A method and system for generating equivalent cloud data of a power battery

Abstract

The application provides a power battery equivalent cloud data generation method and system based on offline detection data, which comprises a power battery offline detection step, a power battery cloud data description establishment step, an equivalent data processing step, an equivalent data correction step and an integration verification step. Based on the power battery offline detection data, the power battery cloud data description is established, and the power battery equivalent cloud data is constructed by screening and simulating the power battery cloud data description. The equivalent data correction processing is simulated, the equivalent condition of the power battery equivalent cloud data is comprehensively evaluated, and the step is cycled until the consistency requirement is met. The method of the application establishes a reliable power battery equivalent cloud data, and can provide a reliable data basis for power battery cloud model and algorithm evaluation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power battery testing, in particular to a power battery equivalent cloud data generation method and system based on offline detection data. BACKGROUND

[0002] Under the global carbon reduction action, the automobile industry has become the industry consensus to transform to new energy. As an important part of carbon peak and carbon neutral, China actively develops new energy vehicle industry. As the core power source of new energy vehicles, the reliability and safety of power batteries are the key to ensure the safe operation of new energy vehicles. The national standard GB / T 32960 stipulates that new energy vehicles need to send necessary real-time vehicle operation data to the cloud monitoring platform to monitor its safe operation state, which includes related operation data of power batteries. With the market popularization of new energy vehicles, the cloud monitoring platform has obtained a large amount of real vehicle operation data related to power batteries. Through these real vehicle operation data, complex battery models can be built in the cloud, and big data algorithms can be designed to fully tap the value of data. However, the current cloud data has some problems such as many breakpoints and large intervals, low accuracy and poor reliability, which makes it very difficult to evaluate the accuracy of models and algorithms developed based on cloud data. Therefore, we need a set of reliable data to verify the accuracy of models and algorithms developed based on cloud data. SUMMARY

[0003] To solve the problems of low accuracy and poor reliability of existing power battery cloud monitoring data, and the difficulty in evaluating the precision of power battery models and algorithms established based on cloud data, the present application proposes a power battery equivalent cloud data generation method based on offline detection data, which is a method of processing offline detection data by describing power battery cloud data to generate high-reliability power battery equivalent cloud data. This method can provide a reliable data basis for power battery cloud model and algorithm evaluation. The present application also relates to a power battery equivalent cloud data generation system based on offline detection data.

[0004] The technical scheme of the present application is as follows:

[0005] A power battery equivalent cloud data generation method based on offline detection data, characterized in that it comprises the following steps:

[0006] The power battery offline detection step acquires detection data by offline detecting power batteries; the power batteries include four different levels of power batteries or power battery systems to be detected, namely power battery monomer, power battery module, power battery pack and whole vehicle power battery system, and the detection contents include battery capacity test, battery internal resistance test, battery dynamic discharge test under wide temperature range condition, large current fast charging test and small current slow charging test.

[0007] The step of establishing the cloud data of the power battery is based on the existing historical running cloud data of the power battery of the real vehicle, the particle filter algorithm is used to clean and repair the data to form an original data set, and then the original data set is analyzed to include the characteristics of data breakpoint distribution, interval, repetition rate and packet loss rate. A power battery model is established to predict the voltage output of the power battery under a given current condition;

[0008] The step of processing equivalent data is to extract offline detection data according to the interval of the original data set after cleaning and repairing, then simulate packet loss and repetition phenomenon by packet loss rate and repetition rate by Monte Carlo method, and randomly generate battery charge and discharge cycle intervals according to breakpoint distribution;

[0009] The step of correcting equivalent data is based on the generated battery charge and discharge cycle interval. For the last frame of message of each charge and discharge stop process, the power battery model is used to predict the voltage output of the power battery at the next time as the last frame of message of the battery charge and discharge cycle interval.

[0010] The step of integrating and verifying integrates all the message data of the generated battery charge and discharge cycle interval including the last frame of message to form an equivalent data set, and verifies the consistency of the equivalent data set data breakpoint distribution, interval, repetition rate and packet loss rate with the original data set. If the consistency is greater than or equal to the set threshold, the message data of the equivalent data set is the equivalent cloud data of the power battery. If the consistency is less than the set threshold, the equivalent data processing step and the equivalent data correction step are repeated until the generated equivalent data set meets the consistency requirement.

