New energy battery safety early warning monitoring method based on multi-parameter fusion

Abstract

The invention discloses a new energy battery safety early warning monitoring method based on multi-parameter fusion, and particularly relates to the technical field of battery safety early warning, and the method comprises the steps: S1, collecting voltage, temperature, air pressure and current data, and calculating a voltage change rate, a temperature change rate and DC internal resistance; s2, time sequence statistical characteristics are extracted based on a sliding window, an abnormal mode is identified through an isolated forest model, an abnormal score is smoothed, and a first early warning index is output; s3, dynamically calculating a voltage lower limit threshold value and an internal resistance upper limit threshold value according to working conditions, performing temperature compensation correction, comparing real-time parameters with the threshold values, counting abnormal conditions, and outputting a second early warning index; s4, voltage-temperature, voltage-air pressure and temperature-air pressure combined entropy is calculated and smoothed, early-stage accurate early warning of thermal runaway of the battery is achieved, working condition interference is effectively eliminated, and the use safety of the battery is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of battery safety early warning technology, and more specifically, to a new energy battery safety early warning and monitoring method based on multi-parameter fusion. Background Technology

[0002] With the escalating global energy crisis and heightened environmental awareness, the new energy vehicle industry has experienced rapid development. Lithium-ion batteries, due to their high energy density, long cycle life, and low self-discharge rate, have become the primary power source for new energy vehicles. However, lithium-ion batteries are highly susceptible to thermal runaway under conditions of overcharging, internal short circuits, mechanical damage, or high temperatures, leading to battery fires or even explosions, seriously threatening the lives and property of occupants. Therefore, developing accurate and reliable safety early warning and monitoring methods for new energy batteries has significant practical importance and application value.

[0003] Currently, mainstream battery safety warning methods mainly employ single-parameter threshold methods or simple multi-parameter logic judgments. Single-parameter threshold methods typically set fixed alarm thresholds for voltage, temperature, or internal resistance; an alarm is triggered when a parameter exceeds the preset threshold. While simple to implement, this method has significant drawbacks in practical applications: firstly, under normal operating conditions such as rapid acceleration and fast charging, battery voltage, temperature, and internal resistance can fluctuate drastically, easily triggering false alarms; secondly, setting thresholds too high to avoid false alarms can lead to missed detections of early-stage hazards such as slowly developing internal micro-short circuits, often triggering alarms only when the battery has already undergone a violent exothermic reaction or is smoking, leaving occupants and maintenance personnel with extremely limited time for escape and response, thus failing to achieve true early warning.

[0004] To address the aforementioned issues, some studies have attempted to employ multi-parameter fusion methods for battery safety status assessment, such as combining voltage and temperature change rates for comprehensive judgment. However, existing methods largely remain at the level of simple logical combinations of parameters, failing to delve into the inherent coupling relationships and evolution patterns between parameters, and lacking dynamic adaptability to changes in battery operating conditions. Throughout the battery's lifespan, as aging progresses and ambient temperature changes, the baseline values ​​and fluctuation ranges of various parameters will drift. Prediction methods based on fixed thresholds and static models struggle to adapt to these dynamic changes, leading to a decline in the accuracy and reliability of warnings over time. Therefore, how to accurately predict the critical point of battery thermal runaway while effectively eliminating interference from normal operating condition fluctuations and battery aging has become a pressing technical challenge in this field. Summary of the Invention

