candidate health state prediction network is determined as the target health state prediction network.

5. The method according to claim 4, characterized in that The training data set includes training power data and training health status values; The step of training the initial health status prediction network according to the training data set to obtain a candidate health status prediction network includes: Inputting the training power data into the initial health state prediction network to obtain a predicted health state value of the training data set; Obtain error data between the predicted health status value of the training data set and the training health status value; If the error data is less than a preset error threshold, it is determined that the initial health status prediction network training is completed, and the candidate health status prediction network is obtained.

6. The method according to claim 4, characterized in that The verification data set includes verification power data and verification health status value; the correlation verification of the candidate health status prediction network is performed according to the verification data set to obtain the correlation verification result, including: Inputting the verification power data into the candidate health state prediction network to obtain a predicted health state value of the verification data set; Obtaining a correlation coefficient between the predicted health status value and the verified health status value; If the correlation coefficient is greater than a preset correlation threshold, determining that the correlation verification result is verified passed; If the correlation coefficient is not greater than the preset correlation threshold, the correlation verification result is determined to be verification failure.

7. The method according to claim 4, characterized in that The method further comprises: When the correlation verification result is verification failure, the battery data set is re-divided to obtain a training data set and a verification data set, and the candidate health status prediction network is trained according to the divided training data set until the correlation verification result of the candidate health status prediction network is verification passed, thereby obtaining the target health status prediction network.

8. A battery health status prediction device, characterized in that: The device comprises: A data acquisition module is used to acquire power data of the battery to be tested under multiple cycle numbers; A state prediction module is used to input the power data into a target health state prediction network to obtain the health state value of the battery to be tested; the target health state prediction network is obtained by training an initial health state prediction network according to a training data set; the training data set is used to train the prediction accuracy of the initial health state prediction network during the training process.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

11. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images