Fault prediction method, device and system for battery module

Abstract

The invention discloses a fault prediction method, device and system for a battery module, and relates to the technical field of batteries, and the method comprises the steps: obtaining the target operation data of the battery module and a unique identifier corresponding to the battery module under the condition of determining that the battery module of an energy storage system does not have a fault; inputting the target operation data into a preset fault prediction model corresponding to the unique identifier to obtain a fault occurrence probability output by the fault prediction model; and when it is determined that the fault occurrence probability is greater than the preset probability value, it is determined that the battery module will have a fault in the future. The method is used for solving the problems that in the prior art, when the battery module is managed, the fault prediction accuracy is low, and fault pre-judgment cannot be achieved.

Description

Technical Field

[0001] The present application relates to the field of battery technology, and in particular to a method, device and system for predicting faults of a battery module. Background Art

[0002] With the increase in user electricity demand and the continuous development of new energy power generation technology, the demand for energy storage technology is getting higher and higher. The application scenarios of energy storage technology include multiple scenarios such as industry, enterprise and home. For enterprise and home scenarios, due to the diversity of demand, different environmental conditions, and geographical dispersion, distributed energy storage is generally used.

[0003] In the prior art, the management of distributed energy storage still adopts a unified management method, in which the fault detection management means determines whether there is a fault in the battery module of the energy storage system by comparing the battery parameters with the preset threshold. However, due to the existence of random characteristics inside the battery cells of the battery module, the battery cells are different in different environments, so the accuracy of the existing fault detection method is low.

[0004] In addition, the existing technology can only ensure timely detection of faults when a battery module fails. However, since the battery module has already failed, the energy storage system cannot operate normally, which may cause serious consequences. Summary of the invention

[0005] In response to the above-mentioned problems and technical needs, the applicant has proposed a battery module fault prediction method, device and system to solve the problems of low fault prediction accuracy and inability to predict faults when managing battery modules in the prior art.

[0006] The present application provides a method for predicting a battery module failure, the method comprising:

[0007] When it is determined that the battery module of the energy storage system has not failed, obtaining target operation data of the battery module and a unique identifier corresponding to the battery module, wherein the target operation data includes: current operation data at the current moment and historical operation data corresponding to a preset time period before the current moment;

[0008] Input the target operation data into a preset fault prediction model corresponding to the unique identifier to obtain the fault occurrence probability output by the fault prediction model, wherein the fault prediction model is trained based on the target operation data sample and the fault occurrence probability sample corresponding to the unique identifier, and the target operation data sample includes: an operation data sample corresponding to time t and a historical operation data sample corresponding to a preset time period before time t;

[0009] When it is determined that the probability of the failure occurring is greater than a preset probability value, it is determined that the battery module will fail in the future.

[0010] According to a method for predicting a fault of a battery module according to an embodiment of the present application, the target operation data includes: a current value, a voltage value, and a temperature value;

[0011] The step of inputting the target operation data into a preset fault prediction model corresponding to the unique identifier to obtain a fault occurrence probability output by the fault prediction model includes:

[0012] Inputting the target operation data into the fault prediction model corresponding to the unique identifier, inputting the current value and the temperature value into the objective function through the fault prediction model, and obtaining the calculated voltage value output by the objective function; calculating the calculated difference between the calculated voltage value and the voltage value; obtaining and outputting the fault occurrence probability based on the calculated difference;

[0013] Wherein, the objective function is a preset mathematical function.

[0014] According to a method for predicting a fault of a battery module according to an embodiment of the present application, calculating the calculated voltage value and the calculated difference between the voltage values ​​includes:

[0015] Inputting the calculated voltage value and the voltage value corresponding to each time point in the target operation data into a preset first deviation formula to obtain a calculated difference value output by the first deviation formula;

[0016] Wherein, the first deviation formula includes:

[0017]

[0018] Among them, D t represents the calculated difference, n represents the total number of time points, V t ’ -i Indicates the calculated voltage value, V t-i Indicates voltage value.

[0019] According to a method for predicting a battery module failure in one embodiment of the present application, after determining that the battery module will fail in the future, the method further includes:

[0020] Add a fault mark to the current job data;

[0021] Based on the fault marker, the fault prediction model is optimized using the target operation data.

