Method for diagnosing micro-short circuit and early thermal runaway of power battery based on simulink simulation

Abstract

The application relates to a power battery micro-short circuit and early thermal runaway diagnosis distinguishing method based on Simulink simulation, and belongs to the field of power battery fault detection. The method comprises the following steps: simulating battery behavior based on Simulink simulation to obtain a simulation voltage curve, and extracting a fault diagnosis feature in a discharge voltage signal feature from current discharge data of a plurality of groups of cycle discharge data of the simulation voltage curve; performing DBSCAN clustering analysis on the extracted fault diagnosis feature to determine a fault battery; and distinguishing the micro-short circuit and thermal runaway of the fault battery by extracting a fault classification feature from historical charging data in the simulation voltage curve of the fault battery. The application can capture slight changes in the battery discharge process by establishing a fault feature combination, and can accurately distinguish the micro-short circuit and thermal runaway in the early stage of the fault, so that 100% diagnosis accuracy of the battery micro-short circuit and thermal runaway fault is realized, and a 12-hour safety warning of the thermal runaway fault is realized.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of power battery fault detection, and relates to a method for diagnosing and distinguishing micro short circuit and early thermal runaway of a power battery based on Simulink simulation. BACKGROUND

[0002] In a battery management system (BMS), early warning of battery thermal runaway has always been a key issue to ensure the safety of electric vehicles and energy storage systems. During use, especially under high temperature, overcharge, external impact and other conditions, power batteries may produce serious faults such as thermal runaway, causing the internal temperature to rise sharply and even cause combustion or explosion, threatening personal safety and equipment integrity. Therefore, early and accurate identification of battery faults, as well as the distinction between thermal runaway faults and micro short circuit faults, and taking effective measures at an early stage of the fault, are crucial to improving the safety and reliability of the battery system.

[0003] Existing early warning methods of thermal runaway usually rely on a single or small number of features, such as the surface temperature of the battery, the rate of voltage change, gas release, etc., but these methods often fail to fully capture the complex fault mechanisms inside the battery, leading to misjudgment and missed judgment, especially in complex environments and large load fluctuations. In addition, these methods are difficult to distinguish between micro short circuit faults and thermal runaway faults, resulting in a lack of effective targeted measures for the system. SUMMARY

[0004] Therefore, the purpose of the present application is to provide a method for diagnosing and distinguishing micro short circuit and early thermal runaway of a power battery based on Simulink simulation.

[0005] To achieve the above purpose, the present application provides the following technical solutions:

[0006] A method for diagnosing and distinguishing micro short circuit and early thermal runaway of a power battery based on Simulink simulation, the method comprising the following steps:

[0007] S1, simulating the behavior of the battery based on Simulink simulation to obtain a simulation voltage curve, the simulation voltage curve including a plurality of sets of cycle discharge data and a plurality of sets of cycle charge data;

[0008] S2, extracting a fault diagnosis feature in a discharge voltage signal feature from the current discharge data of the plurality of sets of cycle discharge data of the simulation voltage curve;

[0009] S3, performing DBSCAN clustering analysis based on the extracted fault diagnosis feature to determine a fault battery;

[0010] S4, extracting a fault classification feature from the historical charge data in the simulation voltage curve of the fault battery;

[0011] S5, distinguishing the micro-short circuit and thermal runaway of the fault battery according to the fault classification feature.

[0012] Further, in step S1, the battery fault behaviors under different working conditions are simulated by using a Simulink simulation experiment to obtain a plurality of groups of simulation voltage curves of the batteries, including the following steps:

[0013] S11, establishing a second-order equivalent circuit model of the single lithium ion battery;

[0014] S12, establishing a thermal model of the single lithium ion battery;

[0015] S13, setting the resistance value of the short-circuit resistance of the fault battery to control the battery fault type;

[0016] S14, connecting a plurality of batteries in series to form a battery pack, and adjusting the resistance value of the short-circuit resistance R S to control the fault condition of the battery pack to simulate, and then obtain a plurality of groups of thermal runaway battery data, micro-short circuit battery data and normal battery data.

[0017] Further, in step S11, the second-order equivalent circuit model includes an open circuit voltage OCV, an ohmic internal resistance R0, polarization resistances R1 and R2, polarization capacitances C1 and C2, and a short-circuit resistance R S , wherein,

[0018] The battery cell outputs a current source The current source passes through, in sequence, a parallel combination of the polarization resistance R2 and the polarization capacitance C2, a parallel combination of the polarization resistance R1 and the polarization capacitance C1, the ohmic internal resistance R0 and the open circuit voltage OCV, and the short-circuit resistance R S is connected in parallel to both ends of the series combination of the parallel combination of the polarization resistance R2 and the polarization capacitance C2, the parallel combination of the polarization resistance R1 and the polarization capacitance C1, the ohmic internal resistance R0 and the open circuit voltage OCV;

[0019] The pulse segment voltage curve of the second-order equivalent circuit model of the single lithium ion battery is obtained by a C / 2 rate GITT discharge test in a long calibration program test, and the parameter identification is performed by the pulse segment voltage curve;

[0020] In the pulse voltage curve, the voltage first decreases due to the discharge of the battery, then rises due to the polarization effect, and finally tends to be flat. In the curve, a plurality of pulse points A, B, C, D and E are selected, wherein the point A is the starting point, the point C is the lowest point, the point E is the terminal point, the point B is the point between the points A and C, and the point D is the point between the points C and E;

[0021] The open circuit voltage (OCV) reflects the terminal voltage of the battery when it reaches a balanced state after being left for a period of time, and is expressed as:

[0022] OCV = V E

[0023] wherein V E is the voltage at point E of the pulse segment;

