A real-time monitoring, early warning and alarming method for electric vehicle fire

Abstract

This invention discloses a real-time monitoring, early warning, and alarm method for electric vehicle fires, relating to the field of electric vehicle technology. The method collects static and real-time data from new energy electric vehicles; preprocesses the collected static and real-time data and converts them into a preliminary battery model for the corresponding vehicle model; performs algorithm analysis based on the preliminary battery model and sets a dynamic early warning threshold based on battery characteristic parameters; iteratively updates the preliminary battery model to form a dynamic battery model for the corresponding vehicle model; calculates the input real-time battery data for the corresponding vehicle model based on the dynamic battery model; and determines whether the battery exhibits any abnormalities. This invention employs dual-voltage data fusion analysis of static OCV and charge / discharge terminal voltage, combined with the dynamic battery model to calculate voltage rise and drop changes. This enables accurate identification of hidden risks such as micro-short circuits and early lithium plating within the battery that cannot be detected by vehicle-side chips, achieving early warning of fire hazards.

Description

Technical Field

[0001] This invention belongs to the field of electric vehicle technology, specifically relating to a method for real-time monitoring, early warning, and alarm of electric vehicle fires. Background Technology

[0002] With the rapid development of the new energy electric vehicle industry, the safety of electric vehicles is particularly important, especially since battery-related fires are relatively frequent. Currently, most electric vehicle fires are related to abnormalities in the battery system, including micro short circuits inside the battery, lithium plating, abnormal voltage differences, abnormal temperature differences, and abnormalities in the battery or vehicle end. Among these, micro short circuits inside the battery are difficult to be directly detected by the vehicle-end chip, becoming a key hidden risk point that can cause fires.

[0003] Currently, most methods for detecting minor internal anomalies in batteries rely on the battery management system to monitor the drop in open-circuit voltage (OVC) under static conditions. By judging the degree of battery self-discharge, potential hazards can be identified. However, this detection method requires the battery to be in a stationary state for a long time to complete effective monitoring. In actual use, mainstream electric vehicles such as commercial vehicles and passenger cars need to be charged and discharged frequently, which makes it difficult to meet the conditions for static monitoring. This results in the inability to detect early hazards such as minor internal short circuits in the battery in a timely manner, and the inability to achieve predictive early warning of fires.

[0004] To address these issues, we propose a real-time monitoring, early warning, and alarm method for electric vehicle fires. This method enables early detection of battery anomalies, helps mitigate risks, and improves the real-time monitoring, early warning, and alarm capabilities for electric vehicle fires. Summary of the Invention

[0005] The purpose of this invention is to provide a method for real-time monitoring, early warning, and alarm of electric vehicle fires, which can detect battery abnormalities in advance, avoid risks, and improve the real-time monitoring, early warning, and alarm capabilities of electric vehicle fires, thereby solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A method for real-time monitoring, early warning, and alarm of electric vehicle fires includes the following steps:

[0008] S1. Collect static and real-time data of new energy electric vehicles;

[0009] S2. Preprocess the collected static and real-time data and convert them into a preliminary battery model for the corresponding vehicle model.

[0010] S3. Based on the preliminary battery model, perform algorithm analysis to calculate battery characteristic parameters, and set dynamic warning thresholds based on the battery characteristic parameters. The battery characteristic parameters include battery aging mode, internal resistance change, SOC-OCV curve, voltage rise change, and voltage drop change.

[0011] S4. Real-time calculation of battery characteristic parameters, iterative updates of the preliminary battery model, and formation of a dynamic battery model for the corresponding vehicle model.

[0012] S5. Based on the dynamic battery model, calculate the real-time data of the corresponding vehicle battery to obtain the battery voltage rise and voltage drop data during the charging and discharging process at the same current under different SOC states and different temperature states.

[0013] S6. Compare the voltage rise data and voltage drop data with the preset dynamic warning threshold and the corresponding data of other vehicles under the same conditions to determine whether there is an abnormality in the battery.

