Uniform marginal membrane state for characterizing battery safety and number network observation method of mean marginal membrane state

Abstract

The invention discloses a mean-edge film state for representing battery safety and a number network observation method thereof. The mean-edge film state comprises five key steps of film resistance circuit design, number network design, mean-edge film resistance estimation, mean-edge film state calculation and safety state prediction. The method capable of quantitatively predicting the safety state of the battery is constructed based on modules such as a mean-edge film resistance circuit, a number channel element, mean-edge film resistance, a mean-edge film state and the safety state of the battery. The insulation resistance based on the battery diaphragm can directly characterize the safety problem of thermal runaway of the battery, the insulation resistance and the equalization resistance are connected in parallel to form the mean-edge resistance, the mean-edge film state of the percentage value between the mean-edge resistance and the equalization resistance is used for directly and quantitatively characterizing the safety state of the battery, and the problem that the safety state of the current battery cannot be directly and quantitatively characterized is solved; neurons are simulated by adopting number meridian elements, and a number meridian network is constructed by adopting the number meridian elements to perform system modeling and control, so that active-while-learning optimization can be realized, and the problem that a current neural network can only be active after learning optimization is solved.

Description

Technical Field

[0001] The present invention proposes a method for characterizing the safety of a battery by using a uniform film state and a numerical network observation method thereof, which is applicable to the field of new energy and smart car lithium-ion power batteries, specifically to estimate the insulation resistance of the battery separator (abbreviated as: insulation film resistance (11), R mi ), on this basis, the insulating film resistor R mi (11) and the balancing resistor R cb (12) Parallel connection to construct uniform film resistance R mib (13), using uniform film resistance R mi (13) and the balancing resistor R cb (12) The percentage value of the uniform membrane state S mib (40) to characterize or predict the safety status S of new energy and smart car power batteries os (50). Background Art

[0002] In recent years, with the warming of the atmosphere and the increasingly harsh global environment, the peak of carbon emissions and carbon neutrality have become extremely urgent. Lithium-ion batteries have the advantages of high energy density, high-power charging and discharging capabilities, and long life. They are widely used in new energy vehicles and energy storage. During use, lithium-ion batteries may have thermal runaway and fire, which brings huge losses to social production and life. Therefore, safety is one of the biggest challenges of lithium-ion batteries. The present invention uses the diaphragm insulation resistance to construct a new uniform edge film state S mib (40) to characterize the battery safety state, and a new mathematical network (2) theory was invented and used to observe the safety state S of lithium-ion batteries. os (50).

[0003] In the field of lithium-ion battery safety risk warning, there are invention contents with application numbers CN202310446506.4, CN202311012840.5, and CN202310852097.8, which are essentially different from the content of the present invention, as described below.

[0004] The invention of application number CN202310446506.4 is named “A power battery safety risk warning method based on safety entropy”. The invention content proposes a method for determining the safety status of a battery, which uses a variety of usage data of the battery before the damage accident to calculate the probability or constant for conversion to obtain safety entropy, and uses this entropy value to determine the safety status of the battery. It can achieve a quantitative description of the safety of the battery system, but it needs to calculate the probability value of each independent event, and the accuracy of each probability value will directly affect the accuracy of the battery safety status judgment. The difference from the content of the present invention is that the present invention directly adopts the homogeneous membrane state S mib(40) is used to characterize the battery safety status, which is essentially different from using entropy to determine the battery safety status.

[0005] The invention of application number CN202311012840.5 is named “A method for safety early warning assessment of electrochemical energy storage batteries”. The invention content proposes a method for evaluating the safety status of batteries, constructs a battery safety model architecture algorithm, and uses multiple factors such as battery current, voltage, temperature and other input information to form different safety event factors as model inputs from multiple safety perspectives. Then, a function expression for calculating the battery safety factor is derived, and this function is used to define the safety status of the battery. The innovation of the invention content is to construct a comprehensive model that takes into account multiple safety factors. Although it is reliable, it requires coupling of multiple factors, which greatly increases the complexity of the calculation and the modeling process is complicated. The innovation of the invention content is to use the homogeneous membrane state S mib (40) is used to characterize the battery safety status, which is essentially different from the functional expression of multiple factors such as battery current, voltage, and temperature.

[0006] The invention with application number CN202310852097.8 is named “Power Battery Safety Early Warning Method and Online Monitoring Device”. The invention proposes a SVR estimation method based on support vector regression. This method does not require the establishment of an accurate battery model, but can accurately determine the battery safety status through autonomous learning based on input multi-dimensional signal data such as current and voltage. This method requires a large amount of reference data for training, and the training data and training method have a great influence on the error of battery safety status judgment. The present invention adopts a uniform film state S mib (40) is used to characterize the battery safety state, and the numerical network model estimation algorithm is used to perform the uniform edge film state S mib (40) Observation is essentially different from the use of multi-dimensional signal data such as input current and voltage and learning algorithms.

