Lithium-ion battery health status prediction method, system, device and storage medium

Abstract

The present invention discloses a method, system, device and storage medium for predicting the health status of a lithium-ion battery. The existing various methods for predicting the health status of a battery have their own defects. The method adopted by the present invention comprises: obtaining the time series data of the average voltage, average current, average temperature, acquisition time and battery health status during the charge and discharge cycle of the lithium-ion battery, and classifying according to the standard of using the battery health status as the target value and the remaining data as the attribute value; establishing an LSTM model, setting the hyperparameters of the LSTM model; using a data set dimension adjustment method based on Choquet integral to perform non-additive dimension adjustment on the training data set; using the adjusted training data set to pre-train and retrain the LSTM model to obtain a retrained LSTM model; and using the retrained LSTM model to predict the battery health status data at the next moment. The present invention reduces the dimension of the data set while ensuring the prediction accuracy of the model as much as possible, thereby greatly reducing the complexity of the calculation.

Description

Technical Field

[0001] The present invention belongs to the technical field of lithium-ion batteries, and specifically provides a method, system, device and storage medium for predicting the health status of lithium-ion batteries based on non-additive dimensionality adjustment and adaptive updating of machine learning models. Background Art

[0002] As the proportion of renewable energy in the power industry gradually increases, especially wind and solar power generation, the instability of these energy sources has brought challenges to the stable operation of the power grid. Therefore, the demand for efficient energy storage solutions is growing. Among the many energy storage technologies, battery energy storage has attracted widespread attention due to its environmental protection, high efficiency and strong adaptability. In particular, lithium-ion batteries have become the leader in the field of energy storage with their advantages such as long life, fast charging capability, light weight and environmental protection. With the advancement of technology, their cost is gradually decreasing and their application scope is constantly expanding.

[0003] The state of health (SOH) of a lithium-ion battery is a key indicator of the remaining service life of the battery. It reflects the ratio of the battery's current available capacity to its initial capacity. Generally speaking, when the battery's SOH drops to 70%-80%, it is considered that the battery is close to or has reached its end of service life (EOF). In a battery pack, a single battery with degraded performance will affect the operation of the entire system, because the increase in its internal impedance will drag down the performance of other healthy batteries, ultimately leading to a decrease in the capacity of the entire system and an increased risk of failure, which may pose a safety hazard. Therefore, accurately evaluating and predicting the battery's SOH is crucial to ensuring the safe and efficient operation of the battery system.

[0004] At present, there are three main methods for evaluating battery SOH: methods based on battery equivalent circuit models, methods based on battery impedance spectrum analysis, and methods based on data-driven methods. The equivalent circuit model method relies on accurate data on the initial state of the battery, but due to the complexity of the chemical reactions inside the battery, it is challenging to build an accurate model; the impedance spectrum analysis method has high requirements for the operating conditions of the battery, and is affected by factors such as temperature and charge and discharge rates, which may reduce the accuracy of the prediction; and although the data-driven method has high accuracy, it requires a large amount of experimental data for model training and has high requirements for computing resources, which increases costs and may limit its deployment in embedded systems that require real-time computing.

[0005] In general, although there are many ways to evaluate the health state (SOH) of batteries at this stage, each technology faces specific limitations and challenges. The next research work should be committed to seeking a comprehensive solution to improve the accuracy of prediction, reduce related costs, and cope with the time-varying characteristics of SOH. Such efforts will help promote the continuous advancement and widespread application of battery energy storage technology. Summary of the invention

[0006] The technical problem to be solved by the present invention is to overcome the defects of the above-mentioned prior art and provide a lithium-ion battery health status prediction method, system, device and storage medium, which are based on non-additive dimensionality adjustment and adaptive update of machine learning models, while ensuring prediction accuracy, reducing calculation pressure through small batch calculations.

[0007] To this end, the present invention adopts the following technical solution.

[0008] In a first aspect, the present invention adopts a lithium-ion battery health status prediction method, which comprises:

[0009] The average voltage, average current, average temperature, acquisition time and battery health status time series data of the lithium-ion battery during the charge and discharge cycle are obtained, and the battery health status is used as the target value and the remaining data is used as the attribute value. The data is classified to obtain the attribute value data containing multi-dimensional data and the target value data containing only the battery health status;

[0010] Establish a long short-term memory network LSTM model and set the hyperparameters of the LSTM model according to the nonlinear relationship between attribute value data and target value data;

[0011] The acquired attribute value data and target value data are used as training data sets, and the data set dimension adjustment method based on Choquet integral is used to perform non-additive dimension adjustment on the attribute value data in the training data set, so as to change the dimension of the attribute value data in the training data set and extract the attribute value data with the greatest influence on the target value;

[0012] Pre-training and re-training the LSTM model using the adjusted training data set to obtain a re-trained LSTM model;

[0013] Use the retrained LSTM model to predict the battery health status data at the next moment.

