Battery data cleaning method and device and vehicle

Abstract

The invention relates to the technical field of vehicles, in particular to a battery data cleaning method and device and a vehicle, and the method comprises the steps: obtaining the battery data of the vehicle, and recognizing the current working condition data of a battery in the battery data; a pre-calibrated first query table is queried based on the current working condition data to obtain a first cleaning threshold value of the battery data, and the first query table is a corresponding relation table of the working condition data and the cleaning threshold value; and association rules and equivalent circuit models of the battery pack of the vehicle are obtained, multi-stage cleaning actions are executed on the battery data based on the first cleaning threshold value, the association rules and the equivalent circuit models, and the data of the current-stage cleaning action is the data obtained after cleaning of the previous-stage cleaning action. Therefore, the problems of static cleaning, insufficient cleaning precision, weak invisible fault identification capability, insufficient data relevance utilization and the like in related technologies are solved.

Description

Technical Field

[0001] This application relates to the field of vehicle technology, and in particular to a battery data cleaning method, apparatus, and vehicle. Background Technology

[0002] With the development of the new energy vehicle industry, operating vehicles continuously upload massive amounts of battery operation data to cloud servers. This data is the core foundation for conducting battery health status assessments. However, due to limitations such as the accuracy of onboard sensors, CAN (Controller Area Network) bus interference, data transmission packet loss, and asynchronous data collection cycles of battery management systems, the raw data generally suffers from problems such as noise, missing values, transition points, and outliers.

[0003] The relevant technologies generally involve parsing the received data packets to extract time series data such as voltage, temperature, current, and SOC (State of Charge); then filtering out anomalies using preset static physical thresholds, supplemented by a few simple logical rules (such as SOC should monotonically increase during charging) for secondary screening; finally, simple processing such as removing or linear interpolation is applied to the abnormal data.

[0004] Although this scheme is simple to implement and has low computational overhead, it has significant drawbacks: First, the electrochemical characteristics of the battery dynamically change with temperature, aging degree, and charge / discharge rate, and a fixed threshold cannot adapt to this nonlinear, time-varying operating state. Second, this method processes data from different dimensions in isolation, ignoring the inherent physical coupling relationships between signals such as voltage, current, and temperature, making it difficult to identify hidden faults that are reasonable in one dimension but contradictory in multiple dimensions. These problems lead to unreliable data quality after cleaning, which restricts the accuracy and safety of upper-level applications such as battery state estimation, life prediction, and safety early warning. Summary of the Invention

[0005] This application provides a battery data cleaning method, apparatus, and vehicle to address the problems of insufficient cleaning accuracy, weak ability to identify hidden faults, and insufficient utilization of data correlation in related technologies.

[0006] The first aspect of this application provides a battery data cleaning method, including the following steps: acquiring battery data of a vehicle and identifying the current operating condition data of the battery in the battery data; querying a pre-calibrated first lookup table based on the current operating condition data to obtain a first cleaning threshold for the battery data, wherein the first lookup table is a correspondence table between operating condition data and cleaning threshold; acquiring the association rules and equivalent circuit model of the vehicle's battery pack; and performing multi-level cleaning actions on the battery data based on the first cleaning threshold, multiple association rules, and multiple equivalent circuit models, wherein the data of the current level cleaning action is the data cleaned by the previous level cleaning action.

[0007] Optionally, based on a first cleaning threshold, association rules, and multiple equivalent circuit models, multi-level cleaning actions are performed on the battery data, including: performing a first-level cleaning action on the battery data according to the first cleaning threshold; performing a second-level cleaning action on the data cleaned by the first-level cleaning action according to the association rules; and performing a third-level cleaning action on the data cleaned by the second-level cleaning action according to the equivalent circuit model.

[0008] Optionally, a first-level cleaning action is performed on the battery data according to a first cleaning threshold, including: identifying the data value corresponding to at least one data point in the battery data; if the data value corresponding to the data point is less than or equal to the first cleaning threshold, then the data point is retained; if the data value corresponding to the data point is greater than the first cleaning threshold, then the data point is removed from the battery data and marked as first abnormal data.

[0009] Optionally, feature data is extracted from the data cleaned by the first-level cleaning action; a second cleaning threshold for the feature data is determined according to the association rules; and a second-level cleaning action is performed on the feature data according to the association rules and the second cleaning threshold.

[0010] Optionally, the second cleaning threshold includes a temperature threshold, a current threshold, a rate of change threshold, a range threshold, and a power threshold. The association rules include a first rule, a second rule, and a third rule. Determining the second cleaning threshold for the feature data based on the association rules includes: obtaining a second lookup table of the association rules and the second cleaning threshold, wherein the second lookup table is a table showing the correspondence between the association rules and the second cleaning threshold; if the association rule is the first rule, then the second lookup table is queried based on the first rule to obtain the second cleaning threshold as a voltage range threshold; if the association rule is the second rule, then the second lookup table is queried based on the second rule to obtain the second cleaning threshold as a current threshold and a rate of change threshold; if the association rule is the third rule, then the second lookup table is queried based on the third rule to obtain the second cleaning threshold as a temperature rise threshold and a power threshold.

[0011] Optionally, the feature data includes temperature data, first current data, first voltage data, and second voltage data. A second-level cleaning action is performed on the feature data according to association rules and a second cleaning threshold, including: when the association rule is the first rule, calculating the voltage range value based on the first voltage data; if the voltage range value is less than or equal to the range threshold, retaining the first voltage data; if the voltage range value is greater than the range threshold, marking the first voltage data as second abnormal data; when the association rule is the second rule, when the first current data is less than the current threshold, if the rate of change of the first voltage data is greater than the rate of change threshold, marking the current first voltage data as second abnormal data; if the rate of change of the first voltage data is less than or equal to the rate of change threshold, retaining the first voltage data; when the association rule is the third rule, calculating the instantaneous heating power based on the first current data and the second voltage data, and calculating the temperature rise rate based on the temperature data; if the instantaneous heating power is greater than the power threshold and the temperature rise rate is less than or equal to the temperature rise threshold, marking the temperature data as second abnormal data; if the instantaneous heating power is greater than the power threshold and the temperature rise rate is greater than the temperature rise threshold, retaining the temperature data.

[0012] Optionally, a third-level cleaning action is performed on the data cleaned by the second-level cleaning action according to the equivalent circuit model, including: extracting the second current data, the third voltage data, and the circuit state data from the data cleaned by the second-level cleaning action; predicting the terminal voltage data at the current moment through the equivalent circuit model based on the second current data and the circuit state data at the previous moment, and calculating the residual between the terminal voltage data at the current moment and the real-time terminal voltage data; comparing the residual with a residual threshold, and if the residual is greater than the residual threshold and the duration of the residual being greater than the residual threshold is greater than the time threshold, then the data segment with the residual greater than the residual threshold is marked as the third abnormal data; if the residual is less than or equal to the residual threshold, or if the residual is greater than the residual threshold and the duration of the residual being greater than the residual threshold is less than the time threshold, then the residual data segment is retained.