[0011] Preferably, in the step of offline testing of the power battery, the data sampling frequency of offline testing is higher than the uploading frequency of cloud running data, which allows different sampling frequencies to be used for different test equipment in offline testing.

[0012] Preferably, in the step of establishing the cloud data of the power battery, the particle filter algorithm for cleaning and repairing the data includes filtering and noise reduction of the data by the particle filter algorithm, sorting and deduplication of the data by logical rules, and identification and detection of data anomalies by fuzzy rules.

[0013] Preferably, analyzing the characteristics of the original data set also includes charge and discharge depth distribution, data granularity, data content, data error, data noise, data error, different judgment conditions for different characteristics of the original data set, for judging data interval distribution, by calculating the time stamp interval of adjacent messages, selecting the mode as the predetermined data interval; for judging the data repetition rate, it is through the consistency or not more than the set first time interval of the time stamp; for the packet loss rate, it is through the adjacent frame message time stamp interval in the continuous segment exceeds the set second time interval; for the charge and discharge depth distribution, it is through judging the single charge or discharge segment record start and end battery capacity of the battery, and calculating the segment length, and then taking the average value of the length of all segments.

[0014] Preferably, in the equivalent data processing step, in addition to the Monte Carlo method, the importance sampling method, the acceptance-rejection sampling method, the decision tree method, the random forest method, and / or the Adaboost algorithm are also used to simulate the packet loss and repetition phenomenon.

[0015] Preferably, the power battery model established in the power battery cloud data description step includes several combinations of equivalent circuit model, electrochemical P2D model, single particle model, heterogeneous model, three-dimensional electrode model, molecular dynamics model, data-driven model, and neural network correction model.

[0016] A power battery equivalent cloud data generation system based on offline detection data, characterized by comprising a power battery offline detection module, a power battery cloud data description module, an equivalent data processing module, an equivalent data correction module and an integration verification module connected in turn;

[0017] The power battery offline detection module detects the power battery offline to obtain detection data; the power battery includes four different levels of power battery or power battery system to be detected, including power battery monomer, power battery module, power battery pack and whole vehicle power battery system, and the detection content includes battery capacity test, battery internal resistance test, battery dynamic discharge test under wide temperature range condition, large current fast charging test and small current slow charging test;

[0018] The power battery cloud data description module is established based on the existing power battery cloud running data of the historical running of the real vehicle, adopts the particle filter algorithm to clean and repair the data to form an original data set, and then analyzes the characteristics of the original data set including data breakpoint distribution, interval, repetition rate and packet loss rate; and a power battery model is established to predict the voltage output of the power battery under a given current condition;

[0019] An equivalent data processing module extracts offline detection data according to the interval of the cloud running data of the original data set after cleaning and repairing, and then simulates packet loss and repetition by means of Monte Carlo method through packet loss rate and repetition rate, and generates a battery charge and discharge cycle interval according to the breakpoint distribution at random;

[0020] An equivalent data correction module, based on the generated battery charge and discharge cycle interval, for each last frame message of the charge and discharge stop process, adopts the power battery model to predict the voltage output of the power battery at the next moment and saves it as the last frame message of the battery charge and discharge cycle interval;

[0021] An integration verification module integrates all the message data of all the generated battery charge and discharge cycle intervals including the last frame message to form an equivalent data set, and verifies the consistency of the data breakpoint distribution, interval, repetition rate and packet loss rate characteristics of the equivalent data set with the original data set, if the consistency is greater than or equal to the set threshold, the message data of the equivalent data set is the equivalent cloud data of the power battery, if the consistency is less than the set threshold, the equivalent data processing module and the equivalent data correction module are repeated until the generated equivalent data set meets the consistency requirement.

[0022] Preferably, in the power battery offline detection module, the data sampling frequency of offline testing is higher than the uploading frequency of cloud running data, allowing different sampling frequencies to be used for different test equipment in offline testing.