[0014] Furthermore, the time series data of the first 1 / 2 length in the training data set is used as pre-training data.

[0015] Furthermore, a dataset dimension adjustment method based on Choquet integral is used to perform non-additive dimension adjustment on the pre-training data in the training dataset, a prediction step size is set, and the adjusted pre-training data is used to pre-train the LSTM model; after the LSTM model is pre-trained, time series data of the prediction step size is taken as re-training data, and a dataset dimension adjustment method based on Choquet integral is used to perform non-additive dimension adjustment on the re-training data, and the LSTM model is re-trained with the adjusted re-training data, thereby realizing adaptive updating of the model.

[0016] Furthermore, after the non-additive dimension adjustment of the training data set, the data set weight calculation and analysis are performed first, and then the correlation analysis is performed to select the data with the greatest impact on the battery health status, and finally the LSTM model is pre-trained and retrained.

[0017] Furthermore, the data set dimension adjustment method based on Choquet integral is as follows:

[0018] The Choquet integral is defined as follows: From the measurable domain Mapped to a measurable function in the field of real numbers, then the measurable function On Non-Additive Interaction Measures is defined as:

[0019] (1)

[0020] In the formula, is the integration variable of the Riemann integral, , is called a measurable function of Split set, if the set , then the Choquet integral can be simply abbreviated as ;

[0021] In the discrete domain, formula (1) is transformed into:

[0022] (2)

[0023] in, , Is a collection A permutation of ;

[0024] Formula (2) can be further transformed into:

[0025] (3)

[0026] Will In binary form it is ,but ,at the same time:

[0027] (4)

[0028] In the formula, Indicates taking the decimal part;

[0029] If the collection Represents the attribute value set in the battery health status prediction process, and its column number is , then according to formula (4), it is converted into the number of columns of matrix.

[0030] Furthermore, a weight calculation method is used to calculate and analyze the weight of the data set to achieve Select from the column attribute value data Column the most influential attribute value data, the specific steps are as follows:

[0031] First, the matrix is ​​calculated according to formula (4) , then according to formula (3), the left side of the equation is the target value, and the right side is It is known that by linear regression, The value of The weight of the attribute value data From formula (5), we get:

[0032] (5)

[0033] in,

[0034] (6)

[0035] Formula (5) and formula (6) show that The weight of each attribute value is determined by a series of non-additive interaction measures If a set of attribute value data contains the first attribute values, then the non-additive interaction measure value corresponding to this set is added when calculating the weight;

[0036] Perform the above calculations for each attribute value and select the one with the highest weight. The attribute value data is taken as the result of dimension reduction, and then The attribute value data is recalculated according to formula (4) to obtain the matrix , the number of columns is .

[0037] Furthermore, we get the matrix After that, the correlation between each column of data and the battery health status is analyzed, and the data sequence with a correlation of 0 with the battery health status is eliminated, and the number of columns is obtained. The matrix Z , .

[0038] In a second aspect, the present invention provides a lithium-ion battery health state prediction system, which is used to implement the lithium-ion battery health state prediction method, and comprises:

[0039] Time series data acquisition module: acquires the time series data of average voltage, average current, average temperature, acquisition time and battery health status during the charge and discharge cycle of lithium-ion batteries, and classifies them according to the standard of battery health status as target value and other data as attribute value, to obtain attribute value data containing multi-dimensional data and target value data containing only battery health status;

[0040] LSTM model building module: build a long short-term memory network LSTM model and set the hyperparameters of the LSTM model according to the nonlinear relationship between attribute value data and target value data;

[0041] LSTM model non-additive dimension adjustment module: the acquired attribute value data and target value data are used as training data sets, and the data set dimension adjustment method based on Choquet integral is used to perform non-additive dimension adjustment on the attribute value data in the training data set, so as to change the dimension of the attribute value data in the training data set and extract the attribute value data with the greatest impact on the target value;

[0042] LSTM model training module: using the adjusted training data set to perform pre-training and re-training of the LSTM model to obtain a re-trained LSTM model;

[0043] Prediction module: Use the retrained LSTM model to predict the battery health status data at the next moment.