[0005] To overcome the above-mentioned deficiencies of the prior art, embodiments of the present invention provide a new energy battery safety early warning and monitoring method based on multi-parameter fusion.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A new energy battery safety early warning and monitoring method based on multi-parameter fusion includes the following steps: S1. Data Acquisition: Real-time acquisition of voltage, temperature, air pressure and current data of individual battery cells, and calculation of voltage change rate, temperature change rate and DC internal resistance after preprocessing; S2. Trend Detection: Based on the sliding window, the time-series statistical features of voltage change rate, temperature change rate and internal resistance are extracted. Abnormal patterns are identified through the isolated forest model, and the first early warning index is output. S3. Threshold Evaluation: Dynamically calculate the lower voltage threshold and the upper internal resistance threshold based on the current operating conditions, compare the real-time parameters with the dynamic thresholds, statistically analyze abnormal situations, and output the second early warning index. S4. Coupling Analysis: Calculate the joint entropy between each pair of the three parameters: voltage, temperature, and air pressure. Monitor the changing trend of the joint entropy. When the joint entropy continues to decrease and the rate of decrease exceeds the threshold, it is determined that the parameter coupling relationship has abruptly changed, and the third early warning index is output. S5. Fusion Early Warning: Input the first, second, and third early warning indices into the fuzzy logic inference engine for comprehensive decision-making and output a safety level early warning signal.

[0007] Specifically, in S2: Training the isolated forest model involves collecting full lifecycle data of batteries within the range of -20℃ to 60℃, 0.1C to 3C, and 20% to 100% SOC. Time-series statistical features are extracted and labeled as normal categories. After normalization, the number of trees is set to 100, the sampling ratio to 256, and the pollution coefficient to 0.01 for training. The abnormal scores output by the model are smoothed by an exponentially weighted moving average with a smoothing coefficient ranging from 0.1 to 0.3. The smoothed abnormal scores are then normalized and used as the first warning index.

[0008] Specifically, in S3: A multidimensional lookup table is constructed by pre-measuring the reference voltage and reference internal resistance of the battery at different temperatures, different SOCs, and different charge / discharge rates; The dynamic voltage lower limit threshold is calculated based on the reference voltage, current current, and SOH, while the dynamic internal resistance upper limit threshold is calculated based on the reference internal resistance, SOH, and current current. The calibration coefficients in the calculation formula were determined by selecting batteries of the same model, conducting hybrid pulse power characteristic tests, and fitting the results using the least squares method.

[0009] Specifically, in S3: Based on the deviation between the current measured temperature and the reference temperature, temperature compensation corrections are made to the lower limit threshold of dynamic voltage and the upper limit threshold of dynamic internal resistance. The voltage compensation coefficient is taken as 0.01 to 0.03 V / ℃, and the internal resistance compensation coefficient is taken as 0.05 to 0.15 mΩ / ℃. The real-time voltage and real-time internal resistance are compared with the corrected threshold, the frequency of anomalies and the magnitude of deviations are statistically analyzed, and the second early warning index is obtained after normalization.

[0010] Specifically, in S4: Within a sliding window, the voltage, temperature, and air pressure data are discretized, and the voltage-temperature joint entropy, voltage-air pressure joint entropy, and temperature-air pressure joint entropy are calculated. The joint entropy sequence is then smoothed using an exponentially weighted moving average with a smoothing coefficient ranging from 0.1 to 0.3.

[0011] Specifically, in S4: When any joint entropy continuously decreases within 3 to 5 consecutive windows and the rate of decrease exceeds the mutation rate threshold, the coupling relationship is determined to be decoupled. The mutation rate threshold is taken as the 95th quantile of the distribution of the rate of change of joint entropy under normal operating conditions, ranging from 15% to 25%; the third early warning index is obtained by quantifying the magnitude and rate of decrease of joint entropy.

[0012] Specifically, in S5: The fuzzy logic inference engine uses a membership function to divide the input variables α, β, and χ into three fuzzy sets: low, medium, and high, and to divide the output variable safety level into three fuzzy sets: attention, warning, and danger. Reasoning is performed based on fuzzy rules developed by experts, and the security level value is obtained by defuzzifying using the center of gravity method.

[0013] Specifically, in S5: The corresponding warning signal is triggered based on the range of the safety level value: The system will prompt you to record data and suggest scheduling an inspection. Warning level: Battery power is limited; Recommendation: Get the battery repaired at the nearest service center. In case of danger, immediately cut off the high voltage and activate the audible and visual alarm.