[0024] The ohmic internal resistance R0 mainly reflects the transient response, and is identified by extracting the voltages of the A-B segment and the C-D segment of the pulse segment, wherein:

[0025]

[0026] wherein V A , V B , V C , and V D are the voltages at points A, B, C, and D of the pulse segment, respectively, and I is the test current of the GITT;

[0027] The polarization resistances R1 and R2 and the polarization capacitances C1 and C2 are polarization parameters, and affect the voltage response at point C-E after the end of the first pulse discharge, which is regarded as a zero input voltage response:

[0028]

[0029] wherein V(t) represents the battery voltage, and V oc represents the open circuit voltage;

[0030] The second-order exponential fitting is performed on the voltage segment, and the fitting formula is:

[0031]

[0032] The fitting function is used to fit the voltage curve, and the coefficients obtained after fitting are used to represent the open circuit voltage OCV, the polarization resistances R1 and R2, and the polarization capacitances C1 and C2 of the battery.

[0033] The open circuit voltage OCV, the polarization resistances R1 and R2, and the polarization capacitances C1 and C2 all consider the influence of aging factors, and the ohmic internal resistance R0 additionally considers the influence of temperature.

[0034] Further, in step S12, the thermal model considers the internal resistance heat and the environmental heat dissipation, and the expression is:

[0035]

[0036] wherein m is the battery mass, C p is the battery heat capacity, represents the derivative of the battery temperature with respect to time, and U tV represents the battery voltage, T represents the battery temperature, T ∞ represents the temperature at which the battery stabilizes after standing, h is the heat exchange coefficient, and A is the battery surface area.

[0037] Further, in step S13, the fault battery short-circuit resistance value setting includes maximum value calculation, slight short-circuit resistance value analysis, moderate short-circuit resistance value analysis, severe short-circuit resistance value analysis, and thermal runaway short-circuit resistance value analysis, wherein, according to the analysis results, the short-circuit resistance value is determined according to the following formula:

[0038] R S = R S0 e -f(t)

[0039] wherein R S0 is the initial resistance value of the fault resistance when the thermal runaway just occurs, e -f(t) represents a function that changes with time.

[0040] Further, in step S2, the fault diagnosis features in the discharge voltage signal features extracted from the cycle discharge data include the time integral value in the voltage interval, the autocorrelation coefficient, and the discrete wavelet transform approximation coefficient energy, wherein,

[0041] The time integral feature in the voltage interval refers to the integral value of the voltage curve in the 3V to 4V voltage interval, which is used to measure the cumulative voltage of the voltage interval, and the extraction process is as follows:

[0042] Extract the voltage data points in the 3V-4V interval from the discharge voltage curve V(t), obtain the reconstructed voltage curve V1(t), and then integrate the time:

[0043]

[0044] wherein t represents the sampling point, sampling every 1 second, t1 represents the total number of sampling points of the reconstructed voltage curve, and V1(t) represents the voltage value of the tth sampling point in the reconstructed voltage curve;

[0045] The autocorrelation coefficient feature is used to capture the lag relationship or periodicity of the signal, and its expression is:

[0046]

[0047] wherein represents the mean value of the voltage signal, N represents the total number of sampling points of the voltage signal, t represents the tth sampling point, and sampling is performed every 1 second; V(t), V(t+1) represent the voltages of the tth and t+1th sampling points, respectively;

[0048] The extracted discrete wavelet transform approximation coefficient energy feature DWT-CA E is expressed as:

[0049]

[0050] The extraction process is:

[0051] The input discrete voltage signal V = {V1, V2,..., V n is decomposed by discrete wavelet transform:

[0052]

[0053] where C A [k] is the kth approximation coefficient, C D [k] is the kth detail coefficient, φ j,k (t) is the scaling function, and ψ j,k (t) is the wavelet function, j is the scale level, and k is the position index of the signal in time;

[0054] The kth approximation coefficient and the detail coefficient are calculated:

[0055]

[0056] where the approximation coefficient C A represents the overall trend or stationary part of the signal, the detail coefficient C D represents the detail change or noise of the signal, and N represents the total number of sampling points of the voltage signal.

[0057] Further, in step S3, the density-based noise application spatial clustering DBSCAN algorithm is used to classify the time integral feature VIT 3-4 , the autocorrelation coefficient feature ACF, and the discrete wavelet transform approximation coefficient energy feature DWT-C A E, and finally the fault battery detection result is obtained:

[0058] First, the mean and variance of all battery data for the three features are calculated, and normalization processing is performed;

[0059] Then, in the DBSCAN algorithm, the minimum sample number and the field radius are set, and the normalized features are clustered and analyzed;

[0060] Through DBSCAN clustering, the features of different batteries are automatically divided into normal battery clusters and outliers, and the outliers are the fault batteries.

[0061] Further, in step S4, the extraction process of the second-order maximum charging voltage difference feature is as follows:

[0062] The maximum charging voltage is calculated:

[0063]

[0064] Calculate the first-order maximum charging voltage differential value:

[0065]

[0066] Calculate the second-order maximum charging voltage differential value:

[0067]

[0068] Further, in step S5, according to the second-order maximum charging voltage differential value feature The logic of distinguishing the fault type of the faulty battery is that:

[0069] The second-order maximum charging voltage differential value of the micro-short-circuit battery is stable;

[0070] The second-order maximum charging voltage differential value of the thermal runaway battery changes with the change rate of the short-circuit resistance, the faster the short-circuit resistance decreases, the smaller the second-order maximum charging voltage differential value of the thermal runaway battery is;

[0071] Therefore, a threshold value is set, when the second-order maximum charging voltage differential value is less than the threshold value, the battery is judged as a thermal runaway battery fault, otherwise it is a micro-short-circuit battery fault.