[0007] In summary, the existing lithium-ion battery safety status assessment method has certain shortcomings. It does not directly use the diaphragm insulation resistance (that is, the insulation film resistance R) of the key insulation resistance inside the battery. mi (11)) to characterize the battery safety status, it is difficult to os (50) Therefore, it is necessary to improve the definition of battery safety status and its observation algorithm on this basis to improve the battery safety status S os (50) is used to determine the accuracy and effectiveness of the judgment. Summary of the Invention

[0008] The present invention proposes a homogeneous membrane state for characterizing battery safety and a numerical network observation method thereof, so as to solve the shortcomings of direct quantitative characterization methods and estimation methods of battery safety status.

[0009] To achieve this object, the present invention adopts the following technical solutions:

[0010] 1. A method for observing the uniform edge film state and its numerical network for characterizing battery safety, characterized by:

[0011] The paper includes five key steps: membrane resistance circuit design (1), data-based network design (2), uniform membrane resistance estimation (3), uniform membrane state calculation (4), and safety state prediction (5), as well as uniform membrane resistance circuit (10), data-based element (8), twin data-based network (23), control data-based network (25), and battery uniform membrane resistance R mib (13) Battery uniform edge film state S mib (40) Battery safety status S os (50) and other modules. Based on the above modules, a method for battery safety status prediction (5) is constructed, which is divided into the following five steps.

[0012] Step 1: Design the membrane resistance circuit (1), establish the battery uniform membrane resistance circuit (10) model, and analyze the battery safety status S os (50) and the homogeneous membrane state S mib (40) (i.e., the insulation resistance of the battery separator).

[0013] Based on the intrinsic electrochemical characteristics of the battery, an approximate solution is established to characterize the insulation resistance of the battery separator (abbreviated as: battery insulation film resistance R mi (11)) Changes in the cell's average insulation resistance R mib (13) Model of the uniform-edge membrane resistance circuit (10).

[0014] Battery insulation film resistance R mi (11) Under normal circumstances, it is infinite. It is not convenient to obtain it through detection during the battery charging and discharging process. Therefore, it is not convenient to directly use the battery insulation resistance R mi (11) to characterize the battery safety state S os (50).

[0015] The present invention uses a battery management system BMS (61) to collect the system cell balancing resistor R in the board BMU (63) cb (12) and the battery insulation film resistance R mi (11) are connected in parallel to obtain a cell uniform film resistance R mib (13), the cell has a uniform film resistance R mib (13) is exactly the equivalent resistance used in the uniform film resistance circuit model (10).

[0016] Battery insulation resistance R mib (13) is a finite value, which is normally slightly smaller than the cell balancing resistance Rcb (12), convenient for balancing the cell resistance R cb (12) Compare and get a percentage to characterize the battery insulation film resistance R mi (11) changes.

[0017] When the battery insulation film resistance R mi (11) When it is infinite or very large, the cell insulation resistance R mib (13) and the cell balancing resistor R cb (12) The values ​​are similar and slightly smaller. During the use of the battery, due to the growth of lithium dendrites inside the battery, they continuously invade the diaphragm, resulting in a decrease in the insulation resistance of the diaphragm, that is, the insulation resistance of the battery insulation film R mi (11) becomes smaller, and then the battery insulation film resistance R mib (13) will become smaller. Therefore, the battery membrane resistance R can be observed through the battery membrane resistance circuit model (10) mib (13), through the battery uniform insulation film resistance R mib (13) to characterize the resistance of the battery insulation film R mi (11) changes, and the changes in the insulation resistance of the battery diaphragm are obtained, which can characterize the safety status of the battery S os (50).

[0018] The battery insulation film resistance circuit model (17) simplifies a battery cell into an equivalent circuit consisting of a capacitor and a resistor in parallel. The capacitor is the equivalent capacitance C of the battery cell. ce (14), this resistance is the cell's dielectric film resistance R mib (13) Battery insulation resistance R mib (13) The battery insulation film R mi (11) and the cell balancing resistor R cb (12) are connected in parallel.

[0019] Step 2: Design of the data-processing network (2). Design of the battery data-processing network (20) to solve the problem of theoretical design of the data-processing network for simulating the dynamics and control of the battery system.

[0020] The battery data network (2) is composed of a twin data network (24) for simulating battery dynamics and a control data network (27) for estimating battery status, and the basic data processing unit is a data element (8).

