Battery health state determination method and device and server

Abstract

The invention provides a battery health state determination method and device and a server, before specific implementation, a particle swarm optimization algorithm and a random forest regression prediction model are jointly used, and a corresponding preset battery health state prediction model is obtained through training. In specific implementation, a target ultrasonic transducer group is deployed on a working surface of a target battery; wherein the target ultrasonic transducer group comprises three ultrasonic transducers, the connecting edges among the ultrasonic transducers form an equilateral triangle, and the center of the equilateral triangle coincides with the center of the working surface of the target battery; a waveform signal about the target battery is acquired by using the target ultrasonic transducer group; according to the waveform signal, obtaining and utilizing corresponding time domain data and frequency domain data to determine a target feature about the target battery; and using a preset battery health state prediction model to determine the health state of the target battery by processing the target features. Therefore, the health state of the target battery can be monitored and determined efficiently and accurately.

Description

Technical Field

[0001] The present specification belongs to the field of electrical data processing technology, and in particular, relates to a method, device and server for determining a battery health status. Background Art

[0002] With the development and promotion of new energy technologies, more and more new energy batteries (for example, sodium-ion batteries) are gradually being used in electric vehicles, energy storage industries, communication base stations and other scenarios.

[0003] In the process of using new energy batteries, it is often necessary to monitor the battery health status. However, based on the existing methods, when determining the battery health status, there are often problems such as poor accuracy, low reference value, and easy errors.

[0004] To address the above problems, no effective solution has been proposed yet. Summary of the invention

[0005] This specification provides a method, device and server for determining the health status of a battery, which can monitor and determine the health status of a target battery more efficiently and accurately, and effectively reduce detection errors.

[0006] This specification provides a method for determining a battery health status, including: Using a target ultrasonic transducer group to transmit an ultrasonic signal on a working surface of a target battery; and collecting a corresponding waveform signal; wherein the target ultrasonic transducer group includes three ultrasonic transducers, the connecting sides of the ultrasonic transducers form an equilateral triangle, and the center of the equilateral triangle coincides with the center of the working surface of the target battery; According to the waveform signal, obtaining corresponding time domain data and frequency domain data; Determine target features of the target battery based on the time domain data and the frequency domain data; The target features are processed using a preset battery health state prediction model to obtain a corresponding target prediction result; wherein the preset battery health state prediction model is obtained by training by jointly using a particle swarm optimization algorithm and a random forest regression prediction model; According to the target prediction results, the health status of the target battery is determined.

[0007] In one embodiment, the target feature includes at least one of the following: average intensity attenuation ratio, average flight time, and average main frequency deviation ratio.

[0008] In one embodiment, when the target feature includes at least an average intensity attenuation ratio, determining the target feature of the target battery according to the time domain data and the frequency domain data includes: Determine the average intensity attenuation ratio of the target battery according to the following formula:

[0009] in, is the average intensity attenuation ratio, A i1 is the signal strength of the signal with the highest signal strength value in the signal data received by the ultrasonic transducer numbered i, A i2 is the signal strength of the signal whose signal strength value ranks second in the signal data received by the ultrasonic transducer numbered i, A i3 is the signal strength of the signal ranked third in signal strength value in the signal data received by the ultrasonic transducer numbered i, and A is the signal strength of the main frequency signal of the ultrasonic signal emitted by the ultrasonic transducer.

[0010] In one embodiment, when the target feature includes at least the average flight time, the target feature about the target battery is determined according to the time domain data and the frequency domain data, including: Determine the average flight time of the target battery using the following formula:

[0011] in, is the average flight time, T i1 T is the flight time of the signal with the highest signal strength value in the signal data received by the ultrasonic transducer numbered i, i2 T is the flight time of the signal with the second highest signal strength value in the signal data received by the ultrasonic transducer numbered i, i3 It is the flight time of the signal with the third highest signal strength value in the signal data received by the ultrasonic transducer numbered i.

[0012] In one embodiment, when the target feature includes at least an average main frequency offset ratio, determining the target feature of the target battery according to the time domain data and the frequency domain data includes: Determine the average main frequency deviation ratio of the target battery according to the following formula: in, is the average main frequency deviation ratio, f i1 is the frequency of the signal with the highest signal strength value in the signal data received by the ultrasonic transducer numbered i, f i2 is the frequency of the signal whose signal strength value ranks second in the signal data received by the ultrasonic transducer numbered i, f i3 is the frequency of the signal with the third highest signal strength value in the signal data received by the ultrasonic transducer numbered i, and f is the frequency of the main frequency signal of the ultrasonic signal emitted by the ultrasonic transducer.

[0013] In one embodiment, the method further comprises: Screening multiple batteries in different health states as sample batteries; According to the preset deployment rules, a corresponding target ultrasonic transducer group is deployed on the working plane of the sample battery, and the target ultrasonic transducer group is used to perform a test experiment to obtain sample data corresponding to the sample battery; Using the sample data, construct a corresponding sample data set; A plurality of computational particles are determined; A plurality of sample data are randomly extracted from the sample data set to generate a plurality of sub-sample data sets; and an initial prediction model based on the random forest regression prediction model is constructed; Multiple computing particles are called to perform multiple rounds of iterative training based on the initial prediction model and the sub-sample data set until the training end conditions are met, and a preset battery health status prediction model that meets the requirements is obtained.

[0014] In one embodiment, multiple computing particles are called to perform multiple rounds of iterative training based on the initial prediction model and the sub-sample data set, including: Calling multiple computing particles to determine the speed parameters of the current round corresponding to the computing particles based on the local solution particle parameter group of the previous round and the global solution particle parameter group of the previous round; wherein the particle parameter group includes at least: pruning threshold, number of decision trees, and test sample ratio; Call multiple computing particles to update their respective prediction models of the previous round based on the speed parameters and sub-sample data sets of the current round, and obtain their respective prediction models of the current round; Call multiple computing particles to update the local solution particle parameter group of the current round according to their respective prediction models of the current round; and update the global solution particle parameter group of the current round according to the local solution particle parameter group of the current round; Check whether the training end conditions are met; When it is determined that the training end conditions are met, a preset battery health status prediction model that meets the requirements is determined based on the global solution particle parameter group of the current round.

[0015] This specification also provides a device for determining a battery health status, including: A collection module, used to transmit ultrasonic signals on the working surface of the target battery using a target ultrasonic transducer group; and collect corresponding waveform signals; wherein the target ultrasonic transducer group includes three ultrasonic transducers, the connecting sides of the ultrasonic transducers form an equilateral triangle, and the center of the equilateral triangle coincides with the center of the working surface of the target battery; An acquisition module, used for acquiring corresponding time domain data and frequency domain data according to the waveform signal; A first determination module, used to determine a target feature of a target battery according to the time domain data and the frequency domain data; A processing module, used to process target features using a preset battery health state prediction model to obtain a corresponding target prediction result; wherein the preset battery health state prediction model is obtained by training by jointly using a particle swarm optimization algorithm and a random forest regression prediction model; The second determination module is used to determine the health status of the target battery according to the target prediction result.

[0016] The present specification also provides a server, comprising a processor and a memory for storing processor executable instructions, wherein the processor implements the steps of the method for determining the battery health status when executing the instructions.

[0017] The present specification also provides a computer-readable storage medium having computer instructions stored thereon, and when the instructions are executed by a processor, the steps of the method for determining the battery health status are implemented.

