Early warning method and model training method for abnormal charging and discharging of lithium-ion batteries

Abstract

The present invention provides an early warning method and a model training method for abnormal charging and discharging of lithium-ion batteries, which can be applied to the field of lithium-ion battery technology. The early warning method includes: obtaining voltage information and current information of a target lithium-ion battery during the charging / discharging process; inputting the voltage information and the current information into a trained target model, and outputting the state of charge probability distribution information of the target lithium-ion battery; wherein the state of charge probability distribution information includes the mean of each state of charge prediction information and the variance of each state of charge prediction information; and generating early warning information based on the mean of each state of charge prediction information and the variance of each state of charge prediction information. The present invention also provides an early warning device and a model training device, equipment, and storage medium for abnormal charging and discharging of lithium-ion batteries.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium-ion batteries, and more specifically to an early warning method and a model training method for abnormal charging and discharging of lithium-ion batteries. Background Art

[0002] Lithium-ion batteries are the main power source for electric vehicles, energy storage power stations, digital products and other equipment. However, lithium-ion batteries are prone to overcharge / over-discharge during use, so it is necessary to monitor the charging and discharging status of lithium-ion batteries in real time.

[0003] In the process of realizing the concept of the present invention, it was found through research that at least in the related technology, there is a certain lag in judging abnormal charging and discharging of lithium-ion batteries, which leads to technical problems such as reduced performance and shortened life of the battery, and even causes safety risks. Summary of the invention

[0004] In view of the above problems, the present invention provides an early warning method, a model training method, a device, a equipment, and a medium for abnormal charging and discharging of a lithium-ion battery.

[0005] According to a first aspect of the present invention, there is provided a method for early warning of abnormal charging and discharging of a lithium-ion battery, comprising:

[0006] The voltage information and current information of the target lithium-ion battery during the charge / discharge process are obtained; the voltage information and current information are input into the trained target model, and the state of charge probability distribution information of the target lithium-ion battery is output; wherein the state of charge probability distribution information includes the mean of each state of charge prediction information and the variance of each state of charge prediction information; and according to the mean of each state of charge prediction information and the variance of each state of charge prediction information, early warning information is generated.

[0007] According to an embodiment of the present invention, warning information is generated based on the mean of each state of charge prediction information and the variance of each state of charge prediction information, including: generating probability interval information under a target confidence level based on the mean of each state of charge prediction information and the variance of each state of charge prediction information; and generating warning information in response to the probability interval information exceeding the predetermined probability interval information.

[0008] According to an embodiment of the present invention, the probability interval information includes an upper limit value of the probability interval and a lower limit value of the probability interval; the predetermined probability interval information includes an upper limit value of the predetermined probability interval and a lower limit value of the predetermined probability interval; in response to the probability interval information exceeding the predetermined probability interval information, warning information is generated, including: during the charging process of the target lithium-ion battery, in response to the upper limit value of the probability interval being greater than the upper limit value of the predetermined probability interval, warning information for characterizing overcharging is generated; during the discharging process of the target lithium-ion battery, in response to the lower limit value of the probability interval being less than the lower limit value of the predetermined probability interval, warning information for characterizing over-discharging is generated.

[0009] According to an embodiment of the present invention, probability interval information under a target confidence level is generated based on the mean of each state of charge prediction information and the variance of each state of charge prediction information, including: generating a variance median value based on the variance and a predetermined parameter; determining the difference between the mean and the variance median value as a probability interval lower limit value; determining the sum of the mean and the variance median value as a probability interval upper limit value; and generating probability interval information based on the probability interval lower limit value and the probability interval upper limit value.

[0010] A second aspect of the present invention provides a model training method, comprising:

[0011] Obtain sample voltage information and sample current information of a sample lithium-ion battery during the charge / discharge process; process the sample voltage information and sample current information to generate sample state-of-charge information; input the sample voltage information and sample current information into an initial model, and output probability distribution information of each sample state-of-charge prediction value; wherein the probability distribution information includes a mean value of each sample state-of-charge prediction value and a variance of each sample state-of-charge prediction value; based on a target loss function, calculate a loss gradient according to the mean value of each sample state-of-charge prediction value, the variance of each sample state-of-charge prediction value and the sample state-of-charge information; based on the loss gradient, use a natural gradient optimization method to adjust the parameters of the initial model until a convergence condition is reached, and obtain a target model, wherein the target model is a trained target model in the above method.

[0012] According to an embodiment of the present invention, the loss gradient includes the loss gradient of the mean and the loss gradient of the variance; based on the loss gradient, the parameters of the initial model are adjusted using the natural gradient optimization method until the convergence condition is reached to obtain the target model, including: based on the loss gradient of the mean and the target matrix, the gradient value of the mean is obtained; wherein the target matrix is ​​generated based on the second-order derivative of the target loss function with respect to the model parameters; based on the loss gradient of the variance and the target matrix, the gradient value of the variance is obtained; and based on the gradient value of the mean, the gradient value of the variance and a predetermined adjustment step size, the model parameters for predicting the mean and the model parameters for predicting the variance in the initial model are updated accordingly until the convergence condition is reached to obtain the target model.

