Lithium battery state prediction method and storage medium

Abstract

The invention discloses a lithium battery state prediction method and a storage medium, and the method comprises the steps: obtaining the battery operation data and a state recognition model of a target lithium battery at a plurality of moments, and enabling the state recognition model to be obtained based on the training of a deep learning model, the state recognition model at least comprises a gating module and a multi-task prediction module, the gating module is used for determining the weight of input features, and the multi-task prediction module is used for predicting more than two battery state parameters; and inputting the battery operation data into a state recognition model to obtain prediction parameters related to the lithium battery operation state, the prediction parameters at least comprising the residual life, the health state and the charge state, the efficiency and accuracy of battery management are improved, optimization of battery use and maintenance is assisted, and thus the service life of the battery is prolonged.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium battery health detection, and particularly relates to a method for predicting the state of a lithium battery and a storage medium. Background Art

[0002] Accurately evaluating and predicting the operating state of a lithium battery faces many challenges. The electrochemical reaction mechanism inside the battery is complex, and the operating state is affected by multiple factors such as temperature, charge and discharge current, aging degree, state of charge, and state of health. Existing related technologies are difficult to comprehensively capture the complex non-linear relationships among the influencing factors of the lithium battery operating state. Traditional physical model-based methods can construct models starting from the internal reaction mechanism of the battery, but due to the complexity of the internal reactions of the battery, it is difficult to accurately obtain and update model parameters in real time, and there are significant limitations in practical applications. To address the above technical problems, it is urgent to develop a more mature method for predicting the state of a lithium battery. Summary of the Invention

[0003] The present invention provides a method for predicting the state of a lithium battery and a storage medium to solve the problem in related technologies that the prediction model can only predict a single battery state parameter.

[0004] According to one aspect of the present invention, a method for predicting the state of a lithium battery is provided, including: Obtaining battery operation data and a state recognition model at multiple moments of a target lithium battery, where the state recognition model is trained based on a deep learning model, the state recognition model at least includes a gating module and a multi-task prediction module, the gating module is used to determine the weights of input features, and the multi-task prediction module is used to predict two or more battery state parameters; Inputting the battery operation data into the state recognition model to obtain prediction parameters related to the lithium battery operation state, where the prediction parameters at least include remaining life, state of health, and state of charge.

[0005] According to another aspect of the present invention, a device for predicting the state of a lithium battery is provided, and the device includes: An obtaining module, configured to obtain battery operation data and a state recognition model at multiple moments of a target lithium battery, the state recognition model is trained based on a deep learning model, the state recognition model at least includes a gating module and a multi-task prediction module, the gating module is used to determine the weights of input features, and the multi-task prediction module is used to predict two or more battery state parameters; A prediction parameter determination module, configured to input the battery operation data into the state recognition model to obtain prediction parameters related to the lithium battery operation state, where the prediction parameters at least include remaining life, state of health, and state of charge.

[0006] According to another aspect of the present invention, there is provided an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the lithium battery state prediction method according to any embodiment of the present invention.

[0007] According to another aspect of the present invention, there is provided a computer-readable storage medium storing computer instructions for implementing the lithium battery state prediction method according to any embodiment of the present invention when executed by a processor.

[0008] The technical solution of the embodiment of the present invention obtains battery operation data and a state recognition model at multiple moments of a target lithium battery. Since the state recognition model is trained based on a deep learning model, the state recognition model at least includes a gating module and a multi-task prediction module. The gating module is used to determine the weights of input features, and the multi-task prediction module is used to predict more than two battery state parameters, which can provide sufficient data support and tools for the state prediction of lithium batteries; inputting the battery operation data into the state recognition model to obtain prediction parameters related to the lithium battery operation state, wherein the prediction parameters at least include remaining life, health state, and state of charge, can accurately predict multiple state parameters of the lithium battery, solves the problem that the prediction model in the related art can only predict a single battery state parameter, improves the efficiency and accuracy of battery management, helps to optimize battery use and maintenance, and thus extends the battery life.

[0009] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present invention, nor is it used to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. Description of the Drawings

