Lithium ion battery thermal runaway modeling method and system

Abstract

The invention relates to the technical field of lithium ion batteries, in particular to a modeling method and system for thermal runaway of a lithium ion battery. The method comprises the following steps: acquiring real-time temperature data and real-time current data of each battery cell in a battery pack; filtering the real-time temperature data to generate standardized temperature data; calculating a temperature difference value between adjacent battery cells according to the standardized temperature data, calculating heat flow in combination with the real-time current data, and generating heat flow distribution data between the battery cells; calculating a heat exchange efficiency index of each adjacent battery cell pair according to the heat flow distribution data between the battery cells; and establishing a heat exchange efficiency reference value based on the historical operation data of the battery pack, comparing the real-time heat exchange efficiency index with the heat exchange efficiency reference value, calculating the deviation degree, and generating heat exchange efficiency deviation characteristic data. According to the method, the state change of the battery cell is subjected to early recognition by utilizing heat exchange efficiency deviation analysis, so that the early warning of the thermal runaway risk is realized, and the occurrence of battery safety accidents is effectively prevented.

Description

Technical Field

[0001] This invention relates to the field of lithium-ion battery technology, and in particular to a modeling method and system for thermal runaway in lithium-ion batteries. Background Technology

[0002] Lithium-ion battery thermal runaway is an uncontrolled exothermic reaction that occurs when a battery is subjected to external or internal factors. External mechanical abuse (such as impact or crushing), electrical abuse (such as overcharging or over-discharging), or thermal abuse (such as high-temperature environments) can all disrupt the battery's internal balance and trigger thermal runaway. During thermal runaway, the SEI film first decomposes, and the negative electrode, losing its protection, reacts with the electrolyte, releasing heat. As the temperature rises, the separator melts, causing direct contact between the positive and negative electrodes, leading to an internal short circuit. At this point, the positive electrode material decomposes at high temperatures, releasing oxygen, which reacts violently with the negative electrode material and electrolyte, generating a large amount of heat and gas. This causes a rapid increase in internal battery pressure, potentially leading to battery ejection, fire, or even explosion, while releasing toxic and harmful gases such as carbon monoxide, carbon dioxide, and hydrogen fluoride. Once thermal runaway begins, it can trigger a chain reaction, releasing enormous amounts of energy and posing a serious threat to personnel and property safety, as well as causing environmental pollution. The thermal runaway problem of lithium-ion batteries is particularly noteworthy in applications such as electric vehicles and energy storage systems.

[0003] However, current modeling methods for thermal runaway in lithium-ion batteries often fail to dynamically reflect parameter fluctuations and adjacent thermal interference during actual operation, lacking the ability to finely characterize heat propagation behavior in complex battery packs. Furthermore, the topological characteristics of heat conduction paths are closely related to the cell arrangement; traditional methods neglect the spatial connections between cells, resulting in weak identification capabilities for heat diffusion directions. Summary of the Invention

[0004] Therefore, it is necessary for the present invention to provide a modeling method and system for thermal runaway of lithium-ion batteries to solve at least one of the above-mentioned technical problems.

[0005] To achieve the above objectives, a modeling method for thermal runaway in lithium-ion batteries includes the following steps: Step S1: Obtain real-time temperature data and real-time current data of each cell in the battery pack; filter the real-time temperature data to generate standardized temperature data; calculate the temperature difference between adjacent cells based on the standardized temperature data, and calculate the heat flow in combination with the real-time current data to generate heat flow distribution data between cells. Step S2: Calculate the heat exchange efficiency index of each adjacent cell pair based on the heat flow distribution data between cells, where the heat exchange efficiency index is the ratio of heat flow to temperature difference; establish the heat exchange efficiency benchmark value of each cell pair under stable operating conditions based on the historical operating data of the battery pack, and compare the real-time heat exchange efficiency index with the heat exchange efficiency benchmark value to calculate the degree of deviation and generate heat exchange efficiency deviation characteristic data. Step S3: Calculate the cell state probability based on the heat exchange efficiency deviation characteristic data, where the cell state includes normal state, warning state, and critical state; perform an influence analysis on adjacent cells on the cell state probability to generate cell state probability distribution data; perform probability weighted calculation of heat transfer parameters based on the cell state probability distribution data to generate dynamic heat transfer parameter data. Step S4: Construct a battery pack heat transfer network model based on dynamic heat transfer parameter data; identify the critical path of heat propagation through the battery pack heat transfer network model, and solve the temperature evolution equation along the critical path using the finite difference method to generate a dynamic prediction model for thermal runaway of lithium-ion batteries.

[0006] Preferably, step S1 includes the following steps: Step S11: Collect real-time temperature and current information of each cell in the battery pack, set the sampling frequency to once every five seconds, and record the timestamp of data collection to obtain real-time temperature data and real-time current data. Step S12: Perform moving average filtering on the real-time temperature data, and use the arithmetic mean of five consecutive sampling points as the filtered temperature value at the current moment to generate standardized temperature data; Step S13: Determine the adjacency relationship of each cell based on the pre-acquired physical layout of the battery pack, and establish a cell adjacency table; Step S14: Calculate the temperature difference between each pair of adjacent cells based on standardized temperature data to obtain the temperature difference between adjacent cells; Step S15: Calculate the instantaneous heat generation power of each cell based on Ohm's law using real-time current data, distribute the instantaneous heat generation power of each cell to adjacent cells based on the heat transfer path, and combine the conduction heat flow generated by the temperature difference between adjacent cells to obtain the total heat flow. Step S16: Organize the total heat flow of each adjacent cell pair based on the spatial layout of the battery pack, and establish a heat flow matrix with cell number as row and column index, where the matrix elements of the heat flow matrix represent the heat flow values ​​between corresponding cell pairs. Step S17: Perform data standardization processing on the heat flow matrix, convert the heat flow values ​​into uniform physical units, and generate heat flow distribution data between battery cells.

[0007] Preferably, step S2 includes the following steps: Step S21: Extract the total heat flow value of each adjacent cell pair from the heat flow distribution data between cells, establish a heat flow data table, and extract the temperature difference value of the corresponding cell pair to establish a temperature difference data table. Step S22: Calculate the heat exchange efficiency based on each pair of adjacent cells according to the heat flow data table and the temperature difference data table, and obtain the heat exchange efficiency index, where the heat exchange efficiency is the ratio of heat flow to temperature difference. Step S23: Arrange and organize the real-time heat exchange efficiency indicators of all adjacent cell pairs according to the cell number order to establish a real-time heat exchange efficiency indicator data table; Step S24: Retrieve the historical temperature and current records of the battery pack over the past thirty days, and filter out stable operating periods where the ambient temperature change is less than 5°C and the charge / discharge current change is less than 10% of the rated current. Step S25: Calculate the historical heat exchange efficiency index of each cell pair under stable operating conditions based on the selected stable operating period data; Step S26: Take the arithmetic mean of multiple historical heat exchange efficiency indices for each pair of adjacent cells to obtain the baseline value of heat exchange efficiency; Step S27: Compare the real-time heat exchange efficiency index with the heat exchange efficiency benchmark value, calculate the degree of deviation, and generate heat exchange efficiency deviation characteristic data.

[0008] The present invention also provides a modeling system for thermal runaway of lithium-ion batteries, used to execute the above-described modeling method for thermal runaway of lithium-ion batteries, wherein the modeling system for thermal runaway of lithium-ion batteries includes: The heat flow data acquisition module is used to acquire real-time temperature data and real-time current data of each cell in the battery pack; filter the real-time temperature data to generate standardized temperature data; calculate the temperature difference between adjacent cells based on the standardized temperature data, and calculate the heat flow in combination with the real-time current data to generate heat flow distribution data between cells. The efficiency deviation analysis module is used to calculate the heat exchange efficiency index of each adjacent cell pair based on the heat flow distribution data between cells. The heat exchange efficiency index is the ratio of heat flow to temperature difference. Based on the historical operating data of the battery pack, a benchmark value of the heat exchange efficiency of each cell pair under stable operating conditions is established. The real-time heat exchange efficiency index is compared with the benchmark value of heat exchange efficiency to calculate the degree of deviation and generate heat exchange efficiency deviation characteristic data. The efficiency deviation analysis module is used to calculate the cell state probability based on heat exchange efficiency deviation characteristic data, where the cell state includes normal state, warning state and critical state; perform adjacent cell influence analysis on the cell state probability to generate cell state probability distribution data; and perform probability weighted calculation of heat transfer parameters based on the cell state probability distribution data to generate dynamic heat transfer parameter data. The dynamic modeling generation module is used to construct a battery pack heat transfer network model based on dynamic heat transfer parameter data; the critical path of heat propagation is identified through the battery pack heat transfer network model, and the temperature evolution equation is solved along the critical path using the finite difference method to generate a dynamic prediction model for thermal runaway of lithium-ion batteries.