[0018] Based on the method, device and server for determining the battery health state provided in this specification, before specific implementation, the particle swarm optimization algorithm and the random forest regression prediction model can be used together to train a preset battery health state prediction model with high accuracy and good effect. In specific implementation, the corresponding target ultrasonic transducer group can be deployed on the working surface of the target battery of interest according to the preset deployment rules; wherein the target ultrasonic transducer group includes 3 ultrasonic transducers, the 3 ultrasonic transducers are equidistant from each other, the connecting edges between the ultrasonic transducers form an equilateral triangle, and the center of the equilateral triangle coincides with the center of the working surface of the target battery. In this way, the above-mentioned target ultrasonic transducer group can be used to simultaneously transmit ultrasonic signals to collect a waveform signal about the target battery that is relatively comprehensive, complete, and has a high reference value. Then, according to the waveform signal, the corresponding time domain data and frequency domain data are obtained and used to determine the target features of the target battery; the target features are processed using the preset battery health state prediction model to obtain the corresponding target prediction results; and the health state of the target battery is determined according to the target prediction results. This enables the health status of the target battery to be monitored and determined more efficiently and accurately, effectively reducing detection errors and ensuring stable and safe operation of the target battery. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of this specification, the drawings required for use in the embodiments will be briefly introduced below. The drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 is a flowchart of a method for determining a battery health status provided by an embodiment of this specification; Figure 2 is a schematic diagram of an embodiment of a method for determining a battery health state provided by an embodiment of this specification, in a scenario example; Figure 3 is a schematic diagram of an embodiment of a method for determining a battery health state provided by an embodiment of this specification, in a scenario example; Figure 4 is a schematic diagram of an embodiment of a method for determining a battery health state provided by an embodiment of this specification, in a scenario example; Figure 5 It is a schematic diagram of the structure of a server provided by an embodiment of this specification; Figure 6 It is a schematic diagram of the structure of a device for determining a battery health status provided by an embodiment of this specification; Figure 7 is a schematic diagram of an embodiment of a method for determining a battery health state provided by an embodiment of this specification, in a scenario example; Figure 8 is a schematic diagram of an embodiment of a method for determining a battery health state provided by an embodiment of this specification, in a scenario example; Fig. 9 This is a schematic diagram of an embodiment of a method for determining a battery health status provided by an embodiment of this specification, in a scenario example. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0022] It should be noted that the user-related information and data involved in the embodiments of this specification are all information and data authorized by the user or fully authorized by relevant parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with relevant laws, regulations and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users or relevant parties to choose to authorize or refuse.

[0023] It should also be noted that in the embodiments of this specification, certain software, components, models and other existing solutions in the industry may be mentioned. They should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of this application, but it does not mean that the applicant has or will necessarily use the solution.

[0024] See also Figure 1 As shown, the embodiment of this specification provides a method for determining the health status of a battery, wherein the method is specifically applied to a server side. When implemented specifically, the method may include the following contents: S101: using a target ultrasonic transducer group to transmit an ultrasonic signal on a working surface of a target battery; and collecting a corresponding waveform signal; wherein the target ultrasonic transducer group includes three ultrasonic transducers, the connecting sides of the ultrasonic transducers form an equilateral triangle, and the center of the equilateral triangle coincides with the center of the working surface of the target battery; S102: Acquire corresponding time domain data and frequency domain data according to the waveform signal; S103: Determine target features of the target battery according to the time domain data and the frequency domain data; S104: Processing target features using a preset battery health state prediction model to obtain a corresponding target prediction result; wherein the preset battery health state prediction model is trained by jointly using a particle swarm optimization algorithm and a random forest regression prediction model; S105: Determine the health status of the target battery according to the target prediction result.

[0025] The above-mentioned state of health (SOH) can be specifically understood as a reflection of the health and life status of the battery, which can be a specific reflection of the battery's power, energy, charge and discharge power, etc.

[0026] The target battery may be a sodium ion battery, which has good low temperature performance, good fast charging performance, and high energy density. Of course, in addition to sodium ion batteries, the target battery may also be other types of batteries such as lead-acid batteries.

[0027] Specifically, the target battery may be a battery installed in a new energy power station (eg, a wind-solar coupled power station) for storing electric energy.

[0028] Before specific implementation, the target ultrasonic transducer group can be deployed on the working surface of the target battery according to the preset deployment rules. Figure 2 The tab in the figure specifically refers to the metal conductor that leads the positive and negative electrodes out of the battery cell of the target battery, which can be understood as the contact point of the positive and negative electrodes when the target battery is charged and discharged.

[0029] The target ultrasonic transducer group may specifically include three identical ultrasonic transducers; and the spacing between the ultrasonic transducers is equal (for example, the spacing between each other is d1=d2=d3=1.5cm), and the connecting sides between different ultrasonic transducers form an equilateral triangle. Furthermore, the center of the equilateral triangle formed by the three ultrasonic transducers coincides with the center of the working surface of the target battery. The working surface of the target battery may specifically be the surface with the largest internal area of ​​the target battery (for example, 10cm * 7cm). When deploying the ultrasonic transducer sensors, try to place the three ultrasonic transducer sensors on the same horizontal plane.

[0030] The ultrasonic transducer mentioned above can be specifically understood as a device that can convert electrical energy into sound energy.

[0031] Specifically, the target ultrasonic transducer group is also connected to the server via a wired or wireless method. Accordingly, the server can control the target ultrasonic transducer group to collect and upload the corresponding waveform signal, and then determine the health status of the target battery according to the waveform signal.

[0032] It should be noted that three ultrasonic transducers are selected here to construct the target ultrasonic transducer group, considering that the equilateral triangle formed by the three ultrasonic transducers can better ensure the uniform distribution of the transmitted ultrasonic signals; and, compared with using a larger number of ultrasonic transducers, the use of three ultrasonic transducers can effectively avoid redundant signals and reduce complex calculations, while also providing sufficient signal coverage and balancing signal reception and calculation efficiency. Specifically, the use of three ultrasonic transducers to form an equilateral triangle can ensure the uniform distribution of signals, so that the ultrasonic signal propagates more comprehensively inside the battery, avoiding signal blind spots and having better balance. In addition, based on the symmetry of the above-mentioned equilateral triangle, the signal capture capability can also be effectively enhanced, helping the ultrasonic transducer to collect relatively more detailed signals, so that potential problems inside the battery (for example, defects, gas accumulation, etc.) can be better detected.

[0033] In this way, the ultrasonic signal emitted by the target ultrasonic transducer group deployed according to the preset deployment rules can more completely and comprehensively cover the entire area of ​​the target battery, avoid omissions, and more accurately detect the overall situation of the target battery.

[0034] Furthermore, the distance between the ultrasonic transducer and the center of the working surface is greater than a first distance threshold and less than or equal to a second distance threshold. The first distance threshold and the second distance threshold can be determined based on the size parameters of the working surface of the target battery and the performance parameters of the ultrasonic transducer. In this way, the deployed target ultrasonic transducer group can, on the one hand, fully and completely cover the entire area of ​​the target battery; on the other hand, it can also reduce the attenuation of the signal when collecting the signal, ensuring that the signal collected by the ultrasonic transducer has a high degree of credibility.

[0035] In specific implementation, the three ultrasonic transducers included in the target ultrasonic transducer group can be controlled to emit the same ultrasonic signal at the same time; and then the three ultrasonic transducers can be controlled to collect the reflected signals of the ultrasonic signals emitted by themselves and the ultrasonic signals emitted by other ultrasonic transducers to obtain corresponding waveform signals.

[0036] The transmission center frequency of the ultrasonic transducer may include: 300KHz-5MHz. Specifically, according to the battery property characteristics of the target battery, a matching frequency may be selected as the transmission center frequency of the ultrasonic transducer.

[0037] For example, when the positive electrode material of the target battery is determined to be a sodium ion soft-pack battery of Prussian blue, the emission center frequency of the ultrasonic transducer can be determined to be 1 MHz; when the positive electrode material of the target battery is determined to be a sodium ion soft-pack battery of layered transition metal oxide, the emission center frequency of the ultrasonic transducer can be determined to be 2.5 MHz; when the positive electrode material of the target battery is determined to be a sodium ion soft-pack battery of sodium iron phosphate, the emission center frequency of the ultrasonic transducer can be determined to be 800 KHz.