[0013] A third aspect of the present invention provides an early warning device for abnormal charging and discharging of a lithium-ion battery, comprising:

[0014] A first acquisition module is used to acquire voltage information and current information of a target lithium-ion battery during a charge / discharge process;

[0015] A prediction module, used to input voltage information and current information into a trained target model, and output the state of charge probability distribution information of the target lithium-ion battery; wherein the state of charge probability distribution information includes the mean of each state of charge prediction information and the variance of each state of charge prediction information; each state of charge prediction information obeys a normal distribution; and

[0016] The early warning module is used to generate early warning information according to the mean of each state of charge prediction information and the variance of each state of charge prediction information.

[0017] A fourth aspect of the present invention provides a model training device, comprising:

[0018] A second acquisition module is used to acquire sample voltage information and sample current information of the sample lithium-ion battery during the charging / discharging process;

[0019] A processing module, used for processing the sample voltage information and the sample current information to generate sample charge state information;

[0020] An output module, used to input sample voltage information and sample current information into an initial model, and output probability distribution information of each sample state of charge prediction value; the probability distribution information includes a mean value of each sample state of charge prediction value and a variance of each sample state of charge prediction value;

[0021] A loss calculation module, used to calculate the loss gradient based on the target loss function, according to the mean of the state of charge prediction values ​​of each sample, the variance of the state of charge prediction values ​​of each sample, and the state of charge information of the sample;

[0022] The adjustment module is used to adjust the parameters of the initial model based on the loss gradient using the natural gradient optimization method until the convergence condition is reached to obtain the target model, wherein the target model is the trained target model in the above method.

[0023] A fifth aspect of the present invention provides an electronic device, comprising: one or more processors; a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the above method.

[0024] The sixth aspect of the present invention further provides a computer-readable storage medium on which a computer program or instruction is stored, and the steps of the above method are implemented when the above computer program or instruction is executed by a processor.

[0025] The seventh aspect of the present invention also provides a computer program product, including a computer program or instructions, which implement the steps of the above method when executed by a processor.

[0026] According to an embodiment of the present invention, by acquiring voltage information and current information of a target lithium-ion battery during a charge / discharge process and inputting the information into a target model, a mean value of each state of charge prediction information and a variance of each state of charge prediction information are output, and warning information is generated according to the mean value and the variance of each state of charge prediction information. Since the state of charge probability distribution information of the target lithium-ion battery is obtained according to the trained target model, a warning is issued when the target lithium-ion battery has not actually reached overcharge / discharge, thereby solving the technical problem that the related technology has a hysteresis in the judgment of overcharge / discharge, that is, the warning information is generated only when the target lithium-ion battery is already in an overcharge / discharge state, thereby maintaining the health of the battery and improving the safety of use. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The above contents and other objects, features and advantages of the present invention will become more apparent through the following description of the embodiments of the present invention with reference to the accompanying drawings, in which:

[0028] Figure 1 A flow chart of a method for early warning of abnormal charging and discharging of a lithium-ion battery according to an embodiment of the present invention is shown.

[0029] Figure 2a The probability interval information of the target lithium-ion battery within 0s to 860s according to an embodiment of the present invention is shown.

[0030] Figure 2b The probability interval information of the target lithium-ion battery within 1000s to 2200s according to an embodiment of the present invention is shown.

[0031] Figure 2c The probability interval information of the target lithium-ion battery within 9000s~10200s according to an embodiment of the present invention is shown.

[0032] Figure 3 A flow chart of a model training method according to an embodiment of the present invention is shown.

[0033] Figure 4 The structure block diagram of the device for early warning of abnormal charging and discharging of a lithium-ion battery according to an embodiment of the present invention is shown.

[0034] Figure 5 A structural block diagram of a model training device according to an embodiment of the present invention is shown.

[0035] Figure 6A block diagram of an electronic device suitable for implementing an early warning method or a model training method for abnormal charging and discharging of a lithium-ion battery according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0036] Below, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present invention. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of embodiments of the present invention. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of concepts of the present invention.

[0037] The terms used herein are only for describing specific embodiments and are not intended to limit the present invention. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0038] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0039] When using expressions such as "at least one of A, B, and C, etc.", they should generally be interpreted according to the meaning of the expression commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0040] In the embodiments of the present invention, the collection, updating, analysis, processing, use, transmission, provision, disclosure, storage and other aspects of the data involved (for example, including but not limited to user personal information) are in compliance with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. In particular, necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain the security of user personal information and network security.

[0041] In the technical solution of the present invention, the user information (including but not limited to user personal information, user image information, user device information, such as location information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved are all information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with relevant laws, regulations and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0042] Overcharging of lithium-ion batteries refers to the situation where the amount of electricity charged into the battery during the charging process exceeds its designed or specified rated charging capacity. Correspondingly, overdischarging of lithium-ion batteries refers to the situation where the battery voltage during the discharge process is lower than the minimum discharge voltage designed for the battery. Battery overdischarging may cause polarization and damage to the battery's electrode materials, thereby reducing the battery capacity and even causing unstable internal battery reactions, leading to safety risks. Long-term overcharging and overdischarging will affect the positive and negative active materials of lithium-ion batteries and accelerate battery aging, causing the battery to experience electrolyte decomposition, gas production and expansion, etc., posing certain safety hazards.