[0023] Preferably, the power battery cloud data description module adopts a particle filter algorithm to clean and repair the data, including filtering and denoising the data by means of the particle filter algorithm, sorting and deduplicating the data by means of logical rules, identifying and detecting data anomalies by means of fuzzy rules; analyzing the characteristics of the original data set also includes charge and discharge depth distribution, data granularity, data content, data error, data noise and data error.

[0024] Preferably, the power battery model includes several combinations of equivalent circuit model, electrochemical P2D model, single particle model, heterogeneous model, three-dimensional electrode model, molecular dynamics model, data-driven model and neural network correction model.

[0025] The beneficial effects of the present application are:

[0026] The application provides a power battery equivalent cloud data generation method based on offline detection data, which is a method for processing offline detection data by describing power battery cloud data, and generating power battery equivalent cloud data with high credibility. The method can select power batteries or power battery systems of different scales for detection to generate the same kind of power battery data results as the cloud data, and has universality. When establishing the cloud data description, the data is cleaned and repaired by using multiple targeted methods to reduce the influence on the cloud data description, and then the cloud data description is comprehensively established from multiple dimensions, so that the cloud data description has accuracy. Based on the offline detection data of the power battery, the cloud data description is used for simulation, different characteristic descriptions are used to ensure the accuracy of data simulation by using suitable simulation methods, so as to generate equivalent cloud data of the power battery. Then, in order to simulate the characteristics of the power battery under the actual vehicle condition, such as red light, user parking or long-term parking, which causes the phenomenon of sudden interruption in the discharge segment, the battery current will be zero at this time, but this frame of data does not exist in the test data, so the power battery model is established to simulate the data of the last frame of data, so the battery model is used to predict the subsequent frames of data of the last frame of data of the battery charging and discharging stop process and save, correct the accuracy of the equivalent data, improve the accuracy of the generated equivalent cloud data, and make it can be truly applied to the industrial production process. Finally, the data characteristics of the equivalent data set and the equivalent degree of the cloud data are verified, if the consistency is poor, the equivalent data processing step and the equivalent data correction step are repeated, and the integration and verification are carried out again through the integration verification step, until the generated equivalent data set meets the consistency requirement, the self-feedback verification and correction method can effectively improve the accuracy of the power battery equivalent data.

[0027] The application provides a power battery equivalent cloud data generation method and system based on offline detection data, which simulates various defects and problems existing in the actual power battery cloud data set, so that the cloud data set in the real environment can be better simulated, and the adaptability of the battery model and algorithm to the existing problem data set can be detected. The method and system make full use of existing offline detection experimental data, without the need of additional experimental tests to simulate the running process of the battery in the actual vehicle environment and to abstract the cloud data again, thereby reducing the demand and loss of experimental data. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 It is a power battery equivalent cloud data generation method based on offline detection data of the application.

[0029] Figure 2is the dynamic working condition test data of a certain 116AH battery based on the BST-4000 battery test machine of the application.

[0030] Figure 3 is a second-order equivalent circuit model diagram.

[0031] Figure 4 is the data generated in a single charge-discharge cycle in the equivalent data processing step of the application.

[0032] Figure 5 is the equivalent cloud data diagram of the power battery obtained by using the method of the application.

[0033] Figure 6 is the preferred flowchart of the equivalent cloud data generation of the application. DETAILED DESCRIPTION

[0034] The application will be described below in conjunction with the accompanying drawings.

[0035] The application relates to a power battery equivalent cloud data generation method based on offline detection data, and a flowchart of the method is shown in Figure 1 The method comprises the following steps:

[0036] The power battery offline detection step is used to perform offline detection on the power battery to obtain detection data; the power battery includes four different levels of power battery or power battery system to be detected, namely power battery monomer, power battery module, power battery pack and whole vehicle power battery system; the detection content refers to the test on the types of power battery cloud data obtained in the cloud monitoring platform, including but not limited to battery capacity test, battery internal resistance test, battery dynamic discharge test under wide temperature range conditions, large current fast charging test and small current slow charging test.