[0014] Specifically, in S5: The maximum value of the output membership function is calculated as the confidence level. When the confidence level is lower than the preset threshold, an alert signal is output along with a message suggesting manual review due to the low confidence level.

[0015] The technical effects and advantages of this invention are as follows: By combining multi-parameter fusion with multi-level analysis, the shortcomings of existing technologies, such as prediction lag and high false alarm rates, are effectively overcome. First, in the trend detection step, an isolated forest model is used to identify anomalies in time-series statistical features, capturing weak anomaly signals such as slowly developing early micro-short circuits. Second, in the threshold evaluation step, a dynamic adaptive threshold mechanism is introduced, combining multiple operating parameters such as current, SOC, SOH, and temperature to calculate and correct thresholds in real time, effectively eliminating interference from normal operating condition fluctuations and battery aging. Finally, in the coupling analysis step, the concept of joint entropy is introduced, accurately capturing the strongest evidence of impending thermal runaway by monitoring abrupt changes in the coupling relationship of voltage, temperature, and air pressure. The three-dimensional warning index is comprehensively decided through a fuzzy logic inference engine, achieving accurate full-cycle warning from early anomalies to critical thermal runaway.

[0016] Optimization measures were introduced in several key stages to further improve the stability and reliability of the early warning system. In trend detection, exponential smoothing filtering was applied to anomaly scores to suppress false triggers caused by single-shot noise; temperature compensation correction was added to threshold evaluation to improve adaptability over a wide temperature range; in coupling analysis, the joint entropy sequence was smoothed to eliminate the influence of sensor noise and discretization errors; in fusion early warning, decision confidence was calculated and a verification prompt was added when the confidence was low, enhancing the interpretability of the early warning results. Through these technical means, this invention can identify thermal runaway risks several minutes to tens of minutes in advance, providing valuable early warning time for occupant escape and maintenance, and significantly improving the safety of new energy battery use. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0018] 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. 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.

[0019] like Figure 1 As shown, the steps of the new energy battery safety early warning and monitoring method based on multi-parameter fusion are as follows: Step S1: Data acquisition. Real-time acquisition of voltage, temperature, air pressure, and current data of individual battery cells. After filtering and noise reduction, the voltage change rate, temperature change rate, and DC internal resistance are calculated to construct a multi-dimensional monitoring dataset for subsequent analysis. The process is as follows: First, a sensor network deployed within the battery module collects real-time data on the voltage (V), surface temperature (T), internal pressure (P), and total current (I) of each individual battery cell at a frequency of at least 10Hz. The raw data is then filtered to remove high-frequency noise using a low-pass filter. Subsequently, derived parameters are calculated. Voltage change rate dV / dt: calculated based on the voltage difference and time difference between the current moment and the previous moment; Temperature change rate dT / dt: calculated in the same way as above; DC internal resistance R: calculated by dividing the voltage difference by the current difference (ΔV / ΔI) at the instant of current change (such as charging and discharging switching); The original parameters (V,T,P) and the derived parameters (dV / dt,dT / dt,R) are combined to form a multidimensional monitoring dataset for use in subsequent steps.

[0020] Step S2: Trend detection. Temporal statistical features of voltage change rate, temperature change rate, and internal resistance are extracted using a sliding window and input into an isolated forest model for anomaly pattern recognition. The anomaly scores output by the model are then subjected to exponential smoothing filtering and normalized to obtain the first warning index α, which characterizes whether the battery exhibits an early, slow-developing abnormal trend. The process is as follows: Constructing feature vectors: Set a sliding window of length L (e.g., 60 seconds), sliding once every 10 seconds. Within each window, extract time series data of voltage change rate (dV / dt), temperature change rate (dT / dt), and internal resistance (R).

[0021] Feature extraction: For the time series of each parameter, extract its statistical features, such as mean, variance, peak value, trough value, kurtosis, and the first-order linear slope of the fitted curve. Concatenate these features into a high-dimensional feature vector.