[0043] Based on this, overcharge and over-discharge fault diagnosis in related technologies is usually based on thresholds. Specifically, the terminal voltage of the battery is sampled and compared with the preset threshold to achieve overcharge and over-discharge diagnosis. However, in the actual process, due to the influence of sampling errors, environmental noise, electromagnetic interference, etc., the measured terminal voltage is inaccurate, resulting in the battery still having charging or discharging current after the warning cut-off signal is issued, making it difficult to stop charging and discharging in time, resulting in overcharge and over-discharge faults.

[0044] In view of this, an embodiment of the present invention provides an early warning method for abnormal charging and discharging of a lithium-ion battery, comprising: obtaining voltage information and current information of a target lithium-ion battery during a charging / discharging process; inputting the voltage information and current information into a trained target model, and outputting state of charge probability distribution information of the target lithium-ion battery; wherein the state of charge probability distribution information includes a mean value of each state of charge prediction information and a variance of each state of charge prediction information; and generating early warning information according to the mean value of each state of charge prediction information and the variance of each state of charge prediction information.

[0045] Figure 1 A flow chart of a method for early warning of abnormal charging and discharging of a lithium-ion battery according to an embodiment of the present invention is shown.

[0046] like Figure 1 As shown, the embodiment 100 includes operations S110 to S130.

[0047] In operation S110 , voltage information and current information of a target lithium-ion battery during a charge / discharge process are acquired.

[0048] In operation S120 , the voltage information and the current information are input into the trained target model, and the state of charge probability distribution information of the target lithium-ion battery is output.

[0049] In operation S130 , warning information is generated according to the mean of each state of charge prediction information and the variance of each state of charge prediction information.

[0050] According to an embodiment of the present invention, the positive electrode active material of the target lithium-ion battery may be nickel-cobalt-manganese, and the negative electrode active material may be graphite.

[0051] According to an embodiment of the present invention, voltage information and current information can be collected in real time using a data acquisition instrument, a Hall current sensor, or other equipment.

[0052] According to an embodiment of the present invention, the target model may include a natural gradient boosting (NGBoost) model. The state of charge probability distribution information includes a mean value of each state of charge prediction information and a variance of each state of charge prediction information.

[0053] According to an embodiment of the present invention, the state of charge is an important parameter for measuring the remaining power of the target lithium-ion battery, and predicting the state of charge is a prerequisite for formulating multiple control strategies. The predicted information of the state of charge is dynamically correlated with the voltage information and the current information.

[0054] According to an embodiment of the present invention, based on the mean value of each state of charge prediction information and the variance of each state of charge prediction information, during the charging process of the target lithium-ion battery, when the state of charge is about to approach 100%, warning information for characterizing overcharging can be generated to avoid overcharging of the target lithium-ion battery; at the same time, during the discharging process of the target lithium-ion battery, when the state of charge of the target lithium-ion battery is about to approach 0%, warning information for characterizing over-discharging can be generated to avoid over-discharging of the battery.

[0055] According to an embodiment of the present invention, by acquiring voltage information and current information of a target lithium-ion battery during a charge / discharge process and inputting the information into a target model, a mean value of each state of charge prediction information and a variance of each state of charge prediction information are output, and warning information is generated according to the mean value and the variance of each state of charge prediction information. Since the state of charge probability distribution information of the target lithium-ion battery is obtained according to the trained target model, a warning is issued when the target lithium-ion battery has not actually reached overcharge / discharge, thereby solving the technical problem that the related technology has a hysteresis in the judgment of overcharge / discharge, that is, the warning information is generated only when the target lithium-ion battery is already in an overcharge / discharge state, thereby maintaining the health of the battery and improving the safety of use.

[0056] According to an embodiment of the present invention, warning information is generated based on the mean of each state of charge prediction information and the variance of each state of charge prediction information, including: generating probability interval information under a target confidence level based on the mean of each state of charge prediction information and the variance of each state of charge prediction information; and generating warning information in response to the probability interval information exceeding the predetermined probability interval information.

[0057] According to an embodiment of the present invention, the target confidence level indicates the credibility of the probability interval information, which may be 95%. The probability interval information may indicate the probability estimation range for the state of charge prediction information, and the probability interval information includes an upper limit value and a lower limit value of the probability interval.

[0058] According to an embodiment of the present invention, the predetermined probability interval information can be pre-set according to the safety specification of the target lithium-ion battery. By comparing the probability interval information with the predetermined probability interval information, the probability interval information is closer to the predetermined probability interval information than the mean value, so that the warning information can be generated in time.

[0059] According to an embodiment of the present invention, probability interval information under a target confidence level is generated based on the mean of each state of charge prediction information and the variance of each state of charge prediction information, including: generating a variance median value based on the variance and a predetermined parameter; determining the difference between the mean and the variance median value as a probability interval lower limit value; determining the sum of the mean and the variance median value as a probability interval upper limit value; and generating probability interval information based on the probability interval lower limit value and the probability interval upper limit value.

[0060] According to an embodiment of the present invention, the predetermined parameter may be a probability parameter of a normal distribution corresponding to the target confidence level, a variance median value and a variance The square root of is related to the standard deviation, and the median value of the variance is different under different target confidence levels.

[0061] According to an embodiment of the present invention, taking a target confidence level of 95% as an example, approximately 95% of the data fall within two standard deviations of the mean. At this time, the lower bound of the probability interval can be expressed as , the upper bound of the probability interval can be expressed as . Accordingly, the probability interval information can be expressed as [ ], where μ represents the mean value of the state of charge prediction information.