[0072] The beneficial effects of the present application are that:

[0073] The present application introduces three important features: 3-4V voltage curve time integral (VTI34), autocorrelation coefficient (ACF) and wavelet transform low-frequency energy (DWT-CAE), and constructs a more comprehensive fault feature combination, which can capture the subtle changes in the battery discharge process; at the same time, based on the evolution characteristics of micro-short-circuit and early thermal runaway faults, a feature is creatively proposed which can accurately distinguish micro-short-circuit and thermal runaway in the early stage of fault, effectively improving the sensitivity and accuracy of fault detection, achieving 100% diagnostic accuracy of battery micro-short-circuit and thermal runaway faults, and 12h safety warning of thermal runaway fault.

[0074] The present application uses the density-based noise application space clustering (DBSCAN) algorithm to cluster analyze the features, which realizes effective diagnosis of abnormal batteries and significantly improves the robustness and accuracy of fault recognition. The DBSCAN algorithm can process data sets containing noise and automatically determine the number of clusters, making the fault detection more reliable.

[0075] The application extracts the second-order maximum charging voltage differential value based on historical charging voltage data, and successfully distinguishes the micro-short circuit fault and the thermal runaway fault through the set second-order differential threshold (-0.003 V). This method utilizes the different performances of short-circuit resistance change characteristics in the two faults, realizes the high-precision judgment of the fault type, and improves the pertinence and practicality of the early warning system.

[0076] Other advantages, objects, and features of the application will be set forth in part in the following specification taken in conjunction with the accompanying drawings, and in part will become apparent to those skilled in the art from a consideration of the following specification and from practice of the application. The objects and other advantages of the application will be realized and attained by means of the instrumentalities and combinations pointed out in the following specification. BRIEF DESCRIPTION OF DRAWINGS

[0077] In order to make the purposes, technical solutions and advantages of the application clearer, the preferred detailed description of the application will be combined with the drawings to describe the application, wherein:

[0078] Figure 1 A simplified flowchart of the power battery micro-short circuit and early thermal runaway diagnosis method based on Simulink simulation of the application;

[0079] Figure 2 A schematic diagram of a second-order equivalent circuit model of a single lithium ion battery;

[0080] Figure 3 A pulse segment voltage curve of the second-order equivalent circuit model;

[0081] Figure 4 A schematic diagram of the voltage curve of the GITT test;

[0082] Figure 5 A schematic diagram of the experimental temperature curve under 0.5C constant current discharge condition at 25°;

[0083] Figure 6 A schematic diagram of the battery temperature change process;

[0084] Figure 7 A schematic diagram of the relationship between short-circuit resistance and time;

[0085] Figure 8 A schematic diagram of a battery pack obtained by connecting a plurality of batteries in series;

[0086] Figure 9 A voltage diagram obtained by Simulink simulation;

[0087] Figure 10 A temperature diagram obtained by Simulink simulation;

[0088] Figure 11 Time integral characteristic VIT3-4 a characteristic value distribution diagram of the eigenvalues of the autocorrelation coefficient feature ACF;

[0089] Figure 12 a characteristic value distribution diagram of the eigenvalues of the autocorrelation coefficient feature ACF;

[0090] Figure 13 a discrete wavelet transform approximation coefficient energy DWT-C A a characteristic value distribution diagram of the eigenvalues of the autocorrelation coefficient feature ACF;

[0091] Figure 14 a fault diagnosis result diagram;

[0092] Figure 15 a second-order maximum charging voltage differential value feature a characteristic value distribution diagram of the eigenvalues of the autocorrelation coefficient feature ACF;

[0093] Figure 16 a micro short circuit and thermal runaway fault distinguishing result diagram;

[0094] Figure 17 a detailed flow diagram of the power battery micro short circuit and early thermal runaway diagnosis distinguishing method based on Simulink simulation of the present application. DETAILED DESCRIPTION

[0095] The present application will be described in more detail by the following specific examples. Other advantages and benefits of the present application will be apparent from this disclosure. The present application can be implemented or applied in other different specific embodiments, and the details in the present specification can be modified or changed based on different views and applications without departing from the spirit of the present application. It should be noted that the diagrams provided in the following examples only illustrate the basic concept of the present application in a schematic manner, and the following examples and features in the examples can be combined with each other without conflict.

[0096] The accompanying drawings are only used for illustrative explanation, and the representation is only a schematic diagram, not a physical diagram, and should not be understood as a limitation of the present application. In order to better illustrate the embodiments of the present application, some components of the drawings may be omitted, enlarged or reduced, and do not represent the actual size of the product. It is understandable to those skilled in the art that some well-known structures and their descriptions in the drawings may be omitted.

[0097] The same or similar reference numerals in the drawings of the embodiments of the present application correspond to the same or similar components; in the description of the present application, it is understood that if the orientations or positional relationships indicated by the terms "upper", "lower", "left", "right", "front", "back", etc. are based on the orientations or positional relationships shown in the drawings, they are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, therefore the terms describing the positional relationship in the drawings are only used for exemplary illustration, and cannot be understood as a limitation on the present application, for those skilled in the art, the specific meanings of the above terms can be understood according to the specific circumstances.

[0098] Please refer to Figures 1-17 , a method for diagnosing and distinguishing micro-short circuit and early thermal runaway of power battery based on Simulink simulation.