[0038] In this way, the target ultrasonic transducer group can be controlled to generate ultrasonic signals at different transmission center frequencies according to different battery property characteristics, so as to obtain a waveform signal with relatively better effect and relatively higher accuracy.

[0039] In specific implementation, the corresponding time domain data and frequency domain data, two different dimensional data, can be first obtained according to the waveform signal; then the time domain data and frequency domain data, two different dimensional data, can be used together to calculate the required signal characteristic parameters of the target battery based on the ultrasonic signal as the target feature; then, the pre-trained preset battery health status prediction model is used to process the above target features to obtain the corresponding target prediction results; and based on the target prediction results, the health status of the target battery is finally determined.

[0040] The above-mentioned preset battery health status prediction model is pre-trained by using a particle swarm optimization algorithm and a random forest regression prediction model, and can automatically predict and output the battery health status based on the relevant target features of the input model. The specific training method of this model will be explained later.

[0041] Based on the above embodiment, before specific implementation, the particle swarm optimization algorithm and the random forest regression prediction model can be used together to train a preset battery health status prediction model with high accuracy and good effect. In specific implementation, a target ultrasonic transducer group can be deployed on the working surface of the target battery of interest; wherein, the target ultrasonic transducer group includes 3 ultrasonic transducers, the 3 ultrasonic transducers are equidistant from each other, the connecting edges form an equilateral triangle, and the center of the equilateral triangle coincides with the center of the working surface of the target battery. In this way, the above target ultrasonic transducer group can be used to simultaneously emit ultrasonic signals to collect a relatively comprehensive and complete waveform signal about the target battery. Then, according to the waveform signal, the corresponding time domain data and frequency domain data are obtained and used to determine the target features of the target battery; the preset battery health status prediction model is used to process the target features to obtain the corresponding target prediction results; and the health status of the target battery is determined according to the target prediction results. Therefore, the health status of the target battery can be monitored and determined more efficiently and accurately, effectively reducing errors and ensuring stable and safe operation of the target battery.

[0042] In specific implementation, the target ultrasonic transducer group can be deployed on the working surface of the target battery in advance, and then according to each specified time interval, the target ultrasonic transducer group can be triggered to start collecting and use the corresponding waveform signal to perform regular monitoring of the health status of the target battery to ensure the safe and stable operation of the target battery.

[0043] In some embodiments, the above-mentioned acquisition of corresponding time domain data and frequency domain data according to the waveform signal may include the following contents during specific implementation: S1: Obtain a corresponding time domain signal by filtering and amplifying the waveform signal; S2: By performing Fourier transform on the time domain signal, the corresponding frequency domain signal is obtained.

[0044] Based on the above embodiment, the waveform signal can be used to process and obtain signal data about the target battery based on different dimensions such as time domain and frequency domain, so that the health status of the target battery can be comprehensively and accurately predicted based on the above signal data.

[0045] In some embodiments, the target feature includes at least one of the following: average intensity attenuation ratio, average flight time, average main frequency deviation ratio, etc.

[0046] During specific implementation, one or more combinations of the above-listed features may be used as target features.

[0047] In some embodiments, when the target feature includes at least the average intensity attenuation ratio, the target feature of the target battery is determined according to the time domain data and the frequency domain data. When implemented specifically, it may include: Determine the average intensity attenuation ratio of the target battery according to the following formula:

[0048] in, is the average intensity attenuation ratio, A i1 is the signal strength of the signal with the highest signal strength value in the signal data received by the ultrasonic transducer numbered i, A i2 is the signal strength of the signal whose signal strength value ranks second in the signal data received by the ultrasonic transducer numbered i, A i3 is the signal strength of the signal ranked third in signal strength value in the signal data received by the ultrasonic transducer numbered i, and A is the signal strength of the main frequency signal of the ultrasonic signal emitted by the ultrasonic transducer.

[0049] In some embodiments, when the target feature includes at least the average flight time, the target feature of the target battery is determined according to the time domain data and the frequency domain data. When implemented specifically, it may include: Determine the average flight time of the target battery using the following formula:

[0050] in, is the average flight time, T i1 T is the flight time of the signal with the highest signal strength value in the signal data received by the ultrasonic transducer numbered i, i2 T is the flight time of the signal with the second highest signal strength value in the signal data received by the ultrasonic transducer numbered i, i3 It is the flight time of the signal with the third highest signal strength value in the signal data received by the ultrasonic transducer numbered i.

[0051] In some embodiments, when the target feature includes at least an average main frequency offset ratio, the target feature of the target battery is determined according to the time domain data and the frequency domain data. When implemented specifically, the following may be included: Determine the average main frequency deviation ratio of the target battery according to the following formula:

[0052] in, is the average main frequency deviation ratio, f i1 is the frequency of the signal with the highest signal strength value in the signal data received by the ultrasonic transducer numbered i, f i2 is the frequency of the signal whose signal strength value ranks second in the signal data received by the ultrasonic transducer numbered i, f i3 is the frequency of the signal with the third highest signal strength value in the signal data received by the ultrasonic transducer numbered i, and f is the frequency of the main frequency signal of the ultrasonic signal emitted by the ultrasonic transducer.

[0053] In some embodiments, during specific implementation, the signal data received by each ultrasonic transducer can be extracted based on the time domain data and the frequency domain data; wherein the above signal data can include multiple groups of signal data corresponding to different frequencies, and each group of signal data at least includes attribute information such as signal strength, flight signal, frequency, etc. Then, according to the signal strength, the signal data received by each ultrasonic transducer is sorted in descending order of signal strength; then the attribute information of the top three signal data is obtained; and according to the attribute information of the top three signal data, the corresponding target characteristics of the target battery are calculated.

[0054] In some embodiments, after determining the health status of the target battery according to the target prediction result, the method may further include the following when implemented: S1: When it is determined that the health status of the target battery is abnormal, determine the abnormal risk level of the target battery; S2: Detect whether the abnormal risk level of the target battery is less than a preset level threshold; S3: When it is determined that the abnormal risk level is less than a preset level threshold, a matching target maintenance strategy is determined from a preset maintenance strategy set according to the abnormal risk level; S4: Automatically maintain the target battery according to the target maintenance strategy.

[0055] The preset maintenance strategy set includes a plurality of preset maintenance strategies; each preset maintenance strategy corresponds to at least one abnormal risk level.

[0056] Before specific implementation, a large number of historical battery maintenance records can be collected; then the historical battery maintenance records are clustered to obtain multiple maintenance processing operation data groups; wherein each maintenance processing operation data group corresponds to at least one abnormal risk level, and contains common maintenance processing operations corresponding to the abnormal risk level; based on each maintenance processing operation data group, a plurality of preset maintenance strategies corresponding to the abnormal risk level are constructed; and a plurality of preset maintenance strategies are combined to obtain a preset maintenance strategy set.

[0057] Based on the above embodiments, when the abnormal risk level of the target battery is small and the risk level is relatively low, a matching target maintenance strategy can be determined and utilized to automatically maintain the target battery, so as to promptly eliminate risk factors that affect the health status of the target battery and continue to operate stably and safely.

[0058] In a specific implementation, after detecting whether the abnormal risk level of the target battery is less than a preset level threshold, the method may further include the following contents when implemented: S1: When it is determined that the abnormal risk level is greater than or equal to a preset level threshold, a matching target maintenance strategy is determined from a preset maintenance strategy set according to the abnormal risk level; S2: calling a preset abnormal cause analysis model to process target features and obtain multiple abnormal causes (e.g., gas accumulation, etc.) whose probability values ​​are greater than or equal to a preset probability threshold; wherein the preset abnormal cause analysis model is an algorithm model obtained in advance by using big data analysis and machine learning training; S3: Combine the target maintenance strategy and the abnormality cause to generate an abnormality analysis and processing report for the target battery; S4: Sending abnormal risk processing prompt information about the target battery to the operation and maintenance end; wherein the abnormal risk processing prompt information at least carries an abnormal analysis and processing report.