[0021] Based on the multi-state coupled observation requirements of batteries and the neural network theory, a battery dynamics mathematical model with multiple states, multiple parameters and multiple algorithms was designed (21).

[0022] The battery second-order equivalent circuit model ECM (24) is used to simulate the equivalent circuit inside the battery, which is the battery electrolyte polarization capacitance Cpl (63) Electrode polarization capacitance C pe (64) Electrolyte polarization resistance R pl (65), electrode polarization resistance R pe (66) Ohmic polarization internal resistance R po (67).

[0023] Due to the ohmic polarization internal resistance R po , electrolyte polarization resistance R pl (65) and the electrode polarization resistance R pe (66) are very small, and are comparable to the cell's dielectric film resistance R mib (13) and the battery insulation film resistance R mi (11) The difference in resistance values ​​is very large, which is ignored in the present invention.

[0024] The estimated battery electrolyte polarization capacitance C pl (63) and electrode polarization capacitance C pe (64) is passed to the membrane resistance circuit model to estimate the cell average membrane resistance R mib (13).

[0025] Electrolyte polarization capacitance C pl (63) and electrode polarization capacitance C pd (64) After connecting in series, the equivalent capacitance C of the cell is formed ce (13), the relationship between them is:

[0026]

[0027] Step 3: Estimation of the mean film resistance (3): Estimation of the mean film resistance R bm (13) is an approximate value to solve the problem of the average film resistance R mib (13) The problem of obtaining values ​​in real time.

[0028] Based on the uniform film resistance circuit (10) model and the twin network model (20), the module-based observer (30) algorithm is used to estimate the uniform film resistance R mib (13) the size of the value,

[0029] Based on the system cell balancing resistor R in the acquisition board BMU (63) of the battery management system BMS (61) cb (12), the battery insulation film resistance R mi (11).

[0030]

[0031] When the battery is in normal state, the battery insulation film resistance R mi(11) is an infinite value. Therefore, formula (3.2) is a very large value under normal circumstances and belongs to an uncertain value state. Therefore, the present invention does not use the battery insulation film to resist R mi (11) Directly characterize the battery safety status S os (50).

[0032] Step 4: Calculate the uniform membrane state (4) and calculate the uniform membrane state S mib (40) Solve the problem of real-time uniform film state S mib (40)The problem of obtaining values.

[0033] According to the estimated battery insulation resistance R mib (13) value, and the cell balancing resistor R in the battery management system (61) acquisition board BMU (63) cb (12) and compare to obtain a percentage value of the battery uniform film state S mib (40),

[0034] During the normal operation of the battery, due to the resistance of the battery insulation film R mi (11) is an infinite value, resulting in the battery's uniform film resistance R mib (13) Very close to the cell balancing resistance R cb (12), therefore, under normal circumstances, the uniform membrane state S mib (40) is a value close to 100%.

[0035] Step 5: Safety status prediction (5): predict the battery safety status and solve the battery safety status prediction problem.

[0036] Based on the uniform edge membrane state safety threshold S tms (51), according to the calculated cell uniform membrane state (40), when S mib ≥S tms , it can be predicted that the battery is in a membrane insulation safe state (53).

[0037] Based on the uniform edge membrane state safety threshold S tms (52) and the critical threshold S of the homogeneous membrane state tmc (52), according to the calculated cell uniform membrane state (40), when S mrs >S mib ≥S mrc , it can be predicted that the battery is in a critical state of thermal runaway (54).

[0038] Based on the critical threshold S of the uniform film state mrc , according to the calculated cell uniform membrane state (40), when S mib mrc ​, it can be predicted that the battery is in a state of thermal runaway risk (55).

[0039] Through the homogeneous membrane state S mir (40) The value predicts the battery safety state S os (50) Whether it is in the membrane insulation safety state (53), the thermal runaway critical state (54), or the thermal runaway risk state (55).

[0040] 2. The uniform-edge film resistor circuit (10) according to claim 1, characterized in that:

[0041] Simplify the battery into the equivalent capacitance C of the cell ce (14) and the insulating film resistance R mi (11) The equivalent circuit in parallel, on this basis, close the cell balancing circuit on the acquisition board BMU (63), and change the cell balancing resistor R cb (12) Introduced into the equivalent circuit, forming the cell equivalent capacitance C ce (14) Battery insulation film resistance R mi (11) and cell balancing resistor R cb (12) A circuit with three electronic components connected in parallel, where the battery insulation film resistor R mi (11) and cell balancing resistor R cb (12) is in parallel state and can be combined and simplified into a uniform film resistor R mib (13);

[0042] During the charging process, the kinetic differential equation model of the uniform film resistance circuit (10) can be established as follows:

[0043]

[0044] Among them, U ct is the cell terminal voltage (68), I cm The current (69) is measured for the cell.