[0022] Isolation Forest Detection: The feature vector is input into a pre-trained isolation forest model. This model is trained using historical data from the entire lifecycle of a normal battery (covering various operating conditions) to learn the boundaries of the "normal" feature space. The model outputs an anomaly score, ranging from 0 to 1. A higher score indicates a more "isolated" combination of time-series features, and is more likely to be an anomaly. The training process for the isolation forest model includes: collecting full lifecycle data from at least 100 batteries within the range of -20℃ to 60℃, 0.1C to 3C, and 20% to 100% SOC; extracting feature vectors and labeling them as "normal"; performing z-score normalization; and setting the number of trees to 100, the sampling ratio to 256, and the contamination coefficient to 0.01 for training. To further suppress anomalous score jumps caused by single noise events, an exponentially weighted moving average smoothing process is applied to the anomalous scores output by the isolated forest. Let the anomalous scores of the isolated forest in the current sliding window be... The anomaly score after smoothing the previous window is The abnormal score after smoothing in the current window. Calculated using the following formula: ; in Assign an anomaly score to the isolated forest in the current sliding window. This is a smoothing coefficient, ranging from 0.1 to 0.3. The specific value can be determined through debugging based on actual data; a typical value is 0.2. (Initial time...) Smoothed outlier scores Used for subsequent normalization.

[0023] Output the first warning index: The abnormal score is normalized and used as the first warning index α (0~1). This index mainly characterizes whether there is a slowly developing abnormal trend inside the battery that does not conform to the normal aging process, such as increased voltage fluctuations caused by micro-short circuits.

[0024] Step S3: Threshold evaluation. Based on the current battery operating conditions (current, SOC, temperature, SOH), dynamically calculate adaptive thresholds for the lower voltage limit and upper internal resistance limit, and perform temperature compensation correction on the thresholds. Compare the real-time parameters with the dynamic thresholds, statistically analyze the frequency of anomalies and the magnitude of deviations, and obtain the second warning index β after normalization, reflecting the degree of deviation of the battery state from normal operating conditions. The process is as follows: Establish a benchmark model: In advance, through experiments, determine the benchmark voltage curve and benchmark internal resistance value of the battery under different temperatures, different states of charge and different charge and discharge rates, and construct a multidimensional lookup table.

[0025] Real-time threshold calculation: Dynamic voltage lower limit threshold: The system reads the current in real time. And the state of charge (SOC), find the corresponding reference voltage from the lookup table. Then, considering the current battery health state (SOH) (the lower the SOH, the higher the internal resistance, and the voltage fluctuation tolerance should be adjusted), the dynamic lower limit threshold is calculated using the formula: .in, This is the dynamic voltage lower limit threshold. As the reference voltage, For the current, For battery health status, and This is the calibration coefficient. During high-current discharge, the lower voltage limit is dynamically lowered to avoid false alarms; as the battery ages, the lower voltage limit will also be adjusted accordingly.

[0026] Dynamic internal resistance upper limit threshold: Similarly, find the reference internal resistance based on SOC and temperature. Calculate the upper limit threshold of internal resistance: .in This is the upper limit threshold of the dynamic internal resistance. As the reference internal resistance, , This is the calibration coefficient.

[0027] Calibration coefficient , , , The experimental data were fitted using the least squares method to determine the following: Ten batteries of the same model were selected for hybrid pulse power characteristic testing. The actual voltage, internal resistance and reference value were recorded under different operating conditions, and the coefficients were obtained by fitting the data. To improve the adaptability of the threshold over a wide temperature range, temperature compensation correction is applied to the dynamic threshold; let the current measured temperature be... The reference temperature is (Typically 25℃) Define temperature deviation The corrected lower threshold of dynamic voltage and upper threshold of internal resistance are calculated as follows: ; ; in This is the lower limit threshold of the dynamic voltage after temperature compensation. This is the uncompensated dynamic voltage lower limit threshold. For temperature deviation, This is the upper limit threshold of the dynamic internal resistance after temperature compensation. The upper limit threshold of the uncompensated dynamic internal resistance. , The temperature compensation coefficient is determined through fitting experimental data, and its typical value range is as follows: , The revised threshold is used in subsequent comparison and scoring steps.