[0022] According to a fault prediction method for a battery module according to an embodiment of the present application, the training process of the fault prediction model includes:

[0023] Acquire the target operation data sample and the fault occurrence probability sample;

[0024] Inputting the target operation data sample and the fault occurrence probability sample into a fault prediction model to obtain a predicted fault occurrence probability sample output by the fault prediction model;

[0025] Calculating the probability prediction difference between the fault occurrence probability sample and the predicted fault occurrence probability sample based on a preset second deviation formula, and calculating the prediction accuracy value of the fault prediction model based on the precision index indication value, the recall index indication value and a preset scoring formula;

[0026] The precision index indicator value is the ratio of the predicted positive sample to the target operation data sample when the probability prediction difference corresponding to the predicted positive sample in the target operation data sample is less than a preset value; the recall index indicator value is the ratio of the actual positive sample to the target operation data sample when the probability prediction difference corresponding to the actual positive sample in the target operation data sample is less than a preset value;

[0027] Based on the probability prediction difference and the prediction accuracy value, the model parameters of the fault prediction model are adjusted until the probability prediction difference is less than the preset value and the prediction accuracy value is greater than the preset accuracy value.

[0028] According to a method for predicting a fault of a battery module according to an embodiment of the present application, obtaining the target operation data sample includes:

[0029] The target operation data samples are extracted from the operation data of multiple battery modules based on a preset sample collection rule.

[0030] The present application also provides a battery module fault prediction device, including:

[0031] An acquisition module, configured to acquire target operation data of the battery module and a unique identifier corresponding to the battery module when it is determined that the battery module of the energy storage system has not failed, wherein the target operation data includes: current operation data at the current moment and historical operation data corresponding to a preset time period before the current moment;

[0032] A prediction module, used for inputting the target operation data into a preset fault prediction model corresponding to the unique identifier, and obtaining the fault occurrence probability output by the fault prediction model, wherein the fault prediction model is trained based on the target operation data sample and the fault occurrence probability sample corresponding to the unique identifier, and the target operation data sample includes: an operation data sample corresponding to time t and a historical operation data sample corresponding to a preset time period before time t;

[0033] The determination module is used to determine that the battery module will fail in the future when it is determined that the probability of the failure is greater than a preset probability value.

[0034] The embodiment of the present application further provides a battery module fault prediction system, comprising: a battery module fault prediction device, a data communication device and a cloud device as described in the above embodiment, wherein the fault prediction device and the cloud device are communicatively connected via the data communication device;

[0035] The fault prediction device is used to add a fault mark to the current operation data when it is determined that the battery module will fail in the future, and send the data added with the fault mark and the unique identifier of the battery module to the cloud device through the data communication device;

[0036] The cloud device is used to obtain target operation data corresponding to current operation data based on the fault mark, and optimize the fault prediction model based on the target operation data.

[0037] According to a fault prediction system for a battery module according to an embodiment of the present application, the fault prediction device is further used to determine a target cloud application of a cloud device that is closest to the battery module;

[0038] The fault prediction device is used to send the data added with the fault mark and the unique identifier of the battery module to the target cloud application through the data communication device.

[0039] According to a fault prediction system for a battery module according to an embodiment of the present application, the target cloud application includes: a model training unit and a model distribution unit;

[0040] The module training unit is used to optimize the fault prediction model based on the target operation data, and send the optimized fault prediction model to the model distribution unit;

[0041] The model distribution unit is used to send the optimized fault prediction device to the fault prediction device with a unique identifier corresponding to the target operation data.

[0042] The battery module fault prediction method, device and system provided in the embodiments of the present application obtain the current operation data of the battery module at the current moment, the historical operation data corresponding to the preset time period before the current moment and the unique identifier corresponding to the battery module when it is determined that the battery module of the energy storage system has not failed; the current operation data and the historical operation data are input into the preset fault prediction model corresponding to the unique identifier to obtain the fault occurrence probability output by the fault prediction model. Different battery modules of the present application correspond to different fault prediction models, and the fault prediction model is trained based on the target operation data samples of the corresponding battery module, which can make good use of the characteristics of the battery module so that each fault prediction model forms a personalized prediction, thereby improving the accuracy of fault prediction. The input data of the present application is the operation data within a period of time, and it can predict whether the battery module will fail in the future based on the changes in the operation data (the operation status of the battery module); and when it is determined that the probability of failure is greater than the preset probability value, it is determined that the battery module will fail in the future, thereby achieving the purpose of battery module failure prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0044] Figure 1 It is a flowchart of a method for predicting a battery module failure provided in an embodiment of the present application;