[0009] This invention, starting with multi-source data fusion, constructs a dynamic, accurate, and predictive temperature evolution and risk assessment system, providing a systematic and intelligent solution for battery safety management. By collecting real-time temperature and current data from the cells during battery pack operation and establishing a high-frequency sampling mechanism, it ensures the ability to capture detailed thermal changes, meeting the accuracy requirements for modeling in complex thermal environments. Standardized temperature data obtained after moving average filtering of the temperature signals not only improves noise immunity but also provides a stable input foundation for subsequent thermal differential and heat flow analysis. A cell adjacency table established using prior physical layout information tightly couples heat flow modeling with spatial structure, ensuring the effective reproduction of the actual heat transfer path in the modeled structure. Based on this, Ohm's law is introduced to quantify the instantaneous heat generation behavior of the cells, and combined with heat conduction theory, the distribution and flow of heat between adjacent cells are accurately simulated, further constructing a heat flow matrix that systematically characterizes the thermal dynamics of the entire battery pack. Subsequently, standardization processing unifies the physical units of the heat flow data, providing data assurance for constructing a horizontally comparable heat conduction analysis system. The heat exchange efficiency between adjacent cells is calculated using the ratio of heat flow to temperature difference. While recording heat exchange efficiency at multiple time points, a structured real-time heat exchange efficiency index data table is simultaneously organized, providing a quantitative basis for subsequent abnormal state identification. The method also proactively incorporates stable operating condition data from historical phases, eliminating interference from ambient temperature fluctuations and drastic changes in charging and discharging current. This ensures that the extracted heat exchange efficiency benchmark value better represents the long-term heat transfer behavior between cells, improving the model's long-term stability assessment capability. During the comparison of real-time heat exchange efficiency values ​​with benchmark values, the system establishes a deviation level classification mechanism, introducing graded evaluation rules to directly link the trend of heat exchange efficiency changes with abnormal state identification, providing an effective approach for early fault warning. Furthermore, by combining the deviation level with the state probability mapping relationship and constructing a cell state probability distribution, the system probabilistically expresses the thermal safety state of each cell. This not only overcomes the shortcomings of static judgment in traditional thermal management strategies but also possesses the ability to dynamically track the evolution of thermal runaway. Building upon this foundation, heat sources are identified for battery cells with high critical state probabilities, establishing a list of heat source cells. Combined with cell heat flow data, the thermal conductivity intensity and risk propagation impact on surrounding cells are precisely calculated. This operation extends local thermal risk prediction to global thermal evolution prediction, constructing a risk amplification model oriented towards propagation mechanisms. Subsequently, the method applies the risk propagation impact value of heat source cells to non-heat source cells, further refining their state probabilities and forming a dynamic state expression influenced by environmental conduction, enabling the system to reflect risk diffusion in the spatial dimension. The cell state probabilities are then used for dynamic weighted calculation of thermal conductivity and specific heat capacity, constructing the dynamic thermal property parameters corresponding to the cell at the current moment.These dynamic parameters not only improve the thermal model's responsiveness to real-world physical changes but also provide a thermophysical foundation for high-precision thermal prediction over time and state evolution. By establishing a network topology framework based on node attributes and connectivity, and assigning dynamic thermal parameters to the connecting edges in the graph structure, the system further forms a heat transfer network model, which is then used for critical path analysis and heat propagation channel identification. The extraction process of priority path identification and critical node groups constructs a dynamic identification mechanism for high-risk propagation channels in space, and based on this result, a spatial finite difference model is constructed. A high-resolution thermal evolution simulation field is built through spatial gridding and temporal discretization. Finally, the system simulates the heat propagation process through a system of difference equations, identifies the temperature evolution trend under specific paths, and makes real-time predictions of thermal runaway risk accordingly. To apply the model to real-time online monitoring and early warning systems, the method also designs a temperature evolution solver with independent computational capabilities and encapsulates it as a modular component for easy integration and deployment. In addition, by establishing a risk assessment rule base, the temperature threshold judgment, risk level classification, and heat propagation laws are standardized, enabling the system to have adjustable and updatable risk discrimination capabilities. Ultimately, the system integrates the temperature solver, risk assessment rules, and dynamic thermal parameters into a complete modeling architecture. This not only enables full-cycle dynamic simulation of the thermal runaway process of battery cells but also allows for automatic updates of thermal property parameters based on operational data, truly constructing an intelligent thermal management model that integrates prediction, judgment, and response. This method is particularly suitable for applications requiring high-reliability thermal safety monitoring, such as new energy vehicle battery packs, stationary energy storage battery arrays, and spacecraft power compartments, demonstrating significant value in reducing the risk of sudden thermal runaway and improving the system's stable operation assurance capabilities. Attached Figure Description

[0010] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a schematic flowchart illustrating the steps of a lithium-ion battery thermal runaway modeling method according to the present invention. Figure 2 for Figure 1 A detailed flowchart of step S1; Figure 3 for Figure 1 A detailed flowchart of step S2. Detailed Implementation

[0011] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0012] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0013] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0014] To achieve the above objectives, please refer to Figures 1 to 3 This invention provides a modeling method for thermal runaway in lithium-ion batteries, the method comprising the following steps: Step S1: Obtain real-time temperature data and real-time current data of each cell in the battery pack; filter the real-time temperature data to generate standardized temperature data; calculate the temperature difference between adjacent cells based on the standardized temperature data, and calculate the heat flow in combination with the real-time current data to generate heat flow distribution data between cells. This invention employs a thermocouple temperature sensor array and a Hall current sensor integrated within the lithium-ion battery module to collect real-time data on the surface temperature of each cell and the operating current flowing through each cell. The acquisition frequency is set to 1Hz to meet thermal dynamic response requirements. Since the temperature sensors may be affected by electromagnetic interference or instantaneous fluctuations during operation, the temperature data needs to be filtered. A sliding window mid-range filtering method is used here, with a window size of 5 seconds, to smooth and denoise the time-series temperature data of each cell, obtaining stable and reliable standardized temperature data. Based on this, for physically adjacent cells within the same module, their temperature differences are calculated sequentially; for example, the standardized temperature difference between cells numbered i and i+1 is calculated. Combining the average operating current of the two cells, and utilizing the positive correlation between heat flow and Joule heat generated by current in the law of thermal convection, the heat transfer per unit time between the cell pair is estimated, further forming inter-cell heat flow distribution data reflecting the overall heat flow conduction relationship within the module. This heat flow data is stored in a two-dimensional matrix with cell pairs as basic units, and each matrix unit contains information such as heat flow, temperature difference, and location identifier.

[0015] Step S2: Calculate the heat exchange efficiency index of each adjacent cell pair based on the heat flow distribution data between cells, where the heat exchange efficiency index is the ratio of heat flow to temperature difference; establish the heat exchange efficiency benchmark value of each cell pair under stable operating conditions based on the historical operating data of the battery pack, and compare the real-time heat exchange efficiency index with the heat exchange efficiency benchmark value to calculate the degree of deviation and generate heat exchange efficiency deviation characteristic data. Based on the heat flow distribution data obtained in S1, this embodiment of the invention calculates the heat exchange efficiency index per unit temperature difference for each pair of physically adjacent cells (e.g., cell i and cell i+1) according to the ratio of heat flow to temperature difference. To evaluate the rationality of this index, historical operating data is introduced to establish a benchmark heat exchange efficiency value under stable operating conditions. Historical data is collected from operating periods with small temperature gradients and stable heat dissipation patterns under long-term constant current charge-discharge conditions. After data filtering, approximately 100 hours of operating data are extracted, and the average heat exchange efficiency of each pair of cells is statistically analyzed as its benchmark value. Then, the real-time heat exchange efficiency is compared with its historical benchmark, and the deviation percentage is used to represent the degree of deviation. The calculation method is to subtract the benchmark value from the real-time value and then divide by the benchmark value. This deviation result is then combined with time series smoothing to finally generate heat exchange efficiency deviation feature data for each pair of cells. This data is used to identify abnormal heat conduction trends in the subsequent modeling stage.

[0016] Step S3: Calculate the cell state probability based on the heat exchange efficiency deviation characteristic data, where the cell state includes normal state, warning state, and critical state; perform an influence analysis on adjacent cells on the cell state probability to generate cell state probability distribution data; perform probability weighted calculation of heat transfer parameters based on the cell state probability distribution data to generate dynamic heat transfer parameter data. This invention uses the heat exchange efficiency deviation characteristic data obtained in the previous step as input, and combines it with a preset multi-state Bayesian probability model to evaluate the state of each battery cell. The model assumes that a battery cell may be in three states: normal state (stable heat conduction, deviation less than 10%), warning state (deviation between 10% and 30%), and critical state (deviation exceeding 30%). By establishing a conditional probability table and combining the heat exchange deviation data between each battery cell and its neighboring cells, the probability of each cell being in each state is calculated. Considering that heat conduction is a mutually influential process, a Markov random field method is used to further correct the state probability of individual battery cells, making it spatially reflect the linkage relationship with the thermal states of surrounding battery cells. Based on this, a state probability distribution map of the entire group of battery cells is constructed. On this basis, key parameters of heat transfer between each battery cell (such as thermal conductivity and heat capacity) are solved using probability weighting, that is, the parameters corresponding to each state are assigned values ​​and averaged according to probability weighting to obtain more dynamic heat transfer parameters reflecting the current thermal state, forming dynamic heat transfer parameter data, providing physical constraints for subsequent modeling.

[0017] Step S4: Construct a battery pack heat transfer network model based on dynamic heat transfer parameter data; identify the critical path of heat propagation through the battery pack heat transfer network model, and solve the temperature evolution equation along the critical path using the finite difference method to generate a dynamic prediction model for thermal runaway of lithium-ion batteries.

[0018] This invention, based on dynamic heat transfer parameter data, abstracts the heat exchange relationship between battery cells into a node-edge structured heat transfer network model. Each battery cell corresponds to a node in the network, and the heat exchange relationship between adjacent cells forms a weighted edge, with the edge weight representing the real-time heat exchange efficiency. During network construction, a directed weighted graph method from graph theory is used to represent the propagation of heat from high to low temperature. Then, path analysis is performed on this heat transfer network to calculate the propagation path with the highest heat flux density and heat exchange efficiency, identifying key channels that may lead to thermal runaway. For these key paths, a finite difference heat conduction model is further established to construct a dynamic evolution equation for the cell temperature over time. Spatially, the grid is divided according to the cell arrangement structure; temporally, the calculation step size is set to 1 second, continuously calculating the temperature change trend for 10 minutes. Combining the dynamic heat capacity and thermal conductivity parameters of each node, the forward Euler method is used to iteratively solve the heat conduction equation, thus forming the final dynamic prediction model for lithium-ion battery thermal runaway. This model can simulate the thermal diffusion trend of battery modules under different initial temperature fields and different discharge rates in real time, providing a predictive basis for the vehicle BMS system to identify thermal runaway risks in advance.

[0019] In this embodiment, the temperature and current data of each individual cell in the battery pack are first collected in real time. The sampling frequency is set to 2Hz, the temperature detection accuracy is ±0.2℃, and the current detection accuracy is ±0.05A. After standardization, the collected multi-point temperature sequences are used to calculate the transient heat flux density distribution on the cell surface using Fourier's law of heat conduction. The basic relationship is as follows: ;in, For heat flux density, The equivalent thermal conductivity of the battery cell material. Let be the temperature distribution function along the radial direction. By calculating the temperature difference at different locations, the temperature gradient between the surface and interior of the battery cell can be obtained, and the local heat flow distribution can be calculated accordingly. To verify the feasibility of this calculation method, an experimental test was conducted on a ternary lithium battery cell with a rated capacity of 2.6Ah under a constant temperature environment of 25℃. The experiment compared four charging rate conditions: 1C, 2C, 3C, and 4C. A total of five thermocouples were arranged on the surface of the battery cell to obtain temperature distribution data, and the corresponding current input was recorded. The typical heat flow distribution results obtained are shown in Table 1.

[0020] Charging rate Maximum temperature gradient (°C / cm) <![CDATA[Heat flux density (W / m 2 )]]> Calculation error (%) 1C 0.8 450 4.2 2C 1.4 780 3.8 3C 2.6 1240 4.5 4C 3.9 1850 4.9 Table 1 The results show that the calculated heat flux density and the measured thermal image results have an error of less than 5%, indicating that the heat flux distribution calculation method in this embodiment can effectively reflect the actual thermal diffusion state of the battery cell surface.

[0021] Preferably, step S1 includes the following steps: Step S11: Collect real-time temperature and current information of each cell in the battery pack, set the sampling frequency to once every five seconds, and record the timestamp of data collection to obtain real-time temperature data and real-time current data. In this embodiment of the invention, a series module consisting of 96 lithium-ion cells is used for thermal runaway modeling data acquisition in a battery pack for a certain type of electric vehicle. Each cell is equipped with a high-precision NTC thermistor temperature sensor (temperature acquisition accuracy ±0.2℃), and the sampling current connected in series with the cell is acquired by a Hall current sensor (current measurement range 0~300A, accuracy ±1%). The acquisition cycle is set to 5 seconds, meaning that the temperature of all cells and the current flowing through the entire group are sampled uniformly every 5 seconds, while simultaneously recording a Unix timestamp for each set of sampled data. This data acquisition is automatically executed by an embedded BMS, without manual operation. The acquired data is automatically written to the embedded system's circular buffer and synchronously transmitted to a remote industrial PC for subsequent processing. Each frame of raw data includes: cell number, cell temperature (℃), sampling current (A), and timestamp (seconds).