[0037] The power battery cloud data description step is established based on the existing power battery cloud operation data (i.e. historical operation data of the real vehicle), the data is cleaned and repaired to form an original data set, then the original data set is analyzed to include the features of data breakpoint distribution, interval, repetition rate and packet loss rate, the cloud data features are described, and a power battery model is established to predict the voltage output of the power battery under a given current working condition; the data cleaning and repairing method preferably includes filtering and noise reduction of the data set by using a particle filtering algorithm, sorting and deduplication of the data set by using a logical rule, and identification and detection of data anomalies by using a fuzzy rule; the description of the cloud data features refers to the comprehensive description and embodiment of the data features of the cloud data set as much as possible, and the features include but are not limited to data breakpoint, data interval, data repetition rate, packet loss rate, charge-discharge depth distribution, data granularity, data content, data error, data noise and data error.

[0038] The equivalent data processing step extracts the offline detection data according to the interval of the cloud running data of the original data set after cleaning and repairing, and then simulates the packet loss and repetition by using the Monte Carlo method through the packet loss rate and repetition rate, and generates the battery charge and discharge cycle interval according to the breakpoint distribution at random. That is, based on the detection data generated by the offline detection step of the power battery, the detection data is screened by using the data characteristics of the cloud data set; first, the equivalent data under the condition of a specific upload frequency is generated, that is, the detection data is extracted according to the data interval of the cloud data to obtain the equivalent data, then the data repetition and packet loss are simulated by using the probability (packet loss rate and repetition rate), that is, the equivalent data is screened frame by frame to simulate the data repetition rate and packet loss rate, and finally the charge and discharge depth is determined based on the statistical frequency, that is, the equivalent data is traversed frame by frame, and the battery charge and discharge cycle interval segment is randomly generated by using the probability (breakpoint distribution) for further screening and limiting.

[0039] The equivalent data modification step, based on the battery charge and discharge cycle interval segment obtained in the equivalent data processing step, for the last frame message of each charge and discharge stop process, uses the power battery model to predict the subsequent output of the power battery, that is, the voltage at the next time, as the last frame message of the battery charge and discharge cycle interval.

[0040] The integration verification step integrates all the message data of the generated battery charge and discharge cycle interval including the last frame message to form an equivalent data set, and verifies the consistency of the data breakpoint distribution, interval, repetition rate and packet loss rate characteristics of the equivalent data set with the original data set. If the consistency is greater than or equal to the set threshold, the message data of the equivalent data set is the equivalent cloud data of the power battery; if the consistency is poor (lower than the set threshold), the equivalent data processing step and the equivalent data modification step are repeated until the generated equivalent data set meets the consistency requirement.

[0041] Preferably, the data sampling frequency of the offline test in the power battery offline detection step is higher than the upload frequency of the cloud running data, which allows different sampling frequencies to be used for different test equipment in the offline test.

[0042] Preferably, in the step of establishing the cloud data description of the power battery, the particle filtering algorithm is used to filter and denoise the data, logical rules are used to sort and de-duplicate the data, fuzzy rules are used to identify and detect data anomalies; analyzing the characteristics of the original data set also includes charge and discharge depth distribution, data granularity, data content, data error, data noise, and data error, and different characteristics of the original data set have different judgment conditions. For determining the data interval distribution, the mode is selected as the predetermined data interval by calculating the timestamp interval of adjacent messages; for determining the data repetition rate, the timestamp is consistent or does not exceed the set first time interval; for the packet loss rate, the timestamp interval of adjacent frame messages in the continuous segment exceeds the set second time interval; for the charge and discharge depth distribution, the single charge or discharge segment record start and end battery capacity of the battery is determined, and the segment length is calculated, and then the average value of the lengths of all segments is taken.

[0043] Preferably, the methods that can be used in the step of simulating data repetition and packet loss through probability in the equivalent data processing step include Monte Carlo method, importance sampling method, acceptance-rejection sampling method, decision tree method, random forest method, and Adaboost algorithm.