[0045] Figure 2 is a structural schematic diagram of a fault prediction device for a battery module provided in an embodiment of the present application;

[0046] Figure 3 is a structural schematic diagram of a fault prediction system for a battery module provided in an embodiment of the present application;

[0047] Figure 4 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] The present application embodiment provides a method for predicting a battery module failure, which can be applied to a smart terminal. Some other descriptions in the present application embodiment are for illustrative purposes only and are not intended to limit the scope of protection of the present application. They will not be described one by one. The specific implementation of the method is as follows: Figure 1 As shown:

[0050] Step 101 : when it is determined that the battery module of the energy storage system has not failed, obtain target operation data of the battery module and a unique identifier corresponding to the battery module.

[0051] The target operation data includes: current operation data at the current moment and historical operation data corresponding to a preset time period before the current moment.

[0052] The unique identifier includes the IP address of the battery module.

[0053] The target operation data include: current value, voltage value and temperature value.

[0054] Step 102: input the target operation data into a preset fault prediction model corresponding to the unique identifier to obtain the fault occurrence probability output by the fault prediction model.

[0055] Among them, the fault prediction model is trained based on the target operation data samples and fault occurrence probability samples corresponding to the unique identifier, and the target operation data samples include: the operation data samples corresponding to time t and the historical operation data samples corresponding to the preset time period before time t.

[0056] Among them, different battery modules correspond to different fault prediction models.

[0057] Step 103 , when it is determined that the probability of the failure occurring is greater than a preset probability value, it is determined that the battery module will fail in the future.

[0058] Specifically, when it is determined that the probability of a fault occurring is less than or equal to a preset probability value, based on the prediction result of the target operation data acquired at the current moment, a fault will not occur in the future.

[0059] Specifically, the essence of this application is to achieve self-diagnosis of battery module failures, with the focus on predicting and intervening before the battery cells of the battery module may fail. The battery cells perform different charging and discharging operations in different environments, and the charging and discharging operations will affect their lifespan. If there are some subtle changes inside the battery cell, if no intervention is made and the charging and discharging operations are continued, it may cause a short circuit problem inside the battery cell, thereby causing irreversible damage to the battery cell or causing a greater risk of thermal runaway.

[0060] Generally, the detection parameters of the battery cell are current, voltage and temperature, so the fault prediction module of this application is trained based on these three parameters. Voltage refers to the result presentation at a specified temperature and current, because the objective function is pre-set. The objective function is shown in formula (1):

[0061] V t ′ =F t (I,T)…………………………………………(1)

[0062] Among them, V t ′ Indicates the calculated voltage value, that is, the voltage at time t is the instantaneous presentation of the specified temperature and the current at time t. I represents current, T represents temperature, and F represents temperature. t Represents a functional relationship.

[0063] The functional relationship is obtained based on Ohm's law and the temperature coefficient of resistance, and is not described in detail in this application.

[0064] The battery module fault prediction method provided in the embodiment of the present application obtains the current operation data of the battery module at the current moment, the historical operation data corresponding to the preset time period before the current moment, and the unique identifier corresponding to the battery module when it is determined that the battery module of the energy storage system has not failed; the current operation data and the historical operation data are input into the preset fault prediction model corresponding to the unique identifier to obtain the fault occurrence probability output by the fault prediction model. Different battery modules of the present application correspond to different fault prediction models, and the fault prediction model is trained based on the target operation data samples of the corresponding battery module, which can make good use of the characteristics of the battery module so that each fault prediction model forms a personalized prediction, thereby improving the accuracy of fault prediction. The input data of the present application is the operation data within a period of time, and it can predict whether the battery module will fail in the future based on the changes in the operation data (the operation status of the battery module); and when it is determined that the probability of failure is greater than the preset probability value, it is determined that the battery module will fail in the future, thereby achieving the purpose of battery module failure prediction.