[0062] In order to better reflect the effect of early warning of overcharge / discharge of the target lithium-ion battery based on the probability interval information and the predetermined probability interval information, the following will be based on Figure 2a~Figure 2c The probability interval information of the target lithium-ion battery with a target confidence level of 95% in different time periods is further explained.

[0063] Figure 2a The probability interval information of the target lithium-ion battery within 0s to 860s according to an embodiment of the present invention is shown.

[0064] Figure 2b The probability interval information of the target lithium-ion battery within 1000s to 2200s according to an embodiment of the present invention is shown.

[0065] Figure 2c The probability interval information of the target lithium-ion battery within 9000s~10200s according to an embodiment of the present invention is shown.

[0066] like Figure 2a~Figure 2c As shown, the horizontal axis represents the sampling time k (in seconds), and the vertical axis represents the charge state prediction information corresponding to the sampling time.

[0067] pass Figure 2a~Figure 2c It can be clearly seen that compared with the mean of each state of charge prediction information, the probability interval information will exceed the state of charge prediction information earlier. Figure 2c The lower bound value of the medium probability interval information reaches 0% earlier than the mean μ, and the over-discharge warning information can be generated earlier.

[0068] According to an embodiment of the present invention, the predetermined probability interval information includes a predetermined probability interval upper limit value and a predetermined probability interval lower limit value; in response to the probability interval information exceeding the predetermined probability interval information, warning information is generated, including: during the charging process of the target lithium-ion battery, in response to the probability interval upper limit value being greater than the predetermined probability interval upper limit value, warning information for characterizing overcharging is generated; during the discharging process of the target lithium-ion battery, in response to the probability interval lower limit value being less than the predetermined probability interval lower limit value, warning information for characterizing over-discharging is generated.

[0069] According to an embodiment of the present invention, during the charging process of the target lithium-ion battery, in response to the upper limit value of the probability interval being greater than the upper limit value of the predetermined probability interval, that is, when the state of charge information of the target lithium-ion battery is about to approach 100%, the upper limit value of the state of charge probability estimation interval approaches 100% earlier than the mean value, that is, when the upper limit value reaches 100%, a warning message for characterizing overcharging is generated to avoid overcharging.

[0070] According to an embodiment of the present invention, during the discharge of a target lithium-ion battery, in response to a lower limit value of a probability interval being less than a lower limit value of a predetermined probability interval, that is, when the state of charge information of the target lithium-ion battery is about to approach 0%, the lower limit value of the state of charge probability estimation interval approaches 0% earlier than the mean value, that is, when the lower limit value reaches 0%, a warning message for characterizing over-discharge is generated, and the discharge of the target lithium-ion battery is stopped.

[0071] According to an embodiment of the present invention, an overcharge / discharge warning is issued through the upper and lower limit values ​​of the probability interval corresponding to the target confidence level, so that the charging / discharging can be stopped as soon as possible when the actual charging / discharging amount does not exceed the predetermined probability interval information, thereby effectively avoiding the occurrence of overcharge / discharge.

[0072] Based on the above-mentioned lithium-ion battery charging and discharging abnormality early warning method, the present invention proposes a model training method for obtaining the target model in the above-mentioned embodiment 100, the method comprising:

[0073] The sample voltage information and the sample current information of the sample lithium-ion battery during the charge / discharge process are obtained; the sample voltage information and the sample current information are processed to generate the sample state of charge information; the sample voltage information and the sample current information are input into the initial model, and the probability distribution information of each sample state of charge prediction value is output; the probability distribution information includes the mean of each sample state of charge prediction value and the variance of each sample state of charge prediction value; based on the target loss function, the loss gradient is calculated according to the mean of each sample state of charge prediction value, the variance of each sample state of charge prediction value and the sample state of charge information; based on the loss gradient, the parameters of the initial model are adjusted by using the natural gradient optimization method until the convergence condition is reached to obtain the target model.

[0074] Figure 3 A flow chart of a model training method according to an embodiment of the present invention is shown.

[0075] like Figure 3 As shown, embodiment 300 includes operations S310 to S350.

[0076] In operation S310, sample voltage information and sample current information of a sample lithium-ion battery during a charge / discharge process are acquired.

[0077] According to an embodiment of the present invention, a charging and discharging machine can be used to charge and discharge a target lithium-ion battery. The sampling circuit of the sample voltage information and the sample current information includes components for collecting current such as a Hall current sensor or a shunt and components for collecting voltage such as a resistor voltage divider circuit. The collected initial information needs to be amplified, filtered, analog-to-digital converted, and communication isolated to obtain the sample voltage information and the sample current information.

[0078] In operation S320 , the sample voltage information and the sample current information are processed to generate sample state of charge information.

[0079] According to an embodiment of the present invention, the charge state prediction information corresponding to the kth moment The state of charge (SOC) can be calculated based on the ampere-hour integral formula shown in the following formula (1).

[0080] (1)

[0081] in, is the Coulomb coefficient, Δt is the sampling period, I(k) represents the sample current information at the kth moment, C n Indicates the rated capacity of the target lithium-ion battery.