[0044] In a third aspect, the present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the lithium-ion battery health status prediction method when executing the computer program.

[0045] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the lithium-ion battery health status prediction method are implemented.

[0046] The beneficial effects of the present invention are as follows: Based on non-additive dimensionality adjustment and adaptive updating of the LSTM model, the present invention reduces the dimension of the data set while ensuring the prediction accuracy of the model as much as possible, thereby greatly reducing the complexity of calculation; not only the influence of each attribute value on the target value is taken into account, but also the mutual influence between different attribute values, thereby realizing non-additive adjustment of the attribute value data dimension. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0048] Figure 1 A flow chart of a method for predicting the health status of a lithium-ion battery according to the present invention;

[0049] Figure 2 A comparison chart of the predicted SOH value and the actual value curve of the experimental battery in the application example of the present invention;

[0050] Figure 3 This is a distribution diagram of the error between the predicted value and the true value of the SOH of the experimental battery in the application example of the present invention;

[0051] Figure 4 It is a structural diagram of the lithium-ion battery health status prediction system of the present invention;

[0052] Figure 5 A schematic diagram of a logical structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] Example 1

[0055] This embodiment provides a method for predicting the health status of a lithium-ion battery. Figure 1 As shown, the steps are as follows:

[0056] S1) obtaining the time series data of average voltage, average current, average temperature, acquisition time and battery health status during the charge and discharge cycle of the lithium-ion battery, and classifying them according to the standard of using the battery health status as the target value and the remaining data as the attribute value, to obtain attribute value data containing multi-dimensional data and target value data containing only the battery health status;

[0057] S2) establishing an LSTM model and setting the hyperparameters of the LSTM model according to the nonlinear relationship between the attribute value data and the target value data;

[0058] S3) The acquired attribute value data and target value data are used as training data sets, and the data set dimension adjustment method based on Choquet integral is used to perform non-additive dimension adjustment on the attribute value data in the training data set, so as to change the dimension of the attribute value data in the training data set and extract the attribute value data with the greatest impact on the target value; then the data set weight calculation and analysis are performed, and finally the correlation analysis is performed to actively eliminate the data irrelevant to the battery health status, so as to facilitate the subsequent pre-training and re-training of the LSTM model;

[0059] S4) pre-training and re-training the LSTM model using the training dataset after non-additive dimension adjustment to obtain a re-trained LSTM model;

[0060] S5) Use the retrained LSTM model to predict the battery health status data at the next moment. The retrained LSTM model can already predict the battery SOH well. i The battery SOH at the moment, just i- The attribute value data at time 1 can be input into the LSTM model after dimension adjustment.

[0061] The process of repeating the model adaptive update and prediction is continuously carried out, and a relatively accurate prediction of the battery SOH at each subsequent moment can be made.

[0062] The specific contents of steps S3) and S4) are as follows: using the time series data of the first 1 / 2 length in the training data set as pre-training data, using the data set dimension adjustment method based on Choquet integral to perform non-additive dimension adjustment on the pre-training data, setting the prediction step size, and using the adjusted pre-training data to pre-train the LSTM model; after the LSTM model is pre-trained, taking the time series data of the prediction step size as re-training data, using the data set dimension adjustment method based on Choquet integral to perform non-additive dimension adjustment on the re-training data, and using the adjusted re-training data to re-train the LSTM model, thereby realizing adaptive updating of the model.

[0063] Specifically, the data set dimension adjustment method based on Choquet integral is as follows:

[0064] In fuzzy mathematics, the definition is in the measurable domain. Fuzzy measure on The following conditions need to be met:

[0065] FM1), ;

[0066] FM2), for any set , both have ;

[0067] FM3) If the collection , satisfy , and , then ;

[0068] Fuzzy measure is essentially a function that maps a set to the real number domain. This function needs to satisfy the above three conditions. If a function If only condition FM1) is satisfied, it is called a non-additive interaction measure; this definition of non-additive interaction measure retains the most basic properties of fuzzy measure, can be applied to general fuzzy integrals, and removes restrictions on signs and sizes.

