Multi-mode heat flow field coupling early-warning cooling decision-making system and method for energy storage battery module

Abstract

The invention relates to the technical field of battery thermal management, in particular to an energy storage battery module multi-mode heat flow field coupling early warning cooling decision system and method, which comprises a multi-sensor acquisition module, an infrared thermal imaging preprocessing module, a battery pack thermal diffusion mechanism model module, a data fusion model module and a cooling decision module, the core innovation lies in that a battery pack thermal diffusion mechanism model is constructed based on a differential geometry theory, a battery pack surface temperature field is regarded as a Riemannian manifold embedded in a three-dimensional space, early prediction of thermal runaway is realized by analyzing topological structure change of a heat flow field, and a system acquires temperature, pressure, vibration and other multi-modal data of an energy storage battery module and performs thermal runaway prediction on the energy storage battery module. The method comprises the following steps: preprocessing infrared thermal imaging data, generating three-dimensional thermal diffusion data, constructing an electric-thermal-flow multi-physical field dynamic model, identifying a critical point and a topological structure in a thermal flow field, generating thermal field characteristic data, fusing multi-source data to generate a thermal diffusion imbalance criterion, and optimizing cooling parameters by adopting a reinforcement learning algorithm.

Description

Technical Field

[0001] This invention relates to the field of battery thermal management technology, and in particular to a multimodal thermal flow field coupling early warning and cooling decision system and method for energy storage battery modules. Background Technology

[0002] With the rapid development of the new energy industry, energy storage battery systems are increasingly being used in electric vehicles, large-scale energy storage power stations, and portable devices. However, energy storage batteries are prone to generating a large amount of heat during high-power charging and discharging. If this heat cannot be dissipated in time, it may lead to thermal runaway, or even cause safety accidents such as fires and explosions. Therefore, an efficient battery thermal management system is crucial to ensuring the safe operation of batteries.

[0003] Traditional battery thermal management systems typically employ point temperature monitoring, relying on distributed temperature sensors to monitor the battery surface temperature and activating the cooling system when the temperature exceeds a preset threshold. This method has several drawbacks: first, it depends on surface temperature measurement, making it difficult to accurately reflect the internal thermal state of the battery; second, thermal runaway detection is delayed, often triggering cooling only after the temperature has already risen significantly, making it difficult to effectively prevent the spread of thermal runaway; and third, the cooling strategy is simple, typically employing global uniform cooling, which is energy-intensive and inefficient.

[0004] In recent years, with the development of sensing and computing technologies, battery thermal management technologies based on infrared thermal imaging and multiphysics coupling have begun to attract attention. However, existing technologies still have shortcomings in thermal field data processing, thermal runaway early warning, and cooling decisions, especially in the lack of accurate modeling and intelligent early warning mechanisms for battery thermal diffusion processes, making it difficult to achieve early prediction and precise cooling of thermal runaway. Summary of the Invention

[0005] The purpose of this invention is to provide a multimodal thermal flow field coupling early warning and cooling decision system and method for energy storage battery modules, which aims to solve the shortcomings of existing battery thermal management systems in thermal runaway prediction and cooling decision-making, and to achieve early warning and precise cooling of battery thermal runaway.

[0006] This invention proposes a multimodal thermal flow field coupled early warning and cooling decision system for energy storage battery modules, comprising: The multi-sensor acquisition module is used to collect temperature, pressure and vibration data of the battery pack, module and individual cells of the energy storage battery module, as well as water pressure data of the cooling pipes in the battery pack, flow data of the inlet and outlet of the cooling pipes and humidity data of the battery pack bottom plate. The infrared thermal imaging preprocessing module is communicatively connected to the multi-sensor acquisition module and is used to process the two-dimensional planar infrared temperature data acquired in real time by the infrared thermal imager and filter out background noise. The battery pack thermal diffusion mechanism model module is communicatively connected to the infrared thermal imaging preprocessing module and the multi-sensor acquisition module. It is used to construct a dynamic model of the battery pack's electro-thermal-fluid multiphysics field based on differential geometry theory. The surface temperature field of the battery pack is regarded as a Riemannian manifold embedded in three-dimensional space. The geometric representation of the thermal flow field is constructed, the critical points and topological structures in the thermal flow field are identified, and thermal field feature data is generated. The data fusion model module is communicatively connected to the battery pack thermal diffusion mechanism model module, and is used to receive the thermal field feature data and generate thermal diffusion imbalance criteria and thresholds based on multi-source data fusion; The cooling decision module is communicatively connected to the data fusion model module. It is used to optimize cooling parameters and generate and execute targeted cooling strategies based on the heat diffusion imbalance criterion and threshold using a reinforcement learning algorithm.