[0082] According to an embodiment of the present invention, the state of charge prediction information, the sample current information and the sample voltage information constitute a training sample set A as shown in the following equation (2).

[0083] (2)

[0084] Where n represents the total number of training sample sets and , x represents the feature vector of the training sample set, y represents the label value of the training sample set, and y i represents the i-th state of charge prediction information, x i =[I(k) I(k-1) U(k) U(k-1)], U(k) is the sample voltage information at the kth moment, I(k-1) is the sample current information at the k-1th moment, and U(k-1) is the sample voltage information at the k-1th moment.

[0085] In operation S330 , sample voltage information and sample current information are input into an initial model, and probability distribution information of each sample state of charge prediction value is output.

[0086] According to an embodiment of the present invention, the initial model may represent an untrained NGBoost model.

[0087] According to an embodiment of the present invention, the sample state of charge prediction value is made to obey the normal distribution, that is, for the training sample set A, the feature vector x is input and the probability distribution information of the label value y is output. The probability distribution information includes the mean of each sample state of charge prediction value and the variance of each sample state of charge prediction value. Probability density function It is expressed as the following formula (3).

[0088] (3)

[0089] Among them, θ represents the model parameters.

[0090] In operation S340 , based on the target loss function, a loss gradient is calculated according to the mean of each sample state of charge prediction value, the variance of each sample state of charge prediction value, and the sample state of charge information.

[0091] According to an embodiment of the present invention, a weak learner may be used to predict the mean and variance of the state of charge prediction value. The weak learning selection may include a decision tree, the mean μ of the i sample state of charge prediction values i It can be expressed as the following formula (4).

[0092] (4)

[0093] in, is the mean parameter of the initial model, is the mean prediction function.

[0094] According to an embodiment of the present invention, the variance of the predicted state of charge values ​​of i samples is It can be expressed as the following formula (5).

[0095] (5)

[0096] in, is the variance parameter of the initial model, is the variance prediction function.

[0097] According to an embodiment of the present invention, the target loss function It can be expressed as the following formula (6), which is used to measure the predictive ability of the model through the negative log-likelihood function.

[0098] (6)

[0099] According to an embodiment of the present invention, the loss gradient can be used to optimize the model, and the loss gradient includes the loss gradient of the mean and the loss gradient of the variance.

[0100] According to an embodiment of the present invention, the loss gradient of the mean It can be expressed as the following formula (7).

[0101] (7)

[0102] According to an embodiment of the present invention, the loss gradient of the variance It can be expressed as the following formula (8).

[0103] (8)

[0104] In operation S350, based on the loss gradient, the parameters of the initial model are adjusted using a natural gradient optimization method until a convergence condition is reached to obtain a target model.

[0105] According to an embodiment of the present invention, the initial mean input of the initial model is μ0=mean(y), and the initial variance input is assumed to be 0.

[0106] According to an embodiment of the present invention, the initial model parameters are adjusted by a natural gradient optimization method to update the mean and variance until a convergence condition is reached. The convergence condition may be a predetermined number of iterations or a predetermined threshold of the model parameters.

[0107] According to an embodiment of the present invention, the acquired sample voltage information and sample current information are input into the initial model, the probability distribution information of the predicted value of each sample state of charge is output, and the loss gradient is calculated based on the target loss function, and then the parameters of the initial model are adjusted until the convergence condition is reached to obtain the target model, the prediction ability of the initial model is optimized, and training is performed based on the sample voltage information and sample current information actually collected, so that it can adapt to different actual charging and discharging conditions, and obtain a target model with higher prediction accuracy for use in early warning of abnormal charging and discharging of lithium-ion batteries, and maintain the health of the target lithium-ion batteries.

[0108] According to an embodiment of the present invention, based on the loss gradient, the parameters of the initial model are adjusted using the natural gradient optimization method until the convergence conditions are reached to obtain the target model, including: obtaining the gradient value of the mean based on the loss gradient and the target matrix of the mean; obtaining the gradient value of the variance based on the loss gradient and the target matrix of the variance; and correspondingly updating the model parameters for predicting the mean and the model parameters for predicting the variance in the initial model based on the gradient value of the mean, the gradient value of the variance and a predetermined adjustment step size, respectively, until the convergence conditions are reached to obtain the target model.

[0109] According to an embodiment of the present invention, the target matrix may be a Fisher information matrix, which is generated based on the second-order derivative of the target loss function with respect to the model parameters. The target matrix F may be expressed as the following formula (9).

[0110] (9)

[0111] Among them, ∂ is the symbol of partial derivative, b and d are different indices of model parameters θ.

[0112] According to an embodiment of the present invention, based on the loss gradient of the mean and the target matrix, the gradient value of the mean is obtained. It can be expressed as the following formula (10).

[0113] (10)

[0114] According to an embodiment of the present invention, F -1 is the inverse matrix of F.

[0115] According to an embodiment of the present invention, based on the loss gradient of the variance and the target matrix, the gradient value of the variance is obtained. It can be expressed as the following formula (11).

[0116] (11)

[0117] According to an embodiment of the present invention, based on the gradient value of the mean and the predetermined adjustment step size, the model parameters used to predict the mean in the initial model are updated accordingly to obtain the target mean parameter It can be expressed as the following formula (12).

[0118] (12)

[0119] Wherein, η is the predetermined adjustment step size.