[0065] In a specific embodiment, the specific implementation of inputting the target operation data into a preset fault prediction model corresponding to the unique identifier to obtain the fault occurrence probability output by the fault prediction model includes:

[0066] The target operation data is input into the fault prediction model corresponding to the unique identifier, and the current value and the temperature value are input into the objective function through the fault prediction model to obtain the calculated voltage value output by the objective function; the calculated voltage value and the calculated voltage value are calculated; and the probability of fault occurrence is obtained and output based on the calculated difference.

[0067] In a specific embodiment, the specific implementation of calculating the calculated voltage value and the calculated difference of the voltage values ​​includes:

[0068] The calculated voltage value and the voltage value corresponding to each time point in the target operation data are input into a preset first deviation formula to obtain a calculated difference value output by the first deviation formula.

[0069] The first deviation formula is shown in formula (2):

[0070]

[0071] Among them, D t represents the calculated difference, n represents the total number of time points, V t ’ -i Indicates the calculated voltage value, V t-i Indicates voltage value.

[0072] Specifically, a correspondence between the calculated difference and the probability of fault occurrence is preset, and the probability of fault occurrence corresponding to the calculated difference is obtained based on the correspondence.

[0073] Specifically, the fault prediction model of the present application is equivalent to a classifier, and its purpose is to determine whether a fault will occur within a certain period of time in the future given a real-time collected value, and specifically by calculating the deviation between the calculated voltage value and the collected voltage value. In other words, the deviation between the trajectory of the calculated voltage value and the trajectory of the collected voltage value. Due to the volatility of the battery cell operation, the accuracy of a single sampling is low. The present application uses a period of data for comparison, which is specifically calculated by formula (2).

[0074] In a specific embodiment, after determining that the battery module will fail in the future, a fault mark is added to the current operation data; based on the fault mark, the target operation data is used to optimize the fault prediction model.

[0075] Specifically, when it is predicted that the battery module will fail in the future, a fault marker is added to the current operation data and used as a negative sample to update the fault prediction model, thereby improving the personalization and prediction accuracy of the fault prediction model.

[0076] In a specific embodiment, the training process of the fault prediction model includes:

[0077] Obtain target operation data samples and fault occurrence probability samples; input the target operation data samples and fault occurrence probability samples into the fault prediction model to obtain predicted fault occurrence probability samples output by the fault prediction model; calculate the probability prediction difference between the fault occurrence probability samples and the predicted fault occurrence probability samples based on a preset second deviation formula, and calculate the prediction accuracy of the fault prediction model based on the precision index indication value, the recall index indication value and the preset scoring formula; adjust the model parameters of the fault prediction model based on the probability prediction difference and the prediction accuracy value until the probability prediction difference is less than the preset value and the prediction accuracy value is greater than the preset accuracy value.

[0078] Among them, the precision index indication value is the ratio of the predicted positive sample to the target operation data sample when the probability prediction difference corresponding to the predicted positive sample in the target operation data sample is less than the preset value; the recall index indication value is the ratio of the actual positive sample to the target operation data sample when the probability prediction difference corresponding to the actual positive sample in the target operation data sample is less than the preset value.

[0079] Specifically, after obtaining the target operation data sample, a preprocessing operation is performed on the data, including denoising and normalization processing.

[0080] Specifically, the fault prediction model obtains the calculated voltage value samples and the calculated difference samples of the voltage value samples corresponding to each time point through formula (2), and obtains the fault occurrence probability samples corresponding to the calculated difference samples based on the preset corresponding relationship.

[0081] The second deviation formula is shown in formula (3):

[0082]

[0083] Among them, P t represents the probability prediction difference, n represents the total number of time points, P t ’ -i represents the predicted failure probability sample, P t-i represents the probability sample of failure occurrence.

[0084] Specifically, since battery modules do not fail frequently, there are many positive samples and relatively few negative samples in the target operation data samples when training the fault prediction model. Therefore, in order to ensure the prediction accuracy of the fault prediction model, the model is scored based on the recall rate (recall index indicator value) and precision rate (precision index indicator value).

[0085] The scoring formula is shown in formula (4):

[0086]

[0087] Among them, P t It represents the prediction accuracy value, W represents the precision index indicator value, and R represents the recall index indicator value.