[0045] 3. The counting unit (8) according to claim 1, characterized in that:

[0046] In order to simulate the human neural network, the receiving dendrite (71) in the human neuron is simulated as an arrow line as the numeric element input (81); the cell body (72) of the neuron is simulated as a circle as the numeric element operator (82); the nucleus (73) in the neuron cell body is simulated as a mathematical operator (83), including arithmetic operators, algebraic operators, logical operators, set operators, matrix operators, calculus operators, etc.; the axon (74) is simulated as a thick solid line to represent the numeric element operation output (84); the myelin sheath (7 5) Simulate a solid circle, a solid ellipse and a hollow circle. The solid ellipse is used as the operation storage node (85) for storing the operation result, the solid circle is used as the data storage node (88) after data weighting processing, and the hollow circle is used as the weight storage node (86); simulate the synapse (76) of the neuron into two types of thin solid arrow lines, one type is the weight input (87) and the other type is the data output (89). The data processed by the neuron is transmitted to the next neuron or effector through the thin solid arrow line. The arrow is directly connected to the circle simulating the neuron cell body.

[0047] The above method simulates the human neural network into a data calculation and storage processing process. The arrow line is responsible for inputting data information, the corresponding operator of the cell nucleus is responsible for performing operations such as addition, subtraction, multiplication and division, the thick solid line axon is responsible for outputting operation data, the solid ellipse point myelin sheath is responsible for storing the operation results, the solid circle point myelin sheath is responsible for data storage, the hollow circle point myelin sheath is responsible for weight parameter storage, and the thin solid arrow line is responsible for weight input and data output. The entire integration is referred to as a data element, and a network composed of many data elements is called a data network.

[0048] The numerical element (8) uses "f'" to represent differential operation, "∫" to represent integral operation, "+" to represent addition operation, "÷" to represent multiplication operation, "-" to represent subtraction operation, and "×" to represent multiplication operation.

[0049] 4. The data network (20) according to claim 1, characterized in that:

[0050] The digital network (20) is composed of a twin digital network (23) and a control digital network (25), wherein the twin digital network (23) is responsible for simulating battery system dynamics and system parameter estimation, and the control digital network (27) is responsible for battery state estimation and prediction.

[0051] The mathematical network (20) uses circle nodes to represent an operation and uses arrow lines to connect the nodes to form a twin mathematical network (24) for simulating system dynamics and a control mathematical network (25) for designing control algorithms.

[0052] The digital network model (22) integrates the uniformly resistive membrane circuit (10) model, the battery equivalent circuit model (24), the model-based observer (30) algorithm, etc. to form a complete battery network model.

[0053] The twin digital network model (23) is composed of a battery equalizing circuit (10) model and a battery equivalent circuit model (24). Parameters in the twin digital network (23) all have initial values, and after the first network update, system control and estimation are performed each time based on the updated parameters.

[0054] The control mathematical network model (25) is composed of an observer algorithm network (31) and an observer parameter network (32) of a model-based observer (30).

[0055] 5. The battery uniform film resistor (13) according to claim 1, characterized in that:

[0056] Based on the cell balancing resistor R in the acquisition board BMU (63) of the battery management system BMS (61) bc (12), and the battery insulation film resistance R mi (11), the cell uniform film resistance R mib The expression of (13) is

[0057]

[0058] Battery insulation resistance R mib In fact, the battery insulation film resistance R mi (i.e., the insulation resistance of the battery diaphragm) and the cell balancing resistance R in the acquisition board BMU (62) for system balancing cb (12) Parallel resistance, which is referred to as the battery membrane resistance R in this invention. mib (13).

[0059] In fact, the battery management system BMS (61) cannot measure the battery insulation film resistance R mi (11), so the cell insulation resistance R cannot be calculated by formula (4.1) mib (13), can only be estimated using the observer algorithm.

[0060] 6. The battery uniform edge film (40) according to claim 1, characterized in that:

[0061] The so-called battery uniform film state S mib (40) is the observed cell average film resistance R mib (13) and the cell balancing resistor R cb (12) is a percentage value. The cell's uniform film resistance R mib (13) is the cell balancing resistor Rcb (12) and the battery insulation film resistance R mi (13) The resistance after parallel connection, therefore, the cell membrane resistance R mib (13) must be smaller than the cell balancing resistance R cb (12).

[0062] Normally, the battery insulation film resistance R mi The value of (11) is approximately infinite, so the cell insulation resistance R mib (12) and the cell balancing resistor R cb (12) are very close in size, that is, under normal circumstances, the battery uniform film state S mib (40) Close to 100%.