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0011] Figure 1 is a flowchart of a lithium battery state prediction method according to Embodiment 1 of the present invention; Figure 2It is a schematic structural diagram of a lithium battery state prediction device provided according to Embodiment 3 of the present invention; Figure 3 It is a schematic structural diagram of a state recognition model in a lithium battery state prediction method provided according to an embodiment of the present invention; Figure 4 It is a schematic diagram of GRU of a state recognition model in a lithium battery state prediction method provided according to an embodiment of the present invention; Figure 5 It is a schematic diagram of a temporal convolutional network of a state recognition model in a lithium battery state prediction method provided according to an embodiment of the present invention; Figure 6(a) is a schematic diagram of the SOH prediction result of a state recognition model in a lithium battery state prediction method provided according to an embodiment of the present invention; Figure 6(b) is a schematic diagram of the SOH prediction result error of a state recognition model in a lithium battery state prediction method provided according to an embodiment of the present invention; Figure 6(c) is a schematic diagram of the SOH prediction result of a long short-term memory network (LSTM) lithium battery state prediction method provided according to an embodiment of the present invention; Figure 6(d) is a schematic diagram of the SOH prediction result error of a long short-term memory network (LSTM) lithium battery state prediction method provided according to an embodiment of the present invention; Figure 6(e) is a schematic diagram of the SOH prediction result of a convolutional long short-term memory fusion network (CNN-LSTM) lithium battery state prediction method provided according to an embodiment of the present invention; Figure 6(f) is a schematic diagram of the SOH prediction result error of a convolutional long short-term memory fusion network (CNN-LSTM) lithium battery state prediction method provided according to an embodiment of the present invention; Figure 6(g) is a schematic diagram of the SOH prediction result of a gated recurrent unit (GRU) lithium battery state prediction method provided according to an embodiment of the present invention; Figure 6(h) is a schematic diagram of the SOH prediction result error of a gated recurrent unit (GRU) lithium battery state prediction method provided according to an embodiment of the present invention; Figure 6(i) is a schematic diagram of the SOH prediction result of a recurrent neural network (RNN) lithium battery state prediction method provided according to an embodiment of the present invention; Figure 6(j) is a schematic diagram of the SOH prediction result error of a recurrent neural network (RNN) lithium battery state prediction method provided according to an embodiment of the present invention; Figure 7(a) is a schematic diagram of the RUL prediction result of a state recognition model in a lithium battery state prediction method provided according to an embodiment of the present invention; Figure 7(b) is a schematic diagram of the RUL prediction result error of a state recognition model in a lithium battery state prediction method provided according to an embodiment of the present invention; Figure 7(c) is a schematic diagram of the RUL prediction result of the long short-term memory network (LSTM) lithium battery state prediction method provided by an embodiment of the present invention; Figure 7(d) is a schematic diagram of the RUL prediction result error of the long short-term memory network (LSTM) lithium battery state prediction method provided by an embodiment of the present invention; Figure 7(e) is a schematic diagram of the RUL prediction result of the convolutional long short-term memory fusion network (CNN-LSTM) lithium battery state prediction method provided by an embodiment of the present invention; Figure 7(f) is a schematic diagram of the RUL prediction result error of the convolutional long short-term memory fusion network (CNN-LSTM) lithium battery state prediction method provided by an embodiment of the present invention; Figure 7(g) is a schematic diagram of the RUL prediction result of the gated recurrent unit (GRU) lithium battery state prediction method provided by an embodiment of the present invention; Figure 7(h) is a schematic diagram of the RUL prediction result error of the gated recurrent unit (GRU) lithium battery state prediction method provided by an embodiment of the present invention; Figure 7(i) is a schematic diagram of the RUL prediction result of the recurrent neural network (RNN) lithium battery state prediction method provided by an embodiment of the present invention; Figure 7(j) is a schematic diagram of the RUL prediction result error of the recurrent neural network (RNN) lithium battery state prediction method provided by an embodiment of the present invention; Figure 8 is a schematic diagram of the SOC prediction result in a lithium battery state prediction method provided by an embodiment of the present invention; Figure 9 is a schematic diagram of the structure of an electronic device implementing the lithium battery state prediction method of an embodiment of the present invention. Detailed implementation manners

[0012] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0013] It should be noted that the terms "first", "second", etc. in the description, claims and above-mentioned drawings of the present invention are used to distinguish similar objects, and do not necessarily need to describe a specific order or sequence. It should be understood that such used data can be interchanged under appropriate circumstances, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0014] It should be noted that the modifications of "one" and "multiple" mentioned in this disclosure are illustrative rather than restrictive. Those skilled in the art should understand that unless clearly specified otherwise in the context, it should be understood as "one or more".

[0015] The names of the messages or information exchanged between multiple devices in the embodiments of this disclosure are only for illustrative purposes and do not limit the scope of these messages or information.

[0016] It can be understood that before using the technical solutions disclosed in the embodiments of this disclosure, the types, usage scopes, usage scenarios, etc. of the personal information involved in this disclosure should be informed to the user and the user's authorization should be obtained in an appropriate manner in accordance with relevant laws and regulations.

[0017] For example, when receiving an active request from a user, a prompt message is sent to the user to clearly prompt the user that the operation requested by the user will require obtaining and using the user's personal information. Thus, the user can autonomously choose whether to provide personal information to software or hardware such as an electronic device, application program, server or storage medium that executes the operations of the technical solutions of this disclosure according to the prompt message.

[0018] As an optional but non-limiting implementation manner, the way of sending a prompt message to the user in response to receiving an active request from the user can be, for example, in the form of a pop-up window. The prompt message can be presented in text in the pop-up window. In addition, the pop-up window can also carry selection controls for the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0019] It can be understood that the above process of notifying and obtaining the user's authorization is only illustrative and does not limit the implementation manners of this disclosure. Other manners that meet relevant laws and regulations can also be applied to the implementation manners of this disclosure.

[0020] It is understood that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of data) should comply with the requirements of corresponding laws, regulations and related provisions.

[0021] Embodiment 1: Figure 1 The figure is a flowchart of a lithium battery state prediction method provided in Embodiment 1 of the present invention. This embodiment is applicable to situations where real-time detection and accurate prediction of the lithium battery state are required. This method can be executed by a lithium battery state prediction device, which can be implemented in the form of hardware and / or software. Optionally, it is implemented through an electronic device, which can be a mobile terminal, a PC or a server, etc.

[0022] As Figure 1 shown, the method may specifically include: S110. Obtain the battery operation data and the state recognition model at multiple moments of the target lithium battery. Among them, the state recognition model is trained based on a deep learning model. The state recognition model at least includes a gating module and a multi-task prediction module. The gating module is used to determine the weights of input features, and the multi-task prediction module is used to predict two or more battery state parameters.