[0028] Comparison and scoring: Real-time voltage data will be compared and scored. and In comparison, if < If a voltage anomaly occurs, record it; record the real-time internal resistance. and In comparison, if > If so, an internal resistance anomaly is recorded.

[0029] Output the second early warning index: Statistically analyze the frequency and deviation of voltage and internal resistance anomalies within the current sliding window, and normalize them to obtain the second early warning index β(0-1). This index mainly characterizes whether the current state of the battery significantly deviates from its expected dynamic operating range.

[0030] Step S4: Coupling analysis, calculating the joint entropy of voltage-temperature, voltage-pressure, and temperature-pressure, and smoothing the joint entropy sequence; monitoring the continuous decreasing trend of the joint entropy, when the decreasing rate exceeds a threshold, determining that the parameter coupling relationship has abruptly changed, and outputting the third warning index χ as the strongest evidence of impending thermal runaway; the process is as follows: Parameter selection: Three parameters with clear physical meaning and strong correlation were selected: voltage (V), temperature (T), and air pressure (P). During normal battery operation and thermal abuse, these three are highly coupled (e.g., internal short circuits usually lead to voltage drop, temperature rise, and air pressure increase).

[0031] Joint entropy calculation: Within the same sliding window, the voltage and temperature data are discretized to construct a two-dimensional histogram, and the voltage-temperature joint entropy H(V,T) is calculated. Similarly, the voltage-pressure joint entropy H(V,P) and the temperature-pressure joint entropy H(T,P) are calculated. The larger the joint entropy, the more complex and uncertain the relationship between the parameters; conversely, the smaller the joint entropy, the more completely one parameter can be determined by the other, i.e., the higher the coupling degree. To eliminate instantaneous fluctuations caused by sensor noise and discretization errors, the calculated joint entropy sequence is smoothed according to the following rules: Let the joint entropy of the current sliding window be... The joint entropy after smoothing in the previous window are respectively The joint entropy after smoothing the current window is calculated using the following formula: ; ; ; in This is the voltage-temperature joint entropy after smoothing within the current window. The voltage-temperature joint entropy calculated for the current window. This is the voltage-temperature joint entropy after smoothing from the previous window. The smoothing coefficient ranges from 0.1 to 0.3, with a typical value of 0.2; voltage-pressure joint entropy. and temperature-pressure joint entropy The formulas are the same, only the parameters are different; Initial time The smoothed joint entropy sequence is used for subsequent trend monitoring and mutation detection.

[0032] Trend Monitoring: Monitors the changing trends of the three joint entropies H(V,T), H(V,P), and H(T,P) over time. During normal battery operation, the joint entropy fluctuates within a certain range due to noise and changes in operating conditions. When irreversible and violent chemical reactions begin to occur inside the battery (precursors to thermal runaway), such as a separator rupture leading to an internal short circuit, the voltage rapidly collapses while the temperature begins to soar. At this point, the previously fuzzy statistical relationship between voltage and temperature is disrupted, evolving into a deterministic and strongly correlated collapse mode.

[0033] Mutation determination: If any joint entropy is detected to continuously decrease within m consecutive windows, and the rate of decrease exceeds the preset mutation rate threshold γ, then the coupling relationship is determined to be decoupled; the number of consecutive windows m is 3-5 (for example, m=4 when sampling 10Hz, window 60 seconds, step size 10 seconds); the mutation rate threshold γ is the 95th quantile of the distribution of the joint entropy change rate under normal operating conditions, with a typical range of 15% to 25% (if the rate of decrease exceeds 20%, it is judged as abnormal). The third warning index is output: based on the magnitude and rate of the decrease in joint entropy, the degree to which it deviates from the normal fluctuation range is quantified, and the third warning index χ(0-1) is output. This index is the strongest evidence to judge that thermal runaway is about to occur.