[0007] Preferably, the battery pack thermal diffusion mechanism model module includes: Multi-level thermal manifold building units are used to map the surface temperature distribution of a single battery cell to a two-dimensional curved surface and the internal temperature field to a three-dimensional manifold. Multiple single-cell manifolds are combined into a module-level manifold through continuous mapping, and multiple module manifolds are integrated into a battery pack-level manifold. Manifold metric building units are used to construct Riemannian metrics that reflect thermal anisotropy based on the spatial distribution of material thermal conductivity, and to establish affine connections that describe changes in thermal gradients. The heat flux topology analysis unit is used to perform vector field decomposition on the heat flux field, identify critical points such as source points, sink points and saddle points in the heat flux field, calculate the topology index of each critical point, and identify the branch lines connecting the critical points.

[0008] Preferably, the infrared thermal imaging preprocessing module uses a neural network to filter and temperature-filter the image, removes imaging noise and bad pixels, and obtains three-dimensional thermal diffusion data through interaction with the battery pack thermal diffusion mechanism model module.

[0009] Preferably, the sensing data in the multi-sensor acquisition module is transmitted to the host computer via an IoT transmitter or wired transmission. The infrared thermal imaging preprocessing module also transmits the preprocessed thermal diffusion data to the host computer, and the host computer controls the infrared thermal imager and the IoT transmitter.

[0010] Preferably, the data processing of the multi-sensor acquisition module includes real-time filtering, amplification, and analog-to-digital conversion of the acquired signals using industrial signal instruments.

[0011] Preferably, the data fusion model module includes: The feature mapping unit is used to map the thermal field feature data output by the battery pack thermal diffusion mechanism model module to a standard feature space. A multi-source data integration unit is used to perform correlation analysis between the data collected by the multi-sensor acquisition module and the thermal field characteristic data. The risk assessment unit is used to construct a thermal runaway risk assessment model based on the output of the multi-source data integration unit, and to generate thermal diffusion imbalance criteria and thresholds.

[0012] Preferably, the heat flow topology analysis unit is also used to calculate the Betti number under different thresholds of the heat flow field, track the generation and disappearance of heat field topology features, calculate the degree of distortion of heat flow lines, and identify abnormal heat flow patterns.

[0013] Preferably, the cooling decision module includes: A thermal runaway risk classification unit is used to classify the thermal runaway risk according to the thermal diffusion imbalance criterion and threshold. The cooling strategy generation unit is used to generate targeted cooling strategies based on the location, risk level, and thermal flow topology of thermal runaway. An execution control unit is used to convert the cooling strategy into control commands for the cooling equipment, controlling the cooling water flow rate, spray system, and fan power. The feedback optimization unit is used to continuously optimize cooling parameters based on the cooling effect and improve the cooling strategy through iterative learning.

[0014] Preferably, the system further includes: The thermal runaway feature library is communicatively connected to the battery pack thermal diffusion mechanism model module and the data fusion model module. It is used to store thermal flow topology features, typical thermal runaway evolution paths, and corresponding risk ratings under various thermal runaway scenarios. The visualization monitoring module is communicatively connected to the data fusion model module and the cooling decision module, and is used to present the thermal flow field distribution, thermal runaway risk rating and cooling strategy in a visual manner.