[0022] Step S12: Perform moving average filtering on the real-time temperature data, and use the arithmetic mean of five consecutive sampling points as the filtered temperature value at the current moment to generate standardized temperature data; This embodiment of the invention performs preliminary smoothing processing on the cell temperature data obtained in step S11 to eliminate occasional instantaneous measurement fluctuations. The system adopts a 5-point moving average method, that is, for any time t, the temperature values ​​at five time points (t, t-5s, t-10s, t-15s, and t-20s) under the current cell number are extracted, and their arithmetic mean is calculated as the standard temperature at the current time. This processing is implemented in a Python script based on the Pandas data structure, with a sliding window length of 5, and the processing method is time series window moving average. The filtered data is uniformly labeled as "standardized temperature data" and used for further temperature difference calculation. The processed temperature data is more stable and can effectively reflect the heat transfer trend rather than measurement noise in dynamic driving scenarios such as acceleration, hill climbing, or fast charging.

[0023] Step S13: Determine the adjacency relationship of each cell based on the pre-acquired physical layout of the battery pack, and establish a cell adjacency table; According to the physical structure diagram of the battery module, the battery cells are arranged in a two-dimensional array of 12 columns × 8 rows. The system parses the spatial adjacency relationships of all battery cells through a preset structure mapping file (JSON format) and constructs a dictionary structure called an "adjacency table," where the key is the battery cell number and the value is an array containing the numbers of its adjacent battery cells. For example, the adjacency table records for battery cell number 35 are [27, 28, 34, 36, 43, 44]. The adjacency criteria are: the center distance between two battery cells is within 10 mm, and they share at least one surface or casing with a contact area exceeding 3 square centimeters. This adjacency table is used to guide the establishment of heat flow direction and conduction path, and is automatically derived by the program without manual calibration. Step S14: Calculate the temperature difference between each pair of adjacent cells based on standardized temperature data to obtain the temperature difference between adjacent cells; In this embodiment of the invention, based on the standardized temperature data obtained in S12 and the adjacency list constructed in S13, the system traverses all adjacent cell pairs, reads their current temperature values ​​for each pair, and calculates the difference (using the absolute value of the higher-temperature cell minus the lower-temperature cell, preserving the thermal gradient direction information). The difference is marked as "". "The data is recorded as a triplet structure: <cell number i, cell number j, Value >. Taking cells 34 and 35 as examples, if their temperatures are 39.2℃ and 41.8℃ respectively, then the pair's The T value is 2.6℃. These temperature differences are used for heat conduction modeling and are important physical variables driving heat flow.

[0024] Step S15: Calculate the instantaneous heat generation power of each cell based on Ohm's law using real-time current data, distribute the instantaneous heat generation power of each cell to adjacent cells based on the heat transfer path, and combine the conduction heat flow generated by the temperature difference between adjacent cells to obtain the total heat flow. In this embodiment of the invention, the heat generation power of each battery cell is calculated using the Ohmic heat estimation formula. The model uses a given value: I represents the real-time current value in S11, and R represents the internal resistance of the battery cell (obtained through factory calibration or an online identification module, typically 1.5 milliohms). For example, if a battery cell has a current of 100A, the heat generation power is 100 × 100 × 0.0015 = 15W. This heat generation power is distributed to adjacent battery cells via the cell's heat conduction path. The distribution ratio for each path is set based on the contact area and structural thermal resistance, and the system has an embedded standard thermal resistance matrix library. On the other hand, the heat flow between the battery cell pairs is calculated based on the temperature difference ΔT in S14. Its thermal conductivity is limited by the contact thermal conductivity coefficient (typically ranging from 2 to 10 W / m·K) and thermal resistance. After merging the two heat sources, the instantaneous total heat flow of the battery cell pair is obtained, expressed in watts (W).

[0025] Step S16: Organize the total heat flow of each adjacent cell pair based on the spatial layout of the battery pack, and establish a heat flow matrix with cell number as row and column index, where the matrix elements of the heat flow matrix represent the heat flow values ​​between corresponding cell pairs. In this embodiment of the invention, the heat flow between all adjacent cell pairs is organized into a two-dimensional heat flow matrix with a size of 96×96, where each row and column corresponds to a cell number. Heat flow values ​​are entered only at positions where there are directly physically adjacent cell pairs (refer to S13), with the remaining positions being empty or set to zero, forming a sparse matrix. Matrix elements are recorded as floating-point values ​​(in W), such as matrix

[34]

[35] =3.6, indicating a heat transfer of 3.6W between cells 34 and 35. This matrix structure is stored using a NumPy array and graphically presented using a heat transfer network visualization tool, providing a structural foundation for the next step of constructing heat transfer paths.

[0026] Step S17: Perform data standardization processing on the heat flow matrix, convert the heat flow values ​​into uniform physical units, and generate heat flow distribution data between battery cells.

[0027] This invention addresses the inconsistency in units used in heat flux calculations. For example, some heat conduction paths originate from structural simulation models (unit: Joules / second), requiring a unified conversion to Watts (1 J / s = 1 W). Simultaneously, all heat flux matrices are standardized to a uniform dimensional range (e.g., through normalization or Z-score processing). The system employs a physical quantity unit conversion interface to unify all heat flux values ​​into Watt units, followed by standard deviation normalization to ensure comparability of heat fluxes between different cells, facilitating subsequent deviation analysis and state identification. The standardized heat flux matrix constitutes the "inter-cell heat flux distribution data," which serves as the direct input for the next stage of heat exchange efficiency modeling and thermal runaway modeling.

[0028] In this embodiment, the thermal runaway prediction model is constructed based on multi-dimensional features such as the cell temperature rise rate, heat flux gradient, and current density change rate. The cell state probability distribution is calculated using Bayesian statistics and time-series regression methods, defined as follows: ;in, This represents the probability that the battery cell is in a stable, overheated, or runaway state. The state transition curve output by the model can reflect the thermal stability of the battery cell in real time.

[0029] In the experiment, measured data at different magnifications were input into the model, and the comparison results between the predicted runaway probability and the measured temperature anomaly time are as follows: Operating conditions Model predicts peak probability of runaway Measured temperature rise rate (°C / s) Forecast lead time (s) 2C charging 0.25 0.12 — 3C charging 0.78 0.35 90 4C charging 0.92 0.52 120 Table 2 As shown in Table 2, at 3C and 4C ratios, the model predicts the probability of runaway from the actual temperature anomaly by about 90 to 120 seconds, indicating that the model has a good early warning capability.

[0030] Based on the comprehensive experimental data, the method described in this embodiment, through a combination of physical modeling and data-driven calculation, can not only accurately reconstruct the heat flow distribution and heat exchange efficiency of the battery cell, but also effectively predict thermal runaway trends. The results are highly consistent with actual temperature measurements and infrared thermography verification, proving that the present invention has a reliable physical basis and is feasible.

[0031] Preferably, step S2 includes the following steps: Step S21: Extract the total heat flow value of each adjacent cell pair from the heat flow distribution data between cells, establish a heat flow data table, and extract the temperature difference value of the corresponding cell pair to establish a temperature difference data table. Based on the inter-cell heat flow distribution data generated in step S17, the system extracts the heat flow value between each pair of physically adjacent cells according to the cell adjacency table (see S13). The extraction method involves reading non-zero elements in the heat flow matrix pair by pair, parsing their row and column indices to obtain the cell number pair, and the value representing the total heat flow (in watts) between that pair. The system stores this result in a table with header fields of "Cell Number 1, Cell Number 2, Total Heat Flow". Simultaneously, the system retrieves the temperature difference value corresponding to each pair of cells from the temperature difference table recorded in S14 and establishes another corresponding "Temperature Difference Data Table" with structure fields of "Cell Number 1, Cell Number 2, Temperature Difference". These two data tables will be used for subsequent heat exchange efficiency calculations, and the data organization facilitates parallel processing and historical comparison.

[0032] Step S22: Calculate the heat exchange efficiency based on each pair of adjacent cells according to the heat flow data table and the temperature difference data table, and obtain the heat exchange efficiency index, where the heat exchange efficiency is the ratio of heat flow to temperature difference. In this embodiment of the invention, the system utilizes the heat flow data table and temperature difference data table generated in step S21. The system iterates through the cell pair records in both tables row by row. For each cell pair, it reads the corresponding heat flow value and temperature difference value, and divides the two to obtain the heat exchange efficiency of that cell pair. The physical meaning of heat exchange efficiency is the heat transfer capacity per unit temperature difference, measured in watts per degree Celsius, reflecting the strength of heat conduction performance. To avoid division by zero when the temperature difference is zero, the system skips the calculation or sets it to an invalid value when the absolute value of the temperature difference is less than 0.1°C. In practical applications, if the cell pair shows significant fluctuations in heat exchange efficiency, it may indicate risks such as loose structural fastening, aging of the thermal interface, or contact failure. All calculation results are recorded in the format "Cell No. 1, Cell No. 2, Heat Exchange Efficiency," and the output is an intermediate result table, laying the foundation for subsequent real-time comparison with a benchmark.

[0033] Step S23: Arrange and organize the real-time heat exchange efficiency indicators of all adjacent cell pairs according to the cell number order to establish a real-time heat exchange efficiency indicator data table; To facilitate the traceability, tracking, and visualization of heat exchange efficiency indicators at different time points, this invention arranges the heat exchange efficiency values ​​of all cell pairs in ascending order of cell number and constructs a "real-time heat exchange efficiency indicator data table." This table is organized by rows, with each row recording information for one cell pair, including cell number 1, cell number 2, heat exchange efficiency, and timestamp. The data table supports high-frequency refresh (e.g., updating every 5 seconds) and has historical timeline tagging functionality. The system can store this data in CSV format in a database, or use it as input for subsequent comparison with reference efficiency under stable operating conditions, identifying abnormal deviations and constructing a thermal runaway early warning model.

[0034] Step S24: Retrieve the historical temperature and current records of the battery pack over the past thirty days, and filter out stable operating periods where the ambient temperature change is less than 5°C and the charge / discharge current change is less than 10% of the rated current. In this embodiment of the invention, to obtain the thermal exchange benchmark behavior of a lithium-ion battery pack under stable operating conditions in the near term, the system automatically retrieves the temperature and current records of the most recent 30 days stored in a historical database. Each historical data entry includes cell temperature, battery pack current, ambient temperature, and a timestamp. The system first divides the data sequence into time windows (each hour is a sample segment), calculates the ambient temperature and current variation within each time period, and filters out time periods where the ambient temperature fluctuation does not exceed 5°C and the current fluctuation is less than 10% of the battery's rated current (e.g., if the rated current is 120A, the current variation does not exceed 12A). These time periods are defined as "stable operating conditions," characterized by stable thermal behavior and minimal external disturbances, making them suitable as a thermal performance benchmark data source. The system then numbers, marks, and extracts these data segments for further calculations.