[0069] There are many types of fuzzy integrals. Among them, the nonlinear characteristics of Choquet integral make it a good information processing tool, which has been confirmed by many experimental results. The definition of Choquet integral is as follows: From the measurable domain Measurable functions mapped to the field of real numbers, the set , is defined in The signed efficiency measure on , then the function can be measured On Non-Additive Interaction Measures is defined as:

[0070] (1)

[0071] In the formula, is the integration variable of the Riemann integral, , is called a measurable function of Split set, if the set , then the Choquet integral can be simply abbreviated as ;

[0072] In the discrete domain, formula (1) is transformed into:

[0073] (2)

[0074] in, , Is a collection A permutation of .

[0075] Formula (2) is convenient for manual calculation, but it is cumbersome to calculate on a computer because it requires comparison of function values. Therefore, Formula (2) is further transformed into:

[0076] (3)

[0077] Will In binary form it is ,but ,at the same time:

[0078] (4)

[0079] In the formula, Indicates taking the decimal part;

[0080] If the collection Represents the attribute value set in the battery health status prediction process, and its column number is , then according to formula (4), it is converted into the number of columns of matrix.

[0081] This embodiment uses a weight calculation method to calculate and analyze the weight of the data set to achieve Select from the column attribute value data Column the most influential attribute value data, the specific steps are as follows:

[0082] First, the matrix is ​​calculated according to formula (4) , then according to formula (3), the left side of the equation is the target value, and the right side is It is known that by linear regression, The value of The weight of each attribute value data is obtained by formula (5):

[0083] (5)

[0084] in,

[0085] (6)

[0086] Formula (5) and formula (6) show that The weight of each attribute value is determined by a series of non-additive interaction measures If a set of attribute value data contains the first attribute values, then the non-additive interaction measure value corresponding to this set is added when calculating the weight;

[0087] Perform the above calculations for each attribute value and select the one with the highest weight. attribute value data as the result of dimensionality reduction, and then The attribute value data is recalculated according to formula (4) to obtain the matrix , the number of columns is . In the matrix After that, the correlation between each column of data and SOH is analyzed, and the data sequence with a correlation of 0 with SOH is eliminated, and the number of columns is obtained. The matrix , This process is non-additive, taking into account not only the impact of each attribute value on the target value, but also the mutual impact between different attribute values, thus achieving non-additive adjustment of the dimension of attribute value data.

[0088] The long short-term memory recursive model (LSTM) in this embodiment is a machine learning model for time series prediction tasks. The concept of memory gate is introduced in this special recursive neural network. The long-term characteristics of the time series are contained in it. At two adjacent moments in the time series, the data corresponding to the long-term memory gate will only change slightly; It is a short-term memory gate, which contains the short-term information of the time series, and there may be a large difference between two adjacent moments.

[0089] Application Examples

[0090] In order to verify the effectiveness of the prediction method of the present invention (i.e., the method described in Example 1), it is necessary to obtain more and more realistic battery data and build an experimental equipment platform. The capacity of the lithium-ion battery used in the experiment is 27Ah, the average voltage is 3.2V, the operating ambient temperature of the battery is 55°C, and the charge and discharge cycle is performed at a current of 1C. 940 cycles are measured to obtain detailed battery voltage, current, temperature, time, and capacity data.

[0091] The hyperparameters of the LSTM model are as follows: input dimension is 54, output dimension is 1, activation function is sigmoid, dropout rate is set to 0, number of hidden variables is 32, batch size is 10, training times are 100 during pre-training, 30 times each time during re-training, sizes (selected hyperparameters) are selected as 50, optimizer is selected as Adam, and learning rate is set to 0.005.

[0092] In order to demonstrate the effectiveness of the prediction method of the present invention, the traditional SOH prediction method containing only the LSTM model and the prediction method proposed by the present invention are applied to the prediction process. The obtained battery SOH prediction data and real data curves are shown as follows: Figure 2 shown.

[0093] Figure 2The red scatter plot is the SOH prediction result of the method proposed in this invention, and the green scatter plot is the SOH prediction result using only the LSTM method. It can be seen that the result calculated using only the LSTM method has a serious overfitting phenomenon, resulting in a large deviation in the prediction; while the method proposed in this invention has a better fit with the actual SOH. The error distribution of further analysis is shown in Figure 3 As shown, it can be found that the prediction errors of the method proposed in the present invention are mostly below 1.5%, which can meet the accuracy requirements of most practical systems.