[0044] Preferably, the power battery model that can be used in the step of equivalent data correction includes several combinations of equivalent circuit model, electrochemical P2D model, single particle model, heterogeneous model, three-dimensional electrode model, molecular dynamics model, data-driven model, and neural network correction model.

[0045] In the following, a specific embodiment is further described.

[0046] We choose a lithium battery as the test battery.

[0047] In this embodiment, a certain 116AH lithium battery system is used for experimental testing. The maximum voltage of the battery pack is 344V, the minimum voltage is 259V, the total capacity is 116AH, and the total energy is 30.3kWh. We carry out aging experiments on the battery system under NEDC cycle conditions as the offline detection step (first step) of the power battery, pulse discharge experiments with 10% SOC as the index to obtain battery aging experiment data, model parameter identification data, and model verification data. The sampling interval of the data is 1s, the voltage index value is 0.001V, the current index value is 0.001A, and the temperature index value is 0.1℃. The dynamic condition test data are shown in Figure 2 , wherein the abscissa is time, the unit is second, the left ordinate is current, the unit is A, and the right ordinate is voltage, the unit is V.

[0048] Then the second step - the establishment of power battery cloud data description step, we have through the existing cloud running data, using particle filtering algorithm for data cleaning and repair, and then analyze the data breakpoint distribution, interval, packet loss rate, repeat rate and other characteristics. In this embodiment, by analyzing the historical running data of a new energy passenger car within 1 year, the repetition rate, packet loss rate, average charge and discharge depth, data interval, charge and discharge characteristics and other information of the data are detected as the cloud data description; wherein the repetition rate determination condition is: the time stamp is consistent or does not exceed 2s; the packet loss rate determination condition is: the time stamp interval of adjacent frames in the continuous segment is more than 15s; the data interval determination condition is: the time stamp interval of adjacent messages is calculated, and the mode is selected as the predetermined data interval; the average length of the discharge segment: the start and end SOC of the single discharge segment record of the battery are determined, the segment length is calculated, and then the average length of all segments is obtained. The data is shown in Table 1.

[0049] Table 1

[0050] Data characteristics Numerical values Packet loss rate 2.3％ Repetition rate 10.2％ Data interval 10s Charge upper SOC 95％ Discharge segment average length 28％

[0051] Secondly, we extract the offline detection data according to the interval of the cloud data, and then use the Monte Carlo method to simulate the packet loss and repetition phenomenon by probability, and generate the battery charge and discharge cycle interval according to the probability, such as Figure 6As shown, first we take the lithium battery offline detection data obtained in the first step as the original data, and take the data sampling interval of the cloud data in the second step as the input, that is, according to the given sampling data interval 10s, take a frame of original data every 10s, and generate a new sampled interval data set 1; Then we simulate the data packet loss rate, use the Monte Carlo method, traverse the new sampled interval data set 1 from the first frame of message, generate a random number for each frame of message, if the generated random number is less than the given packet loss rate 2.3%, discard this frame of message, and generate a new sampled interval data set 2 from all the satisfied data; Secondly, we simulate the data repetition rate, use the Monte Carlo method, traverse the new sampled interval data set 2 from the first frame of message, generate a random number for each frame of message, if the generated random number is less than the given repetition rate 10.2%, discard this frame of message, and generate a new sampled interval data set 3 from all the satisfied data; Thirdly, we start to label the battery state for the new sampled interval data set 3, and label the battery as charging and discharging state according to the current working condition, simulate the charging and discharging depth distribution, and input the charging depth distribution curve, the discharge segment length curve, the charging parking time distribution curve and the discharge parking time distribution curve of the cloud data to simulate the charging and discharging depth distribution comprehensively. Traverse the new sampled interval data set 3 frame by frame, specifically, one is for the charging segment, generate a random number from the first frame of message, select the charging depth and charging parking time according to the random number, and select the number of messages of the segment length from the frame message, mark the last frame of message of the segment as parking state, and increase the time delay of the next frame of message by the parking time; The second is for the discharging segment, generate a random number from the last frame of message, determine the target termination battery capacity value SOC of the discharging process according to the discharging depth, select the discharging segment length and parking time according to the random number, select the number of messages of the segment length from the frame message, mark the last frame of message of the segment as parking state, and increase the time delay of the next frame of message by the parking time.