[0034] Step S5: Fusion Early Warning. The three early warning indices α, β, and χ are input into the fuzzy logic inference engine. A comprehensive decision is made based on the expert rule base to obtain the safety level value. Three-level early warning signals (Attention, Warning, Danger) are output, and the decision confidence level is calculated. When the confidence level is low, a verification prompt is added to improve the credibility and interpretability of the early warning results. The process is as follows: Construct a fuzzy logic inference engine: Define fuzzy sets (e.g., low, medium, high) for input variables α, β, and χ, and define fuzzy sets (e.g., attention, warning, danger) for output variables representing safety levels. Develop fuzzy rules based on expert experience, for example: If α is high, β is medium, and χ is low, a warning is output (this may indicate severe battery aging, but not an emergency thermal runaway); if α is medium, β is medium, and χ is high, a danger is output (this is a typical abrupt change in coupling relationship, and an alarm should be triggered immediately); if α is low, β is low, and χ is low, a caution is output (this may indicate a minor disturbance, requiring continuous observation).

[0035] Defuzzification and Early Warning: Input the current α, β, and χ values ​​into the fuzzy inference engine, and defuzzify using the centroid method to obtain a precise safety level value. Based on the interval in which this value falls, trigger the corresponding early warning signal. Attention (Level 1 Warning): The system records data and suggests arranging an inspection; Warning (Level 2 Warning): Battery power is limited and nearby repairs are recommended; Danger (Level 3 Warning): High voltage is immediately cut off, audible and visual alarms are activated, and passengers / maintenance personnel are notified to evacuate immediately.

[0036] To improve the interpretability of the early warning results, the confidence level of the result is calculated along with the output safety level value. Let the output membership function obtained by fuzzy inference be μ( ), If ∈[0,1] is the domain of security level, then the confidence level is... Defined as the maximum value of the membership function: ; in To output the universe of discourse variable for the security level, To output the membership function, For confidence level, The value ranges from 0 to 1; a larger value indicates a more reliable reasoning result, meaning a higher confidence level in the current warning judgment. When the value is below the preset threshold θ (e.g., θ=0.5), the system will output a warning signal along with a message that "confidence level is low, manual verification is recommended" for maintenance personnel to refer to.

[0037] The above formulas are all dimensionless calculations. Dimensionless calculations can be performed using various methods such as standardization, which will not be elaborated here. The formulas are derived from software simulations based on a large amount of collected data, and the preset parameters in the formulas can be set by those skilled in the art according to the actual situation.

[0038] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, ATA hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state ATA hard disk.

[0039] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0040] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0041] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0042] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0043] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0044] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable ATA hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0045] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A new energy battery safety early warning and monitoring method based on multi-parameter fusion, characterized in that, Includes the following steps: S1. Data Acquisition: Real-time acquisition of voltage, temperature, air pressure and current data of individual battery cells, and calculation of voltage change rate, temperature change rate and DC internal resistance after preprocessing; S2. Trend Detection: Based on the sliding window, the time-series statistical features of voltage change rate, temperature change rate and internal resistance are extracted. Abnormal patterns are identified through the isolated forest model, and the first early warning index is output. S3. Threshold Evaluation: Dynamically calculate the lower voltage threshold and the upper internal resistance threshold based on the current operating conditions, compare the real-time parameters with the dynamic thresholds, statistically analyze abnormal situations, and output the second early warning index. S4. Coupling Analysis: Calculate the joint entropy between each pair of the three parameters: voltage, temperature, and air pressure. Monitor the changing trend of the joint entropy. When the joint entropy continues to decrease and the rate of decrease exceeds the threshold, it is determined that the parameter coupling relationship has abruptly changed, and the third early warning index is output. S5. Fusion Early Warning: Input the first, second, and third early warning indices into the fuzzy logic inference engine for comprehensive decision-making and output a safety level early warning signal.