[0014] When the battery is not abnormal, perform accuracy verification, and if the verification fails, perform threshold calibration until the verification is successful.

[0015] When a battery malfunctions, a fire warning is triggered in real time and an alarm signal is issued.

[0016] Preferably, in step S1, the static data includes basic vehicle information such as license plate number, vehicle type, announcement model, operating unit, and vehicle manufacturer; the basic vehicle information is obtained by retrieving vehicle file management system interface, entering vehicle factory registration information, and integrating operating unit reporting information.

[0017] Preferably, the real-time data includes vehicle login data, vehicle data, drive motor data, ignition switch opening data, brake data, vehicle position data, extreme value data, alarm data, and operating status data of the rechargeable energy storage device, including voltage, current, temperature, temperature difference, pressure difference, and insulation value. The operating status data is forwarded in real time through the vehicle remote information processing terminal.

[0018] Preferably, in step S2, when preprocessing the collected static data, the integrity of the information fields and the coding standardization are checked first, duplicate vehicle registration information is deleted, the missing basic information is supplemented by the data in the vehicle file management system, and finally the data is stored in a unified field format to match the vehicle model.

[0019] When preprocessing the collected dynamic data, 3 The criteria involve outlier detection, deletion of outliers, and correction of missing values ​​using linear interpolation. The corrected data is then standardized.

[0020] Preferably, the preliminary battery model is a basic equivalent battery model applicable to a specified vehicle model, constructed based on preprocessed static data and initial real-time data, and is used to describe the relationship between battery voltage, current, and SOC.

[0021] Preferably, in step S3, when setting the dynamic warning threshold, the characteristic parameter distribution of normal vehicles of the same model is first statistically analyzed, the confidence interval is determined based on the statistical characteristic parameter distribution, the upper limit of the confidence interval is set as the dynamic warning threshold, and the SOC and temperature conditions of the vehicle model are bound according to the set dynamic warning threshold.

[0022] Preferably, in step S4, when the preliminary battery model is iteratively updated, new real-time data is continuously collected and preprocessed. Then, the battery characteristic parameters are recalculated based on the new data. The deviation between the new parameters and the current preliminary battery model parameters is compared. Based on the deviation, an iterative algorithm is used to correct the parameters of the preliminary battery model. The steps of data collection, calculation, deviation comparison and parameter correction are repeated until the prediction error of the preliminary battery model is less than the dynamic warning threshold, thus forming a dynamic battery model.

[0023] Preferably, in step S5, when acquiring battery voltage rise and voltage drop data during the same current charging and discharging process under different SOC states and different temperature states, the current SOC and temperature data are first input into the dynamic battery model, the same charging and discharging current is set, and then the terminal voltage data at adjacent moments under the charging and discharging current are collected. The voltage rise value and voltage drop value are calculated respectively using the voltage rise change estimation formula and the voltage drop change estimation formula.

[0024] Preferably, in step S6, the method for determining whether the battery is abnormal is as follows: if the actual voltage rise and voltage drop under a certain operating condition meet any of the following conditions, the battery is determined to be abnormal; otherwise, the battery is not abnormal:

[0025] 1) The actual pressure rise or drop exceeds the preset dynamic warning threshold;

[0026] 2) The actual pressure rise or drop value deviates more than 2 from the average value of other normal vehicles under the same conditions. , This represents the standard deviation of normal vehicle characteristic parameters for the same model.

[0027] The preferred accuracy verification process is as follows:

[0028] A1. Collect historical desensitization data and actual thermal runaway result files of the battery;

[0029] A2. Based on historical data, a dynamic battery model is used to conduct early warning predictions and obtain early warning prediction results;

[0030] A3. Calculate the recall rate and false alarm rate based on the prediction results;

[0031] A4. Determine whether the verification passes based on the recall rate and false alarm rate. If the verification fails, the dynamic warning threshold needs to be calibrated and the verification repeated until it passes.

[0032] A5. After successful verification, the current dynamic warning threshold will be locked.