[0035] Step S25: Calculate the historical heat exchange efficiency index of each cell pair under stable operating conditions based on the selected stable operating period data; Based on the stable operating condition data selected in S24, this embodiment of the invention reconstructs the cell temperature sequence, the temperature difference sequence of adjacent cell pairs, and the heat flow matrix within that time period. The same method as in S21 to S22 is used to calculate the heat exchange efficiency of all cell pairs in each time segment. Considering the large data sample size and high calculation frequency, a batch processing method based on Python multithreading is used for heat exchange efficiency batch processing. All historical heat exchange efficiency records are appended with timestamps, ambient temperature, and current reference information, and archived as a "Historical Heat Exchange Efficiency Index Library." This library provides data support for subsequent benchmark value extraction and distribution analysis.

[0036] Step S26: Take the arithmetic mean of multiple historical heat exchange efficiency indices for each pair of adjacent cells to obtain the baseline value of heat exchange efficiency; In this embodiment of the invention, the system aggregates historical heat exchange efficiency index data from the database generated in step S25 by cell pair number, and performs an arithmetic mean on the historical heat exchange efficiency data in each group to obtain the average heat exchange capacity of each cell pair. This average value is defined as the "heat exchange efficiency benchmark value" for that cell pair under normal and stable operating conditions, and its physical meaning is the average heat transfer capacity of that cell pair when there are no thermal anomalies. The system outputs the benchmark values ​​of all cell pairs in a table structure, with fields including cell number 1, cell number 2, and heat exchange efficiency benchmark value, and sets a standard deviation range to determine the degree of fluctuation. This benchmark value table is stored in the model system for a long time as a reference line for subsequent online real-time status evaluation.

[0037] Step S27: Compare the real-time heat exchange efficiency index with the heat exchange efficiency benchmark value, calculate the degree of deviation, and generate heat exchange efficiency deviation characteristic data.

[0038] In this embodiment of the invention, the real-time heat exchange efficiency index at the current moment in S23 is compared one-by-one with the benchmark heat exchange efficiency value in S26. The difference and relative deviation rate between the two are calculated to form "heat exchange efficiency deviation characteristic data". For example, if the real-time heat exchange efficiency of a certain cell pair is 1.6W / ℃ and the benchmark value is 2.0W / ℃, then the deviation is -0.4W / ℃, and the relative deviation rate is -20%. The system classifies the deviations into layers: cell pairs with a deviation rate exceeding ±15% are marked as "items of concern", and those exceeding ±30% are marked as "abnormal items". All deviation data are output in a table, with fields including cell pair number, real-time value, benchmark value, deviation value, deviation rate, and risk level identifier. This data is used for cell state identification and thermal diffusion modeling in stage S3, providing a core basis for predicting the thermal runaway path of lithium-ion batteries.

[0039] Preferably, step S27 includes the following steps: Step S271: Compare the real-time heat exchange efficiency index of the same adjacent cell pair with the heat exchange efficiency benchmark value, and establish a pairing comparison data table of real-time value and benchmark value; In this embodiment of the invention, the system has obtained the baseline value of heat exchange efficiency for each pair of adjacent cells in step S26, and the real-time heat exchange efficiency index at the current moment in step S23. Therefore, the goal of this step is to pair and organize the efficiency data of the same pair of cells in two time dimensions. Specifically, the system uses the cell number pair as the primary key, searches for records of the same cell pair in both data tables, and sequentially establishes a pairing comparison data table between the real-time value and the baseline value. This data table structure includes fields such as "Cell Number 1", "Cell Number 2", "Real-time Heat Exchange Efficiency Value", "Baseline Heat Exchange Efficiency Value", and "Timestamp" to ensure consistent pairing and data alignment. This process is implemented in the system through an automatic data association and matching algorithm, requiring no human intervention, and supports updates every sampling period (e.g., every 5 seconds), forming a real-time tracking capability, serving as the basic data structure for subsequent deviation quantification and level identification.

[0040] Step S272: Calculate the deviation for each set of data in the pairing comparison data table. Calculate the deviation for each set of data. Subtract the corresponding heat exchange efficiency benchmark value from the heat exchange efficiency index to obtain the absolute deviation, and then divide by the heat exchange efficiency benchmark value to obtain the relative deviation percentage. Establish a deviation percentage data table. In this embodiment of the invention, the pairing comparison data table generated in S271 is traversed, and deviation calculation is performed on each record. The deviation calculation includes two quantities: first, the absolute deviation, which is the real-time heat exchange efficiency value minus its corresponding reference value, in watts per degree Celsius, used to determine the magnitude of change in heat conduction strength; second, the relative deviation percentage, which is the absolute deviation divided by the reference value and multiplied by 100, expressed as a percentage to facilitate horizontal comparison between different cell pairs. For example, if the real-time efficiency of a cell pair is 1.7 W / ℃ and the reference value is 2.0 W / ℃, then the absolute deviation is -0.3 and the relative deviation is -15%. The calculation process is automatically completed by the system's built-in mathematical processing module, and the calculation results are written into a new "Deviation Percentage Data Table," whose fields include "Cell Number 1," "Cell Number 2," "Absolute Deviation," "Relative Deviation Percentage," and "Timestamp." This table provides basic data for subsequent graded identification.

[0041] Step S273: According to the preset deviation level standard, the values ​​in the deviation percentage data table are classified. The relative deviation is less than 5% and marked as normal level, the relative deviation is between 5% and 15% and marked as concern level, and the relative deviation is greater than 15% and marked as abnormal level. The corresponding deviation level label is assigned to each cell pair. In this embodiment of the invention, the "relative deviation percentage" field in each record of the deviation percentage data table in S272 is used as the classification basis for leveling based on a preset deviation level standard. The classification standard is as follows: deviations within ±5% are considered to have stable thermal conductivity and are marked as "normal level"; deviations between ±5% and ±15% are considered to have a slight tendency for thermal imbalance and are marked as "concern level"; deviations exceeding ±15% indicate a significant change in the heat exchange performance of the battery cell, which may indicate poor internal contact, heat dissipation attenuation, or structural abnormalities, and are marked as "abnormal level". The system automatically labels each record with a level using a rule engine and adds a "deviation level" field to the data table, outputting a "deviation level data table" to form the data support source for the system's early warning module. This process is fully automated and supports custom threshold configuration and dynamic adjustment.

[0042] Step S274: Organize and arrange the deviation percentage data table and deviation level identifier according to the cell number order to form a structured data record containing deviation value and level information, and generate heat exchange efficiency deviation characteristic data.

[0043] In this embodiment of the invention, the deviation grading data table generated in S273 is further organized and archived. It is sorted in ascending order according to the cell number, and the "relative deviation percentage" and "deviation level" fields are paired and combined with the cell number to form "heat exchange efficiency deviation characteristic data." This structured data is presented in tabular format and saved as a JSON object or database record. Fields include "Cell Number 1," "Cell Number 2," "Deviation Percentage," "Deviation Level," and "Update Time." This data can not only be used for subsequent cell status identification but also coupled with a heat conduction network model to analyze the propagation path and critical point of thermal runaway. In practical applications, such as in new energy vehicle thermal management systems or energy storage power station battery pack monitoring systems, this data can be pushed to a host computer every minute for thermal anomaly visualization and trigger fault warnings or active cooling strategies, supporting the implementation of active safety control mechanisms.

[0044] Preferably, the calculation of the cell state probability based on the heat exchange efficiency deviation characteristic data in step S3 includes: The deviation level identifiers of each adjacent cell pair are extracted from the heat exchange efficiency deviation characteristic data, and a state probability mapping lookup table is established. The state probability mapping lookup table includes the following: normal level corresponds to normal state probability 0.9, warning state probability 0.1, and critical state probability 0.0; attention level corresponds to normal state probability 0.3, warning state probability 0.6, and critical state probability 0.1; and abnormal level corresponds to normal state probability 0.1, warning state probability 0.4, and critical state probability 0.5. In this embodiment of the invention, the records for each pair of adjacent cells in the acquired heat exchange efficiency deviation characteristic data include a deviation percentage and a corresponding deviation level label. The system first extracts the deviation level identifier from these records and organizes them using "Cell Number 1 - Cell Number 2" as the primary key to form an input dataset for state probability mapping. Subsequently, the system constructs a lookup table according to a preset state probability mapping standard. This lookup table is used to convert deviation levels (including "Normal Level," "Attention Level," and "Abnormal Level") into probability distributions of three states (Normal State, Warning State, and Critical State). For example, "Normal Level" maps to a normal state probability of 0.9, a warning state probability of 0.1, and a critical state probability of 0; "Attention Level" maps to a normal state probability of 0.3, a warning state probability of 0.6, and a critical state probability of 0.1; and "Abnormal Level" maps to a normal state probability of 0.1, a warning state probability of 0.4, and a critical state probability of 0.5. This lookup table can be built into the system control logic as a static data structure and efficiently accessed through hash matching or a two-dimensional index.

[0045] Based on the cell adjacency list, the distribution of deviation levels of adjacent cell pairs in which each cell participates is statistically analyzed. The state probability value corresponding to each deviation level is obtained through the state probability mapping lookup table. The arithmetic mean of the state probabilities of all adjacent relationships of each cell is calculated to obtain the initial probability values ​​of each cell in the normal state, warning state and critical state. This invention utilizes the cell adjacency table constructed in step S13 (recording which cells form adjacent pairs with each cell). The system can traverse the adjacency relationships of each cell and query the deviation level corresponding to each pair of adjacent cells it participates in. Taking cell A as an example, if it forms three adjacent pairs with B, C, and D, and the deviation levels are "normal level," "concern level," and "abnormal level," respectively, they are mapped to three sets of state probabilities: (0.9, 0.1, 0.0), (0.3, 0.6, 0.1), and (0.1, 0.4, 0.5). After extracting these state probability vectors, the system accumulates the same state items and then divides them by the number of adjacent pairs, thus calculating the arithmetic mean of the three states. This method maps the statistical information of abnormal heat exchange around the cell to the potential thermal state distribution of the cell itself, fully reflecting the influence of the thermal environment on the cell state.

[0046] The initial state probabilities of each cell are normalized, and the sum of the three state probabilities of each cell is verified to be 1. This generates cell state probabilities that include the normal state probability, warning state probability, and critical state probability of each cell.