[0059] Correspondingly, the operation and maintenance end can receive and respond to the abnormal risk processing prompt information, and display the abnormal analysis and processing report to the operation and maintenance personnel, so that the operation and maintenance personnel can use the abnormal analysis and processing report as a reference to manually maintain the target battery to ensure the safe and stable operation of the target battery and extend the service life of the target battery.

[0060] In some embodiments, see Figure 3 As shown, when the method is implemented specifically, it may also include the following contents: S1: Screen multiple batteries in different health states as sample batteries; S2: deploying a corresponding target ultrasonic transducer group on the working plane of the sample battery according to a preset deployment rule, and using the target ultrasonic transducer group to perform a test experiment to obtain sample data corresponding to the sample battery; S3: Using the sample data, construct a corresponding sample data set; S4: Determine multiple computational particles; S5: randomly extract multiple sample data from the sample data set to generate multiple sub-sample data sets; and construct an initial prediction model based on the random forest regression prediction model; S6: Call multiple computing particles to perform multiple rounds of iterative training based on the initial prediction model and the sub-sample data set until the training end condition is met, and a preset battery health status prediction model that meets the requirements is obtained.

[0061] The computing particles mentioned above can be relatively independent computing nodes (e.g., computer nodes) in a distributed cluster. Each computing particle can be regarded as an individual, and is deployed with corresponding local iterative training algorithm rules and is connected to different computing particles. The distributed cluster can be understood as a system composed of multiple computer nodes, which communicate and collaborate with each other through the network to jointly complete a task or provide a service, and has the advantages of high availability, high performance, and scalability. In this way, the performance advantages of the distributed system can be utilized to efficiently complete the relevant training.

[0062] The sample batteries are specifically batteries with known health status. Specifically, the sample batteries may include at least sample batteries with normal health status and sample batteries with abnormal health status.

[0063] When conducting a specific test experiment, the target ultrasonic transducer group can be used to transmit ultrasonic signals on the working surface of the sample battery; and the corresponding waveform signals can be collected; and the waveform signals are then combined with the health status label of the sample battery to obtain sample data corresponding to the sample battery.

[0064] In specific implementation, for the same sample battery, multiple test experiments can be carried out in the above manner to obtain multiple waveform signals; then, based on the multiple waveform signals, a reliable waveform signal is screened out through statistical analysis to construct corresponding sample data.

[0065] In specific implementation, multiple sample data can be combined to construct a corresponding sample data set (or sample database); further, according to the data scale of the sample data set and the state performance of a single computing particle, a matching quantity value can be determined, thereby determining multiple computing particles.

[0066] Based on the above embodiments, the target ultrasonic transducer group can be used to collect sample data with high reference value and relatively comprehensive and complete. Then, the particle swarm optimization algorithm and the random forest regression prediction model can be used in combination to utilize the above sample data to efficiently train a preset battery health status prediction model with high model accuracy and good application effect.

[0067] In a specific implementation, the target ultrasonic transducer group can be used to transmit ultrasonic signals on the working plane of the sample battery; and corresponding waveform signals can be collected at preset time intervals (for example, every 1 second) as sample data corresponding to the sample battery.

[0068] In specific implementation, multiple sample data can be extracted from the sample data set with replacement according to the number of computing particles to generate multiple sub-sample data sets. Different sub-sample data sets can be allocated to different computing particles.

[0069] The above subsample data set can be specifically expressed as: D (b) ={(x i ,y i )|i∈S b}. Among them, D (b) is the subsample data set numbered b, S b is the sample data index set of the sub-sample data set, (x i ,y i ) represents a sample data numbered i, x i represents the sample feature of the sample battery numbered i, y i Indicates the sample health status corresponding to the sample feature numbered i of the sample battery.

[0070] In specific implementation, an initial prediction model based on the random forest regression prediction model can be constructed according to the number of features contained in the sample data in the sample data set.

[0071] Specifically, the formula [log2(M+1)] can be used to determine the initial value m of the number of features involved in a single decision tree in the initial prediction model (i.e., decision attribute features). For example, M takes a value of 3; correspondingly, m takes a value of 2. Then, the initial prediction model based on the random forest regression prediction model can be constructed based on the initial value of the number of features.

[0072] In specific implementation, the initial prediction model can be assigned to each computing particle; then the computing particles are called to perform multiple rounds of iterative training based on the initial prediction model and the corresponding sub-sample data set, and the prediction model is continuously optimized and updated until a preset battery health status prediction model that meets the requirements is obtained.

[0073] In some embodiments, see Figure 4 As shown, the above calls multiple computing particles to perform multiple rounds of iterative training based on the initial prediction model and the sub-sample data set. The specific implementation may include the following: S1: Call multiple computing particles to determine the speed parameters of the current round corresponding to the computing particles based on the local solution particle parameter group of the previous round and the global solution particle parameter group of the previous round; wherein the particle parameter group includes at least: pruning threshold, number of decision trees, and test sample ratio; S2: calling multiple computing particles to update their respective prediction models of the previous round based on the speed parameters and sub-sample data sets of the current round, and obtaining their respective prediction models of the current round; S3: Call multiple computing particles to update the local solution particle parameter group of the current round according to their respective prediction models of the current round; and update the global solution particle parameter group of the current round according to the local solution particle parameter group of the current round; S4: Check whether the training end condition is currently met; S5: When it is determined that the training end condition is met, a preset battery health status prediction model that meets the requirements is determined based on the global solution particle parameter group of the current round.

[0074] Based on the above embodiment, multiple computing particles can be used to continuously adjust and utilize the latest particle parameter group through multiple rounds of iterative training, adjust and update the prediction model of each computing particle, and efficiently train to obtain a preset battery health status prediction model with good globality.

[0075] In specific implementation, a pruning threshold (which can be denoted as ), the number of decision trees (which can be recorded as L), and the proportion of test samples (which can be recorded as X) are combined as a particle parameter group, which is expressed as .

[0076] The above pruning threshold and number of decision trees can be used to optimize and update the prediction model during the model training process. The above test sample ratio can be used to process the subsample data set assigned to the computing particles to obtain the specific training sample set and test sample set in the model training process, so as to perform training update and test update on the prediction model respectively.

[0077] In a specific implementation, the above method determines the speed parameter of the current round corresponding to the calculated particle based on the local solution particle parameter group of the previous round and the global solution particle parameter group of the previous round. In a specific implementation, the following contents may be included: According to the following formula, the speed parameter of the current round of the calculated particle is determined: v id =wv id +c1*rand()*(p id -x id )+c2*rand()*(p gd -x id ) Among them, v id is the speed parameter of the current wheel, w is the inertia weight, c1 is the first acceleration coefficient, c2 is the second acceleration coefficient, rand() represents a random function that varies in the range [0,1], p id is the particle parameter group of the local solution in the previous round (local optimal solution), p gd is the particle parameter group of the global solution of the previous round (global optimal solution), x id is the sample feature in the sample data.

[0078] Specifically, the values ​​of the first acceleration coefficient and the second acceleration coefficient may be 2, and the value of the inertia weight may be 0.8. The values ​​of the first acceleration coefficient, the second acceleration coefficient, and the inertia weight may be determined by performing big data analysis and learning on a large amount of sample data.

[0079] In specific implementation, each computing particle can be called to use the current round's speed parameters and sub-sample data set to update the sample features (for example, update the sample feature x according to the following formula id =x id +v id ), and obtain the updated sample data; then based on the global solution particle parameter group of the previous round, split the updated sample data into the training sample set and the test sample set of the current round, and determine the model training parameters (including: pruning threshold, number of decision trees, etc.). Then use the model training parameters and the training sample set to train and update the prediction model of the previous round to obtain the updated and optimized prediction model; then use the test sample set to test and update the updated and optimized model until the obtained prediction model meets the preset training requirements, and then end the update processing of the current round of calculation particles to obtain the prediction model of the current round of the calculation particle.