[0033] The present invention proposes a real-time monitoring, early warning, and alarm method for electric vehicle fires, which has the following advantages compared with existing technologies:

[0034] 1. This invention uses dual voltage data fusion analysis of static OCV and charge / discharge terminal voltage, combined with dynamic battery model to calculate voltage rise and drop changes, which can accurately identify hidden risks such as micro short circuits inside the battery and early lithium plating that cannot be detected by vehicle-side chips, and achieve early warning of fire hazards.

[0035] 2. This invention constructs a dynamic battery model and compares voltage rise and voltage drop values ​​in two dimensions with benchmark data of other vehicles under the same operating conditions. This avoids misjudgments caused by a single model or fixed threshold. By matching vehicle models with static data and iteratively updating the dynamic model based on real-time data, it can accurately calculate characteristic parameters such as changes in internal resistance, SOC-OCV curve, voltage rise, voltage drop, and battery aging mode. It does not require redeveloping algorithms for specific vehicle models or battery types and is applicable to various new energy commercial vehicles and passenger cars.

[0036] 3. By continuously iterating and updating the initial battery model, this invention can adapt to the characteristic drift caused by battery aging and changes in operating conditions in real time, ensuring stable monitoring accuracy throughout the vehicle's entire life cycle. The dynamic adaptation capability significantly improves the service life and reliability of the monitoring system. Attached Figure Description

[0037] Figure 1 A flowchart illustrating the monitoring, early warning, and alarm process according to an embodiment of the present invention is shown;

[0038] Figure 2 A flowchart illustrating the accuracy verification process according to an embodiment of the present invention is shown. Detailed Implementation

[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] This invention provides, for example Figure 1-2 The method for real-time monitoring, early warning, and alarm of electric vehicle fires, as shown, includes the following steps:

[0041] S1. Collect static and real-time data of new energy electric vehicles;

[0042] The static data includes basic vehicle information such as license plate number, vehicle type, announcement model, operating unit, and vehicle manufacturer; the basic vehicle information is obtained by retrieving it through the vehicle file management system interface, entering vehicle factory registration information, and integrating information reported by the operating unit.

[0043] The vehicle file management system is a core information management system used by car manufacturers, vehicle operating units, or relevant supervisory departments to store and manage basic information of electric vehicles throughout their entire life cycle. It is an offline static data management platform, mainly used to provide basic data for electric vehicle identification, model matching, compliance verification, and subsequent analysis.

[0044] The real-time data includes vehicle login data, vehicle data, drive motor data, ignition switch opening data, brake data, vehicle position data, extreme value data, alarm data, and operating status data of the rechargeable energy storage device, including voltage, current, temperature, temperature difference, pressure difference, and insulation value. The operating status data is forwarded in real time through the vehicle remote information processing terminal. The vehicle remote information processing terminal collects data such as vehicle, motor, battery voltage, battery current, and battery temperature in real time based on the GB32960.2-2016-T standard. The collected real-time data is uploaded to the vehicle manufacturer's remote service and management platform through a 4G / 5G mobile network. The de-identified real-time data is obtained from the vehicle manufacturer's remote service and management platform through an interface, realizing online real-time data collection.

[0045] Data acquisition strictly follows the national standard GB32960.2-2016-T, and real-time data is obtained through the car manufacturer's existing T-Box terminal and remote service platform. No additional hardware sensors are required, which reduces modification costs and implementation difficulty.

[0046] S2. Preprocess the collected static and real-time data and convert them into a preliminary battery model for the corresponding vehicle model.

[0047] When preprocessing the collected static data, first check the integrity of the information fields and the standardization of the coding, delete duplicate vehicle registration information, supplement the missing basic information through the data in the vehicle file management system, and finally store it according to a unified field format to match the vehicle model;

[0048] When preprocessing the collected dynamic data, 3 The criteria involve outlier detection and removal, followed by linear interpolation to impute missing values. The imputed data is then standardized. (3) The detection and judgment formula of the criterion is:

[0049] In the formula, For a single real-time data point, This is the moving average of the real-time data item. Let be the moving standard deviation, if satisfy If the value is not specified, it is considered an outlier and should be deleted; otherwise, it should be retained.