[0047] In this embodiment of the invention, since the generation of state probabilities is based on an arithmetic mean, the sum may deviate from 1 during numerical calculation. The system needs to perform normalization to ensure that the state probabilities of each cell satisfy the physical constraint that the sum is 1. This process is achieved by dividing the three state probabilities by the sum. Taking cell A as an example, if the initial state probabilities are 0.43 for normal state, 0.37 for warning state, and 0.25 for critical state, then the sum is 1.05. After normalization, these are adjusted to approximately 0.41, 0.35, and 0.24, respectively. After performing this operation, the system generates a structured data table containing all cell numbers and their three state probabilities, namely the "Cell State Probability Data Table." This data table is the basic input for subsequent construction of the heat conduction network and prediction of thermal runaway paths. Its fields include "Cell Number," "Normal State Probability," "Warning State Probability," "Critical State Probability," and "Update Timestamp." In the thermal management system of new energy vehicle power battery packs, this data can be pushed to the main control unit in real time to participate in the formulation of active heat dissipation scheduling strategies, and can also serve as the core judgment basis for early warning of thermal runaway.

[0048] Preferably, step S3, which involves analyzing the influence of adjacent cells on the cell state probability, includes: High-risk cells with a critical state probability greater than 0.3 are identified by using the critical state probability value in the cell state probability, and they are marked as heat source cells to establish a list of heat source cells. The adjacent cell list of each heat source cell is determined based on the cell adjacency list, and the heat transfer intensity coefficient of each heat source cell to each adjacent cell is calculated using the heat flow distribution data between cells. By multiplying the critical state probability value of the heat source cell list by the heat transfer intensity coefficient, the risk propagation impact value of the heat source cell on each adjacent cell is calculated. The risk propagation impact values ​​collected from all adjacent heat source cells for each non-heat source cell are summed to obtain the total value of external risk impact received by each cell; The probability correction calculation is performed by using the cell state probability of each cell and the total value of external risk impact. The total value of external risk impact is proportionally allocated to the early warning state probability and the critical state probability, and the normal state probability is reduced to obtain the cell state probability distribution data.

[0049] In this embodiment of the invention, each battery cell corresponds to a set of three probability values ​​for its state in the cell state probability data table generated in the preceding steps. The system uses the critical state probability as a filtering condition and sets a risk identification threshold of 0.3, meaning that batteries with a critical state probability greater than 0.3 are identified as having a high risk of thermal runaway. The system filters out objects that meet this condition from all batteries and records their numbers, corresponding critical state probability values, and timestamps in the "Heat Source Battery Cell List." For example, in a detection cycle, the critical state probabilities of batteries numbered B03, C07, and F11 are 0.35, 0.42, and 0.38, respectively. The system identifies these as heat source batteries and adds them to the list for subsequent propagation modeling. This list is updated in real time and can be shared with the thermal runaway early warning module to achieve probability-based proactive identification of high thermal risks. Based on the battery cell adjacency table established in step S13, the system can quickly query the list of directly adjacent batteries for each heat source battery cell. For each pair of heat source-adjacent battery cells, the system calls the heat flow distribution data and extracts the unit heat flow value between them. In this paper's method, this heat flux value is defined as the "heat transfer intensity coefficient," used to characterize the thermal influence of a heat source cell on adjacent cells, and possesses a certain degree of physical comparability. To prevent measurement errors from affecting modeling stability, the system standardizes the physical units of the heat transfer intensity coefficient (e.g., converting it to the number of joules transferred per unit time) to unify the measure of influence. In the example, the heat transfer intensity between heat source cell B03 and its adjacent cell B04 is 1.7, and between B03 and B02 it is 1.2, both of which will be stored in the "thermal influence coefficient matrix" for use in the next step. The critical state probability value of each heat source cell is multiplied by its heat transfer intensity coefficient for each adjacent cell, and the result is defined as the "risk propagation influence value," representing the risk contribution of a unit heat source risk to the target cell during the propagation process. For example, the critical state probability of heat source B03 is 0.38, its heat transfer intensity to B04 is 1.7, so the influence value is 0.646; and its intensity to B02 is 1.2, so the influence value is 0.456. All calculation results are organized according to the target cell number to form a "Single-Source Risk Propagation Impact Data Table". If a cell is adjacent to multiple heat sources, its final impact value is the sum of the impact values ​​from all sources. All entries for each non-heat source cell in the "Single-Source Risk Propagation Impact Data Table" are summarized, and the risk propagation impact values ​​from all sources are summed to obtain the "Total External Risk Impact Value" for each cell. This total value, using the cell number as the primary key, generates an "External Risk Reception Strength Table". For example, if cell B04 is simultaneously affected by thermal risks from two heat source cells, B03 and C07, its final total external risk impact value is the sum of these two impact values. The establishment of this data table provides quantitative support for subsequent state probability correction and can be displayed in real time on the system interface as an important indicator of the degree of thermal anomaly interaction between cells. The system uses the total external risk impact value and the original state probability of each cell to jointly calculate and construct a dynamic correction model.The specific operation is as follows: The total value of external risk impact is superimposed onto the original state probabilities according to a certain allocation rule (e.g., 30% allocated to the warning state probability and 70% to the critical state probability), while the same total value is subtracted from the normal state probability to keep the total probability of the three states constant. If a probability is negative after correction, the system sets that item to zero and renormalizes it. Taking cell B04 as an example, its original probabilities are 0.6 for normal state, 0.3 for warning state, and 0.1 for critical state. If the received risk impact value is 0.4, its corrected states are 0.2 for normal state, 0.42 for warning state, and 0.38 for critical state. All correction results form a "State Probability Distribution Correction Data Table," which is used to dynamically drive subsequent heat propagation path deduction and control strategy optimization.

[0050] Preferably, step S3, which involves performing a probability-weighted calculation of heat transfer parameters based on the cell state probability distribution data, includes: Establish a reference library of heat transfer parameters for three cell states, where the parameters for the normal state include thermal conductivity. Specific heat capacity The warning parameters include thermal conductivity. Specific heat capacity Critical state parameters include thermal conductivity. Specific heat capacity ; The normal state probability, warning state probability, and critical state probability of each cell are extracted from the cell state probability distribution data to obtain a cell state probability lookup table; Based on the cell state probability lookup table and the thermal transfer parameter benchmark library, the probability weighted parameter calculation of the thermal conductivity coefficient of each cell is performed. The normal state probability is multiplied by the normal state thermal conductivity coefficient, the warning state probability is multiplied by the warning state thermal conductivity coefficient, and the critical state probability is multiplied by the critical state thermal conductivity coefficient. The three are added together to obtain the dynamic thermal conductivity coefficient of the cell. Based on the cell state probability lookup table and the heat transfer parameter benchmark library, the probability weighting parameter of the specific heat capacity of each cell is calculated. The normal state probability is multiplied by the normal state specific heat capacity, the warning state probability is multiplied by the warning state specific heat capacity, and the critical state probability is multiplied by the critical state specific heat capacity. The three are added together to obtain the dynamic specific heat capacity value of the cell. The effectiveness of the dynamic thermal conductivity and dynamic specific heat capacity values ​​of each cell was verified, and dynamic heat transfer parameter data were generated based on the spatial arrangement of the cells.

[0051] This invention establishes a benchmark library of heat transfer parameters for three states of a battery cell: normal state, warning state, and critical state. Specifically, representative values ​​for the thermal conductivity coefficient and specific heat capacity for each state are defined using authoritative experimental data or literature. For example, the thermal conductivity coefficient of a normal state battery cell is set to 2.0 watts per meter Kelvin, indicating high heat transfer efficiency within the cell and between adjacent cells; the specific heat capacity is set to 1100 joules per kilogram Kelvin, meaning the cell can store thermal energy with a high heat capacity. The thermal conductivity coefficient and specific heat capacity for the warning and critical states decrease to 1.6 watts and 900 joules, and 1.2 watts and 700 joules, respectively, reflecting the decline in thermophysical properties of the cell material due to heating or aging. This benchmark library is stored in the form of a structured data table, with each row corresponding to a cell state and columns including the thermal conductivity coefficient and specific heat capacity, ensuring rapid retrieval and compatibility with dynamic state probability data during subsequent calculations. Using the aforementioned heat exchange efficiency deviation analysis and probability mapping algorithm, the probability distribution data of each cell in the battery pack under three states—normal, warning, and critical—is obtained. This data is then aggregated to generate a cell state probability lookup table. The table uses the cell's unique identifier as an index, and each record contains three floating-point values, representing the probability of the cell being in each of the three states. To ensure the accuracy and validity of the probability data, the system automatically checks whether the sum of the three state probabilities equals 1; if a deviation occurs, it is corrected through normalization. Furthermore, the lookup table records the sampling timestamp, supporting time series analysis and dynamic updates to meet the real-time state switching requirements during thermal runaway prediction. For each cell, its state probability is multiplied by the corresponding thermal conductivity coefficient in the benchmark library. Specifically, the current normal state probability is extracted and multiplied by the corresponding 2.0 watts per meter Kelvin thermal conductivity coefficient; the warning state probability is multiplied by 1.6 watts; and the critical state probability is multiplied by 1.2 watts. Subsequently, the three factors are added together to obtain the dynamic thermal conductivity coefficient of the battery cell, which represents a weighted comprehensive reflection of the cell's current heat transfer efficiency. For example, if the three-state probabilities of a battery cell are 0.75, 0.15, and 0.10, its dynamic thermal conductivity coefficient is calculated as (0.75 × 2.0) + (0.15 × 1.6) + (0.10 × 1.2) = 1.5 + 0.24 + 0.12 = 1.86 watts per meter Kelvin. This process is automatically executed in batches within the software system, completing the calculation for all battery cells and reflecting the impact of state changes on thermal conductivity performance in real time. This weighted method avoids rigid assumptions about a single parameter and better reflects the dynamic characteristics of actual battery cell thermal behavior. Similarly, for each battery cell, the specific heat capacity parameter is also calculated using probability weighting. The system first obtains the probabilities of the cell's normal state, warning state, and critical state, then multiplies them by the corresponding specific heat capacity values ​​(1100, 900, and 700 joules per kilogram Kelvin) in the benchmark library, and finally sums the three values ​​together.Taking the state probability of a single battery cell as an example, if the probability of the normal state is 0.6, the probability of the warning state is 0.3, and the probability of the critical state is 0.1, then the dynamic specific heat capacity is (0.6×1100) + (0.3×900) + (0.1×700) = 660 + 270 + 70 = 1000 joules per kilogram of Kelvin. This dynamic value describes the cell's current storage capacity in terms of heat absorption. As the probability of the critical state increases, the overall specific heat capacity decreases, indicating a decrease in the cell's heat capacity and an increase in the risk of thermal runaway. This step involves automatic data calculation without manual intervention, ensuring real-time accuracy under continuous data acquisition. After completing the calculation of the dynamic thermal conductivity coefficient and specific heat capacity, the system performs validity checks on the calculation results for each battery cell. Specifically, this includes determining whether the thermal conductivity coefficient is within the range of 1.0 to 2.5 watts per meter of Kelvin and whether the specific heat capacity is within the range of 600 to 1200 joules per kilogram of Kelvin. If the values ​​exceed the specified range, the system automatically flags the anomaly and replaces them with the historical average or sliding window average from the nearest time period to ensure the physical parameters are reasonable and continuous. Subsequently, based on the predefined spatial coordinates (two-dimensional plane or three-dimensional structure) of the cells within the battery pack, the dynamic parameters are mapped and indexed to form a complete spatially arranged dynamic heat transfer parameter dataset for subsequent heat transfer network model calls. The data structure uses matrix or tensor storage to ensure the model can efficiently read parameters based on spatial adjacency relationships, achieving dynamic coupling simulation of cell state and thermophysical properties.