[0099] Embodiment

[0100] The embodiment gives a specific implementation process of a method for diagnosing and distinguishing micro-short circuit and early thermal runaway of power battery based on Simulink simulation, as shown in Figure 1 , which includes the following steps:

[0101] S1, simulate the behavior of the battery based on Simulink simulation to obtain a simulation voltage curve, the simulation voltage curve including a plurality of sets of cycle discharge data and a plurality of sets of cycle charge data;

[0102] S2, extract a fault diagnosis feature in a discharge voltage signal feature from the current discharge data of the plurality of sets of cycle discharge data of the simulation voltage curve;

[0103] S3, perform DBSCAN clustering analysis based on the extracted fault diagnosis feature to determine a fault battery;

[0104] S4, for the fault battery, extract a fault classification feature according to the historical charge data in the simulation voltage curve thereof;

[0105] S5, distinguish micro-short circuit and thermal runaway of the fault battery according to the fault classification feature.

[0106] In step S1 of the embodiment, the battery fault behavior under different working conditions is simulated by using Simulink simulation experiment to obtain a plurality of sets of simulation voltage curves of the battery, which specifically includes the following steps:

[0107] S11, establish a second-order equivalent circuit model of a single lithium ion battery; as shown in Figure 2 , in the model, including open circuit voltage OCV, ohmic internal resistance R0, polarization resistance R1 and R2, polarization capacitance C1 and C2, and short-circuit resistance R Swherein the battery cell outputs a current source a current source passes through a parallel combination of a polarization resistance R2 and a polarization capacitance C2, a parallel combination of a polarization resistance R1 and a polarization capacitance C1, an ohmic internal resistance R0, and an open circuit voltage OCV in sequence, and a short-circuit resistance R S across a series combination of the parallel combination of the polarization resistance R2 and the polarization capacitance C2, the parallel combination of the polarization resistance R1 and the polarization capacitance C1, the ohmic internal resistance R0, and the open circuit voltage OCV.

[0108] In the present embodiment, as shown in Figure 3 , a pulse segment voltage curve of a second-order equivalent circuit model of a single lithium-ion battery is obtained by a C / 2 rate GITT discharge test in a long calibration procedure test (RPT), and parameter identification is performed by the pulse segment voltage curve, in which the voltage first drops due to battery discharge, then rises due to polarization effect, and finally tends to be flat. In the curve, several pulse points A, B, C, D, and E are selected, wherein A is the starting point, C is the lowest point, E is the terminal point, B is a point between A and C, and D is a point between C and E;

[0109] The open circuit voltage OCV reflects the terminal voltage when the battery reaches an equilibrium state after being left for a period of time, and is expressed as:

[0110] OCV = V E

[0111] wherein V E is the voltage at the pulse segment point E;

[0112] The ohmic internal resistance R0 mainly reflects the transient response, which is identified by extracting the voltages of the A-B segment and the C-D segment of the pulse segment voltage, wherein:

[0113]

[0114] wherein V A , V B , V C , and V D are the voltages at the pulse segment points A, B, C, and D, respectively, and I is the test current 2.5 A of the GITT;

[0115] The polarization resistances R1 and R2 and the polarization capacitances C1 and C2 are polarization parameters, which affect the voltage response of the C-E point after the end of the first pulse discharge, which is regarded as a zero input voltage response:

[0116]

[0117] wherein V(t) represents the battery voltage, Voc represents open circuit voltage;

[0118] The segment voltage is fitted with a second order exponential function, the fitting formula is:

[0119]

[0120] The fitting function is similar to the zero input voltage response formula of the battery, so this function is used to fit the voltage curve. Then the five parameters of the open circuit voltage OCV, polarization resistance R1 and R2, polarization capacitance C1 and C2 of the battery are represented by the coefficients obtained after fitting.

[0121] In this embodiment, the open circuit voltage OCV, the polarization resistance R1 and R2, and the polarization capacitance C1 and C2 all consider the influence of aging factors, and the ohmic resistance R0 additionally considers the influence of temperature.

[0122] In this embodiment, the maximum value of the open circuit voltage OCV max = 4.2V, the maximum value of the ohmic resistance R 0max = 0.1Ω. And the battery type is LG M50T, LG GBM50T21700 cylindrical battery, the nominal capacity is 4.86Ah, and the nominal voltage is set to 3.63V.

[0123] Experimental conditions: For the battery aged in the range of 0-100% SOC, each aging group contains 78 cycles in the entire SOC range (discharged to 2.5V, charged to 4.2V, CV maintained to C / 100), the charge rate is 0.3C, and the discharge rate is 1C. Two RPT procedures are used, respectively after the even and odd aging groups. Model parameterization is mainly based on GITT test in long RPT. As Figure 4 shown, the specific process for obtaining the pulse segment voltage curve of the second order equivalent circuit model of the single lithium ion battery is:

[0124] A), C / 10 and C / 2 discharge charging cycles between 2.5V and 4.2V voltage limits.

[0125] B), GITT_25 test: discharged at 0.5C, containing 25 pulses, each pulse consumes 200mAh of electricity, the pulses are placed for 1 hour, and the lower cutoff voltage is 2.5V.

[0126] C), GITT_5 test: discharged at 0.5C, containing 5 pulses, each pulse consumes 1000mAh (or 1A, considering unit consistency) of electricity, the pulses are placed for 1 hour, and the lower cutoff voltage is also 2.5V.

[0127] S12, a thermal model of a single lithium ion battery is established; the thermal model mainly considers the heat generated by the internal resistance and the heat dissipation of the environment, and its expression is:

[0128]

[0129] where m is the battery mass, C p is the battery heat capacity, is the derivative of battery temperature with respect to time, U t is the battery voltage, T is the battery temperature, T ∞ is the room temperature after the battery is left to stabilize, h is the heat exchange coefficient, and A is the battery surface area. These parameters are fitted based on the experimental temperature curve of 0.5C constant current discharge condition, Figure 5 shows the temperature curve of 0.5C constant current discharge condition at 25°.