[0013] Optionally, after performing multi-level cleaning actions on the battery data based on multiple factors including a first cleaning threshold, association rules, and an equivalent circuit model, the method further includes: acquiring the data after the multi-level cleaning actions, the first abnormal data, the second abnormal data, and the third abnormal data; identifying the abnormal types of the first abnormal data, the second abnormal data, and the third abnormal data; if the abnormal type is an abnormal point type, then using at least one of median filtering and nearest-neighbor linear interpolation to repair the data after the multi-level cleaning actions; if the abnormal type is at least one of an abnormal data segment type and a continuous missing type, then calling the equivalent circuit model to generate simulated voltage data segments for data filling, and marking the filled data segments as simulated data; if the abnormal type is a sensor failure type, then marking the abnormal data as unusable data; and integrating the data after the multi-level cleaning actions, the repaired data, and at least one of the simulated data into the final cleaned data.

[0014] Optionally, identifying the current operating condition data of the battery in the battery data includes: extracting current data from the battery data and calculating real-time rate data based on the current data; extracting temperature data from the battery data and mapping the temperature data to a temperature range to obtain temperature mapping data; extracting battery health status data from the battery data; and combining the current data, temperature data, and battery health status data at each moment into operating condition data.

[0015] A second aspect of this application provides a battery data cleaning apparatus, comprising: an acquisition module for acquiring battery data of a vehicle and identifying current operating condition data of the battery in the battery data; a query module for querying a pre-calibrated first query table based on the current operating condition data to obtain a first cleaning threshold for the battery data, wherein the first query table is a correspondence table between operating condition data and cleaning thresholds; and a cleaning module for acquiring association rules and equivalent circuit models of the vehicle's battery pack, and performing multi-level cleaning actions on the battery data based on the first cleaning threshold, multiple association rules, and multiple equivalent circuit models, wherein the data of the current level cleaning action is the data cleaned by the previous level cleaning action.

[0016] A third aspect of this application provides a vehicle, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the battery data cleaning method of the first aspect.

[0017] Therefore, this application has the following beneficial effects: This application embodiment can acquire vehicle battery data and identify the current operating condition data of the battery within the battery data; based on the current operating condition data, it queries a pre-calibrated first lookup table to obtain a first cleaning threshold for the battery data, wherein the first lookup table is a correspondence table between operating condition data and cleaning thresholds; it acquires the association rules and equivalent circuit model of the vehicle's battery pack, and performs multi-level cleaning actions on the battery data based on the first cleaning threshold, association rules, and multiple equivalent circuit models. The data of the current level cleaning action is the data cleaned by the previous level cleaning action. By dynamically selecting the cleaning threshold, dynamic adaptive cleaning is achieved, improving cleaning accuracy. At the same time, through three levels of cleaning—cleaning threshold cleaning, association rule cleaning, and equivalent circuit model cleaning—deep mining of hidden anomalies is achieved. Thus, it solves the problems of related technologies, such as static cleaning, insufficient cleaning accuracy, weak ability to identify hidden faults, and insufficient utilization of data correlation.

[0018] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0019] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of a battery data cleaning method provided according to an embodiment of this application; Figure 2 This is a schematic diagram of a battery data cleaning system according to an embodiment of this application; Figure 3 This is a schematic diagram of the battery data cleaning apparatus provided according to an embodiment of this application; Figure 4 This is a structural schematic diagram of a vehicle provided according to an embodiment of this application. Detailed Implementation

[0020] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0021] The following describes a battery data cleaning method, apparatus, and vehicle according to embodiments of this application with reference to the accompanying drawings. Addressing the problems mentioned in the background art, such as insufficient cleaning accuracy, weak ability to identify hidden faults, and insufficient utilization of data correlation, this application provides a battery data cleaning method. In this method, battery data from a vehicle is acquired, and the current operating condition data of the battery is identified. Based on the current operating condition data, a pre-calibrated first lookup table is queried to obtain a first cleaning threshold for the battery data. The first lookup table is a correspondence table between operating condition data and cleaning thresholds. Association rules and equivalent circuit models of the vehicle's battery pack are acquired. Based on the first cleaning threshold, multiple association rules, and equivalent circuit models, multi-level cleaning actions are performed on the battery data. The data of the current cleaning action is the data cleaned by the previous cleaning action. By dynamically selecting the cleaning threshold, dynamic adaptive cleaning is achieved, improving cleaning accuracy. Simultaneously, through three levels of cleaning—cleaning threshold cleaning, association rule cleaning, and equivalent circuit model cleaning—deep mining of hidden anomalies is achieved. Thus, the problems of static cleaning, insufficient cleaning accuracy, weak ability to identify hidden faults, and insufficient utilization of data correlation in related technologies are solved.

[0022] Specifically, Figure 1 This is a schematic flowchart of a battery data cleaning method provided in an embodiment of this application.

[0023] like Figure 1 As shown, the battery data cleaning method includes the following steps: In step S101, the vehicle's battery data is acquired, and the current operating condition data of the battery in the battery data is identified.

[0024] Among them, vehicle battery data refers to the operating parameters related to the power battery collected by the vehicle battery management system and uploaded through a communication network, such as the CAN bus, including time-series data such as cell voltage, module temperature, total current, total voltage, SOC, and charging / discharging power.

[0025] It is understood that the embodiments of this application first obtain battery data uploaded by the vehicle (such as voltage, current, temperature, SOC, etc.), and identify the specific operating condition data of the battery at present, thereby providing accurate contextual basis for subsequent battery health assessment, anomaly detection or energy management.

[0026] In this embodiment of the application, identifying the current operating condition data of the battery in the battery data includes: extracting current data from the battery data and calculating real-time rate data based on the current data; extracting temperature data from the battery data and mapping the temperature data to a temperature range to obtain temperature mapping data; extracting battery health status data from the battery data; and combining the current data, temperature data and battery health status data at each moment into operating condition data.

[0027] Among them, current data refers to the real-time charging and discharging current value of the battery during operation; real-time rate data refers to the ratio of the current charging and discharging current to the battery's rated capacity, used to standardize the description of the charging and discharging rate; temperature data refers to the real-time temperature of the battery cell or module, which directly affects the battery's internal resistance, reaction rate, and safety; temperature range and temperature mapping data refers to dividing continuous temperature values ​​into several discrete ranges according to preset rules, such as low temperature range: ≤0℃, normal temperature range: 0~40℃, high temperature range: >40℃, and classifying temperature values ​​into corresponding ranges to form data that is easy to analyze; battery health status data is used to characterize the degree of degradation of the battery's current performance relative to a new battery, usually expressed as a percentage, and is obtained directly from battery data provided by the vehicle battery management system or through capacity estimation algorithms using historical data.