[0063] According to the estimated battery insulation resistance R mib (13) value, and the cell balancing resistor R in the battery management system (61) acquisition board BMU (63) cb (12) and compare to obtain a percentage value of the battery uniform film state S mib (40).

[0064]

[0065] 7. The battery safety state (50) according to claim 1, characterized in that:

[0066] Based on the uniform edge membrane state safety threshold S tms (51), when the calculated cell uniform membrane state S mib (40),

[0067] S mib ≥S tms (5.1)

[0068] The battery can be predicted to be in a membrane insulation safe state (53);

[0069] Based on the uniform edge membrane state safety threshold S tms (51) and the critical threshold S of the homogeneous membrane state tmc (52), when the calculated cell uniform membrane state S mib (40) for,

[0070] S tms >S mib ≥S tmc (5.2)

[0071] It can be predicted that the battery is in a critical state of thermal runaway (54);

[0072] Based on the critical threshold S of the uniform film state tmc , when the calculated battery uniform membrane state Smib (40) for,

[0073] S mib tmc (5.3)

[0074] It can be predicted that the battery is in a state of thermal runaway risk (55). BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 A schematic diagram of a battery data network (20) according to a specific embodiment of the present invention;

[0076] Figure 2 A schematic diagram of a uniform film resistance circuit (10) according to a specific embodiment of the present invention;

[0077] Figure 3 A schematic diagram of a digital element (8) according to a specific embodiment of the present invention;

[0078] Figure 4 Schematic diagram of a battery equivalent circuit model (24) according to a specific embodiment of the present invention.

[0079] Terminology

[0080] The following is a table of terms that explain the symbols and related concepts used in the present invention.

[0081]

[0082]

[0083]

[0084] DETAILED DESCRIPTION

[0085] The present invention will be further described in detail below with reference to the accompanying drawings and examples. It will be understood that the specific embodiments described herein are intended only to illustrate the present invention and are not intended to limit the present invention. It should also be noted that, for ease of description, the accompanying drawings only illustrate portions relevant to the present invention, not all structures.

[0086] The present invention is to propose a method for predicting the safety status of a battery from the perspective of the battery uniform film state. Figure 1 ​As shown in the figure: by establishing a twin digital network TMNN (23) of the battery system, the electrical performance characteristics of the battery system working process are digitally twinned, so as to simulate various voltages in the battery system and the intrinsic parameters of the battery system model in real time; by establishing a control digital network CMNN (27) of the battery system, the state that needs to be observed in the battery system is observed in real time, so as to control the battery system in real time.

[0087] For specific implementation, please refer to Figure 1 、 Figure 2 、 Figure 3 、 Figure 4 .

[0088] See also Figure 1 As shown, the top layer is composed of a battery system consisting of lithium-ion batteries, relays, and a battery control unit. The right side is the power output of the battery system, which can collect the battery current I m and voltage U m , used for various control purposes in battery management systems.

[0089] See also Figure 1 As shown, the second layer from the top is the battery system twin digital network (23). The main function of this network is to simulate the battery system digital twin. The dynamic model used is:

[0090]

[0091] The battery system dynamics model is established according to formula (2.1) and implemented using the twin mathematical network TMNN (23). By establishing the battery system twin mathematical network TMNN (23), the key parameters of the battery system model can be observed: the battery electrolyte polarization capacitance C pl (63) Electrode polarization capacitance C pe (64) Electrolyte polarization resistance R pl (65), electrode polarization resistance R pe (66) Ohmic polarization internal resistance R po (67) etc., for the battery insulation film resistance R im The estimation of (11) provides the basis for the intrinsic parameters of the battery.

[0092] See also Figure 1As shown, the third layer from the top is the battery system module-based observer MBO (30), which is implemented using the control matrix network CMNN (25). By establishing the control matrix network CMNN (25) of the module-based observer MBO (30) algorithm, the state of charge SOC (26) of the battery system can be observed online in real time. At the same time, after the module-based observer MBO (30) is introduced, the deviation obtained by comparing the terminal voltage observed by the twin matrix network TMNN (23) with the actual measured terminal voltage is introduced into the control network to implement iterative control to reduce the deviation, and finally make the system parameters approach the true value.

[0093] See also Figure 1 As shown, the fourth layer from the top is the cell uniform film resistance R mib (13) and is implemented using the control matrix network CMNN (25), with the uniform film resistance R mib The calculation formula of (13) is:

[0094]

[0095] It can be seen from formula (3.1) that the cell membrane resistance R mib (13) is the resistance of the battery insulation film R mi (11) and battery balancing resistor R cb (12) are connected in parallel.