[0120] According to an embodiment of the present invention, based on the gradient value of the variance and the predetermined adjustment step size, the model parameters used to predict the variance in the initial model are updated accordingly to obtain the target variance parameter It can be expressed as the following formula (13).

[0121] (13)

[0122] According to an embodiment of the present invention, the target model after training is used for the newly input sample x new , we can get the mean value of the state of charge prediction information expressed by the following formula (14): .

[0123] (14)

[0124] in, is the mean prediction function of the target model, is the target mean parameter at the kth moment.

[0125] According to an embodiment of the present invention, the target model after training is used for the newly input sample x new, the variance of the state of charge prediction information can be obtained as follows: .

[0126] (15)

[0127] in, is the variance prediction function of the target model, is the target variance parameter at the kth moment, k∈K, and K is the total number of moments.

[0128] According to an embodiment of the present invention, based on the loss gradient, the parameters of the initial model are adjusted using the natural gradient optimization method until the convergence condition is reached to obtain the target model. Since the model parameters for predicting the mean and the model parameters for predicting the variance in the initial model are updated accordingly based on the gradient value of the mean, the gradient value of the variance and the predetermined adjustment step size, the target mean model parameters and the target variance model parameters can more accurately capture the distribution characteristics of the data, thereby effectively improving the generalization and robustness of the initial model.

[0129] According to an embodiment of the present invention, the computing chip implementing the above-mentioned embodiment 100 and embodiment 300 may be a single-chip microcomputer, a field programmable gate array or a digital signal processor, etc., and has key functions such as storage, communication, control, and analog-to-digital conversion.

[0130] According to an embodiment of the present invention, the above-mentioned early warning method also includes real-time display of each charge state prediction information, and the charge state prediction information can be displayed through a display device including a communication interface and an isolated communication transceiver / receiver with a monitoring unit, a display chip, a control chip, a driver chip, etc.

[0131] Based on the above-mentioned lithium-ion battery charging and discharging abnormality warning method, the present invention also provides a lithium-ion battery charging and discharging abnormality warning device. Figure 4 The device is described in detail.

[0132] Figure 4 The structure block diagram of the device for early warning of abnormal charging and discharging of a lithium-ion battery according to an embodiment of the present invention is shown.

[0133] like Figure 4 As shown, the device 400 includes a first acquisition module 410 , a prediction module 420 and an early warning module 430 .

[0134] The first acquisition module 410 is used to acquire voltage information and current information of the target lithium-ion battery during the charging / discharging process. In one embodiment, the first acquisition module 410 can be used to perform the operation S110 described above, which will not be described in detail here.

[0135] The prediction module 420 is used to input the voltage information and the current information into the trained target model, and output the charge state probability distribution information of the target lithium-ion battery. In one embodiment, the prediction module 420 can be used to perform the operation S120 described above, which will not be described in detail here.

[0136] The warning module 430 is used to generate warning information according to the mean of each state of charge prediction information and the variance of each state of charge prediction information. In one embodiment, the warning module 430 can be used to perform the operation S130 described above, which will not be repeated here.

[0137] According to an embodiment of the present invention, the early warning module 430 includes a first generation submodule and a second generation submodule.

[0138] The first generating submodule is used to generate probability interval information under a target confidence level according to a mean value of each state of charge prediction information and a variance of each state of charge prediction information.

[0139] The second generating submodule is used to generate warning information in response to the probability interval information exceeding the predetermined probability interval information.

[0140] According to an embodiment of the present invention, the second generating submodule includes a first generating unit and a second generating unit.

[0141] The first generating unit is used to generate warning information for indicating overcharging in response to an upper limit value of a probability interval being greater than an upper limit value of a predetermined probability interval during charging of the target lithium-ion battery.

[0142] The second generating unit is used to generate warning information for characterizing over-discharging in response to a probability interval lower limit value being less than a predetermined probability interval lower limit value during the discharge process of the target lithium-ion battery.

[0143] According to an embodiment of the present invention, the first generating submodule includes a first generating unit, a first determining unit, a second determining unit and a second generating submodule.

[0144] The first generating unit is used to generate a variance intermediate value according to the variance and a predetermined parameter.

[0145] The first determining unit is used to determine the difference between the mean and the middle value of the variance as the lower limit value of the probability interval.

[0146] The second determining unit is used to determine the sum of the mean value and the variance middle value as the probability interval upper limit value.

[0147] The first second generating unit is used to generate probability interval information according to the probability interval lower limit value and the probability interval upper limit value.

[0148] According to an embodiment of the present invention, any multiple modules of the first acquisition module 410, the prediction module 420 and the early warning module 430 can be combined into one module for implementation, or any one of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present invention, at least one of the first acquisition module 410, the prediction module 420 and the early warning module 430 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or implemented in any one of the three implementation methods of software, hardware and firmware or in any appropriate combination of any of them. Alternatively, at least one of the first acquisition module 410, the prediction module 420 and the early warning module 430 can be at least partially implemented as a computer program module, and when the computer program module is run, the corresponding function can be executed. The first acquisition module 410, the prediction module 420 and the warning module 430 communicate via IIC, RS485, RS232 or CAN.