[0080] Specifically, when each computing particle can use the training sample set and the test sample set to update the model respectively, the training sample set and the test sample set can be split for each decision tree using the feature selection formula; then the split training sample set and the split test sample set are used to update the decision tree, and finally the corresponding regression tree is obtained, and then the prediction model of the current round of the computing particle is obtained.

[0081] Among them, the above feature selection formula can be expressed as follows:

[0082] Among them, j is the selected feature index, m is the number of sample data in the current node, and y i is the true health status of the sample data numbered i, is the predicted value of the health status after splitting according to feature j.

[0083] In specific implementation, each computing particle can respectively calculate and detect the fitting effect of the prediction model according to the evaluation index mean square error (MSE) and R-squared parameter (R-squared) of the prediction model, and then determine whether the prediction model meets the preset training requirements. If the evaluation index mean square error and R-squared parameter calculated based on the prediction model are respectively less than the preset reference value of the evaluation index mean square error and the preset reference value of the R-squared parameter, it can be determined that the prediction model meets the preset training requirements, and the prediction model is determined as the prediction model of the current round of the computing particle. On the contrary, if the evaluation index mean square error calculated based on the prediction model is greater than or equal to the preset reference value of the evaluation index mean square error, and / or the R-squared parameter is greater than or equal to the preset reference value of the R-squared parameter, it can be determined that the prediction model does not meet the preset training requirements, and then the model can continue to be updated according to the training sample set and the test sample set.

[0084] Specifically, the evaluation index mean square error and R square parameter can be calculated according to the following formula:

[0085] Among them, MSE is the mean square error of the evaluation index, R 2 The reference value of the preset evaluation index mean square error and the reference value of the preset R square parameter can be determined by statistical analysis based on historical data.

[0086] In specific implementation, each computing particle can determine the corresponding particle parameter group according to its respective current round prediction model; and compare the fitting effect of the particle parameter group determined based on the evaluation index mean square error and R square parameter with the current local solution particle parameter group of the computing particle (i.e., the current local optimal solution). If the particle parameter group is better than the current local solution particle parameter group, the current local solution particle parameter group of the computing particle is replaced with the particle parameter group; otherwise, the current local solution particle parameter group is kept unchanged, thereby realizing the update of the local particle parameter group of the current round.

[0087] After the current local solution particle parameter group of the computation particle is updated to the particle parameter group, the computation particle is triggered to compare the current local solution particle parameter group with the current global solution particle parameter group (i.e., the current global optimal solution). If the current local solution particle parameter group is better than the current global solution particle parameter group, the current global solution particle parameter group is replaced with the current local solution particle parameter group of the computation particle; otherwise, the current global solution particle parameter group is kept unchanged, thereby realizing the update of the global solution particle parameter group.

[0088] In specific implementation, it is possible to detect whether the current training end condition is met, for example, by detecting whether the number of rounds in the current round exceeds the specified number of rounds, and / or whether the model error of the optimal prediction model trained based on the global solution particle parameter group of the current round is less than a preset model error threshold, to determine whether the training end condition is met.

[0089] When it is determined that the training termination conditions are currently met, a model that meets the requirements can be screened out from the prediction models of the current round of multiple computing particles based on the global solution example parameter group of the current round as a battery health status prediction model that meets the requirements; and the training is terminated.

[0090] On the contrary, when it is determined that the training end condition is not met at present, the above process can be repeated to continue the next round of iterative training until the training end condition is met.

[0091] As can be seen from the above, the method for determining the battery health state provided in the embodiment of this specification can be trained more efficiently to obtain a preset battery health state prediction model with high accuracy and good effect by jointly using the particle swarm optimization algorithm and the random forest regression prediction model before specific implementation. In specific implementation, a target ultrasonic transducer group can be deployed on the working surface of the target battery of interest; wherein, the target ultrasonic transducer group includes 3 ultrasonic transducers, the 3 ultrasonic transducers are equidistant from each other, the connecting edges form an equilateral triangle, and the center of the equilateral triangle coincides with the center of the working surface of the target battery. In this way, the above-mentioned target ultrasonic transducer group can be used to simultaneously emit ultrasonic signals to collect a more comprehensive and complete waveform signal about the target battery. Then, according to the waveform signal, the corresponding time domain data and frequency domain data are obtained and used to determine the target features of the target battery; the target features are processed using the preset battery health state prediction model to obtain the corresponding target prediction results; and the health state of the target battery is determined according to the target prediction results. Thus, the health state of the target battery can be monitored and determined more efficiently and accurately, the error can be effectively reduced, and the stable and safe operation of the target battery can be ensured.

[0092] This specification embodiment provides a server, referring to Figure 5 The server includes a network communication port 501, a processor 502 and a memory 503, and the above structures are connected through internal cables so that each structure can perform specific data interaction.

[0093] The network communication port 501 may be used to receive monitoring instructions about a target battery.

[0094] The processor 502 can be specifically used to respond to monitoring instructions, use the target ultrasonic transducer group to emit ultrasonic signals on the working surface of the target battery; and collect corresponding waveform signals; wherein the target ultrasonic transducer group includes 3 ultrasonic transducers, and the connecting sides of the ultrasonic transducers form an equilateral triangle, and the center of the equilateral triangle coincides with the center of the working surface of the target battery; according to the waveform signal, corresponding time domain data and frequency domain data are obtained; according to the time domain data and the frequency domain data, target features about the target battery are determined; the target features are processed using a preset battery health status prediction model to obtain corresponding target prediction results; wherein the preset battery health status prediction model is obtained by jointly training using a particle swarm optimization algorithm and a random forest regression prediction model; according to the target prediction result, the health status of the target battery is determined.

[0095] The memory 503 may be specifically used to store corresponding instruction programs, as well as target characteristics, a preset battery health status prediction model and other related data.

[0096] Based on the above method, the relevant structural performance of the server can be effectively utilized, the data processing speed of the electronic device can be improved, and the data processing for determining the battery health status can be efficiently realized.

[0097] In this embodiment, the network communication port 501 can be a virtual port that is bound to different communication protocols so that different data can be sent or received. For example, the network communication port can be a port responsible for web data communication, a port responsible for FTP data communication, or a port responsible for email data communication. In addition, the network communication port can also be a physical communication interface or communication chip. For example, it can be a wireless mobile network communication chip, such as GSM, CDMA, etc.; it can also be a Wifi chip; it can also be a Bluetooth chip.

[0098] In this embodiment, the processor 502 may be implemented in any appropriate manner. For example, the processor may take the form of a microprocessor or a processor and a computer-readable medium storing a computer-readable program code (such as software or firmware) executable by the (micro)processor, a logic gate, a switch, an application specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller, etc. This specification does not limit this.

[0099] In this embodiment, the memory 503 may include multiple levels. In a digital system, anything that can store binary data can be a memory; in an integrated circuit, a circuit with a storage function but no physical form is also called a memory, such as RAM, FIFO, etc.; in a system, a storage device with a physical form is also called a memory, such as a memory stick, TF card, etc.

[0100] The embodiment of the present specification also provides a computer-readable storage medium based on the above-mentioned method for determining the health status of the battery, wherein the computer-readable storage medium stores computer program instructions, which, when executed, implement: using a target ultrasonic transducer group to emit an ultrasonic signal on the working surface of the target battery; and collecting a corresponding waveform signal; wherein the target ultrasonic transducer group includes three ultrasonic transducers, and the connecting sides of the ultrasonic transducers form an equilateral triangle, and the center of the equilateral triangle coincides with the center of the working surface of the target battery; according to the waveform signal, obtaining corresponding time domain data and frequency domain data; according to the time domain data and the frequency domain data, determining the target features of the target battery; using a preset battery health status prediction model to process the target features to obtain a corresponding target prediction result; wherein the preset battery health status prediction model is obtained by training using a particle swarm optimization algorithm and a random forest regression prediction model; and determining the health status of the target battery according to the target prediction result.