[0028] It is understood that the embodiments of this application extract current data from battery data and calculate real-time rate data by combining it with the battery's rated capacity; at the same time, temperature data is acquired and classified into predefined temperature ranges (such as low temperature, normal temperature, and high temperature) to form a temperature mapping; then, it is combined with battery health status data that reflects the degree of battery aging; finally, the rate, temperature mapping, and SOH corresponding to each time point are combined to form a ternary label, constructing a working condition data label that can comprehensively characterize the current overall operating status of the battery, providing support for subsequent accurate analysis and decision-making.

[0029] In step S102, a pre-calibrated first lookup table is queried based on the current operating condition data to obtain the first cleaning threshold of the battery data, wherein the first lookup table is a correspondence table between operating condition data and cleaning threshold.

[0030] Among them, the cleaning threshold refers to the reasonable upper and lower limits or rule parameters that are dynamically selected for the current specific operating conditions to determine whether the original battery data is abnormal, such as the allowable range of single cell voltage and the temperature change rate threshold.

[0031] It is understood that, based on the identified current battery operating condition data (including real-time rate, temperature range, and health status), this application embodiment queries a first lookup table established in advance through calibration, namely the "operating condition-cleaning threshold" correspondence table, to dynamically obtain a set of cleaning thresholds applicable to the specific operating condition. These thresholds can more accurately reflect the normal behavior boundary of the battery under the current operating conditions, thereby providing a highly adaptable judgment basis for subsequent anomaly detection and data cleaning.

[0032] In this embodiment of the application, before querying the pre-calibrated first lookup table based on the current operating condition data, the method further includes: obtaining battery data under different operating conditions; determining the normal parameter range of the battery under different operating conditions based on the battery data; determining the cleaning threshold based on the normal parameter range of the battery; and establishing the first lookup table based on the different operating conditions and the cleaning threshold.

[0033] Among them, battery data under different operating conditions refers to the raw operating data of the battery collected under various typical operating conditions, such as low-temperature high-rate discharge, room-temperature quilting, and high-temperature aging charging, including parameters such as voltage, current, temperature, and SOC, used to characterize the battery's behavior in various real-world scenarios; the normal parameter range of the battery refers to the reasonable value range that each parameter should meet under normal conditions, determined by statistical analysis of a large amount of high-quality, fault-free historical or experimental data under specific operating conditions; the cleaning threshold refers to the judgment boundary (such as upper / lower limit, rate of change limit, etc.) set based on the above normal parameter range for data cleaning. Data points exceeding this threshold will be considered abnormal or unreliable and need to be corrected, interpolated, or removed.

[0034] It is understood that before querying the pre-calibrated first lookup table based on the current operating condition data in this embodiment of the application, it is necessary to collect high-quality battery operation data covering a variety of typical operating conditions (such as different charge / discharge rates, temperature environments, and aging degrees); then, for each operating condition, the reasonable range of values ​​for each parameter under normal conditions is determined through statistical analysis or physical modeling, and the corresponding cleaning threshold is set accordingly; finally, the correspondence between "operating condition - cleaning threshold" is organized and stored as a pre-calibrated first lookup table, providing a dynamic and adaptive judgment basis for subsequent real-time data cleaning.

[0035] In step S103, the association rules and equivalent circuit model of the vehicle's battery pack are obtained. Based on the first cleaning threshold, multiple association rules and equivalent circuit models, multi-level cleaning actions are performed on the battery data. The data of the current level cleaning action is the data cleaned by the previous level cleaning action.

[0036] Among them, the association rules of the battery pack refer to the logical or statistical dependencies between various parameters within the battery system, such as "the total voltage should increase with the state of charge (SOC) during charging," "the voltage difference between adjacent cells should not exceed a certain value," and "the current should be close to zero and the voltage should tend to be stable when the battery is at rest." These rules reflect the inherent consistency between multidimensional data of the battery under normal operating conditions and can be used to identify hidden anomalies that violate physical or engineering common sense. The equivalent circuit model is a simplified circuit structure used to simulate the electrochemical behavior of a battery. It is typically composed of components such as an ideal voltage source, internal resistance, and a parallel resistor-capacitor network. It can predict the voltage response based on the input current or based on actual voltage. Voltage measurement can be used to infer state, which can be applied to state estimation and anomaly detection. Multi-level cleaning refers to dividing the data cleaning process into multiple sequentially executed stages, such as the first stage: initial screening of cleaning thresholds, the second stage: association rule verification, and the third stage: equivalent circuit model verification. Each stage uses different methods to progressively refine the data. The data of the current cleaning stage is the data cleaned by the previous stage. This emphasizes the sequential dependency and data flow transitivity of multi-level cleaning. Only data that has been filtered or corrected by the preceding cleaning steps will be sent to the next more complex cleaning logic to avoid noise or outliers interfering with the judgment of the model or rules.

[0037] It is understood that, based on the multi-dimensional association rules and equivalent circuit model of the vehicle battery pack, this application embodiment constructs a multi-level collaborative data cleaning process by combining the cleaning threshold determined adaptively under the current operating conditions. First, the original battery data is initially screened based on the dynamic threshold. Then, the association rules are used to check the logical consistency between the data. Finally, the equivalent circuit model is used to verify whether the measured signal meets the expected physical behavior of the battery. Each level of cleaning uses the processed data output from the previous level as input, improving the data quality layer by layer, and realizing refined cleaning from simple anomaly removal to complex hidden fault identification.

[0038] In this embodiment, based on a first cleaning threshold, association rules, and multiple equivalent circuit models, a multi-level cleaning action is performed on the battery data, including: performing a first-level cleaning action on the battery data according to the first cleaning threshold; performing a second-level cleaning action on the data cleaned by the first-level cleaning action according to the association rules; and performing a third-level cleaning action on the data cleaned by the second-level cleaning action according to the equivalent circuit model.

[0039] It is understood that the embodiments of this application perform a three-level progressive cleaning of battery data: First, based on the cleaning threshold determined by the current operating condition data, the original battery data is cleaned in the first level to remove anomalies that are obviously beyond the physical reasonable range; second, using the association rules that reflect the inherent consistency of multiple battery parameters, the data after the first level of cleaning is verified in the second level to identify and correct hidden anomalies with logical contradictions; finally, based on the equivalent circuit model, the data after the second level of cleaning is verified in the third level. By comparing the consistency between the measured voltage and the model prediction value, residual noise or erroneous data that does not conform to the electrochemical dynamic characteristics of the battery is further filtered out, thereby obtaining high-reliability, high-quality battery operating data.