[0023] Among them, the target lithium battery can be understood as a specific lithium-ion battery as the object of detection or analysis. The target lithium battery can be used for energy storage and power supply and is an important component of electronic devices or electric vehicles, etc. By collecting and analyzing the battery operation data of the target lithium battery, the performance of the lithium battery can be evaluated and the future state can be predicted. The battery operation data can be understood as various data generated during the use of the target lithium battery, including but not limited to voltage, current and temperature, etc. The battery operation data is used to detect the working state of the lithium battery and provide basic data information for subsequent state recognition. The state recognition model can be understood as a model constructed based on deep learning technology, which is used to judge the current state of the battery according to the battery operation data. By analyzing the battery operation data, the state recognition model can predict important parameters such as the health status, remaining life and state of charge of the lithium battery, and help optimize the battery management and maintenance plan. The gating module can be understood as a mechanism in the deep learning model, such as GRU (gated recurrent unit), which is used to control the information flow. The gating module can learn to determine which input features are more important, so as to adjust the weights of the input features, which helps to improve the state recognition model's understanding ability of time series data. The multi-task prediction module can be understood as a model component that can predict multiple related output variables.

[0024] Based on the above solution, optionally, before obtaining the battery operation data of the target lithium battery at multiple moments, it further includes: obtaining the acquisition data of the target lithium battery at multiple moments, performing post-processing on the acquisition data to obtain post-processing analysis data; and determining the battery operation data according to the acquisition data and the post-processing analysis data.

[0025] Among them, the acquisition data can be understood as the data collected from the target lithium battery at different time points, including but not limited to voltage, current, temperature, etc. The post-processing includes methods such as obtaining the maximum value, average value, skewness, and kurtosis.

[0026] Based on the above method, optionally, the obtaining the acquisition data of the target lithium battery at multiple moments includes: obtaining the acquisition data through the sensors of the lithium battery pack; or, transmitting the operation data on the remote device to the central server through wireless communication technology to obtain the acquisition data; or, in response to the data upload operation, obtaining the uploaded acquisition data; or, pulling data from the preset battery operation database to obtain the acquisition data, etc., which is not specifically limited here.

[0027] Based on the above method, optionally, the acquisition data at least includes: charging voltage, discharging voltage, charging current, discharging current, rated voltage, rated capacity, charging capacity (the capacity acceptable under charging conditions), discharging capacity (the capacity that can be released under discharging conditions), charging time, discharging time, and ambient temperature, etc. corresponding to several charge-discharge cycles before the charge-discharge cycle.

[0028] Adopting this technical solution, by performing post-processing on the acquisition data of the target lithium battery, useful information can be determined, thereby accurately determining the battery operation data, which not only improves the accuracy of battery state assessment, but also enhances the reliability of predicting the remaining life and health state, helps to optimize the battery management and maintenance strategy, and extends the battery service life.

[0029] Based on the above solution, optionally, the state recognition model further includes a first feature extraction module and a second feature extraction module; the multi-task prediction module includes a first prediction unit, a second prediction unit, and a third prediction unit; the first feature extraction module, the second feature extraction module, the first prediction unit, the second prediction unit, and the third prediction unit are all connected to the gating module; and the second feature extraction module is connected to the second prediction unit.

[0030] Among them, the first feature extraction module and the second feature extraction module can be understood as part of the state recognition model, responsible for extracting features from the input data. The first prediction unit, the second prediction unit, and the third prediction unit can be understood as sub-components of the multi-task prediction module, and are respectively used to predict the remaining life, health state, and state of charge.

[0031] An optional implementation manner is as Figure 3 shown. The structural compositions of the first prediction unit, the second prediction unit, and the third prediction unit are all the same, and at least include a temporal convolutional network (the structure is as Figure 5 described) and a gated recurrent unit (GRU) (the structure is as Figure 4 shown), etc.

[0032] is the normalized feature set of the health state and the remaining service life, is the normalized feature set of the state of charge, H k , H k+1 , · · ·，H k+w-1 is the feature extracted by the shared layer from the normalized feature set, SOC t-1 is the SOC at the previous moment of the prediction target, SOC t is the SOC at the prediction target moment, SOC t+1 , and is the SOC at the next moment of the prediction target.

[0033] For the temporal convolutional network, assume that we have an input sequence , and hope to predict the corresponding output at each time point. It generates the following mapping:

[0034] To increase the receptive field without increasing the number of parameters, the TCN uses dilated convolution. Dilated convolution, also known as atrous convolution, is a form of convolution in which gaps (i.e., skipping some input units) are inserted between the convolution kernels. This technique allows the model to capture a larger receptive field without increasing the number of parameters, thereby better understanding the context information in the input data. The dilation factor determines the spacing between the elements in the convolution kernel. For example, if the dilation factor is 2, the elements in the convolution kernel will be spaced one input unit apart. The mathematical representation of dilated convolution:

[0035] Among them, f is the convolution kernel, kis the convolutional kernel size, x is the input sequence, d is the dilation rate.

[0036] For gated recurrent units, assume x t is the input information at the current time step, h t-1 is the hidden state at the previous time step. The hidden state acts as the memory of the neural network, which contains information about the data seen by previous nodes, h t is the hidden state passed to the next time step. Then we have:

[0037] where: is the candidate hidden state, is the reset gate, is the update gate, is the sigmoid function, through which data can be transformed into values in the range of 0 - 1. Tanh is the activation function, through which data can be transformed into values in the range of [-1, 1].