[0050] When using linear interpolation to imput missing values, the missing data is calculated based on the valid data from adjacent time points, and then the calculated missing data is added in. The formula for calculating missing data is as follows:

[0051] ,

[0052] In the formula, For missing data to be filled, and For valid data at times adjacent to the missing value, t is the timestamp of the missing value;

[0053] The standardization process uses the min-max standardization formula, which is:

[0054] ,

[0055] In the formula, For standardized data, For data to be standardized, and These are the historical minimum and maximum values ​​of the standardized data item, respectively.

[0056] The preliminary battery model is a basic equivalent battery model applicable to a specified vehicle model, constructed based on preprocessed static data and initial real-time data, and is used to describe the relationship between battery voltage, current, and SOC.

[0057] The formula for the basic equivalent model of a battery is as follows:

[0058] ,

[0059] in, This refers to the battery terminal voltage. The open-circuit voltage is a function of the state of charge (SOC), which is obtained by matching the initial static data with the vehicle's basic OCV-SOC curve. For battery charging and discharging current, This is the initial value of the battery's internal resistance in ohms. This initial value is taken from the factory parameters of the battery of the same vehicle model. This is the polarization voltage, which is a fixed empirical value taken from the same vehicle model;

[0060] S3. Based on the preliminary battery model, perform algorithm analysis to calculate battery characteristic parameters, and set dynamic warning thresholds based on the battery characteristic parameters. The battery characteristic parameters include battery aging mode, internal resistance change, SOC-OCV curve, voltage rise change, and voltage drop change.

[0061] The formula for calculating the SOC-OCV curve is:

[0062] ,

[0063] ,

[0064] In the formula, Let k be the function relationship between open-circuit voltage and SOC at time k. Let SOC be the value at time k. The SOC value at time k-1 For charging and discharging efficiency, Let k be the battery current. The sampling time interval, For the battery's rated capacity, The initial value of the battery's internal ohmic resistance. Let k be the terminal voltage. Let k be the polarization voltage at time k;

[0065] The formula for calculating the change in internal resistance is:

[0066] ,

[0067] in, Let be the Ohmic internal resistance at time k. Let k be the battery current. Let k be the terminal voltage. Let k be the polarization voltage at time k. Let k be the function relationship between the open-circuit voltage and the state of charge (SOC) at time k.

[0068] The formula for calculating voltage rise is:

[0069] ,in, Let be the voltages at adjacent times t+1 and t when charging with the same current. This represents the voltage rise between adjacent time points;

[0070] The formula for calculating voltage drop change is:

[0071] ,in, Let be the voltages at adjacent times t and t+1 when charging with the same current. The voltage drop values ​​at adjacent time points;

[0072] The formula for calculating aging patterns is:

[0073] ,in, Let k be the actual capacity at time k. For the battery's rated capacity, The aging rate is obtained by fitting historical capacity data. For vehicle running time;

[0074] When setting the dynamic warning threshold, first statistically analyze the distribution of characteristic parameters of normal vehicles of the same model, determine the confidence interval based on the statistical characteristic parameter distribution, set the upper limit of the confidence interval as the dynamic warning threshold, and bind the SOC and temperature conditions of the vehicle model according to the set dynamic warning threshold.

[0075] The confidence interval is determined using the statistical confidence interval method. The algorithm formula for the statistical confidence interval method is as follows:

[0076] ,in, This is the dynamic early warning threshold under a certain working condition. This represents the average of normal vehicle characteristic parameters for the same model. The standard deviation of normal vehicle characteristic parameters of the same model is used. Taking the mean and twice the standard deviation can ensure that 95% of normal data are within the threshold, reducing false alarms.

[0077] S4. Real-time calculation of battery characteristic parameters, iterative updates of the preliminary battery model, and formation of a dynamic battery model for the corresponding vehicle model.