[0088] In a specific embodiment, the specific implementation of obtaining the target job data sample includes:

[0089] Target operation data samples are extracted from the operation data of multiple battery modules based on preset sample collection rules.

[0090] Specifically, when a battery cell fails, its operating data does not conform to a linear law, and the data closer to time t is more valuable, while data far away from time t may also have potential features. Therefore, this application extracts data for a period of time as target operating data samples.

[0091] For example, the time S hours before the fault occurs is taken as the farthest distance point of the fault, and the period from the time point corresponding to the farthest distance point to time t is divided into M areas. The number of sampling points in each area is different, and the closer the area is to time t, the larger the number of sampling points.

[0092] For example, S is 8 hours, which is divided into 4 areas, including: the first 8 hours to the first 1 hour, the first 1 hour to the first 10 minutes, the first 10 minutes to the first 2 minutes, and the first 2 minutes to the first 1 second. The number of sampling points in the four areas is distributed as 5%, 10%, 25% and 60%. Of course, this is only an example, and the distribution ratio can be changed. Users can set it according to their actual needs.

[0093] In addition, the length of the time period can also be increased or decreased. Increasing the time length can mine more data features, but the prediction accuracy will be relatively reduced. This application does not impose any restrictions on the length of the time period, and users can set it according to their actual needs.

[0094] In addition, the present application pre-trains a general fault prediction model, deploys and starts the general fault prediction model, and in the subsequent prediction process, uses the operating conditions of each battery module to optimize its own fault prediction model to ensure the personalization of the model and the prediction accuracy.

[0095] The present application also provides a battery module fault prediction device. The specific implementation of the device can refer to the description of the battery module fault prediction method, and the repeated parts will not be repeated. Figure 2 As shown, the device comprises:

[0096] The acquisition module 201 is used to acquire the target operation data of the battery module and the unique identifier corresponding to the battery module when it is determined that the battery module of the energy storage system has not failed, wherein the target operation data includes: the current operation data at the current moment and the historical operation data corresponding to the preset time period before the current moment;

[0097] The prediction module 202 is used to input the target operation data into a preset fault prediction model corresponding to the unique identifier, and obtain the fault occurrence probability output by the fault prediction model, wherein the fault prediction model is trained based on the target operation data sample and the fault occurrence probability sample corresponding to the unique identifier, and the target operation data sample includes: the operation data sample corresponding to time t and the historical operation data sample corresponding to the preset time period before time t;

[0098] The determination module 203 is used to determine that the battery module will fail in the future when it is determined that the probability of failure is greater than a preset probability value.

[0099] In a specific embodiment, the target operation data includes: current value, voltage value and temperature value; the prediction module 202 is used to input the target operation data into the fault prediction model corresponding to the unique identifier, input the current value and temperature value into the objective function through the fault prediction model, and obtain the calculated voltage value output by the objective function; calculate the calculated voltage value and the calculated voltage value; obtain and output the probability of fault occurrence based on the calculated difference; wherein the objective function is a pre-set mathematical function.

[0100] In a specific embodiment, the prediction module 202 is used to input the calculated voltage value and the voltage value corresponding to each time point in the target operation data into a preset first deviation formula to obtain a calculated difference value output by the first deviation formula;

[0101] Among them, the first deviation formula includes:

[0102]

[0103] Among them, D t represents the calculated difference, n represents the total number of time points, V t ’ -i Indicates the calculated voltage value, V t-i Indicates voltage value.

[0104] In a specific embodiment, the device also includes a training module for adding fault tags to current operation data; based on the fault tags, the fault prediction model is optimized using target operation data.

[0105] In a specific embodiment, the training module is also used to obtain target operation data samples and fault occurrence probability samples; input the target operation data samples and fault occurrence probability samples into the fault prediction model to obtain the predicted fault occurrence probability samples output by the fault prediction model; calculate the probability prediction difference between the fault occurrence probability samples and the predicted fault occurrence probability samples based on a preset second deviation formula, and calculate the prediction accuracy of the fault prediction model based on the precision index indicator value, the recall index indicator value and the preset scoring formula; wherein the precision index indicator value is the ratio of the predicted positive sample to the target operation data sample when the probability prediction difference corresponding to the predicted positive sample in the target operation data sample is less than the preset value; the recall index indicator value is the ratio of the actual positive sample to the target operation data sample when the probability prediction difference corresponding to the actual positive sample in the target operation data sample is less than the preset value; based on the probability prediction difference and the prediction accuracy value, adjust the model parameters of the fault prediction model until the probability prediction difference is less than the preset value and the prediction accuracy value is greater than the preset accuracy value.