[0038] Adopting this technical solution, through the structured model design, each component can handle specific prediction tasks, thereby improving the performance and prediction accuracy of the overall model.

[0039] On the basis of the above solution, optionally, the objective loss in the training process of the state recognition model includes a first objective loss and a second objective loss. The first objective loss is used to adjust the model parameters of the first prediction unit and the second prediction unit, and the second objective loss is used to adjust the model parameters of the third prediction unit.

[0040] Among them, the objective loss can be understood as a metric for quantifying the error magnitude of model prediction, which is the objective to be optimized during the training process. The training process of the state recognition model aims to minimize the objective loss. The objective loss at least includes the first objective loss and the second objective loss, etc. The first objective loss can be understood as the loss function set for the first prediction unit (such as the state of charge SOC prediction) and the second prediction unit (such as the state of health SOH prediction), which is used to evaluate the accuracy of the prediction results of the two units. The second objective loss can be understood as the loss function set for the third prediction unit, which is used to evaluate the accuracy of the prediction results of the third prediction unit. The model parameters can be understood as the adjustable variables in the multi-task prediction module, and these variables determine how the multi-task prediction module converts the input into the output. During the training process, by adjusting the model parameters, the objective loss is reduced, so that the state recognition model can better fit the training data and improve the prediction accuracy.

[0041] With this technical solution, by designing the objective loss in a hierarchical manner, the state recognition model is allowed to be optimized separately for different prediction tasks, improving the pertinence and accuracy of the overall prediction performance of the state recognition model. It can not only improve the prediction effect of a single prediction parameter, but also enhance the state recognition model's understanding and adaptation ability to complex battery state changes.

[0042] Based on the above solution, optionally, the first objective loss is determined in the following manner: respectively determine a first loss related to the remaining useful life and a second loss related to the health state; determine a fourth weight and a fifth weight according to the first loss and the second loss, wherein the fourth weight and the fifth weight are positively correlated with the values of the first loss and the second loss; determine the first objective loss according to the first objective loss, the second objective loss, the fourth weight, and the fifth weight.

[0043] Among them, the first loss can be understood as a metric for measuring the accuracy of the state recognition model's prediction result for the remaining useful life (RUL). The second loss can be understood as a metric for measuring the accuracy of the state recognition model's prediction result for the state of health (SOH). The fourth weight can be understood as a weight coefficient associated with the first loss, used to balance the influence of the first loss in the first objective loss. The value of the fourth weight is positively correlated with the first loss. In the case where the first loss is large, a higher weight is determined to be assigned to emphasize the improvement requirement of the first loss. The fifth weight can be understood as a weight coefficient associated with the second loss, used to balance the influence of the second loss in the first objective loss. The value of the fifth weight is positively correlated with the second loss. In the case where the second loss is large, a higher weight is determined to be assigned to emphasize the improvement requirement of the first loss.

[0044] The values of the first loss and the second loss are determined according to the ratio to the result of the previous iteration. The loss of each prediction parameter is compared with the result of its previous iteration to calculate the ratio. In the case where the ratio is high, it indicates that the prediction parameter is more challenging to optimize. Through this method, the optimization difficulty of each prediction parameter's task can be quantitatively evaluated. In the backpropagation stage, according to the evaluation result, a larger weight factor is multiplied by the loss corresponding to the prediction task with greater optimization difficulty, thereby enhancing its role in parameter update. Correspondingly, a smaller weight factor is multiplied by the loss corresponding to the prediction task with smaller optimization difficulty to reduce its influence on parameter update.

[0045] Based on the above solution, optionally, the determination of the first objective loss according to the first objective loss, the second objective loss, the fourth weight, and the fifth weight can be determined based on the following formula:

[0046] ; wherein, represents the first target loss, represents the fourth weight, represents the first loss, represents the fifth weight, represents the second loss.

[0047] Adopting this technical solution, by dynamically adjusting the weights of the losses of different prediction parameters, the state recognition model can more flexibly handle different prediction tasks, and automatically adjust the degree of attention to each prediction task, thereby improving the overall prediction performance of the state recognition model.

[0048] Based on the above solution, optionally, the second target loss is determined based on the following method: determining a plurality of third losses corresponding to the state of charge at a plurality of moments adjacent to and earlier than the prediction moment; respectively determining the sixth weights of the plurality of third losses, and determining the second target loss according to the plurality of third losses and their corresponding sixth weights, wherein the sixth weight has a positive correlation with the value of the third loss.

[0049] Wherein, the third loss can be understood as a metric for measuring the accuracy of the prediction result of the state recognition model for the state of charge (SOC), mainly relative to a plurality of time points adjacent to and earlier than the prediction moment. It is calculated by comparing the difference between the predicted state of charge value and the actual value of the state recognition model, and helps to optimize the part related to the state of charge prediction in the model. Each time point has a corresponding third loss. The sixth weight can be understood as a weight coefficient associated with each third loss, used to balance the influence of the third losses at each time point on the second target loss. The sixth weight has a positive correlation with the corresponding third loss, that is, in the case where the third loss at a certain time point is larger, a larger weight is assigned.