[0040] In this embodiment of the application, a first-level cleaning action is performed on the battery data according to a first cleaning threshold, including: identifying the data value corresponding to at least one data point in the battery data; if the data value corresponding to the data point is less than or equal to the first cleaning threshold, then the data point is retained; if the data value corresponding to the data point is greater than the first cleaning threshold, then the data point is removed from the battery data and the data point is marked as first abnormal data.

[0041] Among them, a data point refers to the specific observed value of a parameter (such as the voltage of a single cell, the temperature at a certain time) in the battery data at a certain moment, which is the basic unit of time series data; the first abnormal data refers to the data points that are judged to exceed the first cleaning threshold range in the first level of cleaning, and are regarded as preliminary suspicious or invalid data, which need to be processed later, such as removal, correction or marking.

[0042] It is understood that in the first-level cleaning of this application embodiment, each data point (such as voltage, temperature, etc.) in the battery data is compared one by one with the first cleaning threshold of its corresponding parameter. If the value of a data point exceeds the reasonable range defined by the threshold, it is marked as the first abnormal data. If its value is within the threshold range, it is marked as the first cleaned data. This process realizes the preliminary screening of the original data and effectively removes extreme abnormal values ​​that obviously violate the physical boundary or operating condition characteristics, laying the foundation for subsequent more refined cleaning steps.

[0043] In this embodiment of the application, a second-level cleaning action is performed on the data cleaned by the first-level cleaning action according to the association rules, including: extracting feature data from the data cleaned by the first-level cleaning action; determining a second cleaning threshold for the feature data according to the association rules; and performing the second-level cleaning action on the feature data according to the association rules and the second cleaning threshold.

[0044] Among them, the characteristic data are the temperature data, current data, total voltage data and individual unit voltage data in the first cleaning data.

[0045] It is understood that in the second-level cleaning of this application embodiment, feature data is first extracted from the data retained after the first-level cleaning, including the individual voltage range, voltage change rate, heating power and temperature rise rate, etc. Then, according to the preset association rules, a corresponding second cleaning threshold is matched for each type of feature data. Finally, the second-level cleaning action is performed according to the feature data and the second cleaning threshold. Through multi-dimensional constraints, hidden problems where individual variables are normal but the joint state is abnormal are effectively identified, thereby improving the cleaning accuracy.

[0046] In this embodiment, the second cleaning threshold includes a temperature rise threshold, a current threshold, a rate of change threshold, a range threshold, and a power threshold. The association rules include a first rule, a second rule, and a third rule. Determining the second cleaning threshold for the feature data based on the association rules includes: obtaining a second lookup table of the association rules and the second cleaning threshold, wherein the second lookup table is a table showing the correspondence between the association rules and the second cleaning threshold; if the association rule is the first rule, then the second lookup table is queried based on the first rule to obtain the second cleaning threshold as the range threshold; if the association rule is the second rule, then the second lookup table is queried based on the second rule to obtain the second cleaning threshold as the current threshold and the rate of change threshold; if the association rule is the third rule, then the second lookup table is queried based on the third rule to obtain the second cleaning threshold as the temperature rise threshold and the power threshold.

[0047] Among these, the consistency rule reflects the requirement that the state of individual cells within the battery pack should remain relatively consistent; for example, the voltage difference between individual cells should not be too large. The current-voltage correlation rule is based on electrochemical principles; under a specific current state (such as when stationary), the voltage should exhibit corresponding behavior. The energy-thermal coupling rule is based on energy conservation; the battery should generate Joule heat when operating at high current, leading to a temperature rise. If there is significant heat generation but the temperature does not rise, there may be a faulty temperature sensor or abnormal data. The range data refers to the difference between the maximum and minimum values ​​of a group of individual cell voltages, used to measure the consistency of the battery pack. An excessively large range may indicate performance degradation or measurement abnormalities in individual cells. The range threshold is used to determine whether the range of voltages of all cells within the same module is too large; if it is too large, it indicates inconsistency between the states of individual cells within the battery pack. The current threshold is used to determine the upper limit of the absolute current value for determining whether the battery is in a stationary state; for example, if the absolute current value is <0.5... A. When stationary, theoretically there is no current, and the voltage should tend to stabilize. The rate of change threshold refers to the maximum allowable rate of change of the unit voltage per unit time. If the voltage changes too much when stationary, it may indicate problems such as micro-short circuits or sensor drift. The temperature rise rate refers to the speed at which the temperature rises per unit time. It is used to determine whether the temperature changes synchronously with the heating power. The temperature rise threshold and the power threshold are used to determine whether the current rate of temperature rise has a reasonable correlation with the power. The second abnormal data refers to the abnormal data points identified in the second-level cleaning according to the association rules. After being marked, they will not participate in subsequent analysis or need to be further corrected.

[0048] It is understood that in the second-level cleaning process of this application embodiment, a pre-built second query table, namely a mapping table between association rules and second cleaning thresholds, dynamically obtains threshold parameters applicable to the current verification logic: when the first rule is applied, the table is queried to obtain the corresponding range threshold; when the second rule is applied, the current threshold and rate of change threshold are obtained; when the third rule is applied, the temperature rise threshold and power threshold are obtained. Through this rule-driven threshold query mechanism, it is ensured that each type of association verification uses a criterion that strictly matches its physical logic, thereby achieving accurate and adaptive multi-dimensional anomaly identification.

[0049] In this embodiment, the feature data includes temperature data, first current data, first voltage data, and second voltage data. A second-level cleaning action is performed on the feature data according to association rules and a second cleaning threshold, including: when the association rule is the first rule, calculating the voltage range value based on the first voltage data; if the voltage range value is less than or equal to the range threshold, retaining the first voltage data; if the voltage range value is greater than the range threshold, marking the first voltage data as second abnormal data; when the association rule is the second rule, when the first current data is less than the current threshold, if the rate of change of the first voltage data is greater than the rate of change threshold, marking the current first voltage data as second abnormal data; if the rate of change of the first voltage data is less than or equal to the rate of change threshold, retaining the first voltage data; when the association rule is the third rule, calculating the instantaneous heating power based on the first current data and the second voltage data, and calculating the temperature rise rate based on the temperature data; if the instantaneous heating power is greater than the power threshold and the temperature rise rate is less than or equal to the temperature rise threshold, marking the temperature data as second abnormal data; if the instantaneous heating power is greater than the power threshold and the temperature rise rate is greater than the temperature rise threshold, retaining the temperature data.

[0050] Among them, the first current data is real-time current data; the first voltage data is individual unit voltage data; and the second voltage data is total voltage data.