[0052] Thirdly, we simulate the working condition of the battery at the end of each charging or discharging, and the preferred embodiment adopts the equivalent circuit method to simulate the subsequent 2-3 frames of messages of the end frame and integrate the segments into a whole cloud equivalent data set.

[0053] In this embodiment, an equivalent circuit model of the lithium ion battery is established, the relaxation curve after the end of the discharging segment of the battery is obtained through the pulse discharging data of the battery, the model parameter identification is performed by using the particle swarm optimization algorithm, and the voltage output of the battery under a given current condition can be predicted by establishing the lithium battery model. The power battery model established in this embodiment is as shown in Figure 3 The core formula of the power battery model is as follows:

[0054]

[0055]

[0056] U L,k = U ocv,k -U p1,k -U p2,k -U o,k +v (3)

[0057] where U o is the voltage drop due to the ohmic resistance; I L is the time domain input current; R o is the ohmic resistance; U p1 is the concentration polarization voltage; R p1 is the concentration polarization resistance; C p1 is the concentration polarization capacitance; U p2 is the diffusion polarization voltage; R p2 is the diffusion polarization resistance; C p2 is the diffusion polarization capacitance; U ocv is the open circuit voltage; U L is the terminal voltage.

[0058] According to the interval of cloud data, offline detection data is extracted, then the Monte Carlo method is used to simulate the phenomena such as packet loss and repetition through probability (packet loss rate and repetition rate), and the battery charge-discharge cycle interval is randomly generated according to the breakpoint distribution, and the generated single charge-discharge cycle data is as shown in Figure 4 , wherein the abscissa is time, the unit is second, the left ordinate is current, the unit is A, and the right ordinate is voltage, the unit is V.

[0059] Finally, the consistency of data breakpoint, interval, packet loss rate and other characteristics with the original data set is verified, that is, the charge-discharge characteristics, data loss rate, repetition rate and the like in the data set are recalculated, the equivalent degree of the data is evaluated, if the consistency is poor, the equivalent data set is regenerated until the consistency requirement is met. The data of the equivalent data set generated by the method of the application is as shown in Figure 5 , wherein the abscissa is time, the unit is second, the left ordinate is current, the unit is A, and the right ordinate is voltage, the unit is V, which is a set of reliable equivalent cloud data of power batteries.

[0060] The application also relates to a power battery equivalent cloud data generation system based on offline detection data, which corresponds to the power battery equivalent cloud data generation method based on offline detection data, and can be understood as a system for realizing the power battery equivalent cloud data generation method based on offline detection data, and comprises a power battery offline detection module, a power battery cloud data description module, an equivalent data processing module, an equivalent data correction module and an integration verification module connected in sequence.

[0061] The power battery line offline detection module is used for offline detection of the power battery to obtain detection data; the power battery includes four different levels of power battery or power battery system to be detected, i.e., a power battery monomer, a power battery module, a power battery pack and a whole vehicle power battery system, and the detection content includes battery capacity testing, battery internal resistance testing, battery dynamic discharge testing under a wide temperature range condition, large current fast charging testing and small current slow charging testing; a power battery cloud data description module is established, based on the existing power battery cloud running data of the historical operation of the real vehicle, a particle filtering algorithm is used to clean and repair the data to form an original data set, and then the original data set is analyzed to include the features of data breakpoint distribution, interval, repetition rate and packet loss rate; a power battery model is established to predict the voltage output of the power battery under a given current condition; an equivalent data processing module is used to extract the offline detection data according to the interval of the cloud running data of the original data set after cleaning and repairing, and then a Monte Carlo method is used to simulate packet loss and repetition phenomenon through the repetition rate and the repetition rate, and the breakpoint distribution is randomly generated to generate a battery charge and discharge cycle interval; an equivalent data correction module is used to generate a battery charge and discharge cycle interval, and for the last frame of the packet of each charge and discharge stop process, the power battery model is used to predict the voltage output of the power battery at the next time and saved as the last frame of the packet of the battery charge and discharge cycle interval; an integration verification module is used to integrate all the packet data of all the generated battery charge and discharge cycle intervals including the last frame of the packet to form an equivalent data set, and to verify the consistency of the features of the data breakpoint distribution, interval, repetition rate and packet loss rate of the equivalent data set with the original data set, if the consistency is greater than or equal to a set threshold, the packet data of the equivalent data set is the equivalent cloud data of the power battery, and if the consistency is lower than the set threshold, the equivalent data processing module and the equivalent data correction module are repeated until the generated equivalent data set meets the consistency requirement.