[0096] As with the state of charge (SOC) (26) estimation algorithm, the control matrix network (CMNN) (25) of the model-based observer (MBO) (30) algorithm can be used to estimate the insulation film resistance R of the battery system. mib (11) Online real-time observation is performed. At the same time, after the model-based observer MBO (30) is introduced, the deviation obtained by comparing the observed terminal voltage with the actual measured terminal voltage is imported into the control network to realize iterative control to reduce the deviation, and finally the insulation film resistance R of the battery system is reduced. mib (11) approaches the true value.

[0097] See also Figure 1 As shown, the fourth layer from the top, the lower left corner is the battery system uniform membrane state S mib (40) and the numerical network MNN (2) was used to calculate the uniform edge membrane state S mib (40) is calculated using the following formula:

[0098]

[0099] Battery uniform film state S mib (40) is the cell insulation resistance R mib (13) and the battery balancing resistor R cb(12) ratio, under normal circumstances, the battery insulation film resistance R mib (13) must be greater than the battery balancing resistor R cb (12) is small because the cell membrane resistance R mib (13) is the battery balancing resistor R cb (12) and the battery insulation film resistance R mi (11) Value after parallel connection.

[0100] See also Figure 2 As shown, it is a schematic diagram of the battery membrane resistance circuit. Figure 2 Neutron diagram (a) shows the battery balancing resistor R cb (12) and the battery insulation film resistance R mi (11) Separation state: In fact, in the battery system, the diaphragm inside the battery is made of polymer material, which has insulation resistance to electrons, resulting in the battery not being able to short-circuit internally, and electrons can only bypass through the external circuit. The present invention simplifies the diaphragm insulation resistance to the insulation film resistance R mi (11). At the same time, in the battery system, in order to achieve battery balancing, a balancing resistor is usually connected in parallel to each cell. When the cell voltage is too high, the excess electrical energy is released through this balancing resistor, thereby achieving battery balancing. The present invention refers to this balancing cell as the battery balancing resistor R cb (12).

[0101] See also Figure 2 (b) shows the battery membrane resistance circuit diagram, which uses the battery balancing resistor R cb (12) and the battery insulation film resistance R mi (11) Combined into a battery with uniform film resistance R mib (13). Before merging, the battery insulation film resistance R mi (11) is infinite in theory and difficult to detect or estimate using an observer in an actual battery system because the present invention reduces the resistance of the battery insulation film to R mi (11) and the battery balancing resistor R cb (12) Connect in parallel to form a battery with uniform film resistance R mib (13) Battery insulation resistance R mib (13) is a finite value and is definitely greater than the battery balancing resistance R cb (12) is small. When the battery insulation film resistance R mi (11) When it is infinite, the cell insulation resistance R mib (13) is infinitely close to the battery balancing resistance R cb (12). When the cell's uniform film resistance R mib (13) When the resistance of the battery membrane decreases to hundreds of megohms or tens of megohms, the resistance of the battery membrane R mib(13) With the battery balancing resistor R cb (12) has a very obvious difference. If we can estimate the cell's average film resistance R mib (13), so the battery insulation film resistance R can be calculated by reverse calculation. mi (11) is an actual approximate value, through which it is possible to estimate whether the insulation state of the battery separator is normal.

[0102] See also Figure 3 As shown, neurons are simplified into a diagram of neurons. Figure 3 The top part is a typical diagram of a human neuron. Figure 3 The lower part is a diagram of the digital element simplified by the present invention based on human neurons. Figure 3 In the diagram, the dendrites (71) of the neuron are simplified into the nucleus (82) of the neuron, into the nucleus (83) of the neuron, into the nucleus (84) of the neuron, into the nucleus (85) of the neuron, into the nucleus (86) of the neuron, into the nucleus (87) of the neuron, into the nucleus (88) of the neuron, into the nucleus (89) of the neuron, into the nucleus (88) of the neuron, into the nucleus (87) of the neuron, into the nucleus (88) of the neuron, into the nucleus (89) of the neuron, into the nucleus (87) of the neuron, into the nucleus (83) of the neuron, into the nucleus (83) of the neuron, into the nucleus (84) of the neuron, into the nucleus (84) of the neuron, into the nucleus (84) of the neuron, into the nucleus (87) of the neuron, into the nucleus (89) of the neuron, into the nucleus (89) of the neuron, into the nucleus (87) of the neuron, into the nucleus (87) of the neuron, into the nucleus (89) of the neuron, into the nucleus (89) of the neuron, into the nucleus (83 ...

[0103] Current neural networks have four major shortcomings: "inability to learn while working, inability to adapt to computing power, unknown underlying logic, and inability to mimic functionalities." The human body system has four inherent features: "self-learning, self-growth, self-anti-interference (or self-immunity), and self-repair."