[0078] When the preliminary battery model is iteratively updated, new real-time data is continuously collected and preprocessed. Battery characteristic parameters are then recalculated based on the new data. The deviations between the new parameters and the current preliminary battery model parameters are compared. Based on these deviations, an iterative algorithm is used to correct the parameters of the preliminary battery model. This process of data collection, calculation, deviation comparison, and parameter correction is repeated until the prediction error of the preliminary battery model is less than the dynamic warning threshold, at which point a dynamic battery model is formed. The iterative algorithm formula is as follows:

[0079] ,

[0080] ,

[0081] ,

[0082] in, Let k be the model parameter vector. This is the model parameter vector at time k-1. Here is the gain matrix. Let k be the measured terminal voltage. Let k be the observation vector at time k. Forgetting factor, Let be the covariance matrix at time k, used to reflect the accuracy of parameter estimation. For battery charging and discharging current, The observation vector at time k transpose, The covariance matrix at time k-1;

[0083] The formula for the dynamic battery model is as follows:

[0084] ,

[0085] in, Let be the battery terminal voltage at time t. The OCV curve is dynamically changed with SOC and runtime t. The ohmic internal resistance is dynamically updated over time t. As SOC and current ,temperature Dynamically changing polarization voltage The charging and discharging current of the battery;

[0086] S5. Based on the dynamic battery model, calculate the real-time data of the corresponding vehicle battery to obtain the battery voltage rise and voltage drop data during the charging and discharging process at the same current under different SOC states and different temperature states.

[0087] When acquiring battery voltage rise and voltage drop data during the same current charging and discharging process under different SOC and temperature conditions, the current SOC and temperature data are first input into the dynamic battery model. The same charging and discharging current is set, such as the average current under the current operating condition. Then, the terminal voltage data at adjacent moments under the charging and discharging current are collected. The voltage rise and voltage drop values ​​are calculated respectively using the voltage rise change estimation formula and the voltage drop change estimation formula.

[0088] S6. Compare the voltage rise data and voltage drop data with the preset dynamic warning threshold and the corresponding data of other vehicles under the same conditions to determine whether there is an abnormality in the battery.

[0089] The method for determining whether a battery is abnormal is as follows: if the actual voltage rise and voltage drop under a certain operating condition meet any of the following conditions, the battery is considered abnormal; otherwise, the battery is not abnormal:

[0090] 1) The actual pressure rise or drop exceeds the preset dynamic warning threshold;

[0091] 2) The actual pressure rise or drop value deviates more than 2 from the average value of other normal vehicles under the same conditions. , This represents the standard deviation of normal vehicle characteristic parameters for the same model.

[0092] When the battery is not abnormal, perform accuracy verification, and if the verification fails, perform threshold calibration until the verification is successful.

[0093] like Figure 2 As shown, the accuracy verification process is as follows:

[0094] A1. Collect historical desensitized battery data and actual thermal runaway result files; the actual thermal runaway result files are a record of the actual thermal runaway or spontaneous combustion events that occurred in the vehicle, and are used as a real label to compare with the algorithm warning results; the historical desensitized battery data are historical files that remove privacy and sensitive information and retain only the battery operation-related data, and are used to anonymize the vehicle identity information while fully retaining the battery operation data.

[0095] A2. Based on historical data, a dynamic battery model is used to conduct early warning predictions and obtain early warning prediction results;

[0096] A3. Calculate the recall rate and false alarm rate based on the prediction results;

[0097] The formula for calculating recall is:

[0098] In the formula, Here, C represents the anomaly recall rate, and A represents the number of actual problematic vehicles detected.

[0099] The formula for calculating the false alarm rate is:

[0100] False= In the formula, False is the false alarm rate, N is the total number of samples, B is the number of abnormal vehicles detected, and C is the number of real problem vehicles detected.

[0101] A4. Determine whether the verification passes based on the recall rate and false alarm rate. If the verification fails, the dynamic warning threshold needs to be calibrated and the verification repeated until it passes.

[0102] The verification is judged as follows: if the recall rate is greater than 0.8 and the false alarm rate is less than 0.05, the verification passes; otherwise, the verification fails.