[0130] S13, set the resistance value of the short-circuit resistance of the fault battery to control the battery fault type; specifically, the Chinese national standard GB / T31486-2015 requires that the LIB module maintains 85% of its initial capacity after 28 days of open circuit storage. Therefore, the RISC of a normal LIB battery that meets this standard should satisfy the following relationship:

[0131]

[0132] Therefore, the short-circuit resistance of the battery in this embodiment is set to 700Ω (micro-short circuit), 70Ω (moderate short circuit), and 7Ω (severe short circuit). In this embodiment, the resistance value of the short-circuit resistance is calculated as follows:

[0133] R S = R S0 e -f(t)

[0134] where R S0 is the initial resistance value of the fault resistance when the set thermal runaway just occurs, e -f(t) represents a function that changes with time. In this embodiment, R S0 is taken as 700Ω.

[0135] By capturing H2 gas, the detected lithium dendrites can reach the micron level, which can provide an early warning for battery overcharging (triggering thermal runaway) before the dendrite body pierces the SEI film. It takes about 1600 seconds for the temperature to start rising to the thermal runaway critical point (120°C).

[0136] Therefore, the time taken for the battery temperature in this embodiment to start rising (7Ω, battery temperature: 44°C) to the thermal runaway critical point (0.7Ω, battery temperature: 120°C) is 2100 seconds, as shown in Figure 6 .

[0137] The set parameters of the short-circuit resistance of the fault battery are shown in Table 1 as follows:

[0138] Table 1

[0139]

[0140] In this embodiment, the short-circuit resistance versus time curve is obtained as shown in Figure 7 , wherein R 2 represents the determination coefficient, and RMSE represents the root mean square error; and further, the thermal runaway short-circuit resistance is set as follows:

[0141]

[0142] In this embodiment, the resistance of the micro-short-circuit resistance is set to 600 and 500Ω.

[0143] S14, a plurality of batteries are connected in series to form a battery pack, and the failure condition of the battery pack is simulated by adjusting the resistance value of the short-circuit resistance R S , and then a plurality of sets of thermal runaway battery data, micro-short-circuit battery data and normal battery data are obtained;

[0144] In this embodiment, as shown in Figure 8 , the battery pack is formed by connecting 4 single batteries in series, and 15 sets of data are simulated, including 2 sets of thermal runaway batteries, 2 sets of micro-short-circuit batteries and 11 sets of normal batteries. And the single battery in each set is set to be inconsistent: that is, the SOC is inconsistent. The inconsistency setting is shown in Table 2:

[0145] Table 2

[0146]

[0147] The specific process includes:

[0148] ① Set the battery pack formed by connecting 4 single batteries in series

[0149] ② Working condition setting: 0.8C charging-UDDS discharging, ambient temperature 25℃;

[0150] ③ Charging and discharging strategy: when the maximum voltage of the 4 batteries is 4.2V and the minimum voltage is 2.5V, switch the charging and discharging;

[0151] ④ Battery inconsistency setting: the initial SOC of the 4 batteries is different, and the inconsistency is about 4%;

[0152] ⑤ Fault injection: a resistance is connected in parallel with the target battery, and the fault is injected at 36000s.

[0153] Thermal runaway fault (short-circuit resistance changes rapidly with time): a 700Ω short-circuit resistance (micro-short-circuit occurs) is connected in parallel, and it is reduced to a 0.7Ω short-circuit resistance (the smaller the resistance, the greater the reduction rate) with time, causing the temperature to rise rapidly, leading to thermal runaway.

[0154] Micro short circuit fault (short circuit resistance changes slowly over time, or remains unchanged): parallel a 600Ω / 500Ω short circuit resistance.

[0155] The simulation voltage curve and temperature graph of the thermal runaway resistance group 1 are shown in FIGS. 1 and 2, respectively. Figure 9 Figure 10 The voltage curve changes of four single cells in the thermal runaway battery group 1 are shown in FIG. 3, the straight upward part is the charging voltage curve of each cycle, the fluctuating downward part is the UDDS operating discharge voltage curve of each cycle, and the red curve is the voltage curve change of the thermal runaway fault battery. Figure 9 The temperature changes of four single cells in the thermal runaway battery group 1 are shown in FIG. 4, and the red curve is the temperature change of the thermal runaway battery, which sharply rises in the later stage of thermal runaway. Figure 10 Because there is a large difference between the initial voltage of the battery less than 4.2V at the beginning of the simulation and the charging and discharging strategy, that is, the voltage curve is significantly different from the subsequent cycles. To exclude interference, the discharge data around 22000 seconds in this embodiment is taken as the discharge voltage curve data of the first cycle, and the subsequent charging curve is the charging voltage curve data of the first cycle, and the charging and discharging data of 6 cycles are extracted backward to conduct early warning research on battery thermal runaway.

[0156] In step S2 of this embodiment, the fault diagnosis features in the discharge voltage signal features extracted from the cycle discharge data include time integral features of voltage intervals, autocorrelation coefficients, and discrete wavelet transform approximation coefficient energy features. The time integral features of voltage intervals refer to the integral value of the voltage curve in the 3V-4V voltage interval, which is used to measure the cumulative voltage of the voltage interval. The extraction process is as follows:

[0157] The voltage data points in the 3V-4V interval are extracted from the discharge voltage curve V(t) to obtain the reconstructed voltage curve V1(t), and then the time is integrated:

[0158]

[0159]

[0160] Where t represents the sampling point, sampling is performed every 1 second, t1 represents the total number of sampling points of the reconstructed voltage curve, and V1(t) represents the voltage value of the tth sampling point in the reconstructed voltage curve. In the cycle discharge data curve, the characteristic value distribution of the time integral feature VIT 3-4 is shown in FIG. 5. Figure 11