[0101] In this embodiment, the storage medium includes but is not limited to a random access memory (RAM), a read-only memory (ROM), a cache, a hard disk (HDD) or a memory card. The memory may be used to store computer program instructions. The network communication unit may be an interface for network connection communication set in accordance with the standard specified by the communication protocol.

[0102] In this embodiment, the functions and effects specifically implemented by the program instructions stored in the computer-readable storage medium can be explained in comparison with other implementations and will not be described in detail here.

[0103] The embodiments of the present specification also provide a computer program product, which at least includes a computer program, and when the computer program is executed by a processor, the following method steps are implemented: using a target ultrasonic transducer group to emit an ultrasonic signal on the working surface of a target battery; and collecting corresponding waveform signals; wherein the target ultrasonic transducer group includes 3 ultrasonic transducers, and the connecting sides of the ultrasonic transducers form an equilateral triangle, and the center of the equilateral triangle coincides with the center of the working surface of the target battery; according to the waveform signal, obtaining corresponding time domain data and frequency domain data; according to the time domain data and the frequency domain data, determining the target characteristics of the target battery; using a preset battery health status prediction model to process the target characteristics to obtain a corresponding target prediction result; wherein the preset battery health status prediction model is obtained by training using a particle swarm optimization algorithm and a random forest regression prediction model; according to the target prediction result, determining the health status of the target battery.

[0104] See also Figure 6 As shown, the embodiment of this specification also provides a device for determining the health status of a battery, which may specifically include the following structural modules: The acquisition module 601 can be specifically used to transmit ultrasonic signals on the working surface of the target battery using the target ultrasonic transducer group; and collect corresponding waveform signals; wherein the target ultrasonic transducer group includes three ultrasonic transducers, the connecting sides of the ultrasonic transducers form an equilateral triangle, and the center of the equilateral triangle coincides with the center of the working surface of the target battery; The acquisition module 602 may be specifically configured to acquire corresponding time domain data and frequency domain data according to the waveform signal; The first determination module 603 may be specifically configured to determine a target feature of a target battery according to the time domain data and the frequency domain data; The processing module 604 may be specifically used to process the target feature using a preset battery health state prediction model to obtain a corresponding target prediction result; wherein the preset battery health state prediction model is obtained by training by jointly using a particle swarm optimization algorithm and a random forest regression prediction model; The second determination module 605 may be specifically configured to determine the health status of the target battery according to the target prediction result.

[0105] In some embodiments, the target feature may specifically include at least one of the following: average intensity attenuation ratio, average flight time, average main frequency deviation ratio, etc.

[0106] In some embodiments, when the target feature includes at least an average strength attenuation ratio, the first determination module 603 may be implemented to determine the average strength attenuation ratio of the target battery according to the following formula:

[0107] in, is the average intensity attenuation ratio, A i1 is the signal strength of the signal with the highest signal strength value in the signal data received by the ultrasonic transducer numbered i, A i2 is the signal strength of the signal whose signal strength value ranks second in the signal data received by the ultrasonic transducer numbered i, A i3 is the signal strength of the signal ranked third in signal strength value in the signal data received by the ultrasonic transducer numbered i, and A is the signal strength of the main frequency signal of the ultrasonic signal emitted by the ultrasonic transducer.

[0108] In some embodiments, when the target feature includes at least the average flight time, the first determination module 603 may be implemented according to the time domain data and the frequency domain data to determine the average flight time of the target battery according to the following formula:

[0109] in, is the average flight time, T i1 T is the flight time of the signal with the highest signal strength value in the signal data received by the ultrasonic transducer numbered i, i2 T is the flight time of the signal with the second highest signal strength value in the signal data received by the ultrasonic transducer numbered i, i3 It is the flight time of the signal with the third highest signal strength value in the signal data received by the ultrasonic transducer numbered i.

[0110] In some embodiments, when the target feature includes at least an average main frequency offset ratio, the first determination module 603 may be implemented to determine the average main frequency offset ratio of the target battery according to the following formula:

[0111] in, is the average main frequency deviation ratio, f i1 is the frequency of the signal with the highest signal strength value in the signal data received by the ultrasonic transducer numbered i, f i2 is the frequency of the signal whose signal strength value ranks second in the signal data received by the ultrasonic transducer numbered i, f i3 is the frequency of the signal with the third highest signal strength value in the signal data received by the ultrasonic transducer numbered i, and f is the frequency of the main frequency signal of the ultrasonic signal emitted by the ultrasonic transducer.

[0112] In some embodiments, when the device is implemented, it can also be used to: screen multiple batteries with different health states as sample batteries; deploy a corresponding target ultrasonic transducer group on the working plane of the sample battery according to a preset deployment rule, and use the target ultrasonic transducer group to obtain sample data corresponding to the sample battery by conducting a test experiment; use the sample data to construct a corresponding sample data set; determine multiple computing particles; randomly extract multiple sample data from the sample data set to generate multiple sub-sample data sets; and construct an initial prediction model based on a random forest regression prediction model; call multiple computing particles, and perform multiple rounds of iterative training based on the initial prediction model and the sub-sample data set until the training end condition is met to obtain a preset battery health state prediction model that meets the requirements.

[0113] In some embodiments, when the device is implemented, multiple computing particles can be called in the following manner to perform multiple rounds of iterative training based on the initial prediction model and the sub-sample data set: multiple computing particles are called to determine the speed parameters of the current round corresponding to the computing particles based on the local solution particle parameter group of the previous round and the global solution particle parameter group of the previous round; wherein the particle parameter group includes at least: pruning threshold, number of decision trees, and test sample ratio; multiple computing particles are called to update the prediction models of the previous round based on the speed parameters and sub-sample data sets of the current round to obtain the prediction models of the current round; multiple computing particles are called to update the local solution particle parameter groups of the current round according to the prediction models of the current round; and the global solution particle parameter groups of the current round are updated according to the local solution particle parameter groups of the current round; it is detected whether the training end conditions are met at present; when it is determined that the training end conditions are met, a preset battery health status prediction model that meets the requirements is determined according to the global solution particle parameter group of the current round.

[0114] It should be noted that the units, devices or modules described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. For the convenience of description, the above devices are described separately by functions divided into various modules. Of course, when implementing this specification, the functions of each module can be implemented in the same or more software and / or hardware, or the modules that implement the same function can be implemented by a combination of multiple sub-modules or sub-units. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0115] It can be seen from the above that the battery health status determination device provided based on the embodiments of this specification can monitor and determine the health status of the target battery more efficiently and accurately, effectively reduce errors, and ensure stable and safe operation of the target battery.

[0116] In a specific scenario example, the method for determining the battery health state provided in this specification can be applied to implement sodium ion battery health state monitoring. The specific implementation process may include the following contents.

[0117] In this scenario example, considering that the battery is usually subjected to ultrasonic water immersion detection, the angle spectrum of the lithium-ion battery reflection coefficient under different states of charge (SOC) is obtained, and then the mapping relationship between the angle spectrum and the state of charge of the lithium-ion battery is established, and the two peak spacings of the angle spectrum are used to characterize the state of charge (SOC) of the lithium-ion battery, thereby determining the health status of the battery. However, when building an ultrasonic water immersion detection platform, conventional methods mostly use an angle clamp to fix the angle between the two ultrasonic probes to complete the arrangement of the ultrasonic transducer. This layout of ultrasonic transducers will cause the ultrasonic signal to only pass through part of the battery area and cannot completely cover the entire area of ​​the battery, which will result in only partial measurement of the battery and cannot explore the overall situation of the battery, affecting the accuracy of the determined battery health status. In addition, the eigenvalues ​​collected based on the ultrasonic probe are highly random, and different eigenvalues ​​also affect each other, and often show nonlinear coupling, which increases the difficulty of regression prediction, and thus also leads to the regression prediction model trained based on conventional methods. The evaluation accuracy and precision of the battery health status are low.