[0050] Based on the above solution, optionally, the determining the second target loss according to the plurality of third losses and their corresponding sixth weights can be determined based on the following formula: ; wherein, represents the second target loss, represents the weight coefficient of the two moments before the prediction moment, represents the losses of the two moments before the prediction moment, represents the weight coefficient of the moment before the prediction moment, represents the loss of the moment before the prediction moment, represents the weight coefficient of the prediction moment, Represents the loss at the prediction moment.

[0051] Adopting this technical solution, by assigning different weights to the losses at different moments, the state recognition model not only focuses on the accuracy at the current prediction moment but also on the losses along the entire time path, improving the model's adaptability to dynamic changes. Especially for parameters such as the state of charge that fluctuate greatly over time, it helps to improve the overall prediction accuracy and reliability of the state recognition model.

[0052] S120. Input the battery operation data into the state recognition model to obtain prediction parameters related to the lithium battery operation state, where the prediction parameters at least include remaining useful life, state of health, and state of charge.

[0053] Among them, the remaining useful life (RUL) can be understood as the time or number of cycles from the current working state of the lithium battery until it can no longer meet the specified performance standards, which can provide information for users or management systems on when to replace the lithium battery and is helpful for maintaining and replacing the battery. The state of health (SOH) can be understood as an indicator describing the overall condition of the lithium battery, reflecting the degree of lithium battery aging. The state of charge (SOC) can be understood as representing the ratio of the current stored power of the battery to the maximum capacity. The state of charge (SOC) can be used to manage the charging and discharging process of the lithium battery, avoid overcharging and over-discharging, and directly affect the safety of the lithium battery.

[0054] Based on the above solution, optionally, when the prediction parameter is the remaining useful life, the battery operation data includes voltage data; the inputting the battery operation data into the state recognition model to obtain prediction parameters related to the lithium battery operation state includes: inputting the voltage data into the first feature extraction module for feature extraction to obtain first operation feature data; inputting the voltage data into the gating module to determine the first weight of the first operation feature data; inputting the result of weighted summation of the first operation feature data and the first weight into the first prediction unit to obtain the remaining useful life related to the lithium battery operation state.

[0055] Among them, the first operation feature data can be understood as the feature information obtained by the first feature extraction module processing the voltage data. The first operation feature data can be used to reflect the state of the battery, such as the change of internal chemical reactions. The first weight can be understood as the weight of the first operation feature data determined by the gating module.

[0056] By adopting this technical solution, first extracting the first operation characteristic data from the voltage data, then determining the weight of the first operation characteristic data through the gating module, and finally inputting the weighted characteristics into the first prediction unit to predict the remaining life of the lithium battery, not only the prediction accuracy is improved, but also the battery characteristics under different working conditions can be better adapted.

[0057] On the basis of the above solution, optionally, when the prediction parameter is the state of health, the battery operation data includes voltage data; the inputting the battery operation data into the state recognition model to obtain the prediction parameter related to the operation state of the lithium battery includes: inputting the voltage data into the first feature extraction module for feature extraction to obtain the second operation characteristic data; inputting the voltage data into the gating module to determine the second weight of the second operation characteristic data; inputting the result of weighted summation of the second operation characteristic data and the second weight into the second prediction unit to obtain the state of health related to the operation state of the lithium battery.

[0058] Among them, the second operation characteristic data can be understood as the characteristic information obtained after the first feature extraction module processes the voltage data and is used for the evaluation of the state of health. The second weight can be understood as the weight of the second operation characteristic data determined by the gating module.

[0059] By adopting this technical solution, by extracting the second operation characteristic data from the voltage data, then determining the weight of the second operation characteristic data through the gating module, and finally inputting the weighted characteristics into the second prediction unit to predict the state of health of the lithium battery, not only the prediction accuracy is improved, but also the battery characteristics under different working conditions can be better adapted, which helps to timely detect the battery aging problem and take corresponding maintenance measures.

[0060] On the basis of the above solution, optionally, when the prediction parameter is the state of charge, the battery operation data includes voltage data, current data and temperature data; the inputting the battery operation data into the state recognition model to obtain the prediction parameter related to the operation state of the lithium battery includes: inputting the voltage data, the current data, the temperature data and the state of health into the second feature extraction module for feature extraction to obtain the third operation characteristic data; inputting the voltage data, the current data, the temperature data and the state of health into the gating module to determine the third weight of the third operation characteristic data; inputting the result of weighted summation of the third operation characteristic data and the third weight into the third prediction unit to obtain the state of charge related to the operation state of the lithium battery.

[0061] Among them, the third operating characteristic data can be understood as the characteristics obtained after the second feature extraction module processes multi-source input data (such as voltage data, the current data, temperature data, and health status, etc.), which are used for the evaluation of the state of charge, reflecting the current charging level of the battery and its change trend over time, and helping to more accurately predict the state of charge. The third weight can be understood as the weight of the third operating characteristic data determined by the gating module.

[0062] An optional implementation manner is that for the voltage data and the current data, the voltage / current characteristic values, including the maximum value, the average value, the skewness, and the kurtosis, can be obtained through the feature extraction module. Therefore, for the charging voltage, the maximum charging voltage, the average charging voltage, the skewness of the charging voltage, or / and the kurtosis of the charging voltage can be selected; for the discharging voltage, the maximum discharging voltage, the average discharging voltage, the skewness of the discharging voltage, or / and the kurtosis of the discharging voltage can be selected; for the charging current, the maximum charging current, the average charging current, the skewness of the charging current, or / and the kurtosis of the charging current can be selected; for the discharging current, the maximum discharging current, the average discharging current, the skewness of the discharging current, or / and the kurtosis of the discharging current can be selected.