[0103] A5. After successful verification, the current dynamic warning threshold is locked to facilitate subsequent real-time monitoring;

[0104] When a battery malfunctions, a fire alarm is triggered in real time and an alarm signal is issued.

[0105] Fire warning methods include local pop-up warnings, push warning messages, and push warning prompts. Local pop-up warnings display abnormal vehicle information, abnormality type, and risk level. Push warning messages send warning messages containing license plate number, location, and abnormal parameters to the operating unit or vehicle manufacturer's management platform. Push warning prompts send warning prompts to the vehicle's remote information processing terminal, which displays a battery abnormality warning and reminds the user to pay attention to safety.

[0106] Alarm signals include remote alarm signals, local alarm signals, and emergency linkage signals. Remote alarm signals send alarm notifications to management personnel via SMS and APP push. Local alarm signals are triggered by the vehicle terminal to activate the sound and light alarm. Emergency linkage automatically pushes the vehicle location and hazard information to the nearest rescue organization when the risk level is high.

[0107] This invention employs dual-voltage data fusion analysis of static OCV and charge / discharge terminal voltage, combined with dynamic battery model calculation of voltage rise and drop changes, which can accurately identify hidden risks such as micro short circuits inside the battery and early lithium plating that cannot be detected by vehicle-side chips, thus achieving early warning of fire hazards.

[0108] By constructing a dynamic battery model and combining it with benchmark data from other vehicles under the same operating conditions for a two-dimensional comparison of voltage rise and voltage drop, misjudgments caused by a single model or fixed threshold are avoided. By matching vehicle models with static data and iteratively updating the dynamic model based on real-time data, it can accurately calculate characteristic parameters such as changes in internal resistance, SOC-OCV curve, voltage rise, voltage drop, and battery aging mode. There is no need to redevelop algorithms for specific vehicle models or battery types, making it suitable for various new energy commercial vehicles and passenger cars.

[0109] By continuously iterating and updating the initial battery model, it is possible to adapt to the characteristic drift caused by battery aging and changes in operating conditions in real time, ensuring stable monitoring accuracy throughout the vehicle's entire life cycle. The dynamic adaptation capability significantly improves the service life and reliability of the monitoring system.

[0110] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for real-time monitoring, early warning, and alarm of electric vehicle fires, characterized in that: Includes the following steps: S1. Collect static and real-time data of new energy electric vehicles; S2. Preprocess the collected static and real-time data and convert them into a preliminary battery model for the corresponding vehicle model. S3. Based on the preliminary battery model, perform algorithm analysis to calculate battery characteristic parameters, and set dynamic warning thresholds based on the battery characteristic parameters. The battery characteristic parameters include battery aging mode, internal resistance change, SOC-OCV curve, voltage rise change, and voltage drop change. S4. Real-time calculation of battery characteristic parameters, iterative updates of the preliminary battery model, and formation of a dynamic battery model for the corresponding vehicle model. S5. Based on the dynamic battery model, calculate the real-time data of the corresponding vehicle battery to obtain the battery voltage rise and voltage drop data during the charging and discharging process at the same current under different SOC states and different temperature states. S6. Compare the voltage rise data and voltage drop data with the preset dynamic warning threshold and the corresponding data of other vehicles under the same conditions to determine whether there is an abnormality in the battery. When the battery is not abnormal, perform accuracy verification, and if the verification fails, perform threshold calibration until the verification is successful. When a battery malfunctions, a fire warning is triggered in real time and an alarm signal is issued.

2. The method for real-time monitoring, early warning, and alarm of electric vehicle fires according to claim 1, characterized in that: In step S1, the static data includes basic vehicle information such as license plate number, vehicle type, announcement model, operating unit, and vehicle manufacturer; the basic vehicle information is obtained by retrieving vehicle file management system interface, entering vehicle factory registration information, and integrating operating unit reporting information.