[0062] Further, in the power battery offline detection module, the data sampling frequency of offline testing is higher than the uploading frequency of cloud running data, and different sampling frequencies are allowed for different test equipment in offline testing.

[0063] Further, the power battery cloud data description module uses a particle filtering algorithm to clean and repair the data, including filtering and noise reduction of the data by using the particle filtering algorithm, sorting and deduplication of the data by using logical rules, and identification and detection of data anomalies by using fuzzy rules; the analysis of the features of the original data set also includes charge and discharge depth distribution, data granularity, data content, data error, data noise and data error.

[0064] Further, the power battery model comprises a combination of equivalent circuit model, electrochemical P2D model, single particle model, heterogeneous model, three-dimensional electrode model, molecular dynamics model, data-driven model, and neural network correction model.

[0065] It should be noted that the above specific embodiments can enable those skilled in the art to more fully understand the present application, but in no way limit the present application. Therefore, although the present application has been described in detail with reference to the drawings and embodiments, those skilled in the art should understand that modifications or equivalent replacements can still be made to the present application, and in any case, all technical solutions and improvements that do not deviate from the spirit and scope of the present application should be covered in the protection scope of the patent of the present application.

Claims

1. A method for generating equivalent cloud data of a power battery based on offline detection data, characterized in that, The method comprises the following steps: The power battery offline detection step is used to detect the power battery offline to obtain detection data; the power battery includes four different levels of power battery or power battery system to be detected, i.e., power battery monomer, power battery module, power battery pack and whole vehicle power battery system, and the detection content includes battery capacity test, battery internal resistance test, battery dynamic discharge test under wide temperature range condition, large current fast charging test and small current slow charging test; The power battery cloud data description step is used to establish a power battery model for predicting the voltage output of the power battery under a given current condition; The equivalent data processing step is used to extract the offline detection data according to the interval of the original data set after cleaning and repairing, and then simulate the packet loss and repetition phenomenon by using the Monte Carlo method according to the packet loss rate and repetition rate, and randomly generate the battery charge and discharge cycle interval according to the breakpoint distribution; The equivalent data correction step is used to correct the equivalent data based on the generated battery charge and discharge cycle interval, and for the last frame of the battery charge and discharge cycle interval, the power battery model is used to predict the voltage output of the power battery at the next time and save it as the last frame of the battery charge and discharge cycle interval; The integration and verification step is used to integrate all the generated battery charge and discharge cycle interval data to form an equivalent data set, and verify the consistency of the equivalent data set data breakpoint distribution, interval, repetition rate and packet loss rate with the original data set, if the consistency is greater than or equal to the set threshold, the equivalent data set is the equivalent cloud data of the power battery; if the consistency is less than the set threshold, the equivalent data processing step and the equivalent data correction step are repeated until the generated equivalent data set meets the consistency requirement. In the power battery offline detection step, the data sampling frequency of offline test is higher than the uploading frequency of cloud running data, and different sampling frequencies are allowed for different test equipment in offline test.

2. The method of claim 1, wherein, In the power battery cloud data description step, the particle filtering algorithm is used to clean and repair the data, including filtering and noise reduction of the data by using the particle filtering algorithm, sorting and deduplication of the data by using logical rules, and identification and detection of data anomalies by using fuzzy rules.