[0063] Adopting this technical solution, by comprehensively analyzing various data such as voltage, current, temperature, and health status, extracting the third operating characteristic data related to the state of charge therefrom, then determining the weight of the third operating characteristic data through the gating module, and finally inputting the weighted features into the third prediction unit to predict the state of charge of the lithium battery, not only improves the prediction accuracy, but also better adapts to the battery characteristics under different working conditions, and helps to achieve more efficient and safe battery management.

[0064] On the basis of the above solution, optionally, the training method of the state recognition model includes: obtaining the historical operation detection data of the lithium battery; preprocessing the historical operation detection data of the lithium battery, and the preprocessing at least includes data cleaning, missing value filling, and abnormal data processing, etc.; using the historical operation detection data of the lithium battery to form a feature set, and at the same time making battery state labels, and the feature set and the corresponding labels form a training data set; dividing the training data set into a training set and a test set according to a preset ratio; performing normalization processing on the data in the training set and the test set respectively; training the state prediction model using the training set and obtaining the target loss of the state prediction model, and optimizing the parameters of the state prediction model according to the target loss; using the test set data to verify and test the trained state prediction model respectively.

[0065] The technical solution of the embodiment of the present invention obtains the battery operation data at multiple moments of the target lithium battery and the state recognition model. Since the state recognition model is trained based on a deep learning model, the state recognition model at least includes a gating module and a multi-task prediction module. The gating module is used to determine the weights of the input features, and the multi-task prediction module is used to predict more than two battery state parameters, which can provide sufficient data support and tools for the state prediction of lithium batteries. Inputting the battery operation data into the state recognition model to obtain prediction parameters related to the lithium battery operation state, where the prediction parameters at least include remaining life, health state, and state of charge, can achieve accurate prediction of multiple state parameters of lithium batteries, solve the problem that the prediction model in the related technology can only predict a single battery state parameter, improve the efficiency and accuracy of battery management, help optimize battery use and maintenance, and thus extend the battery life.

[0066] Example 2: A battery made of LiNi 0.83 Co 0.11 Mn 0.06 O2 (NCM) (NCM) was predicted using the method according to Example 1. The specifications of the measured NCM battery are shown in Table 1. The charging process includes a combination of constant current and constant voltage (CC and CV) modes at a charging rate of 0.5C, followed by a 30-minute relaxation period and a discharge at 1C, and the ambient temperature is 25°C.

[0067] Table 1 Detailed specifications of the test samples

[0068] Table 2 shows the statistical errors of different models for predicting the SOH of the battery. The mean absolute error (MAE) of the lithium battery state prediction method is 0.00243, and the root mean square error (RMSE) is 0.00312, both of which are lower than the results of other models. Therefore, the lithium battery state prediction method proposed by the present invention shows its robustness and excellent prediction accuracy in batteries with different operating temperatures, indicating its suitability for battery SOH prediction. Figures 6(a)-6(j) more intuitively show the prediction results of SOH.

[0069] Table 2 Statistical errors of battery SOH prediction under different models

[0070] Table 3 Comparison of RUL estimation results of different models

[0071] As can be seen from Table 3, the estimated values of the four single-task learning models are quite close, and the MAE errors are all greater than 3; while the MAE and RMSE errors of the present invention are significantly smaller than those of other models, indicating that the lithium battery state prediction method proposed by the present invention has high accuracy and strong adaptability. Figures 7(a) - 7(j) more intuitively show the prediction results of RUL.

[0072] Table 4 Comparison of SOC Estimation Results of Different Models

[0073] Table 4 shows the statistical errors of different models in predicting the battery SOC. The mean absolute error (MAE) of the lithium battery state prediction method is 0.00241, and the root mean square error (RMSE) is 0.00338, both of which are lower than the results of other models. Figure 8 More intuitively shows the prediction results of SOC. As shown in (a) of Figure 8, during the entire service life of the battery, the model of the present invention with SOH as one of the input features shows excellent prediction ability. The orange line (predicted by the model of the present invention with SOH as a feature) almost coincides with the blue line (actual SOC). However, the model of the present invention without SOH performs slightly worse during the entire service life of the battery, as shown by the green line. More importantly, Figure 8 In (b), it clearly shows that the time required to reach a specific charging state when SOH is 0.85 is shorter than when SOH is 0.90, which can be seen from the fact that the capacity when SOH is 0.90 (red line) is always higher than when SOH is 0.85 (purple line). Figure 8 In (c), it proves that taking SOH as an input feature can significantly improve the accuracy of SOC estimation.

[0074] Embodiment 3: Figure 2 It is a schematic structural diagram of a lithium battery state prediction device provided in Embodiment 3 of the present invention. As Figure 2 shown, the device includes: an acquisition module 210 and a prediction parameter determination module 220. Among them, the acquisition module 210 is used to acquire the battery operation data and the state recognition model at multiple moments of the target lithium battery. The state recognition model is trained based on a deep learning model. The state recognition model at least includes a gating module and a multi-task prediction module. The gating module is used to determine the weights of the input features, and the multi-task prediction module is used to predict two or more battery state parameters; the prediction parameter determination module 220 is used to input the battery operation data into the state recognition model to obtain prediction parameters related to the lithium battery operation state, where the prediction parameters at least include remaining life, health state, and charge state.