[0118] In view of the above problems (signal action area limitation and characteristic value randomness) and the root causes of the above problems, the applicant proposes a non-destructive, efficient, low-cost and high-precision sodium ion battery health status monitoring method to achieve real-time and accurate assessment of the health status of sodium ion batteries, see Figure 7 As shown, the following steps may be included.

[0119] Step S1, selecting sodium ion batteries with different battery health conditions as the data source of the training set and the test set for ultrasonic testing, arranging three ultrasonic transducers equidistant from each other in the central area of ​​the working surface of the battery to be tested (i.e., deploying a target ultrasonic transducer group); wherein the geometric center of the transducer coincides with the geometric center of the surface to be tested, ensuring that the sensors are located in the same horizontal plane and parallel to the bottom surface of the working surface, and the transducers are connected to a computer.

[0120] Step S2, the three ultrasonic transducers simultaneously transmit ultrasonic signals of specific frequencies and intensities, and simultaneously receive signals transmitted by themselves and the other two transducers, the waveform signals received by the ultrasonic transducers are filtered and amplified to obtain time domain signals, and the frequency domain signals are obtained by Fourier transform, and the characteristic values ​​(for example, target characteristics) of the processed time domain signals and frequency domain signals are calculated: average intensity attenuation ratio, average flight time, average main frequency offset ratio, and the battery health status data are combined as a test data set (for example, a sample data set).

[0121] Step S3: establish a random forest regression prediction model, and optimize the random forest parameters through the particle swarm algorithm to reduce the model training time and improve the model accuracy. Use the training set to train the random forest model (obtain a preset battery health status prediction model), and verify the model accuracy through the test set.

[0122] Step S4: input the real-time ultrasonic signal characteristic value of the sodium-ion battery to be tested into the random forest regression prediction model optimized by the particle swarm algorithm to realize the real-time SOH prediction of the sodium-ion battery to be tested, and then determine the battery health state.

[0123] Specifically, the battery is a soft-pack sodium ion battery.

[0124] Furthermore, for soft-pack sodium-ion batteries, the working surface is the surface with the largest area of ​​the battery, and its size is 10 cm * 7 cm.

[0125] Furthermore, for the sodium ion soft-pack battery described in this embodiment, its positive electrode material is Prussian blue.

[0126] Furthermore, in step S2, the calculation method of each eigenvalue is: Average intensity attenuation ratio Among them, A i1 , A i2 and A i3 is the intensity corresponding to the three highest frequencies of the i-th ultrasonic transducer, and A is the intensity of the main frequency emitted by the ultrasonic transducer. Because each ultrasonic transducer will receive signals emitted by three transducers, the signal received by each transducer has three main frequencies.

[0127] Average flight time Among them, T i1 、T i2 and T i3 is the signal flight time corresponding to the three frequencies with the highest intensity of the i-th ultrasonic transducer.

[0128] Average frequency deviation ratio Among them, f i1 、f i2 and f i3are the three frequencies with the highest intensity of the i-th ultrasonic transducer, and f is the main frequency of the ultrasonic transducer transmitting signal.

[0129] Furthermore, in step S2, the signal received by the ultrasonic transducer is collected once every 1 second.

[0130] Furthermore, the empirical function of the initial value of the number m of random feature attributes of the random forest regression prediction model optimized by the particle swarm algorithm is [log2(M+1)], where M is the total number of attributes and its value is 3, so the initial value of m is 2.

[0131] For specific implementation, see Figure 8 As shown in the figure, the training process of the random forest regression prediction model optimized by the particle swarm algorithm can include the following: (1) Sample extraction: Multiple samples are randomly extracted from the original data set with replacement to generate multiple sub-data sets. The sample extraction formula is: D (b) ={(x i ,y i )|i∈S b} Among them, D (b) is the b-th sub-dataset. b is the sample index set of the bth sub-data set.

[0132] (2) Decision tree training: For each decision tree, a randomly selected feature set is used for splitting to generate a regression tree. The feature selection formula is:

[0133] Among them, j is the feature index, m is the number of samples in the current node, and y i is the true value of the ith sample, is the predicted value after splitting according to feature j.

[0134] (3) Particle swarm algorithm optimization parameters: set the pruning threshold , the number of decision trees L, and the pre-test sample rate X form a space vector As a particle in the particle swarm algorithm, the particle swarm optimization algorithm is used to iteratively optimize and select relatively good particles, thereby selecting appropriate parameters to improve the final classification accuracy of the weighted random forest model. Among them, the particle speed and vector update formula are as follows: v id =wv id +c1*rand()*(p id -x id )+c2*rand()*(p gd -x id ) x id =x id +v id Where w is the inertia weight, c1 and c2 are acceleration constants, where c1= c2=2. rand() is a random function that varies in the range [0,1], p id is the optimal position of the individual, p gd is the global optimal position, x id is the position of the particle, v id is the velocity of the particle.

[0135] Preferably, w=0.8, c1= c2=2.

[0136] (4) Model prediction: Use the random forest model parameters obtained by the particle swarm optimization algorithm in (3) to generate multiple decision trees for combination. The prediction function is expressed as:

[0137] Among them, k represents the kth decision tree, x represents the input sample, and J k Indicates the number of leaf nodes of the decision tree, c kj Represents the predicted value of the jth leaf node of the decision tree, R kj represents the sample set of the leaf node of the decision tree, and K represents the number of decision trees.

[0138] (5) Model evaluation: Combined use of mean square error (MSE) and R-squared (R 2 ) indicator to detect the fitting effect of the model, and its calculation formula is as follows:

[0139] Where n represents the number of samples, y i represents the true value of the i-th sample, represents the predicted value of the i-th sample, Represents the average of all true sample values.

[0140] Specifically, the distance between any two of the three ultrasonic transducers in step S1 is 1.5 cm, and the transmitting center frequency of the ultrasonic transducers is 1 MHz.

[0141] Specifically, the prediction results of the random forest regression prediction model test set optimized by the particle swarm algorithm can be found in Fig. 9 shown.

[0142] Among them, the model evaluation index mean square error (MSE) is 0.63, R 2It is 0.9851. The predicted SOH is very close to the actual SOH in a large range, which shows that the model has high accuracy and high stability.

[0143] Through the above scenario examples, the method for determining the battery health status provided in this specification is verified. First, the deployment method of the ultrasonic transducer is improved: three ultrasonic transducers are arranged at equal distances from each other in the central area of ​​the working surface of the battery to be tested, and the geometric center of the transducer coincides with the geometric center of the surface to be tested, ensuring that the sensor is located in the same horizontal plane and parallel to the bottom of the working surface. This arrangement allows each ultrasonic transducer to simultaneously receive the signal emitted by itself and the other two transducers, expand the signal coverage range, better identify the internal information of the battery, and improve the stability and reliability of the characteristic values ​​of the collected signal. Secondly, the characteristic values ​​used are improved: the average intensity attenuation ratio, the average flight time, and the average main frequency offset ratio are calculated and used in combination for prediction. Furthermore, the model and algorithm are improved: the particle swarm algorithm is introduced and used to optimize the random forest model, specifically, multiple samples are randomly extracted from the original data set with replacement to generate multiple sub-data sets; for each decision tree, the randomly extracted feature set is used for splitting to generate a regression tree; and then the pruning threshold is set. , the number of decision trees L, and the pre-test sample rate X form a space vector As a particle in the particle swarm optimization algorithm, relatively better particles are selected through iterative optimization of the particle swarm optimization algorithm, so as to select appropriate parameters to improve the final classification accuracy of the weighted random forest model.