3. The method of claim 1, wherein, ​ 4. The method of claim 1 to 3, wherein, The analysis of the original data set also includes charge and discharge depth distribution, data granularity, data content, data error, data noise, and data error. Different characteristics of the original data set have different judgment conditions. For judging the interval distribution of data, the mode is selected as the predetermined data interval by calculating the timestamp interval of adjacent messages. For judging the data repetition rate, the timestamp is consistent or does not exceed the first time interval. For the packet loss rate, the timestamp interval of adjacent frame messages in the continuous segment exceeds the second time interval. For the charge and discharge depth distribution, the start and end battery capacity of a single charge or discharge segment of the battery is determined, and the segment length is calculated, and then the average length of all segments is calculated.

5. The method of claim 1, wherein, In the equivalent data processing step, in addition to the Monte Carlo method, the importance sampling method, the acceptance-rejection sampling method, the decision tree method, the random forest method, and / or the Adaboost algorithm are also used to simulate packet loss and repetition.

6. The method of claim 1, wherein, The power battery model established in the power battery cloud data description step includes several combinations of equivalent circuit model, electrochemical P2D model, single particle model, heterogeneous model, three-dimensional electrode model, molecular dynamics model, data-driven model, and neural network correction model.

7. A power battery equivalent cloud data generation system based on offline detection data, characterized in that, It includes a power battery offline detection module, a power battery cloud data description module, an equivalent data processing module, an equivalent data correction module, and an integration verification module connected in sequence. The power battery offline detection module detects the power battery offline to obtain detection data. The power battery includes power battery monomer, power battery module, power battery pack, and whole vehicle power battery system, which are four different levels of power battery or power battery system to be detected. The detection content includes battery capacity test, battery internal resistance test, battery dynamic discharge test under wide temperature range condition, large current fast charging test, and small current slow charging test. The power battery cloud data description module is based on the existing power battery cloud running data of the historical operation of the real vehicle. The particle filter algorithm is used to clean and repair the data to form an original data set, and then analyze the characteristics of the original data set, including data breakpoint distribution, interval, repetition rate, and packet loss rate. A power battery model is established to predict the voltage output of the power battery under a given current condition. The equivalent data processing module extracts the offline detection data according to the interval of the cleaned and repaired original data set cloud running data, and then uses the Monte Carlo method to simulate packet loss and repetition by packet loss rate and repetition rate, and randomly generates the battery charge and discharge cycle interval according to the breakpoint distribution. The equivalent data correction module is based on the generated battery charge and discharge cycle interval. For the last frame message of each charge and discharge stop process, the power battery model is used to predict the voltage output of the power battery at the next time and save it as the last frame message of the battery charge and discharge cycle interval. The integration verification module integrates all the generated battery charge and discharge cycle interval data including all the last frame messages to form an equivalent data set, and verifies the consistency of the equivalent data set data breakpoint distribution, interval, repetition rate, and packet loss rate characteristics with the original data set. If the consistency is greater than or equal to the set threshold, the message data of the equivalent data set is the equivalent cloud data of the power battery; if the consistency is less than the set threshold, the equivalent data processing module and the equivalent data correction module are repeated until the generated equivalent data set meets the consistency requirement.

8. The power battery equivalent cloud data generation system according to claim 7, characterized in that, In the power battery offline detection module, the data sampling frequency of offline testing is higher than the cloud operation data upload frequency, allowing different sampling frequencies to be used for different test equipment in offline testing.

9. The power battery equivalent cloud data generation system according to claim 7, wherein, The power battery cloud data description module uses a particle filter algorithm to clean and repair the data, including filtering and noise reduction of the data using a particle filter algorithm, sorting and deduplication of the data using logical rules, and identification and detection of data anomalies using fuzzy rules. The analysis of the characteristics of the original data set also includes charge and discharge depth distribution, data granularity, data content, data error, data noise, and data error.

10. The power battery equivalent cloud data generation system according to any one of claims 7 to 9, characterized in that, The power battery model includes several combinations of equivalent circuit model, electrochemical P2D model, single particle model, heterogeneous model, three-dimensional electrode model, molecular dynamics model, data-driven model, and neural network correction model.

Images