[0075] The technical solution of the embodiment of the present invention obtains the battery operation data and the state recognition model at multiple moments of the target lithium battery through the acquisition module 210. Since the state recognition model is trained based on a deep learning model, the state recognition model at least includes a gating module and a multi-task prediction module. The gating module is used to determine the weights of the input features, and the multi-task prediction module is used to predict two or more battery state parameters, which can provide sufficient data support and tools for the state prediction of lithium batteries; the prediction parameter determination module 220 inputs the battery operation data into the state recognition model to obtain prediction parameters related to the lithium battery operation state. Among them, the prediction parameters at least include remaining life, health state, and state of charge, which can achieve accurate prediction of multiple state parameters of lithium batteries, solve the problem that the prediction model in the related art can only predict a single battery state parameter, improve the efficiency and accuracy of battery management, help optimize battery use and maintenance, and thus extend the battery life.

[0076] On the basis of the above solution, optionally, the state recognition model further includes a first feature extraction module and a second feature extraction module; the multi-task prediction module includes a first prediction unit, a second prediction unit, and a third prediction unit; the first feature extraction module, the second feature extraction module, the first prediction unit, the second prediction unit, and the third prediction unit are all connected to the gating module; and the second feature extraction module is connected to the second prediction unit.

[0077] On the basis of the above solution, optionally, the prediction parameter includes remaining life; the battery operation data includes voltage data; the prediction parameter determination module includes: a remaining life prediction unit. Among them, the remaining life prediction unit is used to input the voltage data into the first feature extraction module for feature extraction to obtain first operation feature data; input the voltage data into the gating module to determine the first weight of the first operation feature data; input the result of weighted summation of the first operation feature data and the first weight into the first prediction unit to obtain the remaining life related to the lithium battery operation state.

[0078] On the basis of the above solution, optionally, the prediction parameter includes health state; the battery operation data includes voltage data; the prediction parameter determination module includes: a health state prediction unit. Among them, the health state prediction unit is used to input the voltage data into the first feature extraction module for feature extraction to obtain second operation feature data; input the voltage data into the gating module to determine the second weight of the second operation feature data; input the result of weighted summation of the second operation feature data and the second weight into the second prediction unit to obtain the health state related to the lithium battery operation state.

[0079] Based on the above solution, optionally, the prediction parameter includes the state of charge; the battery operation data includes voltage data, current data, and temperature data; the prediction parameter determination module includes: a state of charge prediction unit. Among them, the state of charge prediction unit is used to input the voltage data, the current data, the temperature data, and the health state into the second feature extraction module for feature extraction to obtain third operation feature data; input the voltage data, the current data, the temperature data, and the health state into the gating module to determine the third weight of the third operation feature data; input the result of weighted summation of the third operation feature data and the third weight into the third prediction unit to obtain the state of charge related to the operation state of the lithium battery.

[0080] Based on the above solution, optionally, the target loss in the training process of the state recognition model includes a first target loss and a second target loss. The first target loss is used to adjust the model parameters of the first prediction unit and the second prediction unit, and the second target loss is used to adjust the model parameters of the third prediction unit.

[0081] Based on the above solution, optionally, the first target loss is determined based on the following method: respectively determine a first loss related to the remaining life and a second loss related to the health state; determine a fourth weight and a fifth weight according to the first loss and the second loss, where the fourth weight and the fifth weight are positively correlated with the values of the first loss and the second loss; determine the first target loss according to the first target loss, the second target loss, the fourth weight, and the fifth weight.

[0082] Based on the above solution, optionally, the second target loss is determined based on the following method: determine a plurality of third losses corresponding to the state of charge at a plurality of moments adjacent to and earlier than the prediction moment; respectively determine the sixth weights of the plurality of third losses, and determine the second target loss according to the plurality of third losses and their corresponding sixth weights, where the sixth weight is positively correlated with the value of the third loss.

[0083] Based on the above solution, optionally, the lithium battery state prediction device further includes: a battery operation data determination module. Among them, the battery operation data determination module is used to obtain the acquisition data of the target lithium battery at a plurality of moments before obtaining the battery operation data of the target lithium battery at a plurality of moments, perform post-processing on the acquisition data to obtain post-processing analysis data; determine the battery operation data according to the acquisition data and the post-processing analysis data.

[0084] The lithium battery state prediction device provided by the embodiments of the present invention can execute the lithium battery state prediction method provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0085] Embodiment 4: Figure 9 FIG. shows a schematic structural diagram of an electronic device 10 that can be used to implement the embodiments of the present invention. The electronic device is intended to represent various forms of digital computers, such as, for example, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, for example, personal digital processors, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0086] As Figure 9 shown, the electronic device 10 includes at least one processor 11, and a memory communicatively connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc. Among them, the memory stores a computer program executable by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. The input / output (I / O) interface 15 is also connected to the bus 14.

[0087] Multiple components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0088] The processor 11 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a lithium battery state prediction method.

[0089] In some embodiments, a lithium battery state prediction method can be implemented as a computer program tangibly embodied in a computer-readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the lithium battery state prediction method described above can be executed. Alternatively, in other embodiments, the processor 11 can be configured to execute a lithium battery state prediction method by any other suitable means (e.g., by means of firmware).