[0149] It should be noted that the early warning device for abnormal charging and discharging of a lithium-ion battery in the embodiment of the present invention corresponds to the early warning method for abnormal charging and discharging of a lithium-ion battery in the embodiment of the present invention. The description of the early warning device for abnormal charging and discharging of a lithium-ion battery specifically refers to the early warning method for abnormal charging and discharging of a lithium-ion battery, which will not be repeated here.

[0150] Based on the above model training method, the present invention also provides a model training device. Figure 5 The device is described in detail.

[0151] Figure 5 A structural block diagram of a model training device according to an embodiment of the present invention is shown.

[0152] like Figure 5 As shown, the device 500 includes a second acquisition module 510 , a processing module 520 , an output module 530 , a loss calculation module 540 and an adjustment module 550 .

[0153] The second acquisition module 510 is used to acquire sample voltage information and sample current information of the sample lithium-ion battery during the charging / discharging process. In one embodiment, the second acquisition module 510 can be used to perform the operation S310 described above, which will not be described in detail here.

[0154] The processing module 520 is used to process the sample voltage information and the sample current information to generate sample state of charge information. In one embodiment, the processing module 520 can be used to perform the operation S320 described above, which will not be described in detail here.

[0155] The output module 530 is used to input the sample voltage information and the sample current information into the initial model, and output the probability distribution information of each sample state of charge prediction value. In one embodiment, the output module 530 can be used to perform the operation S330 described above, which will not be repeated here.

[0156] The loss calculation module 540 is used to calculate the loss gradient based on the target loss function according to the mean of each sample state of charge prediction value, the variance of each sample state of charge prediction value and the sample state of charge information. In one embodiment, the loss calculation module 540 can be used to perform the operation S340 described above, which will not be repeated here.

[0157] The adjustment module 550 is used to adjust the parameters of the initial model based on the loss gradient using the natural gradient optimization method until the convergence condition is reached to obtain the target model. In one embodiment, the adjustment module 550 can be used to perform the operation S350 described above, which will not be repeated here.

[0158] According to an embodiment of the present invention, any multiple modules among the second acquisition module 510, the processing module 520, the output module 530, the loss calculation module 540 and the adjustment module 550 can be combined into one module for implementation, or any one of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present invention, at least one of the second acquisition module 510, the processing module 520, the output module 530, the loss calculation module 540 and the adjustment module 550 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or implemented in any one of the three implementation methods of software, hardware and firmware or in a proper combination of any of them. Alternatively, at least one of the second acquisition module 510, the processing module 520, the output module 530, the loss calculation module 540 and the adjustment module 550 can be at least partially implemented as a computer program module, and when the computer program module is executed, the corresponding function can be performed.

[0159] It should be noted that the model training device in the embodiment of the present invention corresponds to the model training method in the embodiment of the present invention. The description of the model training device part specifically refers to the model training method part, which will not be repeated here.

[0160] Figure 6 A block diagram of an electronic device suitable for implementing an early warning method or a model training method for abnormal charging and discharging of a lithium-ion battery according to an embodiment of the present invention is shown.

[0161] like Figure 6 As shown, the electronic device 600 according to an embodiment of the present invention includes a processor 601, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 602 or the program loaded from the storage part 608 to the RAM (Random Access Memory). The processor 601 may include, for example, a general-purpose microprocessor (such as a CPU), an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (for example, an application-specific integrated circuit (ASIC)), etc. The processor 601 may also include an onboard memory for caching purposes. The processor 601 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present invention.

[0162] In RAM 603, various programs and data required for the operation of electronic device 600 are stored. Processor 601, ROM 602 and RAM 603 are connected to each other via bus 604. Processor 601 performs various operations of the method flow according to the embodiment of the present invention by executing the program in ROM 602 and / or RAM 603. It should be noted that the program can also be stored in one or more memories other than ROM 602 and RAM 603. Processor 601 can also perform various operations of the method flow according to the embodiment of the present invention by executing the program stored in the one or more memories.

[0163] According to an embodiment of the present invention, the electronic device 600 may further include an input / output (I / O) interface 605, which is also connected to the bus 604. The electronic device 600 may further include one or more of the following components connected to the input / output (I / O) interface 605: an input portion 606 including a keyboard, a mouse, etc.; an output portion 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 608 including a hard disk, etc.; and a communication portion 609 including a network interface card such as a LAN card, a modem, etc. The communication portion 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the input / output (I / O) interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed, so that a computer program read therefrom is installed into the storage portion 608 as needed.

[0164] The present invention also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiment; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present invention is implemented.

[0165] According to an embodiment of the present invention, the computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, it may include but is not limited to: a portable computer disk, a hard disk, a RAM (Random Access Memory), a ROM (Read Only Memory), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, the computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, an apparatus or a device. For example, according to an embodiment of the present invention, the computer-readable storage medium may include the ROM 602 and / or RAM 603 described above and / or one or more memories other than ROM 602 and RAM 603.

[0166] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0167] It will be appreciated by those skilled in the art that the features described in the various embodiments of the present invention may be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in the present invention. In particular, without departing from the spirit and teachings of the present invention, the features described in the various embodiments of the present invention may be combined and / or combined in various ways. All of these combinations and / or combinations fall within the scope of the present invention.