[0104] "Cannot learn while working" means that current neural networks can only work after learning large-scale data. In fact, humans have the self-learning function of learning while working. The present invention will construct a digital network with "self-learning" function based on the digital element bionic theory.

[0105] "Inability to adapt computing power" is the biggest drawback of current neural networks, which means that they cannot adapt to appropriate computing power according to computing needs, that is, the network does not have the ability to self-grow. This invention will construct a digital network with "self-growth" function based on the digital element bionic theory.

[0106] "Underlying logic is unknown" means that the internal operating mechanism of the current neural network is a "black box", and some internal reasoning logic is unknown, resulting in weak self-interference rejection capability. The present invention will construct a digital network with "self-interference rejection" function based on the digital element bionic theory.

[0107] "Functional bionics are not relevant" means that currently a single neuron is difficult to perform such complex mathematical operations as activation functions, and does not have self-repair capabilities, so functional bionics is not appropriate. This invention will construct a digital network with "self-repair" capabilities based on the digital bionics theory.

[0108] See also Figure 4 As shown, the equivalent circuit model (24) of the battery system is used to simulate the electrical characteristics of the battery system in order to obtain the parameters of the battery model. The simplified components and parameters of the equivalent circuit model (24) are: battery open circuit voltage U oc (27), ohmic polarization internal resistance R po (67), electrolyte polarization resistance R pl (65), electrode polarization resistance R pe (66), electrolyte polarization capacitance C pl (63), electrode polarization capacitance C pl (64), cell terminal voltage U ct (68), the cell measurement current I m (69).

[0109] The advantages of using the homogeneous film state and its numerical network observation method proposed in the present invention for battery safety characterization or prediction are:

[0110] (1) The insulation resistance of the battery separator is used to directly characterize the safety of the battery. The fundamental reason for battery thermal runaway is the destruction of the insulation capacity of the separator. If the insulation resistance of the separator can be monitored in real time, it is equivalent to predicting battery thermal runaway in real time.

[0111] (2) Using the uniform insulation state to characterize the battery safety status. Under normal circumstances, the insulation resistance of the battery diaphragm is infinite, which is very difficult to detect in real time and difficult to calculate using software algorithms. Therefore, the present invention constructs a battery uniform insulation resistance composed of an insulating film resistance and a balancing resistance in parallel. Although the insulating film resistance is infinite, the balancing resistance is a finite and determined value, and the uniform insulation resistance after parallel connection is also a finite and determined value. When the insulating film resistance is infinite, the uniform insulation resistance is infinitely close to the balancing resistance. When the insulating film resistance becomes smaller, the uniform insulation resistance begins to decrease, and the measured uniform insulation resistance value can be used to reversely calculate the specific value to which the insulating film resistance drops. The present invention uses the percentage value between the uniform insulation resistance and the balancing resistance to specifically quantify the battery safety status, and can predict in real time the changes in the battery's thermal runaway safety level.

[0112] (3) Using mathematical elements that can only perform simple operations to simulate human neurons. Human neurons are the most important cells for human thinking and reasoning. However, the ability of a single neuron is not as complex as that of neurons in current neural networks, which can perform such complex activation functions. The present invention only considers that human neurons can only perform simple mathematical operations such as addition, subtraction, multiplication, division, exponentiation, square root, and logarithm. Therefore, the present invention simplifies human neurons into a type of mathematical element that can only perform simple mathematical operations.

[0113] (4) Using mathematical networks to simulate system modeling and control. By using mathematical elements that can perform basic mathematical operations to establish the network model and control algorithm of the controlled system, all the parameters of the controlled system are distributed in this network model. The controlled system can learn and work at the same time, and the results are fed back for further learning iteration, which solves the problem that existing neural networks can only learn first and then work.

Claims

1. A method for observing the uniform edge film state and its numerical network for characterizing battery safety, characterized by: The method includes five key steps, namely, membrane resistance circuit design (1), data network design (2), uniform membrane resistance estimation (3), uniform membrane state calculation (4), and safety state prediction (5), as well as modules including uniform membrane resistance circuit (10), data element (8), uniform membrane resistance (13), uniform membrane state (40), and safety state (50). Based on the above modules, a method for battery safety state prediction (5) is constructed, which is divided into the following five steps: Step 1: Design the membrane resistance circuit (1), establish a battery uniform membrane resistance circuit (10) model, and deconstruct the relationship between the battery safety state (50) and the uniform membrane state (40); Step 2: Design of the data network (2). Design of the battery data network (20) based on the battery uniform film resistance circuit (10) model to solve the problem of theoretical design of the data network for simulating the dynamics and control of the battery system. Step 3, average film resistance estimation (3), using a mathematical network (20) to estimate the approximate value of the battery average film resistance (13), solving the problem of real-time acquisition of the average film resistance (13) value that characterizes the insulation resistance of the battery separator; Step 4, average film state calculation (4), calculates the average film state (40) according to the average film resistance (13), and solves the problem of real-time acquisition of the average film state (40) value of the battery; Step 5, safety state prediction (5), inferring the battery safety state (50) according to the uniform membrane state (40), solving the problem of real-time prediction of battery safety state; The value of the uniform film state (40) is used to predict whether the battery safety state (50) is in a film insulation safety state (53), a thermal runaway critical state (54), or a thermal runaway risk state (55).