[0106] In a specific embodiment, the acquisition module 201 is used to extract target operation data samples from the operation data of multiple battery modules based on a preset sample collection rule.

[0107] The present application also provides a battery module fault prediction system. The specific implementation of the system can refer to the description of the battery module fault prediction method and the battery module fault prediction device. The repeated parts will not be repeated. Figure 3 As shown, the system includes: a fault prediction device 301 for a battery module, a data communication device 302 and a cloud device 303 , and the fault prediction device 301 and the cloud device 303 are connected for communication via the data communication device 302 .

[0108] The fault prediction device 301 is used to add a fault mark to the current operation data when it is determined that the battery module will fail in the future, and send the data with the fault mark and the unique identifier of the battery module to the cloud device 303 through the data communication device 302.

[0109] The cloud device 303 is used to obtain target operation data corresponding to the current operation data based on the fault mark, and optimize the fault prediction model based on the target operation data.

[0110] Specifically, the fault prediction device 301 also includes a battery management unit for managing the battery module. The data communication device 302 includes wireless, Bluetooth, or wired communication methods, and can send the collected data to the cloud device 303, or transmit the data sent from the cloud to the fault prediction device 301. The data packet loss, network connection, and network switching are all completed by the data communication device 302 to ensure data integrity and connection reliability.

[0111] In a specific embodiment, the fault prediction device 301 is also used to determine the target cloud application of the cloud device 303 that is closest to the battery module; the fault prediction device 301 is used to send the data with the fault mark added and the unique identifier of the battery module to the target cloud application through the data communication device 302.

[0112] In a specific embodiment, the target cloud application includes: a model training unit and a model distribution unit. The model training unit is used to optimize the fault prediction model based on the target operation data and send the optimized fault prediction model to the model distribution unit; the model distribution unit is used to send the optimized fault prediction device to the fault prediction device with a unique identifier corresponding to the target operation data.

[0113] Specifically, the amount of calculation in the entire model training process is relatively large, so the present application uses the cloud to train the model, and through the collaborative collaboration of the terminal and the cloud, the model training and the prediction of the probability of failure are achieved.

[0114] First, the fault prediction model is trained based on a large number of target operation data samples through cloud services in the cloud, and the fault prediction model is sent to each fault prediction device.

[0115] Then, the fault prediction device collects the operating data at different time points, and uses the stored fault prediction device and operating data to predict the probability of fault occurrence, and transmits the operating data to the target cloud application corresponding to the battery module in real time. In addition, a fault mark is added to the operating data at the time point when a fault is predicted to occur in the future, and it is also sent to the target cloud application corresponding to the battery module. Of course, the device physical information of the battery module can also be uploaded to the target cloud application at the same time as uploading. Alternatively, the unique identifier, operating data or operating data with a fault mark can also be transmitted to the cloud, and the cloud service on the cloud forwards the data to the target cloud application closest to the unique mark based on the deployment of each cloud application.

[0116] Finally, the target cloud application receives the new job data, updates the fault prediction model based on the new job data, and sends the updated fault prediction model to the corresponding fault prediction device. The updated fault prediction model can be sent to the fault prediction device by setting the number of iterations. The number of iterations can be set by the user according to his actual needs, and this application does not impose any restrictions.

[0117] The terminal and the cloud of this application do not affect each other. The terminal predicts the probability of fault occurrence based on the fault prediction model sent down, and the cloud optimizes the fault prediction model based on the uploaded data. The two are carried out simultaneously, which can ensure both the timely prediction of the fault prediction model and the accuracy of its prediction. In addition, a general model is initially adopted, and with the rapid deployment and startup of the model, as the battery module runs, the cloud will train and update the model according to the specific situation of each battery module itself, and gradually form a personalized model.