[0168] The embodiments of the present invention are described above. However, these embodiments are only for the purpose of illustration, and are not intended to limit the scope of the present invention. Although each embodiment is described above, it does not mean that the measures in each embodiment cannot be used in combination advantageously. Without departing from the scope of the present invention, those skilled in the art may make various substitutions and modifications, which should all fall within the scope of the present invention.

Claims

1. A lithium-ion battery charging and discharging abnormality early warning method, characterized in that: The method comprises: Obtain voltage information and current information of the target lithium-ion battery during the charge / discharge process; Input the voltage information and the current information into a trained target model, and output the state of charge probability distribution information of the target lithium-ion battery; wherein the state of charge probability distribution information includes the mean of each state of charge prediction information and the variance of each state of charge prediction information; and generating warning information according to the mean of each state of charge prediction information and the variance of each state of charge prediction information; The target model is trained according to the following steps: Obtaining sample voltage information and sample current information of a sample lithium-ion battery during a charge / discharge process; Processing the sample voltage information and the sample current information to generate sample state of charge information; Input the sample voltage information and the sample current information into an initial model, and output probability distribution information of each sample state of charge prediction value; wherein the probability distribution information includes the mean of each sample state of charge prediction value and the variance of each sample state of charge prediction value; Based on the target loss function, a loss gradient is calculated according to the mean of the sample state of charge prediction values, the variance of the sample state of charge prediction values ​​and the sample state of charge information; the loss gradient includes the loss gradient of the mean and the loss gradient of the variance; Based on the loss gradient of the mean and the target matrix, a gradient value of the mean is obtained; wherein the target matrix is ​​generated based on the second-order derivative of the target loss function with respect to the model parameters; Based on the loss gradient of the variance and the target matrix, obtaining a gradient value of the variance; and Based on the gradient value of the mean, the gradient value of the variance and the predetermined adjustment step, the model parameters for predicting the mean and the model parameters for predicting the variance in the initial model are correspondingly updated until a convergence condition is reached to obtain a target model.

2. The early warning method according to claim 1, characterized in that: The generating early warning information according to the mean of each state of charge prediction information and the variance of each state of charge prediction information includes: generating probability interval information at a target confidence level according to the mean of each state of charge prediction information and the variance of each state of charge prediction information; and In response to the probability interval information exceeding predetermined probability interval information, the warning information is generated.

3. The early warning method according to claim 2, characterized in that: The probability interval information includes an upper limit value of the probability interval and a lower limit value of the probability interval; the predetermined probability interval information includes an upper limit value of the predetermined probability interval and a lower limit value of the predetermined probability interval; The generating the warning information in response to the probability interval information exceeding the predetermined probability interval information comprises: During the charging process of the target lithium-ion battery, in response to the upper limit value of the probability interval being greater than the upper limit value of the predetermined probability interval, generating warning information for indicating overcharging; During the discharge of the target lithium-ion battery, in response to the lower limit value of the probability interval being less than the lower limit value of the predetermined probability interval, warning information for characterizing over-discharge is generated.

4. The early warning method according to claim 2, characterized in that: The generating the probability interval information under the target confidence level according to the mean of each state of charge prediction information and the variance of each state of charge prediction information includes: generating a variance intermediate value according to the variance and a predetermined parameter; Determine the difference between the mean and the middle value of the variance as the lower limit value of the probability interval; Determine the sum of the mean and the middle value of the variance as the upper limit of the probability interval; The probability interval information is generated according to the probability interval lower boundary value and the probability interval upper boundary value.

5. A lithium-ion battery charging and discharging abnormality warning device, characterized in that: The device comprises: A first acquisition module is used to acquire voltage information and current information of a target lithium-ion battery during a charge / discharge process; A prediction module is used to input the voltage information and the current information into a trained target model, and output the probability distribution information of the state of charge of the target lithium-ion battery; wherein the probability distribution information of the state of charge includes the mean of each state of charge prediction information and the variance of each state of charge prediction information; each state of charge prediction information obeys a normal distribution, wherein the target model is trained according to the following steps: obtaining sample voltage information and sample current information of a sample lithium-ion battery during a charge / discharge process; processing the sample voltage information and the sample current information to generate sample state of charge information; inputting the sample voltage information and the sample current information into an initial model, and outputting the probability distribution information of each sample state of charge prediction value; wherein the probability distribution information includes the mean of each sample state of charge prediction value value and the variance of each sample state of charge prediction value; based on the target loss function, calculate the loss gradient according to the mean of each sample state of charge prediction value, the variance of each sample state of charge prediction value and the sample state of charge information; the loss gradient includes the loss gradient of the mean and the loss gradient of the variance; based on the loss gradient of the mean and the target matrix, obtain the gradient value of the mean; wherein the target matrix is ​​generated based on the second-order derivative of the target loss function with respect to the model parameters; based on the loss gradient of the variance and the target matrix, obtain the gradient value of the variance; and based on the gradient value of the mean, the gradient value of the variance and the predetermined adjustment step size, respectively, correspondingly update the model parameters for predicting the mean and the model parameters for predicting the variance in the initial model until the convergence condition is reached to obtain the target model; and The early warning module is used to generate early warning information according to the mean of each state of charge prediction information and the variance of each state of charge prediction information.

6. An electronic device comprising: one or more processors; a memory for storing one or more computer programs, It is characterized in that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program or instruction stored thereon, characterized in that: When the computer program or instruction is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images