2. The uniform-edge film resistor circuit (10) according to claim 1, characterized in that: The battery is simplified into an equivalent circuit in which a cell equivalent capacitor (14) and an insulating film resistor (11) are connected in parallel; The cell balancing resistor (12) in the battery management system is introduced into the equivalent circuit to form the cell equivalent capacitance C ce (14), an insulating film resistor circuit (17) in which three electronic components, namely, a battery insulating film resistor (11) and a cell balancing resistor (12), are connected in parallel; The battery insulating film resistor (11) and the cell balancing resistor (12) are in parallel and are combined to simplify into a cell balancing film resistor (13) to form the cell equivalent capacitance C ce (14) A uniform-interference film resistor circuit (10) connected in parallel with the uniform-interference film resistor (13).

3. The counting unit (8) according to claim 1, characterized in that: The receiving dendrite (71) in a human neuron is simulated as an arrow line as a data element input (81); The cell body (72) of a human neuron is simulated as a circle as the operator (82) of the mathematical unit; Simulating the nucleus (73) in the cell body of a human neuron into a mathematical operator (83), including arithmetic operators, algebraic operators, logical operators, set operators, matrix operators, calculus operators, etc.; The axon of a human neuron (74) is simulated as a thick solid line to represent the output of the mathematical operation (84); The myelin sheath (75) of a human neuron is simulated as a solid circle, a solid ellipse, and a hollow circle. The solid ellipse is used as a calculation storage node (85) for storing calculation results, the solid circle is used as a data storage node (88) after data weighting processing, and the hollow circle is used as a weight storage node (86). The synapses (76) of human neurons are simulated as two types of thin solid arrow lines, one for weight input (87) and the other for data output (89).

4. The data network (20) according to claim 1, characterized in that: The data network (20) is composed of a twin data network (23) and a control data network (25), wherein the twin data network (23) is responsible for simulating battery system dynamics and system parameter estimation, and the control data network (27) is responsible for battery state estimation and prediction; The digital network model (22) integrates the uniform film resistance circuit (10) model, the battery equivalent circuit model (24), the model-based observer (30) algorithm, etc. to form a complete battery network model; The twin digital network model (23) is composed of a battery equalizing circuit (10) model and a battery equivalent circuit model (24); The parameters in the twin numerical network (23) all have initial values, and after the first network update, the system control and estimation are performed each time according to the updated parameters; The control mathematical network model (25) is composed of an observer algorithm network (31) and an observer parameter network (32) of a model-based observer (30).

5. The battery uniform film resistor (13) according to claim 1, characterized in that: Based on the cell balancing resistor R in the acquisition board (63) of the battery management system (61) cb (12), and the battery insulation film resistance R mi (11) In parallel, the cell's uniform film resistance R mib (13), its definition expression is, R mib =R cb R mi / (R cb +R mi ).

6. The battery uniform edge film (40) according to claim 1, characterized in that: The so-called battery uniform film state S mib (40) is the cell average film resistance R obtained by observing the numerical network (20) mib (13) and the cell balancing resistor R cb (12) is a percentage value, and its definition expression is, S mib =(R mib / R cb )×100%.

7. The battery safety status (50) according to claim 1, characterized in that: Based on the uniform edge membrane state safety threshold S tms (51), when the calculated cell uniform membrane state S mib (40) Satisfy S mib ≥S tms , it can be predicted that the battery is in a membrane insulation safe state (53); Based on the uniform edge membrane state safety threshold S tms (51) and the critical threshold S of the homogeneous membrane state tmc (52), when the calculated cell uniform membrane state S mib (40) Satisfy S tms >S mib ≥S tmc , it can be predicted that the battery is in a critical state of thermal runaway (54); Based on the critical threshold S of the film resistance state tmc , when the calculated battery uniform membrane state S mib (40) Satisfy S mib tmc , it can be predicted that the battery is in a state of thermal runaway risk (55).​

Images