[0051] It is understood that in the second-level cleaning of this application embodiment, based on the first cleaning data retained after the first-level cleaning, multi-dimensional parameters such as temperature data, first current data, first voltage data, and second voltage data are extracted, and deep verification is performed according to three types of association rules: First, a second lookup table is queried according to the type of association rule to obtain the required second cleaning threshold. Through consistency rules and the first voltage data, the range of all individual cell voltages is calculated. If the range threshold is exceeded, the maximum and minimum values ​​of the individual cell voltage at that moment are marked as second abnormal data. Second, according to the current-voltage correlation rule, when the absolute value of the current is lower than the threshold value, the voltage is further checked. When the current threshold (i.e., the battery should be in a static state) is exceeded but the rate of change of the single cell voltage still exceeds the rate of change threshold, the corresponding voltage data is marked as the second abnormal data. Finally, according to the energy-thermal coupling rule, the instantaneous heat generation power is calculated by combining the first current data and the second voltage data. At the same time, the temperature rise rate is calculated from the temperature data. If the heat generation power is greater than the power threshold, but the temperature rise rate is less than or equal to the temperature rise threshold, that is, the power is high but the temperature does not rise or even falls, then the temperature data is judged to be abnormal and marked as the second abnormal data. This process effectively identifies hidden faults or sensor abnormalities that cannot be detected by the first-level cleaning through multivariate physical consistency constraints.

[0052] In the embodiments of this application, The third-level cleaning action is performed on the data after the second-level cleaning action based on the equivalent circuit model. This includes: extracting the second current data, third voltage data, and circuit state data from the data after the second-level cleaning action; predicting the terminal voltage data at the current moment using the equivalent circuit model based on the second current data and the circuit state data at the previous moment, and calculating the residual between the terminal voltage data at the current moment and the real-time terminal voltage data; comparing the residual with a residual threshold. If the residual is greater than the residual threshold and the duration of the residual being greater than the residual threshold is greater than the time threshold, then the data segment with the residual greater than the residual threshold is marked as the third abnormal data. If the residual is less than or equal to the residual threshold, or if the residual is greater than the residual threshold and the duration of the residual being greater than the residual threshold is less than the time threshold, then the residual data segment is retained.

[0053] Among them, the second current data is the real-time current data, which is the actual charging and discharging current value obtained from the battery system at the current sampling moment, and is the input excitation signal of the equivalent circuit model; the real-time terminal voltage data is the total battery voltage (or single cell voltage) data actually measured at the current moment; the circuit state data at the previous moment refers to the value of the internal state variables in the equivalent circuit model at the previous moment, such as the capacitor voltage in the RC (resistor-capacitor) branch, which is used to recursively calculate the current state; the predicted terminal voltage data at the current moment refers to the theoretical terminal voltage calculated by using the equivalent circuit model, combined with the state at the previous moment and the current input current, through the state equation and output equation of the model. Voltage value; residual refers to the difference between the measured terminal voltage and the model predicted terminal voltage, reflecting the degree of deviation between the actual observation and the physical model expectation; residual threshold is the maximum allowable absolute value of residual calibrated based on model accuracy, sensor noise and historical data, used to determine whether the deviation is abnormal; time threshold refers to the minimum continuous time for the residual to exceed the residual threshold, used to eliminate instantaneous interference and focus on persistent abnormal behavior, and is specifically set according to actual needs, without specific limitations here; third abnormal data refers to the data segment identified in the third level of cleaning that deviates significantly and continuously from the prediction of the equivalent circuit model, which may be due to sensor drift, sudden internal failure or communication error.

[0054] It is understood that in the third-level cleaning of this application embodiment, the measured current and terminal voltage at the current moment are extracted from the second cleaning data, and the theoretical terminal voltage at the current moment is recursively calculated by combining the internal state of the equivalent circuit model at the previous moment; then, the predicted value is compared with the actual measured terminal voltage to obtain the residual; if the absolute value of the residual exceeds the preset residual threshold, and the duration of this over-limit state exceeds the set time threshold, it is determined that there is a deep-level anomaly in the data within this time period, and the whole data is marked as the third abnormal data. The data is dynamically consistent through the battery physical model, effectively capturing the hidden faults or measurement deviations that are difficult to detect in the first two cleaning stages and violate the electrochemical laws.

[0055] In this embodiment, after performing multi-level cleaning actions on battery data based on a first cleaning threshold, association rules, and an equivalent circuit model, the method further includes: acquiring the data after the multi-level cleaning actions, first abnormal data, second abnormal data, and third abnormal data; identifying the abnormal types of the first abnormal data, second abnormal data, and third abnormal data; if the abnormal type is an abnormal point type, then at least one of median filtering and nearest neighbor linear interpolation is used to repair the data after the multi-level cleaning actions; if the abnormal type is at least one of an abnormal data segment type and a continuous missing type, then the equivalent circuit model is invoked to generate simulated voltage data segments for data filling, and the filled data segments are marked as simulated data; if the abnormal type is a sensor failure type, then the abnormal data is marked as unusable data; and integrating the data after the multi-level cleaning actions, the repaired data, and at least one of the simulated data into the final cleaned data.

[0056] Among them, anomaly types are classifications of anomalous data forms, mainly including: anomaly point types, i.e., isolated, single or a small number of outlier data points; anomalous data segment types or continuous missing types, i.e., data anomalies or missing data at multiple consecutive time points, but the system as a whole is still in a modelable state; sensor failure types, i.e., large segments of data due to sensor failure, such as long-term (e.g., set to more than 30 seconds) invalid, constant, abrupt, or completely missing data, indicating that the sensor may be faulty and the data is unreliable; median filtering is a nonlinear filtering method that uses the median of the data within a sliding window to replace the center point; neighboring point linear interpolation uses normal data points before and after the anomaly point to perform linear fitting to estimate intermediate missing or outlier values; equivalent circuit model generating simulated voltage data segments refers to using an equivalent circuit model to generate a voltage sequence that conforms to physical laws under the premise of known current input and battery state, used to fill continuous anomalous segments; unusable data refers to data that is completely unreliable due to sensor failure or other reasons, which is not repaired and does not participate in subsequent analysis, and is usually marked as invalid in the dataset.

[0057] It is understood that, after completing multi-level cleaning, the embodiments of this application classify and repair the abnormal data identified at each level, namely the first abnormal data, the second abnormal data, and the third abnormal data. First, the type of abnormal data is determined. If it is an isolated abnormal point, median filtering or linear interpolation of nearby normal points is used for repair. If it is a continuous abnormal segment or a missing segment, an equivalent circuit model is called, and voltage data that conforms to physical laws is generated based on the measured current and battery state through forward simulation to fill in the gaps. Such filling results are clearly marked as simulated data, and the filled data is the repaired data, ensuring the credibility and transparency of subsequent analysis. If the abnormality is a large segment of invalid data caused by sensor failure, it is determined to be unreliable, marked as unusable data, and not repaired. Finally, all the repaired data is integrated with the normal data retained at each level of cleaning to form a complete and highly reliable clean dataset for subsequent advanced applications such as battery state estimation and safety warning.