[0052] Preferably, step S4 includes the following steps: Step S41: Extract the dynamic thermal conductivity coefficient and dynamic specific heat capacity of each cell from the dynamic heat transfer parameter data, construct a network basic framework including node attributes and connection relationships, and generate network framework data; This invention extracts the thermal conductivity and specific heat capacity of each battery cell from the generated dynamic heat transfer parameter data as key attributes of the nodes, constructing an initial heat transfer network framework expressed in the form of "node-edge". Here, a "node" represents each battery cell in the battery pack, and the "node attribute" is its dynamic thermal conductivity and dynamic specific heat capacity; the "connection relationship" is determined by the previously established cell adjacency list, with each pair of physically adjacent cells defined as having a connecting edge. This network framework is expressed in the form of an adjacency list or adjacency matrix, and is accompanied by structured metadata records, such as node number, spatial coordinates, and cell type, enabling subsequent heat propagation path analysis to consider both topological structure and physical attributes. The entire framework's data structure is stored as a graph-structured data object, and automatic organization, retrieval, and updating are achieved using a graph theory modeling library (such as NetworkX or a custom data structure) to obtain the complete network framework data.

[0053] Step S42: Calculate the heat transfer impedance value of each connection edge based on the network framework data and dynamic heat transfer parameter data, and assign the impedance value as the edge weight to the corresponding network connection to generate network weight allocation data; This invention calculates the thermal transfer impedance of a connecting edge based on the dynamic thermal conductivity, specific heat capacity, and geometric distance between each pair of adjacent battery cells. Thermal transfer impedance represents the "impedance" to heat flow during heat transfer; a higher value indicates greater difficulty in heat transfer from one battery cell to another. The specific calculation logic is as follows: First, the average thermal conductivity of the two battery cells corresponding to the connecting edge is determined; then, the thermal resistance per unit heat flow path is calculated by combining the center-to-center distance and contact area between the two cells (obtainable through a predefined battery cell geometry model); finally, this impedance is assigned as the "weight" attribute of the edge to the network. In engineering implementation, the system traverses all edges in the entire network, automatically assigning a corresponding impedance value to each edge, thus forming complete network weight allocation data. This process requires no human intervention, the calculation results are updated in real time, and the thermal resistance matrix can be exported and visualized.

[0054] Step S43: Construct a mathematical structure containing node thermal characteristics, connection topology, and transmission impedance based on network framework data and network weight allocation data to generate a battery pack heat transfer network model; This embodiment of the invention, based on the network framework data generated in S41 and the network weight allocation data obtained in S42, merges the two to construct a complex network model with physical thermal characteristics. Specifically, the model is mathematically based on a graph structure, embedding two dynamic thermal attributes (thermal conductivity and specific heat capacity) in each node and thermal transfer impedance in each edge. Furthermore, the network model records time step information and boundary condition configurations to support subsequent finite difference method solutions. To achieve operability and computability, the system encapsulates the network model using a hybrid matrix and attribute table approach; for example, the adjacency matrix records the connection topology, and the attribute table records the thermal parameters of each node and edge. The model possesses high abstraction and versatility, adapting to battery packs with different arrangements (such as square arrays, cellular arrays, or customized modules). The final generated battery pack heat transfer network model possesses all the necessary data for dynamic temperature evolution simulation.

[0055] Step S44: Identify the critical path of heat propagation through the battery pack thermal transfer network model, and solve the temperature evolution equation along the critical path using the finite difference method to generate thermal runaway risk prediction data.

[0056] After constructing the network model, this embodiment of the invention performs a global analysis of the thermal resistance of each connection edge in the model to identify the optimal heat propagation path (i.e., the dominant path for heat to diffuse from high-risk nodes to the periphery). This path is the key channel in the battery thermal runaway propagation chain. The path identification method uses a weighted shortest path algorithm (such as Dijkstra's or Bellman-Ford's algorithm), with the weights being the aforementioned heat transfer impedance. After identifying the key path, the system constructs a one-dimensional or two-dimensional temperature field distribution model along the path and uses the finite difference method (FDM) to discretely solve the temperature evolution equation. The finite difference method divides the time and space dimensions into grids and iteratively updates the cell temperature value at each time step to simulate the heat diffusion process in the time dimension. During the temperature evolution process, the system receives dynamic heat transfer parameter data in real time to dynamically adjust the thermal conductivity and specific heat capacity, thereby ensuring that the simulation results accurately reflect the cell state fluctuations. The final generated thermal runaway risk prediction data includes temperature distribution maps, heat propagation direction maps, and early warning indicators for abnormal temperature growth in each cell of the battery pack at multiple time steps, providing decision support for the thermal safety management of the battery pack.

[0057] Preferably, step S44 includes the following steps: Step S441: Extract the heat transfer impedance value of each connection edge from the battery pack heat transfer network model, sort the impedance values ​​from smallest to largest, and establish an impedance sorting list. This invention extracts the thermal transfer impedance values ​​of all connected edges from the already constructed battery pack thermal transfer network model. The thermal transfer impedance value reflects the degree of resistance encountered when heat is transferred between adjacent cells, and is measured in K / W. Its calculation was completed in the preceding step S42. Subsequently, the system sorts all connected edge impedance values ​​in ascending order, i.e., the smaller the impedance value, the higher the ranking. The sorting process uses a stable sorting algorithm (such as merge sort or heap sort) to maintain the original network structure. The sorted data is organized into a four-field data structure: "edge number + starting cell number + ending cell number + thermal resistance value," generating an impedance sorting list in the format of a two-dimensional structured array or key-value table for easy subsequent querying and extraction. This sorting operation lays the foundation for extracting the preferred heat propagation path and is a prerequisite for thermal dynamic simulation calculations in the entire modeling method.

[0058] Step S442: Select the top 30% of connection edges with the smallest impedance values ​​according to the impedance sorting list, and mark the top 30% of connection edges as preferred propagation paths, generating preferred path identification data; In this embodiment of the invention, based on the impedance sorting list generated in step S441, the system selects the top 30% of connection edges with the lowest thermal transfer impedance values ​​in this step. This percentage (30%) can be set and adjusted in the configuration file according to different battery pack structures. The selection process uses a list slicing mechanism to directly locate the top 30% of connection edge entries and sets a "priority path" Boolean identifier field for them. The marking information is recorded in the same edge table format as the network structure, generating priority path identifier data. This data not only retains the original connection information and thermal resistance value, but also adds an identifier bit, with the structure being "edge number + starting cell number + ending cell number + thermal resistance value + whether it is a priority path". The completion of this step allows the subsequent model to focus on the main heat diffusion channels for evolution modeling, improving computational efficiency and reducing unnecessary thermal channel processing costs.

[0059] Step S443: Based on the priority path identification data and the topology of the battery pack heat transfer network model, identify the cell node groups connecting the priority paths and establish key node group data; Based on the priority path identification data obtained in S442, this embodiment of the invention analyzes the cell nodes connected to these priority paths, extracts the cell numbers of all cells participating in the low-resistivity heat transfer channel, and removes duplicates, ultimately obtaining a set of key nodes for heat diffusion. To enhance the structure of the analysis, this set is organized into key node group data, including multiple attribute fields such as node number, module region, spatial coordinates, current temperature, and state probability, facilitating the fusion of physical thermal parameters and topological positions during modeling. In actual deployment, for example, in a battery pack composed of 96 cells, this step may identify 24-30 key nodes. These nodes are considered the core control units for heat diffusion and are the core participating units in the subsequent prediction model equation establishment.

[0060] Step S444: Based on the key node group data and priority path identification data, perform computational model construction processing to generate difference equation system data; This invention, based on key node group data, treats cell nodes as grid nodes in a finite difference model, constructing a difference expression for heat transfer channels based on priority path identification data. First, a difference equation with time as the step size and temperature as the variable is established for each pair of key nodes. The equation includes the node's dynamic thermal conductivity, specific heat capacity, and thermal resistance with neighboring key nodes. The system uses a time-forward, spatially central difference-based difference scheme for discretization and supports heat flow modeling of three-dimensional node structures. To ensure numerical stability, the CFL stability criterion is automatically invoked to adjust and control the time step Δt. The entire model construction process is completed by automatically traversing key nodes and their neighboring nodes, without relying on manually written equations, ultimately generating a set of difference equations expressed in matrix form for subsequent model solving and integration.

[0061] Step S445: Perform model integration and verification processing based on the differential equation set data to generate a dynamic prediction model for lithium-ion battery thermal runaway.

[0062] The embodiment of the present invention is the core integration link of this modeling process. The system first inputs the differential equation set established in S444 into a numerical solution engine (such as a custom finite difference solver or an integrated PDE solver module) to solve the dynamic evolution of the temperature field, and the solution result outputs the predicted temperature values of each key battery cell at each time step. To verify the model accuracy, the system will call historical test data for result comparison, and use statistical indicators such as maximum error and mean square error to evaluate the accuracy of the prediction model. At the same time, the regression verification of the early warning time point is combined with known thermal runaway event samples. If the model accuracy meets the preset requirements (such as the error is less than 2K, and the early warning time for thermal runaway identification is not less than 300s), then the model is encapsulated and output as the "dynamic prediction model for lithium-ion battery thermal runaway" for deployment to the BMS system or simulation platform to support real-time prediction and linkage response of thermal runaway risks.

[0063] Particularly importantly, step S444 includes the following steps: Determine the range of battery cells to be calculated using the key node group data, and extract the current temperature values of the corresponding battery cells from the standardized temperature data to generate the initial temperature data of the calculation domain; The main objective of the embodiment of the present invention is to determine the initial temperature conditions required for finite difference calculation. First, the system extracts the battery cell numbers to be simulated from the key node group data generated in the previous stage as the core objects for subsequent temperature calculation, that is, the "battery cells in the calculation domain". These battery cells are the nodes on the heat priority propagation path and play a leading role in the evolution of thermal runaway. Subsequently, from the generated standardized temperature data, retrieve the temperature values matching these battery cell numbers. This data is the result after real-time sampling and moving average filtering processing, with the unit of degrees Celsius and the sampling accuracy of ±0.1°C. The system organizes these temperature values into a table in the order of battery cell numbers, and the fields include "battery cell number + current temperature + time stamp". This table is the initial temperature data of the calculation domain. This data will participate in the solution as the initial condition at time t = 0 in the initialization stage of the differential model.