[0161] The autocorrelation coefficient feature is used to capture the lag relationship or periodicity of the signal, and its expression is:

[0162] ​​​

[0163] wherein, represents the mean value of the voltage signal, N represents the total number of sampling points of the voltage signal, t represents the tth sampling point, and sampling is performed every 1 second; V(t), V(t+1) represent the voltage at the tth and t+1th sampling points, respectively. In the cyclic discharge data curve, the eigenvalue distribution of the autocorrelation coefficient feature ACF is as shown in Figure 12

[0164] Discrete wavelet transform approximation coefficient energy DWT-C A The extraction method of E feature is:

[0165] Haar wavelet decomposition is performed through discrete wavelet transform (DWT) to decompose into a low-frequency part (approximation coefficient C A : representing the overall trend or stationary part of the signal) and a high-frequency part (detail coefficient C D : representing the detail change or noise of the signal) to capture the trend and rapid change of the signal.

[0166] DWT decomposition is achieved by convolution and down-sampling of the signal. In this embodiment, the input signal is a discrete voltage signal V = {V1, V2,..., V n}, then:

[0167]

[0168] wherein C A [k] is the kth approximation coefficient, C D [k] is the kth detail coefficient, φ j,k (t) is a scaling function, and ψ j,k (t) is a wavelet function, j is a scale level, and k is the position index of the signal in time.

[0169] Further, the kth approximation coefficient and the detail coefficient are calculated:

[0170]

[0171] wherein the scaling function and the wavelet function are respectively represented as:

[0172]

[0173] The final discrete wavelet transform approximation coefficient energy DWT-C A E is calculated by the formula:

[0174]

[0175] In the cyclic discharge data curve, the discrete wavelet transform approximation coefficient energy DWT-C A ​The eigenvalue distribution of E is shown in Figure 13

[0176] In step S3 of the embodiment, based on the current cycle full battery discharge voltage curve, three features are extracted: 3-4V voltage curve time integral feature VIT 3-4 , autocorrelation coefficient feature ACF, and discrete wavelet transform approximation coefficient energy DWT-C A E, a fault battery is detected using a density-based noise application spatial clustering (DBSCAN) algorithm. The process of detecting a fault battery based on the DBSCAN algorithm is as follows:

[0177] First, the mean and variance of the three features of all battery data are calculated respectively, and normalized to eliminate the influence of feature dimension. Then, in the DBSCAN algorithm, the minimum sample number is set to 5 and the field radius is set to 0.5, and the normalized features are analyzed by clustering

[0178] Through DBSCAN clustering, the features of different batteries can be automatically divided into normal battery clusters and outliers. Since the features of normal batteries have small differences, their feature points are usually concentrated in areas with high density, and DBSCAN can easily identify these points as core points or boundary points to form normal battery clusters. The features of fault batteries are significantly different from those of normal batteries, and their feature points are usually distributed in areas with low density and cannot meet the clustering conditions, so they are identified as outliers by the DBSCAN algorithm. Based on this, outliers are defined as fault batteries, thereby achieving automatic diagnosis and differentiation of fault batteries. The final fault battery detection result is shown in Figure 14 , where the blue points represent normal battery features and the red points represent fault battery features. As can be seen from the figure, the features of normal batteries are clustered together to form dense clusters, while the features of fault batteries are identified as outliers and successfully distinguished, with an accuracy of 100%. This shows the effectiveness and accuracy of the DBSCAN algorithm in fault battery detection and the superiority and rationality of the selected feature combination.

[0179] In step S4 of the embodiment, for the historical charging data of the fault battery, the second-order maximum charging voltage difference feature is extracted, which reflects the change in the maximum voltage change rate during the battery charging process. Short-circuit resistance is a key factor affecting the charging rate of the battery. The smaller the short-circuit resistance, the slower the charging rate of the battery, and the smaller the maximum voltage during the charging process. The extraction process of the second-order maximum charging voltage difference feature is as follows:

[0180] Calculate the maximum charging voltage:

[0181]

[0182] ​Calculate the first-order maximum charging voltage difference value:

[0183]

[0184] Calculate the second-order maximum charging voltage difference value:

[0185]

[0186] In the above process, because the fault is diagnosed based on the current cycle, the historical charging data of the fault is extracted for only 6 cycles. The 6 maximum charging voltages, the latter minus the former, V2-V1; V3-V2…V6-V5, can only produce 5 values. Similarly, the second-order difference produces 4 values.

[0187] The extracted second-order maximum charging voltage difference value feature Δ 2 V Cmax(n) As shown in Figure 15 , the purple line segment in the figure represents the second-order maximum charging voltage difference value of 2 micro-short circuit batteries, and the red line segment represents the second-order maximum charging voltage difference value of 2 thermal runaway batteries. As can be seen from the figure, the second-order maximum voltage difference value of the micro-short circuit battery changes slowly and is close to 0, while the second-order maximum voltage difference value of the thermal runaway battery does not change in the first few cycles, and significantly decreases at the sixth cycle.

[0188] In step S5 of the embodiment, the second-order maximum charging voltage difference value feature Δ is used to distinguish the fault type of the fault battery, and the logic is as follows:

[0189] For a micro-short circuit battery, the short circuit resistance changes relatively slowly over time, resulting in a slower charging rate for the micro-short circuit battery relative to a normal battery, but the change rate of the charging rate is relatively small. Therefore, the second-order maximum charging voltage difference value of the micro-short circuit battery is relatively stable.