[0058] According to the battery data cleaning method proposed in this application, the current operating condition data of the battery can be identified by acquiring the battery data of the vehicle; a first cleaning threshold of the battery data can be obtained by querying a pre-calibrated first lookup table based on the current operating condition data, wherein the first lookup table is a correspondence table between the operating condition data and the cleaning threshold; the association rules and equivalent circuit model of the vehicle's battery pack can be acquired; and multi-level cleaning actions can be performed on the battery data based on the first cleaning threshold, multiple association rules and equivalent circuit models. The data of the current level cleaning action is the data cleaned by the previous level cleaning action. By dynamically selecting the cleaning threshold, dynamic adaptive cleaning is achieved, thereby improving the cleaning accuracy. At the same time, through three levels of cleaning—cleaning threshold cleaning, association rule cleaning, and equivalent circuit model cleaning—deep mining of hidden anomalies is achieved.

[0059] The battery data cleaning method is further described below through a specific embodiment.

[0060] The overall system architecture of this embodiment is as follows: Figure 2 As shown, the module is described in detail below: Module 1: Operating Condition Sensing Module The input to the operating condition sensing module is the temperature, current, total voltage, etc. in the raw data stream.

[0061] The processing flow of the operating condition sensing module is as follows: ① Real-time rate data calculation: Real-time rate data = |current| / battery rated capacity.

[0062] ② Temperature range mapping: Map the real-time temperature to a preset range (e.g., low temperature: <0℃; normal temperature: 0-45℃; high temperature: >45℃).

[0063] ③ Health status data acquisition: Read directly from the data reported by the vehicle battery management system, or obtain it through the capacity estimation algorithm using historical data in the cloud.

[0064] The operating condition sensing module outputs operating condition labels for each data moment {temperature mapping data, real-time scaling factor, and health status data}.

[0065] Module 2: Cleaning Threshold Generation Module The cleaning threshold generation module pre-establishes a first lookup table of normal battery parameters under different operating conditions or fits an empirical function based on a large amount of battery bench test data.

[0066] For example, for the operating condition {low temperature, real-time rate data = 0.5C, health status data = 90%}, the reasonable voltage range can be found from the first lookup table as [2.7V, 3.95V].

[0067] Based on the real-time tags output by the operating condition sensing module, the cleaning threshold (voltage upper / lower limit, maximum temperature rise rate, etc.) under the operating condition is dynamically queried or calculated, and a set of dynamically changing thresholds is output and sent to the cleaning engine.

[0068] Module 3: Multi-dimensional Fusion Cleaning Engine The engine employs a three-level progressive cleaning strategy, refining the process layer by layer to balance computational efficiency and cleaning accuracy. Specifically, it includes: First-level cleaning action (cleaning threshold filtering): Process: Each data point (e.g., cell voltage V_cell) is compared with the allowable range [V_min_dynamic, V_max_dynamic] output by the cleaning threshold module under the current operating conditions. Data that significantly exceeds the range is quickly marked as anomalies; this step efficiently removes major outliers.

[0069] Second-level cleaning action (association rule validation): Principle: Based on the battery pack consistency principle and the electro-thermal coupling relationship, a series of rules are established to validate the data that has passed the initial cleaning process. Example of the rule base: Rule 1 (Consistency Rule): Calculate the range ΔV = max(V_cells) - min(V_cells) of all individual cell voltages within the same module. If ΔV exceeds the range threshold (which can vary with SOC and temperature), then mark the individual cell data with the highest and lowest voltages as anomalies.

[0070] Rule 2 (Current-Voltage Correlation Rule): During the static stage (|Current| < I_rest), the voltage changes of all monomers should tend to be consistent and slow. If the voltage change rate |dV / dt| of a certain monomer is significantly higher than that of other monomers, it is marked as abnormal (suspected micro-short circuit).

[0071] Rule 3 (Energy-Thermal Coupling Rule): According to the current I and the total voltage V, calculate the instantaneous heat generation power P_loss ≈ I * (V - OCV). This power should have a reasonable correlation with the temperature rise rate of the battery pack. If the heat generation power is large while the temperature rise rate is zero or negative, mark the temperature data as abnormal.

[0072] The third-level cleaning action (residual analysis based on the equivalent circuit model): The process of the third-level cleaning action is as follows: ① Establish a second-order RC equivalent circuit model for the battery, and its model parameters (R0, R1, C1, R2, C2) are functions of SOC, temperature, and SOH (pre-calibrated).

[0073] ② Use the equivalent circuit model to predict the terminal voltage V_pred(k) at the current moment with the real-time current I(k) as the input and the state at the previous moment as the initial condition.

[0074] ③ Compare the predicted voltage V_pred(k) output by the model with the measured voltage V_meas(k) collected by the sensor, and calculate the absolute residual e(k) = |V_meas(k) - V_pred(k)|.

[0075] ④ Dynamically set a residual threshold e_threshold based on the statistical information (such as mean and variance) of the historical residual sequence (for example: μ + 3σ).

[0076] ⑤ If the residual e(k) continuously exceeds e_threshold, it indicates that the measured data has deviated from the normal behavior predicted by the model, and mark this data segment as abnormal. This method is extremely effective for detecting hidden faults such as micro-short circuits and abnormal growth of internal resistance.

[0077] Module 4: Data Repair Module Adopt different repair strategies according to the type and context of the abnormal data: For isolated and short-term abnormal points (such as pulse noise): Use median filtering or linear interpolation of neighboring points for repair.

[0078] For continuously missing or data segments marked as abnormal: Call the battery model in the advanced cleaning, and run the model to generate a simulated voltage data segment for filling under the given current and temperature inputs, and mark the filled data as "simulated data" specifically for reference in subsequent analysis.

[0079] For large chunks of data where the entire sensor has failed: mark them as "unavailable" and do not recommend repairing them.

[0080] Next, the battery data cleaning apparatus of this application is described with reference to the accompanying drawings.

[0081] Figure 3 This is a schematic diagram of the battery data cleaning apparatus provided in the embodiments of this application, as shown below. Figure 3 As shown, the battery data cleaning device 10 includes: an acquisition module 201, a query module 202, and a cleaning module 203.