[0094] Example 2

[0095] This embodiment provides a lithium-ion battery health status prediction system. Figure 4 As shown, the method for implementing the method described in Example 1 includes:

[0096] Time series data acquisition module: acquires the time series data of average voltage, average current, average temperature, acquisition time and battery health status during the charge and discharge cycle of lithium-ion batteries, and classifies them according to the standard of battery health status as target value and other data as attribute value, to obtain attribute value data containing multi-dimensional data and target value data containing only battery health status;

[0097] LSTM model building module: builds the LSTM model and sets the hyperparameters of the LSTM model according to the nonlinear relationship between the attribute value data and the target value data;

[0098] LSTM model non-additive dimension adjustment module: the acquired attribute value data and target value data are used as training data sets, and the data set dimension adjustment method based on Choquet integral is used to perform non-additive dimension adjustment on the attribute value data in the training data set, so as to change the dimension of the attribute value data in the training data set and extract the attribute value data with the greatest impact on the target value;

[0099] LSTM model training module: using the adjusted training data set to perform pre-training and re-training of the LSTM model to obtain a re-trained LSTM model;

[0100] Prediction module: Use the retrained LSTM model to predict the battery health status data at the next moment.

[0101] The retrained LSTM model is now able to predict the battery SOH better. i The battery SOH at the moment, just i- The attribute value data at time 1 can be input into the LSTM model after dimension adjustment.

[0102] The process of repeating the model adaptive update and prediction is continuously carried out, and a relatively accurate prediction of the battery SOH at each subsequent moment can be made.

[0103] It should be noted that each module in the above-mentioned lithium-ion battery health status prediction system can be fully or partially implemented by software, hardware and their combination. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules. For the specific definition of a lithium-ion battery health status prediction system, please refer to the definition of a lithium-ion battery health status prediction system method above. The two have the same functions and effects, which will not be repeated here.

[0104] Example 3

[0105] This embodiment provides an electronic device, comprising: at least one processor; and a memory connected to the at least one processor in communication. The memory stores a computer program executable by the at least one processor, and when the computer program is executed by the at least one processor, the electronic device executes the method according to Embodiment 1 of the present invention.

[0106] Example 4

[0107] This embodiment provides a non-transitory computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor of a computer, is used to cause the computer to execute the method according to Embodiment 1 of the present invention.

[0108] refer to Figure 5 , a block diagram of an electronic device 400 that can be used as a server or client of the present invention will now be described, which is an example of a hardware device that can be applied to various aspects of the present invention. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0109] like Figure 5As shown, the electronic device 400 includes a computing unit 401, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 402 or a computer program loaded from a storage unit 408 into a random access memory (RAM) 403. In the RAM 403, various programs and data required for the operation of the electronic device 400 can also be stored. The computing unit 401, the ROM 402, and the RAM 403 are connected to each other via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0110] A plurality of components in the electronic device 400 are connected to the I / O interface 405, including: an input unit 406, an output unit 407, a storage unit 408, and a communication unit 409. The input unit 406 may be any type of device capable of inputting information to the electronic device 400, and the input unit 406 may receive input digital or character information, and generate key signal inputs related to user settings and / or function control of the electronic device. The output unit 407 may be any type of device capable of presenting information, and may include, but is not limited to, a display, a speaker, a video / audio output terminal, a vibrator, and / or a printer. The storage unit 408 may include, but is not limited to, a disk, an optical disk. The communication unit 409 allows the electronic device 400 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks, and may include, but is not limited to, a modem, a network card, an infrared communication device, a wireless communication transceiver, and / or a chipset, such as a Bluetooth™ device, a WiFi device, a WiMax device, a cellular communication device, and / or the like.

[0111] The computing unit 401 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 401 performs the various methods and processes described above. For example, in some embodiments, the aforementioned metal corrosion prediction method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 408. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 400 via the ROM 402 and / or the communication unit 409. In some embodiments, the computing unit 401 may be configured to perform the aforementioned metal corrosion prediction method in any other appropriate manner (e.g., by means of firmware).

[0112] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0113] In the context of the present invention, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0114] As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, apparatus, and / or device (e.g., disk, optical disk, memory, programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.

[0115] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0116] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0117] A computer system may include clients and servers. Clients and servers are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship to each other.