[0118] Figure 4 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 4 As shown, the electronic device may include: a processor 401, a communication interface 402, a memory 403 and a communication bus 404, wherein the processor 401, the communication interface 402 and the memory 403 communicate with each other through the communication bus 404. The processor 401 may call the logic instructions in the memory 403 to execute the fault prediction method of the battery module, the method comprising: obtaining the target operation data of the battery module and the unique identifier corresponding to the battery module when it is determined that the battery module of the energy storage system has not failed, wherein the target operation data comprises: the current operation data at the current moment and the historical operation data corresponding to the preset time period before the current moment; inputting the target operation data into the preset fault prediction model corresponding to the unique identifier to obtain the fault occurrence probability output by the fault prediction model, wherein the fault prediction model is trained based on the target operation data sample and the fault occurrence probability sample corresponding to the unique identifier, and the target operation data sample comprises: the operation data sample corresponding to the moment t and the historical operation data sample corresponding to the preset time period before the moment t; when it is determined that the fault occurrence probability is greater than the preset probability value, determining that the battery module will fail in the future.

[0119] In addition, the logic instructions in the above-mentioned memory 403 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0120] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the battery module fault prediction method provided by the above methods, the method including: when it is determined that the battery module of the energy storage system has not failed, obtaining target operation data of the battery module and a unique identifier corresponding to the battery module, wherein the target operation data includes: current operation data at the current moment and historical operation data corresponding to a preset time period before the current moment; inputting the target operation data into a preset fault prediction model corresponding to the unique identifier to obtain a fault occurrence probability output by the fault prediction model, wherein the fault prediction model is trained based on the target operation data samples and fault occurrence probability samples corresponding to the unique identifier, and the target operation data samples include: operation data samples corresponding to time t and historical operation data samples corresponding to the preset time period before time t; when it is determined that the fault occurrence probability is greater than the preset probability value, determining that the battery module will fail in the future.

[0121] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the processor executes the battery module fault prediction method provided by the above-mentioned embodiments, the method comprising: when it is determined that the battery module of the energy storage system has not failed, obtaining target operation data of the battery module and a unique identifier corresponding to the battery module, wherein the target operation data comprises: current operation data at the current moment and historical operation data corresponding to a preset time period before the current moment; inputting the target operation data into a preset fault prediction model corresponding to the unique identifier to obtain a fault occurrence probability output by the fault prediction model, wherein the fault prediction model is trained based on the target operation data samples and the fault occurrence probability samples corresponding to the unique identifier, and the target operation data samples comprise: operation data samples corresponding to time t and historical operation data samples corresponding to the preset time period before time t; when it is determined that the fault occurrence probability is greater than the preset probability value, determining that the battery module will fail in the future.

[0122] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0123] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0124] Finally, it should be noted that the above is only the preferred implementation of the present application, and the present application is not limited to the above embodiments. It is understood that other improvements and changes directly derived or associated by those skilled in the art without departing from the spirit and concept of the present application should be considered to be included in the scope of protection of the present application.

Claims

1. A method for predicting a battery module failure, characterized in that: The method comprises: When it is determined that the battery module of the energy storage system has not failed, obtaining target operation data of the battery module and a unique identifier corresponding to the battery module, wherein the target operation data includes: current operation data at the current moment and historical operation data corresponding to a preset time period before the current moment; Input the target operation data into a preset fault prediction model corresponding to the unique identifier to obtain the fault occurrence probability output by the fault prediction model, wherein the fault prediction model is trained based on the target operation data sample and the fault occurrence probability sample corresponding to the unique identifier, and the target operation data sample includes: an operation data sample corresponding to time t and a historical operation data sample corresponding to a preset time period before time t; When it is determined that the probability of the failure occurring is greater than a preset probability value, it is determined that the battery module will fail in the future.

2. The method for predicting battery module failure according to claim 1, characterized in that: The target operation data includes: current value, voltage value and temperature value; The step of inputting the target operation data into a preset fault prediction model corresponding to the unique identifier to obtain a fault occurrence probability output by the fault prediction model includes: Inputting the target operation data into the fault prediction model corresponding to the unique identifier, inputting the current value and the temperature value into the objective function through the fault prediction model, and obtaining the calculated voltage value output by the objective function; calculating the calculated difference between the calculated voltage value and the voltage value; obtaining and outputting the fault occurrence probability based on the calculated difference; Wherein, the objective function is a preset mathematical function.