[0190] In contrast, the short circuit resistance of a thermal runaway battery changes relatively quickly over time, which makes the change rate of the charging rate of the thermal runaway battery larger, and the charging rate gradually slows down. Therefore, the second-order maximum charging voltage difference value of the thermal runaway battery will change with the change rate of the short circuit resistance, and the faster the short circuit resistance decreases, the smaller the second-order maximum charging voltage difference value of the thermal runaway battery.

[0191] Based on the above characteristics, the embodiment selects the second-order maximum charging voltage difference value as a feature to distinguish between micro-short circuit faults and thermal runaway faults, and sets a threshold value of -0.003 V. When the second-order maximum charging voltage difference value is less than this threshold value, the battery is considered to be a thermal runaway battery, and the final fault distinguishing result is as shown in Figure 16As shown, the fault battery is further distinguished to accurately divide it into thermal runaway battery and micro-short circuit battery, the fault distinguishing accuracy is 100%, and the sixth cycle charge-discharge data is 12 hours ahead of thermal runaway, realizing 12 hours of thermal runaway early warning.

[0192] The application proposes a power battery thermal runaway early warning method based on Simulink simulation experiment, which combines more comprehensive fault feature combination and designs a two-step diagnosis process, such as Figure 17 The detailed process diagram of the power battery micro-short circuit and early thermal runaway diagnosis distinguishing method based on Simulink simulation of the application is as follows:

[0193] Fault battery detection: based on the discharge voltage curve (feature graph Cycle6) in the current cycle, three important features are extracted: 3-4V voltage curve-time integral (VTI34), autocorrelation coefficient (ACF), and wavelet transform low-frequency energy (DWT-CAE). These features can comprehensively reflect the subtle changes in the battery discharge process, especially suitable for capturing early voltage abnormalities caused by internal micro-short circuit or thermal runaway. The DBSCAN algorithm based on density is used for clustering analysis of these features, and the abnormal battery detection with an accuracy of 100% is realized.

[0194] Fault type distinguishing: after detecting the fault battery, based on its historical charge voltage data, the second-order maximum charge voltage difference value is extracted, and the fault type is judged through the set second-order difference threshold (-0.003V). Research shows that the short-circuit resistance of micro-short circuit battery changes little, while the short-circuit resistance of thermal runaway battery increases significantly during the fault process. This difference leads to a significant difference in the second-order maximum charge voltage difference value between the two - the difference value of thermal runaway battery decreases significantly as the short-circuit resistance decreases, while the difference value of micro-short circuit battery is small and close to 0. Therefore, if the difference value is less than -0.003V, it can be judged as thermal runaway fault; otherwise, if the difference value is greater than -0.003V and within the micro-short circuit range, it is judged as micro-short circuit fault.

[0195] The innovation of this method is to propose a more comprehensive feature combination (VTI 34 , ACF, DWT-C A E), and improve the robustness and accuracy of fault recognition through DBSCAN algorithm.

[0196] At the same time, based on the short-circuit resistance change characteristics, the method realizes the distinguishing of micro-short circuit fault and thermal runaway fault through the threshold analysis of second-order difference, so that the early warning system has higher precision and pertinence.

[0197] In the present embodiment, the discharge voltage data of the 6th cycle, i.e. the data between 140000 seconds and 165000 seconds (thermal runaway occurred at 209100 seconds), and the charging voltage history data of 1 to 6 cycles are utilized to successfully detect all fault signals (accuracy 100%) and distinguish micro-short circuit fault from thermal runaway fault, achieving a 12-hour safety warning of thermal runaway fault in advance.

[0198] Finally, it should be pointed out that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced equivalently without departing from the purpose and scope of the technical solutions, which should be covered in the scope of the claims of the present application.

Claims

1. A method for distinguishing between micro-short circuit and early thermal runaway of a power battery based on Simulink simulation, characterized in that: The method comprises the following steps: S1, simulating battery behavior based on Simulink simulation to obtain a simulation voltage curve, the simulation voltage curve comprising a plurality of sets of cyclic discharge data and a plurality of sets of cyclic charge data; S2, extracting a fault diagnosis feature in a discharge voltage signal feature from current discharge data of the plurality of sets of cyclic discharge data of the simulation voltage curve; S3, performing DBSCAN clustering analysis on the extracted fault diagnosis feature to determine a fault battery; S4, for the fault battery, extracting a fault classification feature according to historical charge data in the simulation voltage curve thereof; S5, distinguishing micro short circuit and thermal runaway of the fault battery according to the fault classification feature; In step S2, the fault diagnosis feature in the discharge voltage signal feature extracted from the cyclic discharge data comprises a time integral value in a voltage interval, an autocorrelation coefficient and a discrete wavelet transform approximation coefficient energy; In step S4, the second-order maximum charging voltage differential value feature is extracted as follows: Calculate the maximum charge voltage: Calculate the first-order maximum charge voltage differential value: Calculate the second-order maximum charge voltage differential value: In step S5, the second order maximum charge voltage differential value characteristic The logic to distinguish the fault type of the faulty battery lies in: The second-order maximum charge voltage differential value of the micro short circuit battery is stable; The second-order maximum charge voltage differential value of the thermal runaway battery changes with the change rate of the short circuit resistance, the faster the short circuit resistance decreases, the smaller the second-order maximum charge voltage differential value of the thermal runaway battery is; Therefore, a threshold value is set, when the second-order maximum charge voltage differential value is less than the threshold value, the battery is judged as a thermal runaway battery fault, otherwise as a micro short circuit battery fault.