[0118] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A method for predicting the health status of a lithium-ion battery, characterized in that: include: The average voltage, average current, average temperature, acquisition time and battery health status time series data of the lithium-ion battery during the charge and discharge cycle are obtained, and the battery health status is used as the target value and the remaining data is used as the attribute value. The data is classified to obtain the attribute value data containing multi-dimensional data and the target value data containing only the battery health status; Establish a long short-term memory network LSTM model and set the hyperparameters of the LSTM model according to the nonlinear relationship between attribute value data and target value data; The acquired attribute value data and target value data are used as training data sets, and the data set dimension adjustment method based on Choquet integral is used to perform non-additive dimension adjustment on the attribute value data in the training data set, so as to change the dimension of the attribute value data in the training data set and extract the attribute value data with the greatest influence on the target value; Use the adjusted training data set to pre-train and retrain the LSTM model to obtain a retrained LSTM model; Use the retrained LSTM model to predict the battery health status data at the next moment; The data set dimension adjustment method based on Choquet integral is as follows: The Choquet integral is defined as follows: From the measurable domain Mapped to a measurable function in the field of real numbers, then the measurable function On Non-Additive Interaction Measures is defined as: (1) In the formula, is the integration variable of the Riemann integral, , is called a measurable function of Partition set; if the set , then the Choquet integral can be simply abbreviated as ; In the discrete domain, formula (1) is transformed into: (2) in, , Is a collection A permutation of ; Formula (2) can be further transformed into: (3) Will In binary form it is ,but ,at the same time: (4) In the formula, Indicates taking the decimal part; If the collection Represents the attribute value set in the battery health status prediction process, and its column number is , then according to formula (4), it is converted into the number of columns of matrix; A weight calculation method is used to calculate and analyze the weight of the data set to achieve Select from the column attribute value data Column the most influential attribute value data, the specific steps are as follows: First, the matrix is ​​calculated according to formula (4) , then according to formula (3), the left side of the equation is the target value, and the right side is It is known that by linear regression, The value of The weight of the attribute value data From formula (5), we get: (5) in, (6) Formula (5) and formula (6) show that The weight of each attribute value is determined by a series of non-additive interaction measures If a set of attribute value data contains the first attribute values, then the non-additive interaction measure value corresponding to this set is added when calculating the weight; Perform the above calculations for each attribute value and select the one with the highest weight. The attribute value data is taken as the result of dimensionality reduction, and then The attribute value data is recalculated according to formula (4) to obtain the matrix , the number of columns is ; In getting the matrix After that, the correlation between each column of data and the battery health status is analyzed, and the data sequence with a correlation of 0 with the battery health status is eliminated, and the number of columns is obtained. The matrix Z , .

2. The method for predicting the health status of a lithium-ion battery according to claim 1, characterized in that: The time series data of the first 1 / 2 length in the training data set is used as pre-training data.

3. The method for predicting the health status of a lithium-ion battery according to claim 1 or 2, characterized in that: A dataset dimension adjustment method based on Choquet integral is used to perform non-additive dimension adjustment on pre-training data in a training dataset, a prediction step size is set, and the adjusted pre-training data is used to pre-train an LSTM model; after the LSTM model is pre-trained, time series data of the prediction step size is taken as re-training data, a dataset dimension adjustment method based on Choquet integral is used to perform non-additive dimension adjustment on the re-training data, and the adjusted re-training data is used to re-train the LSTM model.

4. The method for predicting the health status of a lithium-ion battery according to claim 3, characterized in that: After adjusting the non-additive dimensions of the training data set, the data set weights are calculated and analyzed first, and then correlation analysis is performed to select the data with the greatest impact on the battery health status. Finally, the LSTM model is pre-trained and retrained.

5. A lithium-ion battery health status prediction system, used to implement the method according to any one of claims 1 to 4, characterized in that: include: Time series data acquisition module: used to obtain the time series data of average voltage, average current, average temperature, acquisition time and battery health status during the charge and discharge cycle of lithium-ion batteries, and classify them according to the standard of battery health status as target value and other data as attribute value, so as to obtain attribute value data containing multi-dimensional data and target value data containing only battery health status; LSTM model building module: used to build a long short-term memory network LSTM model and set the hyperparameters of the LSTM model according to the nonlinear relationship between attribute value data and target value data; LSTM model non-additive dimension adjustment module: the acquired attribute value data and target value data are used as training data sets, and the data set dimension adjustment method based on Choquet integral is used to perform non-additive dimension adjustment on the attribute value data in the training data set, so as to change the dimension of the attribute value data in the training data set and extract the attribute value data with the greatest impact on the target value; LSTM model training module: using the adjusted training data set to perform pre-training and re-training of the LSTM model to obtain a re-trained LSTM model; Prediction module: Use the retrained LSTM model to predict the battery health status data at the next moment.

6. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.

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

Images