[0082] The acquisition module 201 is used to acquire vehicle battery data and identify the current operating condition data of the battery in the battery data; the query module 202 is used to query a pre-calibrated first query table based on the current operating condition data to obtain a first cleaning threshold for the battery data, wherein the first query table is a correspondence table between operating condition data and cleaning threshold; the cleaning module 203 is used to acquire the association rules and equivalent circuit model of the vehicle's battery pack, and perform multi-level cleaning actions on the battery data based on the first cleaning threshold, multiple association rules and equivalent circuit models, wherein the data of the current level cleaning action is the data cleaned by the previous level cleaning action.

[0083] In this embodiment, the cleaning module 203 is further configured to: perform a first-level cleaning action on the battery data according to a first cleaning threshold; perform a second-level cleaning action on the data after the first-level cleaning action according to association rules; and perform a third-level cleaning action on the data after the second-level cleaning action according to an equivalent circuit model.

[0084] In this embodiment of the application, the cleaning module 203 is further configured to: identify the data value corresponding to at least one data point in the battery data; if the data value corresponding to the data point is less than or equal to a first cleaning threshold, then retain the data point; if the data value corresponding to the data point is greater than the first cleaning threshold, then remove the data point from the battery data and mark the data point as the first abnormal data.

[0085] In this embodiment, the cleaning module 203 is further configured to: extract feature data from the data cleaned by the first-level cleaning action; determine a second cleaning threshold for the feature data according to association rules; and perform a second-level cleaning action on the feature data according to the association rules and the second cleaning threshold.

[0086] In this embodiment, the cleaning module 203 is further configured to: The second cleaning threshold includes a temperature threshold, a current threshold, a rate of change threshold, a range threshold, and a power threshold; the association rules include a first rule, a second rule, and a third rule; determining the second cleaning threshold for the feature data based on the association rules includes: obtaining a second lookup table of the association rules and the second cleaning threshold, the second lookup table being a table of correspondence between the association rules and the second cleaning threshold; if the association rule is the first rule, then querying the second lookup table based on the first rule to obtain the second cleaning threshold as a voltage range threshold; if the association rule is the second rule, then querying the second lookup table based on the second rule to obtain the second cleaning threshold as a current threshold and a rate of change threshold; if the association rule is the third rule, then querying the second lookup table based on the third rule to obtain the second cleaning threshold as a temperature rise threshold and a power threshold.

[0087] In this embodiment, the cleaning module 203 is further configured to: perform a second-level cleaning action on the feature data, which includes temperature data, first current data, first voltage data, and second voltage data, according to an association rule and a second cleaning threshold. This includes: when the association rule is the first rule, calculating the voltage range value based on the first voltage data; if the voltage range value is less than or equal to the range threshold, retaining the first voltage data; if the voltage range value is greater than the range threshold, marking the first voltage data as second abnormal data; when the association rule is the second rule, when the first current data is less than the current threshold, if the rate of change of the first voltage data is greater than the rate of change threshold, marking the current first voltage data as second abnormal data; if the rate of change of the first voltage data is less than or equal to the rate of change threshold, retaining the first voltage data; when the association rule is the third rule, calculating the instantaneous heating power based on the first current data and the second voltage data, and calculating the temperature rise rate based on the temperature data; if the instantaneous heating power is greater than the power threshold and the temperature rise rate is less than or equal to the temperature rise threshold, marking the temperature data as second abnormal data; if the instantaneous heating power is greater than the power threshold and the temperature rise rate is greater than the temperature rise threshold, retaining the temperature data.

[0088] In this embodiment, the cleaning module 203 is further configured to: extract the second current data, the third voltage data, and the circuit state data from the data after the second-level cleaning action; predict the terminal voltage data at the current moment using an equivalent circuit model based on the second current data and the circuit state data at the previous moment, and calculate the residual between the terminal voltage data at the current moment and the real-time terminal voltage data; compare the residual with a residual threshold; if the residual is greater than the residual threshold and the duration of the residual being greater than the residual threshold is greater than a time threshold, then mark the data segment with the residual greater than the residual threshold as the third abnormal data; if the residual is less than or equal to the residual threshold, or if the residual is greater than the residual threshold and the duration of the residual being greater than the residual threshold is less than a time threshold, then retain the residual data segment.

[0089] In this embodiment, a repair module is further included. The repair module is used to: after performing multi-level cleaning actions on the battery data based on a first cleaning threshold, association rules, and an equivalent circuit model, acquire the data after the multi-level cleaning actions, the first abnormal data, the second abnormal data, and the third abnormal data, and identify the abnormal types of the first abnormal data, the second abnormal data, and the third abnormal data; if the abnormal type is an abnormal point type, then at least one of median filtering and nearest-neighbor linear interpolation is used to repair the data after the multi-level cleaning actions; if the abnormal type is at least one of an abnormal data segment type and a continuous missing type, then the equivalent circuit model is called to generate a simulated voltage data segment for data filling, and the filled data segment is marked as simulated data; if the abnormal type is a sensor failure type, then the abnormal data is marked as unusable data; and integrate the data after the multi-level cleaning actions, the repaired data, and at least one of the simulated data into the final cleaned data.

[0090] It should be noted that the foregoing explanation of the battery data cleaning method embodiment also applies to the battery data cleaning device of this embodiment, and will not be repeated here.

[0091] According to the battery data cleaning device proposed in the embodiments of this application, the current operating condition data of the battery can be identified by acquiring the battery data of the vehicle; a first cleaning threshold of the battery data can be obtained by querying a pre-calibrated first lookup table based on the current operating condition data, wherein the first lookup table is a correspondence table between the operating condition data and the cleaning threshold; the association rules and equivalent circuit model of the vehicle's battery pack can be acquired; and multi-level cleaning actions can be performed on the battery data based on the first cleaning threshold, multiple association rules and equivalent circuit models. The data of the current level cleaning action is the data after the previous level cleaning action. By dynamically selecting the cleaning threshold, dynamic adaptive cleaning is achieved, thereby improving the cleaning accuracy. At the same time, through three levels of cleaning—cleaning threshold cleaning, association rule cleaning, and equivalent circuit model cleaning—deep mining of hidden anomalies is achieved.

[0092] Figure 4 A schematic diagram of the structure of a vehicle provided in an embodiment of this application. The vehicle may include: The memory 301, the processor 302, and the computer program stored on the memory 301 and capable of running on the processor 302.

[0093] When the processor 302 executes the program, it implements the battery data cleaning method provided in the above embodiments.

[0094] Furthermore, the vehicle also includes: Communication interface 303 is used for communication between memory 301 and processor 302.

[0095] The memory 301 is used to store computer programs that can run on the processor 302.

[0096] The memory 301 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.

[0097] If the memory 301, processor 302, and communication interface 303 are implemented independently, then the communication interface 303, memory 301, and processor 302 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 4 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0098] Optionally, in a specific implementation, if the memory 301, processor 302, and communication interface 303 are integrated on a single chip, then the memory 301, processor 302, and communication interface 303 can communicate with each other through an internal interface.