[0144] Therefore, by adopting multiple ultrasonic characteristic values ​​in the time domain and frequency domain, including the average intensity attenuation ratio, the average flight time, and the average main frequency offset ratio, and using the particle swarm algorithm to iteratively optimize the parameters of the random forest algorithm model, the shortcomings of the conventional method of the random forest algorithm model, that is, the small number of features and the unoptimized model parameters, are overcome, so that the battery SOH can be predicted more accurately; at the same time, by adopting three ultrasonic transducers arranged equidistantly at the geometric center of the battery test surface, and having the functions of transmitting and receiving signals, the signal covers the entire battery test surface, thus overcoming the defect that the conventional method can only measure the local signal of the battery, and realizing the accurate prediction of the overall SOH of the battery.

[0145] Although the present specification provides method operation steps as described in the embodiments or flow charts, more or fewer operation steps may be included based on conventional or non-creative means. The order of steps listed in the embodiments is only one way of executing the order of many steps and does not represent the only order of execution. When the device or client product in practice is executed, it can be executed in sequence or in parallel according to the method shown in the embodiments or the drawings (for example, a parallel processor or a multi-threaded processing environment, or even a distributed data processing environment). The terms "include", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, product or device including a series of elements includes not only those elements, but also other elements that are not explicitly listed, or also includes elements inherent to such a process, method, product or device. In the absence of more restrictions, it is not excluded that there are other identical or equivalent elements in the process, method, product or device including the elements. The words first, second, etc. are used to represent the name, and do not represent any particular order.

[0146] Those skilled in the art also know that, in addition to implementing the controller in a purely computer-readable program code, the controller can be made to implement the same function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, such a controller can be considered as a hardware component, and the devices for implementing various functions included therein can also be considered as structures within the hardware component. Or even, the devices for implementing various functions can be considered as both software modules for implementing the method and structures within the hardware component.

[0147] This specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, classes, etc. that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer-readable storage media including storage devices.

[0148] Through the description of the above embodiments, it can be known that those skilled in the art can clearly understand that the present specification can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present specification can essentially be embodied in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a mobile terminal, a server, or a network device, etc.) to execute the methods described in each embodiment of the present specification or some parts of the embodiments.

[0149] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. This specification can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld devices or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.

[0150] Although the present specification is described through embodiments, those skilled in the art will appreciate that there are many modifications and changes to the present specification without departing from the spirit of the present specification, and it is intended that the appended claims include these modifications and changes without departing from the spirit of the present specification.

Claims

1. A method for determining a battery health status, characterized in that: include: Using a target ultrasonic transducer group to transmit an ultrasonic signal on a working surface of a target battery; and collecting a corresponding waveform signal; wherein the target ultrasonic transducer group includes three ultrasonic transducers, the connecting sides of the ultrasonic transducers form an equilateral triangle, and the center of the equilateral triangle coincides with the center of the working surface of the target battery; According to the waveform signal, obtaining corresponding time domain data and frequency domain data; Determine target features of the target battery based on the time domain data and the frequency domain data; The target features are processed using a preset battery health state prediction model to obtain a corresponding target prediction result; wherein the preset battery health state prediction model is obtained by training by jointly using a particle swarm optimization algorithm and a random forest regression prediction model; According to the target prediction results, the health status of the target battery is determined.

2. The method according to claim 1, characterized in that The target feature includes at least one of the following: average intensity attenuation ratio, average flight time, and average main frequency deviation ratio.

3. The method according to claim 2, characterized in that In the case where the target feature includes at least the average intensity attenuation ratio, the target feature of the target battery is determined according to the time domain data and the frequency domain data, including: Determine the average intensity attenuation ratio of the target battery according to the following formula: in, is the average intensity attenuation ratio, A i1 is the signal strength of the signal with the highest signal strength value in the signal data received by the ultrasonic transducer numbered i, A i2 is the signal strength of the signal whose signal strength value ranks second in the signal data received by the ultrasonic transducer numbered i, A i3 is the signal strength of the signal ranked third in signal strength value in the signal data received by the ultrasonic transducer numbered i, and A is the signal strength of the main frequency signal of the ultrasonic signal emitted by the ultrasonic transducer.

4. The method according to claim 2, characterized in that: When the target feature includes at least the average flight time, the target feature of the target battery is determined based on the time domain data and the frequency domain data, including: Determine the average flight time of the target battery using the following formula: in, is the average flight time, T i1 T is the flight time of the signal with the highest signal strength value in the signal data received by the ultrasonic transducer numbered i, i2 T is the flight time of the signal with the second highest signal strength value in the signal data received by the ultrasonic transducer numbered i, i3 It is the flight time of the signal with the third highest signal strength value in the signal data received by the ultrasonic transducer numbered i.

5. The method according to claim 2, characterized in that: When the target feature includes at least the average main frequency offset ratio, the target feature of the target battery is determined according to the time domain data and the frequency domain data, including: Determine the average main frequency deviation ratio of the target battery according to the following formula: in, is the average main frequency deviation ratio, f i1 is the frequency of the signal with the highest signal strength value in the signal data received by the ultrasonic transducer numbered i, f i2 is the frequency of the signal whose signal strength value ranks second in the signal data received by the ultrasonic transducer numbered i, f i3 is the frequency of the signal with the third highest signal strength value in the signal data received by the ultrasonic transducer numbered i, and f is the frequency of the main frequency signal of the ultrasonic signal emitted by the ultrasonic transducer.

6. The method according to claim 1, characterized in that The method further comprises: Screening multiple batteries in different health states as sample batteries; According to the preset deployment rules, a corresponding target ultrasonic transducer group is deployed on the working plane of the sample battery, and the target ultrasonic transducer group is used to perform a test experiment to obtain sample data corresponding to the sample battery; Using the sample data, construct a corresponding sample data set; A plurality of computational particles are determined; A plurality of sample data are randomly extracted from the sample data set to generate a plurality of sub-sample data sets; and an initial prediction model based on the random forest regression prediction model is constructed; Multiple computing particles are called to perform multiple rounds of iterative training based on the initial prediction model and the sub-sample data set until the training end conditions are met, and a preset battery health status prediction model that meets the requirements is obtained.

7. The method according to claim 6, characterized in that Call multiple computing particles to perform multiple rounds of iterative training based on the initial prediction model and sub-sample datasets, including: Calling multiple computing particles to determine the speed parameters of the current round corresponding to the computing particles based on the local solution particle parameter group of the previous round and the global solution particle parameter group of the previous round; wherein the particle parameter group includes at least: pruning threshold, number of decision trees, and test sample ratio; Call multiple computing particles to update their respective prediction models of the previous round based on the speed parameters and sub-sample data sets of the current round, and obtain their respective prediction models of the current round; Call multiple computing particles to update the local solution particle parameter group of the current round according to their respective prediction models of the current round; and update the global solution particle parameter group of the current round according to the local solution particle parameter group of the current round; Check whether the training end conditions are met; When it is determined that the training end conditions are met, a preset battery health status prediction model that meets the requirements is determined based on the global solution particle parameter group of the current round.

8. A device for determining a battery health status, characterized in that: include: A collection module, used to transmit ultrasonic signals on the working surface of the target battery using a target ultrasonic transducer group; and collect corresponding waveform signals; wherein the target ultrasonic transducer group includes three ultrasonic transducers, the connecting sides of the ultrasonic transducers form an equilateral triangle, and the center of the equilateral triangle coincides with the center of the working surface of the target battery; An acquisition module, used for acquiring corresponding time domain data and frequency domain data according to the waveform signal; A first determination module, used to determine a target feature of a target battery according to the time domain data and the frequency domain data; A processing module, used to process target features using a preset battery health state prediction model to obtain a corresponding target prediction result; wherein the preset battery health state prediction model is obtained by training by jointly using a particle swarm optimization algorithm and a random forest regression prediction model; The second determination module is used to determine the health status of the target battery according to the target prediction result.

9. A server, characterized in that: The method comprises a processor and a memory for storing processor-executable instructions, wherein the processor implements the steps of the method according to any one of claims 1 to 7 when executing the instructions.

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

Images