[0064] Extract the dynamic thermal conductivity and dynamic specific heat capacity parameters of the battery cells corresponding to the key node group data from the battery pack heat transfer network model to establish a parameter data table for the calculation domain; In this embodiment of the invention, after obtaining the cell number within the computational domain, the system continues to extract the dynamic thermal conductivity coefficient and dynamic specific heat capacity parameters related to these cells from the battery pack heat transfer network model. The "dynamic thermal conductivity coefficient" and "dynamic specific heat capacity" mentioned here have already been obtained in the previous steps by weighting the probabilities of the three states (normal state, warning state, and critical state), respectively... and The units are [number]. After extraction, the system matches these two parameter fields for each cell with its number to create a "computational domain parameter data table," whose structure includes "cell number + dynamic thermal conductivity coefficient + dynamic specific heat capacity." This table serves as the core source of the difference coefficient matrix, ensuring that each cell uses its current state's thermophysical properties in the difference equations instead of fixed values, thus guaranteeing the model's adaptability to the dynamic evolution of thermal runaway.

[0065] Based on the priority path identification data, the critical propagation path is divided into calculation units of equal length, and the spatial grid spacing is set to 1 / 10 of the cell spacing to generate spatial grid partitioning data. To achieve spatial discretization of the finite difference method, this embodiment of the invention requires the system to divide key heat propagation paths into several computational units. First, based on the cell connection edges listed in the priority path identification data, the system obtains the physical length of each heat propagation channel, which is equal to the actual center-to-center distance between cells, for example, 20 mm. Next, the system divides each path into 10 equally spaced computational grids, resulting in a spatial grid spacing of 2 mm. Each grid cell represents a heat evolution computation node, which the system numbers and records in each grid cell its path location, cell number, spatial coordinates, and associated start and end nodes. Finally, spatial grid partitioning data is generated, with a data structure including "grid number + path number + grid location index + grid coordinates + adjacent grid relationship," which supports iterative calculations of the spatial dimensions of the difference equation.

[0066] Using the initial temperature data of the computational domain, the computational domain parameter data table, and the spatial grid division data, finite difference temperature evolution equations for each grid cell are established, with a time step of 0.1 seconds, to generate a set of difference equations.

[0067] In this embodiment of the invention, after initial temperature extraction, thermal parameter extraction, and mesh generation, the system enters the finite difference modeling stage. First, the system treats each mesh cell as a thermodynamic calculation node, calling its initial temperature, dynamic thermal conductivity, and dynamic specific heat capacity, and constructing a temperature update equation by combining the thermal resistance information of adjacent meshes. The difference equation uses a first-order forward time difference and a second-order central spatial difference, representing the change in node temperature per unit time as the difference in heat flux between the node and its neighboring nodes divided by its own heat capacity. The time step is set to 0.1 seconds to satisfy the CFL numerical stability condition. The equations of all mesh cells are summarized into a set of difference equations, stored in a sparse matrix format (CSR or COO format), with boundary condition definitions, such as a fixed boundary for the heat source cell temperature or a natural convection boundary. This set of equations is ultimately used for temperature field simulation of the thermal runaway evolution path, providing accurate basic data for subsequent thermal risk level classification and control strategy formulation.

[0068] Of particular importance, step S445 includes the following steps: Step S4451: Establish a temperature evolution solver based on the difference equation system data. The finite difference calculation method, time step setting, and spatial grid configuration are encapsulated into a reusable calculation engine to generate temperature evolution solver component data. This invention aims to encapsulate the difference equation set data generated in the preceding steps into a repeatedly invoked computational module. The system first reads the established finite difference temperature evolution equation set, where each computational unit node contains information such as the corresponding spatial grid number, current temperature value, neighboring node relationships, dynamic thermal conductivity, and specific heat capacity. Subsequently, a numerical computation framework (such as the C++-based numerical library Eigen or Python-based NumPy and SciPy) is invoked to construct a time-progression function with a fixed time step of 0.1 seconds and a spatial grid spacing of 1 / 10 of the cell center-to-center spacing (e.g., 2 mm). The temperature field is then updated using an explicit difference scheme. During this process, the system encapsulates the core computational interface, integrating the temperature update function, boundary condition manager, and thermal parameter reader as sub-modules, and reserving parallel acceleration interfaces for the model (such as OpenMP thread control or CUDA parallel computing). Finally, the system exports the encapsulated module as independent temperature evolution solver component data. Its inputs are initial temperature data, heat transfer parameters and mesh configuration, and its output is a multi-step time temperature distribution sequence, which can be directly used for real-time thermal evolution simulation or risk assessment triggering.

[0069] Step S4452: Establish a risk assessment rule base based on priority path identification data and key node group data, organize temperature threshold judgment criteria, risk level classification rules and propagation path analysis into a structured assessment system, and generate risk assessment rule component data; This invention constructs a structured risk identification mechanism corresponding to temperature evolution data. Based on the thermal runaway critical temperature characteristics of lithium-ion batteries, the system presets multiple graded risk thresholds. For example, temperatures above 60°C but below 80°C are marked as "risk of concern," above 80°C but below 100°C as "high risk," and above 100°C are directly considered "critical risk." Simultaneously, the system combines priority path identification data and key node group data to define the direction and intensity of heat propagation and sets progressive rules for risk transmission along critical paths. For example, "a temperature rise greater than 5°C at three consecutive nodes along a priority path within 10 seconds" is considered an accelerated heat propagation phenomenon. Furthermore, the evaluation system also includes dynamic risk level switching thresholds and node cluster risk judgment rules (such as triggering regional risk escalation if more than two high-risk nodes appear in the same heat propagation cluster). All risk rules are structured into a rule dictionary, with fields including "judgment dimension + condition + risk level output," and encapsulated as risk assessment rule component data for parsing and judging the temperature solution results.

[0070] Step S4453: Integrate the temperature evolution solver component data, risk assessment rule component data, and dynamic heat transfer parameter data to construct a complete modeling architecture that includes real-time computing capabilities, risk prediction functions, and parameter adaptive update mechanisms, and generate a dynamic prediction model for thermal runaway of lithium-ion batteries.

[0071] This invention integrates temperature evolution solving capabilities and risk identification capabilities into a unified thermal runaway modeling platform. First, the system integrates the temperature evolution solver component data encapsulated in S4451 and the risk assessment rule component data generated in S4452, and synchronously accesses the dynamic heat transfer parameter data generated in the preceding steps to achieve state awareness and adaptive updating of thermophysical parameters. The system transmits the results of each time step of the solver's operation to the risk assessor in real time via a programming interface. The assessor then feeds back the analysis results to the scheduling control module for storage and scheduling of the next round of calculation. Simultaneously, the modeling architecture supports a dynamic parameter change mechanism; that is, when a cell's state switches from a warning state to a critical state, its dynamic thermal conductivity coefficient and specific heat capacity are automatically updated to the corresponding critical state values, and the differential calculation coefficients are automatically updated. Finally, the system outputs a multi-dimensional data structure containing a time-series temperature field evolution map, node risk level change curves, and dynamic trends of heat propagation paths. This modeling architecture can be embedded in a BMS system or battery safety testing platform to form a dynamic prediction model for lithium-ion battery thermal runaway with real-time prediction, adaptive parameter updating, and graded response capabilities.

[0072] The present invention also provides a modeling system for thermal runaway of lithium-ion batteries, used to execute the above-described modeling method for thermal runaway of lithium-ion batteries, wherein the modeling system for thermal runaway of lithium-ion batteries includes: The heat flow data acquisition module is used to acquire real-time temperature data and real-time current data of each cell in the battery pack; filter the real-time temperature data to generate standardized temperature data; calculate the temperature difference between adjacent cells based on the standardized temperature data, and calculate the heat flow in combination with the real-time current data to generate heat flow distribution data between cells. The efficiency deviation analysis module is used to calculate the heat exchange efficiency index of each adjacent cell pair based on the heat flow distribution data between cells. The heat exchange efficiency index is the ratio of heat flow to temperature difference. Based on the historical operating data of the battery pack, a benchmark value of the heat exchange efficiency of each cell pair under stable operating conditions is established. The real-time heat exchange efficiency index is compared with the benchmark value of heat exchange efficiency to calculate the degree of deviation and generate heat exchange efficiency deviation characteristic data. The efficiency deviation analysis module is used to calculate the cell state probability based on heat exchange efficiency deviation characteristic data, where the cell state includes normal state, warning state and critical state; perform adjacent cell influence analysis on the cell state probability to generate cell state probability distribution data; and perform probability weighted calculation of heat transfer parameters based on the cell state probability distribution data to generate dynamic heat transfer parameter data. The dynamic modeling generation module is used to construct a battery pack heat transfer network model based on dynamic heat transfer parameter data; the critical path of heat propagation is identified through the battery pack heat transfer network model, and the temperature evolution equation is solved along the critical path using the finite difference method to generate a dynamic prediction model for thermal runaway of lithium-ion batteries.

[0073] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0074] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method of modeling thermal runaway of a lithium-ion battery, the method comprising: The method comprises the following steps: Step S1: obtaining real-time temperature data and real-time current data of each cell in the battery pack; Step S2: calculating the heat exchange efficiency index of each adjacent cell pair according to the inter-cell heat flow distribution data, wherein the heat exchange efficiency index is the ratio of the heat flow and the temperature difference; establishing the heat exchange efficiency baseline value of each cell pair under stable working conditions based on the historical operation data of the battery pack, and comparing the real-time heat exchange efficiency index with the heat exchange efficiency baseline value to calculate the deviation degree and generate heat exchange efficiency deviation feature data; Step S3: calculating the cell state probability based on the heat exchange efficiency deviation feature data, wherein the cell state includes normal state, warning state and critical state; performing adjacent cell influence analysis on the cell state probability to generate cell state probability distribution data; and performing probability weighted calculation on the heat transfer parameters according to the cell state probability distribution data to generate dynamic heat transfer parameter data; Step S4: constructing a battery pack heat transfer network model based on the dynamic heat transfer parameter data; identifying the key path of heat propagation through the battery pack heat transfer network model, and solving the temperature evolution equation along the key path by using the finite difference method to generate a lithium ion battery thermal runaway dynamic prediction model. Step S1 comprises the following steps:

2. The method of claim 1, wherein the thermal runaway of the lithium-ion battery is modeled by, Step S11: collecting real-time temperature and current information of each cell in the battery pack, setting the sampling frequency to be once every five seconds, and recording the time stamp of data collection to obtain real-time temperature data and real-time current data; Step S12: performing moving average filtering on the real-time temperature data, using the arithmetic mean value of the consecutive five sampling points as the filtered temperature value at the current time to generate standardized temperature data; Step S13: determining the adjacent relationship of each cell according to the pre-obtained physical layout of the battery pack, and establishing a cell adjacency table; Step S14: calculating the temperature difference of each pair of adjacent cells based on the standardized temperature data to obtain the temperature difference between adjacent cells; Step S15: using the real-time current data to calculate the instantaneous heat generation power of each cell based on Ohm's law, distributing the instantaneous heat generation power of each cell to adjacent cells based on the heat transfer path, and combining the conduction heat flow generated between adjacent cells due to the temperature difference to obtain the total heat flow; Step S16: organizing the total heat flow of each adjacent cell pair based on the spatial layout of the battery pack to establish a heat flow matrix with cell numbers as row and column indexes, wherein the matrix elements of the heat flow matrix represent the heat flow values between the corresponding cell pairs; Step S17: performing data standardization processing on the heat flow matrix to convert the heat flow values to a unified physical unit to generate inter-cell heat flow distribution data. Step S2 comprises the following steps:

3. The method of claim 2, wherein the thermal runaway of the lithium-ion battery is modeled by, Step S21: extracting the total heat flow value of each adjacent cell pair from the inter-cell heat flow distribution data, establishing a heat flow data table, and extracting the temperature difference value of the corresponding cell pair, establishing a temperature difference data table; Step S22: calculating the heat exchange efficiency index of each adjacent cell pair according to the total heat flow value and the temperature difference value of the corresponding cell pair, and establishing a heat exchange efficiency index data table; Step S23: establishing the heat exchange efficiency baseline value of each cell pair under stable working conditions based on the historical operation data of the battery pack, and comparing the real-time heat exchange efficiency index with the heat exchange efficiency baseline value to calculate the deviation degree and generate heat exchange efficiency deviation feature data; Step S24: calculating the cell state probability based on the heat exchange efficiency deviation feature data, wherein the cell state includes normal state, warning state and critical state; performing adjacent cell influence analysis on the cell state probability to generate cell state probability distribution data; and performing probability weighted calculation on the heat transfer parameters according to the cell state probability distribution data to generate dynamic heat transfer parameter data; Step S25: constructing a battery pack heat transfer network model based on the dynamic heat transfer parameter data; identifying the key path of heat propagation through the battery pack heat transfer network model, and solving the temperature evolution equation along the key path by using the finite difference method to generate a lithium ion battery thermal runaway dynamic prediction model. Step S22: Calculate the heat exchange efficiency index based on the heat flow data table and the temperature difference data table, wherein the heat exchange efficiency is the ratio of the heat flow and the temperature difference; Step S23: Arrange all the real-time heat exchange efficiency indexes of adjacent cell pairs in sequence based on the cell number to establish a real-time heat exchange efficiency index data table; Step S24: Retrieve the historical temperature and current records of the battery pack in the past thirty days, and filter out stable operation period data with an ambient temperature change of less than 5°C and a charging and discharging current change of less than 10% of the rated current; Step S25: Calculate the historical heat exchange efficiency index of each cell pair under stable working conditions based on the filtered stable operation period data; Step S26: Take the arithmetic mean of multiple historical heat exchange efficiency indexes of each adjacent cell pair to obtain a heat exchange efficiency reference value; Step S27: Compare the real-time heat exchange efficiency index with the heat exchange efficiency reference value, calculate the deviation degree, and generate heat exchange efficiency deviation feature data.

4. The method of modeling thermal runaway of a lithium-ion battery of claim 3, wherein, Step S27 includes the following steps: Step S271: Compare the real-time heat exchange efficiency index of the same adjacent cell pair with the heat exchange efficiency reference value to establish a paired comparison data table of real-time values and reference values; Step S272: Calculate the deviation of each group of data in the paired comparison data table, the heat exchange efficiency index minus the corresponding heat exchange efficiency reference value to obtain the absolute deviation, and then divided by the heat exchange efficiency reference value to obtain the relative deviation percentage, and establish a deviation percentage data table; Step S273: According to the preset deviation level standard, each value in the deviation percentage data table is classified, wherein the relative deviation less than 5% is marked as normal level, the relative deviation between 5% and 15% is marked as attention level, and the relative deviation greater than 15% is marked as abnormal level, and each cell pair is assigned a corresponding deviation level identifier; Step S274: Organize and arrange the deviation percentage data table and the deviation level identifier based on the cell number sequence to form a structured data record containing deviation value and level information, and generate heat exchange efficiency deviation feature data.

5. The method of modeling thermal runaway of a lithium-ion battery of claim 4, wherein, The calculation of the cell state probability based on the heat exchange efficiency deviation feature data in step S3 includes: Extract the deviation level identifier of each adjacent cell pair from the heat exchange efficiency deviation feature data to establish a state probability mapping lookup table, wherein the state probability mapping lookup table includes normal state probability 0.9, pre-warning state probability 0.1, and critical state probability 0.0 for normal level, normal state probability 0.3, pre-warning state probability 0.6, and critical state probability 0.1 for attention level, and normal state probability 0.1, pre-warning state probability 0.4, and critical state probability 0.5 for abnormal level; According to the cell adjacency table, the deviation level distribution of the adjacent cell pairs participated by each cell is calculated, the state probability values corresponding to each deviation level are obtained through the state probability mapping lookup table, and the state probabilities of all adjacent relationships of each cell are calculated by arithmetic mean to obtain the initial probability values of each cell in normal state, pre-warning state and critical state; The initial state probability of each battery cell is normalized, the sum of the three state probabilities of each battery cell is verified to be 1, and the battery cell state probability including the normal state probability, the early warning state probability and the critical state probability of each battery cell is generated.

6. The method of modeling thermal runaway of a lithium-ion battery of claim 5, wherein, The adjacent battery cell influence analysis of the battery cell state probability in step S3 includes: The critical state probability value in the battery cell state probability is used to identify the high-risk battery cell with the critical state probability greater than 0.3, which is marked as a heat source battery cell, and a heat source battery cell list is established; According to the battery cell adjacency list, the adjacent battery cell list of each heat source battery cell is determined, and the heat transfer intensity coefficient of the heat source battery cell to each adjacent battery cell is calculated by using the heat flow distribution data between the battery cells; The critical state probability value of the heat source battery cell list is multiplied by the heat transfer intensity coefficient to calculate the risk propagation influence value of the heat source battery cell to each adjacent battery cell; The risk propagation influence values from all adjacent heat source battery cells are accumulated for each non-heat source battery cell to obtain the total external risk influence value received by each battery cell; The battery cell state probability and the total external risk influence value are used for probability correction calculation, the total external risk influence value is proportionally distributed to the early warning state probability and the critical state probability, and the normal state probability is reduced to obtain the battery cell state probability distribution data.

7. The method of modeling thermal runaway of a lithium-ion battery of claim 6, wherein, The probability weighted calculation of the heat transfer parameters according to the battery cell state probability distribution data in step S3 includes: Establish a thermal transfer parameter benchmark library for three states of the battery cell, wherein the normal state parameters include thermal conductivity , specific heat capacity , the pre-warning state parameters include thermal conductivity , specific heat capacity , and the critical state parameters include thermal conductivity , specific heat capacity ; The normal state probability, the early warning state probability and the critical state probability value of each battery cell are extracted from the battery cell state probability distribution data to obtain a battery cell state probability query table; Based on the battery cell state probability query table and the heat transfer parameter benchmark library, the probability weighted parameter calculation of the thermal conductivity coefficient of each battery cell is performed, the normal state probability is multiplied by the normal state thermal conductivity coefficient, the early warning state probability is multiplied by the early warning state thermal conductivity coefficient, and the critical state probability is multiplied by the critical state thermal conductivity coefficient, and the sum of the three is added to obtain the dynamic thermal conductivity coefficient of the battery cell; Based on the battery cell state probability query table and the heat transfer parameter benchmark library, the probability weighted parameter calculation of the specific heat capacity of each battery cell is performed, the normal state probability is multiplied by the normal state specific heat capacity, the early warning state probability is multiplied by the early warning state specific heat capacity, and the critical state probability is multiplied by the critical state specific heat capacity, and the sum of the three is added to obtain the dynamic specific heat capacity value of the battery cell; The validity of the dynamic thermal conductivity coefficient and the dynamic specific heat capacity value of each battery cell is verified, and the dynamic heat transfer parameter data is generated based on the spatial position arrangement and organization of the battery cells.

8. The method of modeling thermal runaway of a lithium-ion battery of claim 7, wherein, Step S4 includes the following steps: Step S41: Extract the dynamic thermal conductivity coefficient and the dynamic specific heat capacity value of each battery cell from the dynamic heat transfer parameter data, construct a network basic framework containing node attributes and connection relationships, and generate network framework data; Step S42: Calculate the heat transfer impedance value of each connection edge based on the network framework data and the dynamic heat transfer parameter data, and assign the impedance value as the edge weight to the corresponding network connection to generate network weight allocation data; Step S43: Construct a mathematical structure containing node thermal characteristics, connection topology and transfer impedance based on the network framework data and the network weight allocation data to generate a battery pack heat transfer network model; Step S44: identifying a key path of heat propagation through the battery pack heat transfer network model, and solving a temperature evolution equation along the key path using a finite difference method to generate thermal runaway risk prediction data.

9. The method of modeling thermal runaway of a lithium-ion battery of claim 8, wherein, Step S44 includes the following steps: Step S441: extracting the heat transfer impedance values of each connection edge from the battery pack heat transfer network model, numerically sorting the impedance values from small to large, and establishing an impedance sorting list; Step S442: selecting the top 30% connection edges with the smallest impedance values according to the impedance sorting list, and marking the top 30% connection edges as priority propagation paths to generate priority path identification data; Step S443: identifying the cell node group connected to the priority path based on the priority path identification data and the topology structure of the battery pack heat transfer network model, and establishing key node group data; Step S444: performing calculation model construction processing based on the key node group data and the priority path identification data to generate difference equation set data; Step S445: performing model integration and verification processing according to the difference equation set data to generate a lithium ion battery thermal runaway dynamic prediction model.

10. A system for modeling thermal runaway of a lithium-ion battery, the system comprising: A modeling method for lithium ion battery thermal runaway as claimed in claim 1, the modeling system for lithium ion battery thermal runaway comprising: a heat flow data acquisition module for acquiring real-time temperature data and real-time current data of each cell in the battery pack; filtering the real-time temperature data to generate standardized temperature data; calculating the temperature difference between each adjacent cell based on the standardized temperature data, and calculating the heat flow based on the real-time current data to generate heat flow distribution data between cells; an efficiency deviation analysis module for calculating the heat exchange efficiency index of each adjacent cell pair based on the heat flow distribution data between cells, wherein the heat exchange efficiency index is the ratio of heat flow to temperature difference; establishing a heat exchange efficiency benchmark value of each cell pair under stable working conditions based on the battery pack historical operation data, and comparing the real-time heat exchange efficiency index with the heat exchange efficiency benchmark value to calculate the deviation degree, and generating heat exchange efficiency deviation feature data; an efficiency deviation analysis module for calculating the cell state probability based on the heat exchange efficiency deviation feature data, wherein the cell state includes normal state, warning state and critical state; performing adjacent cell influence analysis on the cell state probability to generate cell state probability distribution data; performing probability weighted calculation of heat transfer parameters based on the cell state probability distribution data to generate dynamic heat transfer parameter data; a dynamic modeling generation module for constructing a battery pack heat transfer network model based on the dynamic heat transfer parameter data; identifying a key path of heat propagation through the battery pack heat transfer network model, and solving a temperature evolution equation along the key path using a finite difference method to generate a lithium ion battery thermal runaway dynamic prediction model.

Images