2. The method of claim 1, wherein the method is based on a Simulink simulation. In step S1, the battery fault behavior under different working conditions is simulated by Simulink simulation experiment to obtain a plurality of sets of simulation voltage curves of the batteries, which comprises the following steps: S11, establishing a second-order equivalent circuit model of a single lithium ion battery; S12, establishing a thermal model of a single lithium ion battery; S13, setting the resistance value of the fault battery short circuit resistance to control the battery fault type; S14, a plurality of batteries are connected in series to form a battery pack, and a short-circuit resistance is adjusted to control the fault condition of the battery pack The resistance value of the short-circuit resistance is adjusted to control the fault condition of the battery pack, and a plurality of groups of thermal runaway battery data, micro-short circuit battery data and normal battery data are obtained.

3. The method of claim 2, wherein the method is based on a Simulink simulation. In step Sll, the second order equivalent circuit model comprises an open circuit voltage , an ohmic internal resistance , a polarization resistance and a polarization capacitance and as well as a short circuit resistance , wherein, Battery cell at output current source , current source through a polarization resistance and a parallel combination of a polarization resistance and a polarization capacitance , an ohmic internal resistance and an open circuit voltage , and a short circuit resistance in parallel with the parallel combination of the polarization resistance and the polarization capacitance , a polarization resistance and a parallel combination of a polarization resistance and a polarization capacitance , a series combination of an ohmic internal resistance and an open circuit voltage ; The pulse segment voltage curve of the second-order equivalent circuit model of the single lithium ion battery is obtained by C / 2 rate GITT discharge test in the long calibration program test, and the parameter identification is performed through the pulse segment voltage curve; In the pulse voltage curve, the voltage first drops due to battery discharge, then rises due to polarization effect, and finally tends to be flat, and several pulse points are selected in the curve Wherein, A point is the starting point, C point is the lowest point of decline, E point is the end point, B point is the point between A and C points, and D point is the point between C and E points. open-circuit voltage The open-circuit voltage is the voltage of the battery when it is not under load. It is the voltage that is measured when the battery is disconnected from the load. wherein Vp is the voltage at the point E of the pulse segment; ohmic internal resistance The main reflection of the transient response is identified by extracting the voltage of the A-B section and the voltage of the C-D section of the pulse segment voltage, wherein: wherein the voltage at the pulse segment point the voltage at the pulse segment point is the test current for GITT; polarization resistance and polarization capacitance and As a polarization parameter, the voltage response of the cell at the point C-E after the end of a single pulse discharge is affected, which is considered as a zero input voltage response: wherein, represents the battery voltage, represents the open circuit voltage; The voltage segment is fitted by a second-order exponential function, and the fitting formula is: The fitting function is used to fit the voltage curve, and the coefficients obtained after fitting are used to represent the open-circuit voltage of the battery , polarization resistance and , polarization capacitance and parameters; open circuit voltage , polarization resistance and , polarization capacitance and ohmic internal resistance additional temperature influence.

4. The method of claim 3, wherein the method is based on a Simulink simulation. In step S12, the thermal model considers the internal resistance heat and the environmental heat dissipation, and its expression is: where m is the battery mass, is the battery heat capacity, is the derivative of the battery temperature with respect to time, is the battery voltage, is the battery temperature, is the temperature at which the battery stabilizes after being left to rest, i.e. the room temperature, is the heat exchange coefficient, is the battery surface area.

5. The method for diagnosing and distinguishing micro-short circuit and early thermal runaway of power battery based on Simulink simulation according to claim 4, characterized in that: In step S13, the resistance value setting of the fault battery short circuit resistance comprises maximum value calculation, micro short circuit resistance value analysis, moderate short circuit resistance value analysis, severe short circuit resistance value analysis and thermal runaway short circuit resistance value analysis, wherein according to the analysis result, the resistance value of the short circuit resistance is determined according to the following formula: wherein is the initial resistance of the fault resistance at the moment when the set thermal runaway has just occurred, denotes a function of time.

6. The method of claim 2, wherein the method is based on a Simulink simulation. Wherein, The time integral feature in the voltage interval refers to the integral value of the voltage curve in the 3V-4V voltage interval, which is used to measure the cumulative voltage of the voltage interval, and the extraction process is as follows: The voltage data points in the 3V-4V interval are extracted from the discharge voltage curve V(t) to obtain the reconstructed voltage curve V1(t), and then the time is integrated: Wherein, t represents a sampling point, sampling is performed every 1 second, t1 represents the total number of sampling points of the reconstructed voltage curve, represents the voltage value of the tth sampling point in the reconstructed voltage curve; The autocorrelation coefficient feature is used to capture the lag relationship or periodicity of the signal, and its expression is: wherein, represents the mean of the voltage signal, represents the total number of sampling points of the voltage signal, represents the first sampling point, sampled every 1 second; respectively represent the voltage of the first and the first sampling point. Extracted discrete wavelet transform approximation coefficient energy features is represented as: The extraction process is as follows: discrete voltage signal inputted discrete wavelet transform decomposition is performed: wherein, is the kth approximation coefficient, is the kth detail coefficient, is a scaling function, is a wavelet function, is a scale level, is a position index of the signal in time; Calculate the Approximation coefficients and detail coefficients: wherein the approximation coefficients represent the overall trend or smooth part of the signal, the detail coefficients represent the detailed variations or noise of the signal, represent the total number of sampling points of the voltage signal.

7. The method of claim 6, wherein the method is based on a Simulink simulation. In step 3, the DBSCAN algorithm based on density is used to apply spatial clustering to the time integration features , the autocorrelation coefficient features , the discrete wavelet transform approximation coefficient energy features Classification is performed, and finally the fault battery detection result is obtained: Firstly, the mean and variance of the three features of all battery data are calculated respectively, and normalized processing is performed; Then, the minimum sample number and the field radius are set in the DBSCAN algorithm, and the normalized features are subjected to cluster analysis. Through DBSCAN clustering, the features of different batteries are automatically divided into a normal battery cluster and outliers, and the outliers are fault batteries.

Images