3. The method for predicting battery module failure according to claim 2, characterized in that: Calculating the calculated voltage value and the calculated difference of the voltage value includes: Inputting the calculated voltage value and the voltage value corresponding to each time point in the target operation data into a preset first deviation formula to obtain a calculated difference value output by the first deviation formula; Wherein, the first deviation formula includes: Among them, D t represents the calculated difference, n represents the total number of time points, V t ’ -i Indicates the calculated voltage value, V t-i Indicates voltage value.

4. The method for predicting a battery module failure according to any one of claims 1 to 3, characterized in that: After determining that the battery module will fail in the future, the method further includes: Add a fault mark to the current job data; Based on the fault marker, the fault prediction model is optimized using the target operation data.

5. The method for predicting a battery module failure according to any one of claims 1 to 3, characterized in that: The training process of the fault prediction model includes: Acquire the target operation data sample and the fault occurrence probability sample; Inputting the target operation data sample and the fault occurrence probability sample into a fault prediction model to obtain a predicted fault occurrence probability sample output by the fault prediction model; Calculating the probability prediction difference between the fault occurrence probability sample and the predicted fault occurrence probability sample based on a preset second deviation formula, and calculating the prediction accuracy value of the fault prediction model based on the precision index indication value, the recall index indication value and a preset scoring formula; The precision index indicator value is the ratio of the predicted positive sample to the target operation data sample when the probability prediction difference corresponding to the predicted positive sample in the target operation data sample is less than a preset value; the recall index indicator value is the ratio of the actual positive sample to the target operation data sample when the probability prediction difference corresponding to the actual positive sample in the target operation data sample is less than a preset value; Based on the probability prediction difference and the prediction accuracy value, the model parameters of the fault prediction model are adjusted until the probability prediction difference is less than the preset value and the prediction accuracy value is greater than the preset accuracy value.

6. The method for predicting battery module failure according to claim 5, characterized in that: Acquiring the target job data sample includes: The target operation data samples are extracted from the operation data of multiple battery modules based on a preset sample collection rule.

7. A battery module fault prediction device, characterized in that: The device comprises: An acquisition module, used to acquire target operation data of the battery module and a unique identifier corresponding to the battery module when it is determined that the battery module of the energy storage system has not failed, wherein the target operation data includes: current operation data at the current moment and historical operation data corresponding to a preset time period before the current moment; A prediction module, used for inputting the target operation data into a preset fault prediction model corresponding to the unique identifier, and obtaining the fault occurrence probability output by the fault prediction model, wherein the fault prediction model is trained based on the target operation data sample and the fault occurrence probability sample corresponding to the unique identifier, and the target operation data sample includes: an operation data sample corresponding to time t and a historical operation data sample corresponding to a preset time period before time t; The determination module is used to determine that the battery module will fail in the future when it is determined that the probability of the failure is greater than a preset probability value.

8. A battery module fault prediction system, characterized in that: The system comprises: a fault prediction device for a battery module according to claim 7, a data communication device and a cloud device, wherein the fault prediction device and the cloud device are communicatively connected via the data communication device; The fault prediction device is used to add a fault mark to the current operation data when it is determined that the battery module will fail in the future, and send the data added with the fault mark and the unique identifier of the battery module to the cloud device through the data communication device; The cloud device is used to obtain target operation data corresponding to current operation data based on the fault mark, and optimize the fault prediction model based on the target operation data.

9. The battery module failure prediction system according to claim 8, characterized in that: The fault prediction device is further used to determine a target cloud application of a cloud device that is closest to the battery module; The fault prediction device is used to send the data added with the fault mark and the unique identifier of the battery module to the target cloud application through the data communication device.

10. The battery module failure prediction system according to claim 9, characterized in that: The target cloud application includes: a model training unit and a model distribution unit; The module training unit is used to optimize the fault prediction model based on the target operation data, and send the optimized fault prediction model to the model distribution unit; The model distribution unit is used to send the optimized fault prediction device to the fault prediction device with a unique identifier corresponding to the target operation data.

Images