3. The method for real-time monitoring, early warning, and alarm of electric vehicle fires according to claim 2, characterized in that: The real-time data includes vehicle login data, vehicle data, drive motor data, ignition switch opening data, brake data, vehicle position data, extreme value data, alarm data, and operating status data of the rechargeable energy storage device, including voltage, current, temperature, temperature difference, pressure difference, and insulation value. The operating status data is forwarded in real time through the vehicle remote information processing terminal.

4. The method for real-time monitoring, early warning, and alarm of electric vehicle fires according to claim 3, characterized in that: In step S2, when preprocessing the collected static data, the integrity of the information fields and the standardization of the coding are checked first, duplicate vehicle registration information is deleted, the missing basic information is supplemented by the data in the vehicle file management system, and finally the data is stored in a unified field format to match the vehicle model. When preprocessing the collected dynamic data, 3 The criteria involve outlier detection, deletion of outliers, and correction of missing values ​​using linear interpolation. The corrected data is then standardized.

5. The method for real-time monitoring, early warning, and alarm of electric vehicle fires according to claim 4, characterized in that: The preliminary battery model is a basic equivalent battery model applicable to a specified vehicle model, constructed based on preprocessed static data and initial real-time data, and is used to describe the relationship between battery voltage, current, and SOC.

6. The method for real-time monitoring, early warning, and alarm of electric vehicle fires according to claim 5, characterized in that: In step S3, when setting the dynamic warning threshold, the characteristic parameter distribution of normal vehicles of the same model is first statistically analyzed. The confidence interval is determined based on the statistical characteristic parameter distribution. The upper limit of the confidence interval is set as the dynamic warning threshold. The SOC and temperature conditions of the vehicle model are then bound according to the set dynamic warning threshold.

7. The method for real-time monitoring, early warning, and alarm of electric vehicle fires according to claim 6, characterized in that: In step S4, when the preliminary battery model is iteratively updated, new real-time data is continuously collected and preprocessed. Then, the battery characteristic parameters are recalculated based on the new data. The deviation between the new parameters and the current preliminary battery model parameters is compared. Based on the deviation, an iterative algorithm is used to correct the parameters of the preliminary battery model. The steps of data collection, calculation, deviation comparison and parameter correction are repeated until the prediction error of the preliminary battery model is less than the dynamic warning threshold, thus forming a dynamic battery model.

8. The method for real-time monitoring, early warning, and alarm of electric vehicle fires according to claim 7, characterized in that: In step S5, when acquiring battery voltage rise and voltage drop data during the same current charging and discharging process under different SOC states and different temperature states, the current SOC and temperature data are first input into the dynamic battery model, the same charging and discharging current is set, and then the terminal voltage data at adjacent moments under the charging and discharging current are collected. The voltage rise value and voltage drop value are calculated respectively using the voltage rise change estimation formula and the voltage drop change estimation formula.

9. A method for real-time monitoring, early warning, and alarm of electric vehicle fires according to claim 8, characterized in that: In step S6, the method for determining whether the battery is abnormal is as follows: if the actual voltage rise and voltage drop under a certain operating condition meet any of the following conditions, the battery is determined to be abnormal; otherwise, the battery is not abnormal: 1) The actual pressure rise or drop exceeds the preset dynamic warning threshold; 2) The actual pressure rise or drop value deviates more than 2 from the average value of other normal vehicles under the same conditions. , This represents the standard deviation of normal vehicle characteristic parameters for the same model.

10. A method for real-time monitoring, early warning, and alarm of electric vehicle fires according to claim 9, characterized in that: The accuracy verification process is as follows: A1. Collect historical desensitization data and actual thermal runaway result files of the battery; A2. Based on historical data, a dynamic battery model is used to conduct early warning predictions and obtain early warning prediction results; A3. Calculate the recall rate and false alarm rate based on the prediction results; A4. Determine whether the verification passes based on the recall rate and false alarm rate. If the verification fails, the dynamic warning threshold needs to be calibrated and the verification repeated until it passes. A5. After successful verification, the current dynamic warning threshold will be locked.

Images