[0099] Processor 302 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of this application.

[0100] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0101] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0102] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0103] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any of the following techniques known in the art, or a combination thereof: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (FPGAs), field-programmable gate arrays (FPGAs), etc.

[0104] Those skilled in the art will understand that all or part of the steps of the methods implementing the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0105] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A battery data cleaning method, characterized in that, Includes the following steps: Acquire vehicle battery data and identify the current operating condition data of the battery in the battery data; Based on the current operating condition data, a pre-calibrated first lookup table is queried to obtain the first cleaning threshold of the battery data, wherein the first lookup table is a correspondence table between operating condition data and cleaning threshold. The association rules and equivalent circuit model of the vehicle's battery pack are obtained. Based on the first cleaning threshold, the association rules, and multiple equivalent circuit models, multi-level cleaning actions are performed on the battery data. The data of the current cleaning action is the data cleaned by the previous cleaning action.

2. The battery data cleaning method according to claim 1, characterized in that, The process of performing multi-level cleaning actions on the battery data based on the first cleaning threshold, the association rules, and the equivalent circuit model includes: Perform a first-level cleaning action on the battery data according to the first cleaning threshold; According to the association rules, a second-level cleaning action is performed on the data cleaned by the first-level cleaning action; Based on the equivalent circuit model, a third-level cleaning action is performed on the data cleaned by the second-level cleaning action.

3. The battery data cleaning method according to claim 2, characterized in that, The step of performing a first-level cleaning action on the battery data according to the first cleaning threshold includes: Identify the data value corresponding to at least one data point in the battery data; If the data value corresponding to the data point is less than or equal to the first cleaning threshold, then the data point is retained; If the data value corresponding to the data point is greater than the first cleaning threshold, then the data point is removed from the battery data and marked as the first abnormal data.

4. The battery data cleaning method according to claim 2, characterized in that, The step of performing a second-level cleaning action on the data cleaned by the first-level cleaning action according to the association rule includes: Feature data is extracted from the data after the first-level cleaning action; A second cleaning threshold for the feature data is determined based on the association rule; The feature data is subjected to a second-level cleaning action based on the association rule and the second cleaning threshold.

5. The battery data cleaning method according to claim 4, characterized in that, The second cleaning threshold includes a temperature threshold, a current threshold, a rate of change threshold, a range threshold, and a power threshold. The association rules include a first rule, a second rule, and a third rule. Determining the second cleaning threshold for the feature data based on the association rules includes: Obtain a second query table of the association rule and the second cleaning threshold, wherein the second query table is a correspondence table between the association rule and the second cleaning threshold; If the association rule is the first rule, then the second lookup table is queried based on the first rule to obtain the second cleaning threshold as the voltage range threshold; If the association rule is the second rule, then the second lookup table is queried based on the second rule to obtain the second cleaning threshold as a current threshold and a rate of change threshold; If the association rule is the third rule, then the second query table is queried based on the third rule to obtain the second cleaning threshold as a temperature rise threshold and a power threshold.

6. The battery data cleaning method according to claim 5, characterized in that, The feature data includes temperature data, first current data, first voltage data, and second voltage data. The step of performing a second-level cleaning action on the feature data according to the association rule and the second cleaning threshold includes: When the association rule is the first rule, the voltage range value is calculated based on the first voltage data. If the voltage range value is less than or equal to the range threshold, the first voltage data is retained. If the voltage range value is greater than the range threshold, the first voltage data is marked as second abnormal data. When the association rule is the second rule, if the first current data is less than the current threshold, and the rate of change of the first voltage data is greater than the rate of change threshold, then the current first voltage data is marked as the second abnormal data; if the rate of change of the first voltage data is less than or equal to the rate of change threshold, then the first voltage data is retained. When the association rule is the third rule, the instantaneous heating power is calculated based on the first current data and the second voltage data, and the temperature rise rate is calculated based on the temperature data. If the instantaneous heating power is greater than the power threshold and the temperature rise rate is less than or equal to the temperature rise threshold, the temperature data is marked as second abnormal data. If the instantaneous heating power is greater than the power threshold and the temperature rise rate is greater than the temperature rise threshold, the temperature data is retained.

7. The battery data cleaning method according to claim 2, characterized in that, The step of performing a third-level cleaning action on the data after the second-level cleaning action based on the equivalent circuit model includes: Extract the second current data, third voltage data, and circuit status data from the data after the second-level cleaning action; Based on the second current data and the circuit state data at the previous moment, the terminal voltage data at the current moment is predicted through the equivalent circuit model, and the residual between the terminal voltage data at the current moment and the real-time terminal voltage data is calculated. The residual is compared with a residual threshold. If the residual is greater than the residual threshold and the duration of the residual being greater than the residual threshold is greater than a time threshold, the data segment with the residual greater than the residual threshold is marked as third abnormal data. If the residual is less than or equal to the residual threshold, or if the residual is greater than the residual threshold and the duration of the residual being greater than the residual threshold is less than a time threshold, the data segment with the residual is retained.

8. The battery data cleaning method according to claim 1, characterized in that, After performing multi-level cleaning actions on the battery data based on the first cleaning threshold, the association rules, and the equivalent circuit model, the process further includes: The system acquires data after performing multi-level cleaning actions on the battery data, including first abnormal data, second abnormal data, and third abnormal data, and identifies the abnormal types of the first abnormal data, second abnormal data, and third abnormal data. If the anomaly type is an anomaly point type, then at least one of median filtering and nearest neighbor linear interpolation is used to repair the data after performing multi-level cleaning on the battery data; If the anomaly type is at least one of the abnormal data segment type and the continuous missing type, then the equivalent circuit model is invoked to generate an analog voltage data segment for data filling, and the filled data segment is marked as analog data. If the anomaly type is a sensor failure type, then the abnormal data is marked as unusable data; The data obtained after performing multi-level cleaning on the battery data, the repaired data, and at least one of the simulated data are integrated into the final cleaned data.

9. A battery data cleaning device, characterized in that, include: The acquisition module is used to acquire vehicle battery data and identify the current operating condition data of the battery in the battery data; The query module is used to query a pre-calibrated first query table based on the current operating condition data to obtain a first cleaning threshold for the battery data, wherein the first query table is a correspondence table between operating condition data and cleaning threshold. The cleaning module is used to obtain the association rules and equivalent circuit model of the vehicle's battery pack, and perform multi-level cleaning actions on the battery data based on the first cleaning threshold, the association rules and the equivalent circuit model. The data of the current cleaning action is the data cleaned by the previous cleaning action.

10. A vehicle, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the battery data cleaning method according to any one of claims 1-8.

Images