2. The new energy battery safety early warning and monitoring method based on multi-parameter fusion according to claim 1, characterized in that, In S2: Training the isolated forest model involves collecting full lifecycle data of batteries within the range of -20℃ to 60℃, 0.1C to 3C, and 20% to 100% SOC. Time-series statistical features are extracted and labeled as normal categories. After normalization, the number of trees is set to 100, the sampling ratio to 256, and the contamination coefficient to 0.01 for training. The abnormal scores output by the model are smoothed by an exponentially weighted moving average with a smoothing coefficient ranging from 0.1 to 0.

3. The smoothed abnormal scores are then normalized and used as the first warning index.

3. The new energy battery safety early warning and monitoring method based on multi-parameter fusion according to claim 1, characterized in that, In S3: A multidimensional lookup table is constructed by pre-measuring the reference voltage and reference internal resistance of the battery at different temperatures, different SOCs, and different charge / discharge rates; The dynamic voltage lower limit threshold is calculated based on the reference voltage, current current, and SOH, while the dynamic internal resistance upper limit threshold is calculated based on the reference internal resistance, SOH, and current current. The calibration coefficients in the calculation formula were determined by selecting batteries of the same model, conducting hybrid pulse power characteristic tests, and fitting the results using the least squares method.

4. The new energy battery safety early warning and monitoring method based on multi-parameter fusion according to claim 3, characterized in that, In S3: Based on the deviation between the current measured temperature and the reference temperature, temperature compensation corrections are made to the lower limit threshold of dynamic voltage and the upper limit threshold of dynamic internal resistance. The voltage compensation coefficient is taken as 0.01 to 0.03 V / ℃, and the internal resistance compensation coefficient is taken as 0.05 to 0.15 mΩ / ℃. The real-time voltage and real-time internal resistance are compared with the corrected threshold, the frequency of anomalies and the magnitude of deviations are statistically analyzed, and the second early warning index is obtained after normalization.

5. The new energy battery safety early warning and monitoring method based on multi-parameter fusion according to claim 1, characterized in that, In S4: Within a sliding window, the voltage, temperature, and air pressure data are discretized, and the voltage-temperature joint entropy, voltage-air pressure joint entropy, and temperature-air pressure joint entropy are calculated. The joint entropy sequence is then smoothed using an exponentially weighted moving average with a smoothing coefficient ranging from 0.1 to 0.

3.

6. The new energy battery safety early warning and monitoring method based on multi-parameter fusion according to claim 5, characterized in that, In S4: When any joint entropy continuously decreases within 3 to 5 consecutive windows and the rate of decrease exceeds the mutation rate threshold, the coupling relationship is determined to be decoupled. The mutation rate threshold is taken as the 95th quantile of the distribution of the rate of change of joint entropy under normal operating conditions, ranging from 15% to 25%; the third early warning index is obtained by quantifying the magnitude and rate of decrease of joint entropy.

7. The new energy battery safety early warning and monitoring method based on multi-parameter fusion according to claim 1, characterized in that, In S5: The fuzzy logic inference engine uses a membership function to divide the input variables α, β, and χ into three fuzzy sets: low, medium, and high, and to divide the output variable safety level into three fuzzy sets: attention, warning, and danger. Reasoning is performed based on fuzzy rules developed by experts, and the security level value is obtained by defuzzifying using the center of gravity method.

8. The new energy battery safety early warning and monitoring method based on multi-parameter fusion according to claim 7, characterized in that, In S5: The corresponding warning signal is triggered based on the range of the safety level value: The system will prompt you to record data and suggest scheduling an inspection. Warning level: Battery power is limited; Recommendation: Get the battery repaired at the nearest service center. In case of danger, immediately cut off the high voltage and activate the audible and visual alarm.

9. The new energy battery safety early warning and monitoring method based on multi-parameter fusion according to claim 7, characterized in that, In S5: The maximum value of the output membership function is calculated as the confidence level. When the confidence level is lower than the preset threshold, an alert signal is output along with a message suggesting manual review due to the low confidence level.

Images