[0090] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuitry, integrated circuit systems, field programmable gate arrays (FPGA), application specific integrated circuits (ASIC), application specific standard products (ASSP), systems on a chip (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special-purpose or general-purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0091] The computer programs for implementing the methods of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to the processors of a general-purpose computer, a special-purpose computer, or other programmable data processing devices such that when the computer programs are executed by the processors, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer programs can be executed entirely on the machine, partially on the machine, as a stand-alone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0092] In the context of the present invention, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. The computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium can be a machine-readable signal medium. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0093] To provide for interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the electronic device. Other kinds of devices can also be used to provide for interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0094] The systems and techniques described herein can be implemented in a computing system that includes backend components (such as, for example, a data server), or a computing system that includes middleware components (such as, for example, an application server), or a computing system that includes frontend components (such as, for example, a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (such as, for example, a communication network). Examples of communication networks include: local area network (LAN), wide area network (WAN), blockchain network, and the Internet.

[0095] A computing system may include a client and a server. The client and the server are generally far from each other and usually interact via a communication network. The client-server relationship is created by computer programs running on respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system, solving the defects of difficult management and weak business scalability existing in traditional physical hosts and VPS services.

[0096] It should be understood that various forms of the processes shown above can be used, with steps reordered, added, or deleted. For example, the steps recited in the present invention can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved, and no limitation is imposed herein.

[0097] The above specific embodiments do not constitute a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for predicting the state of a lithium battery, characterized in that, Including: Obtain the battery operation data and the state recognition model at multiple moments of the target lithium battery. Among them, the state recognition model is trained based on a deep learning model. The state recognition model at least includes a gating module and a multi-task prediction module. The gating module is used to determine the weights of input features, and the multi-task prediction module is used to predict more than two battery state parameters; Input the battery operation data into the state recognition model to obtain prediction parameters related to the lithium battery operation state. Among them, the prediction parameters at least include remaining life, health state, and state of charge.

2. The method according to claim 1, characterized in that, The state recognition model further includes a first feature extraction module and a second feature extraction module; the multi-task prediction module includes a first prediction unit, a second prediction unit, and a third prediction unit; the first feature extraction module, the second feature extraction module, the first prediction unit, the second prediction unit, and the third prediction unit are all connected to the gating module; and the second feature extraction module is connected to the second prediction unit.

3. The method according to claim 2, characterized in that, The prediction parameter includes remaining life; the battery operation data includes voltage data; inputting the battery operation data into the state recognition model to obtain prediction parameters related to the lithium battery operation state includes: Input the voltage data into the first feature extraction module for feature extraction to obtain first operation feature data; Input the voltage data into the gating module to determine the first weight of the first operation feature data; Input the result of weighted summation of the first operation feature data and the first weight into the first prediction unit to obtain the remaining life related to the lithium battery operation state.

4. The method according to claim 2, characterized in that The prediction parameter includes health state; the battery operation data includes voltage data; inputting the battery operation data into the state recognition model to obtain prediction parameters related to the lithium battery operation state includes: Input the voltage data into the first feature extraction module for feature extraction to obtain second operation feature data; Input the voltage data into the gating module to determine the second weight of the second operation feature data; Input the result of weighted summation of the second operation feature data and the second weight into the second prediction unit to obtain the health state related to the lithium battery operation state.

5. The method according to claim 4, characterized in that, The prediction parameter includes state of charge; the battery operation data includes voltage data, current data, and temperature data; inputting the battery operation data into the state recognition model to obtain prediction parameters related to the lithium battery operation state includes: Input the voltage data, the current data, the temperature data, and the health state into the second feature extraction module for feature extraction to obtain third operation feature data; Input the voltage data, the current data, the temperature data, and the health state into the gating module to determine the third weight of the third operation feature data; Input the result of weighted summation of the third operation feature data and the third weight into the third prediction unit to obtain the state of charge related to the lithium battery operation state.

6. The method according to claim 2, characterized in that, The target loss in the training process of the state recognition model includes a first target loss and a second target loss. The first target loss is used to adjust the model parameters of the first prediction unit and the second prediction unit, and the second target loss is used to adjust the model parameters of the third prediction unit.

7. The method according to claim 6, wherein The first target loss is determined based on the following method: Determine a first loss related to the remaining life and a second loss related to the health state respectively; Determine a fourth weight and a fifth weight according to the first loss and the second loss, where the fourth weight and the fifth weight are positively correlated with the values of the first loss and the second loss; Determine the first target loss according to the first target loss, the second target loss, the fourth weight and the fifth weight.

8. The method according to claim 6, characterized in that, The second target loss is determined based on the following method: Determine a plurality of third losses corresponding to the state of charge at a plurality of moments adjacent to and earlier than the prediction moment; Determine the sixth weights of the plurality of third losses respectively, and determine the second target loss according to the plurality of third losses and their corresponding sixth weights, where the sixth weight is positively correlated with the value of the third loss.

9. The method according to claim 1, wherein Before obtaining the battery operation data of the target lithium battery at multiple moments, it further includes: Obtain the acquisition data of the target lithium battery at multiple moments, and perform post-processing on the acquisition data to obtain post-processing analysis data; Determine the battery operation data according to the acquisition data and the post-processing analysis data.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the processor to implement the lithium battery state prediction method according to any one of claims 1-9 when executed.

Images