[0015] A multimodal thermal flux coupling early warning and cooling decision-making method for energy storage battery modules, including: Collect temperature, pressure and vibration data of the battery pack, module and individual cells of the energy storage battery module, as well as water pressure data of the cooling pipes in the battery pack, flow data of the inlet and outlet of the cooling pipes and humidity data of the battery pack bottom plate; The two-dimensional planar infrared temperature data acquired in real time by the infrared thermal imager is processed to filter out background noise; Based on differential geometry theory, a multi-physics dynamic model of the battery pack is constructed, which considers the surface temperature field of the battery pack as a Riemannian manifold embedded in three-dimensional space. The geometric representation of the thermal flow field is constructed, the critical points and topological structures in the thermal flow field are identified, and thermal field characteristic data are generated. Based on the aforementioned thermal field characteristic data, a thermal diffusion imbalance criterion and threshold are generated through multi-source data fusion. Based on the aforementioned thermal diffusion imbalance criterion and threshold, a reinforcement learning algorithm is used to optimize cooling parameters, generate and execute targeted cooling strategies.

[0016] This invention constructs a thermal diffusion mechanism model for battery packs by introducing differential geometry theory, treating the battery pack as a Riemannian manifold embedded in three-dimensional space. By combining infrared thermal imaging and multi-sensor data, it achieves accurate characterization and analysis of the thermal flow field. By identifying changes in the topological structure of the thermal flow field, it can predict the risk of thermal runaway in advance. Combined with reinforcement learning algorithms, it generates targeted cooling strategies to achieve precise cooling.

[0017] The present invention has the following beneficial effects: 1. Improved predictive capability: Based on the analysis of changes in the thermal flow field topology, thermal runaway can be predicted 5 to 10 minutes in advance, which is more than 60% longer than the warning time of traditional systems, providing sufficient time for the cooling system to respond.

[0018] 2. Improved cooling efficiency: By accurately locating the source of thermal runaway and implementing targeted cooling, cooling efficiency is improved by 40% to 60%, energy consumption is reduced by more than 30%, and battery life is effectively extended.

[0019] 3. Enhanced system safety: Through multimodal data fusion and early warning of thermal runaway, the incidence of thermal runaway events is reduced by more than 85%, significantly improving the safety and reliability of energy storage systems.

[0020] 4. Improved system adaptability: The manifold learning-based method can adapt to different battery types and operating conditions, greatly improving system robustness and expanding its applicability. Attached Figure Description

[0021] Figure 1 This is a structural block diagram of the multimodal thermal flow field coupling early warning and cooling decision system for energy storage battery modules of the present invention; Figure 2 This is a structural block diagram of the battery pack thermal diffusion mechanism model module of the present invention; Figure 3 This is a structural block diagram of the data fusion model module of the present invention; Figure 4 This is a structural block diagram of the cooling decision module of the present invention; Figure 5 This is a flowchart of the system of the present invention; Figure 6 This is a flowchart of the battery thermal runaway early warning and cooling decision-making process in this invention. Detailed Implementation

[0022] Please refer to Figure 1 - Figure 6The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are for illustrative and explanatory purposes only and are not intended to limit the scope of the invention.

[0023] like Figure 1 As shown, the multimodal thermal flow field coupling early warning and cooling decision system for energy storage battery modules provided by the present invention includes: a multi-sensor acquisition module 1, an infrared thermal imaging preprocessing module 2, a battery pack thermal diffusion mechanism model module 3, a data fusion model module 4, and a cooling decision module 5.

[0024] The multi-sensor acquisition module 1 is used to collect temperature, pressure, and vibration data from the battery pack, modules, and individual cells of the energy storage battery module, as well as water pressure data from the cooling pipes within the battery pack, flow rate data from the inlet and outlet of the cooling pipes, and humidity data from the battery pack base plate. In a preferred embodiment of the present invention, the temperature sensor is a PT100 platinum resistance temperature sensor with a measurement range of -50℃ to 150℃ and an accuracy of ±0.1℃; the pressure sensor is a piezoresistive pressure sensor with a measurement range of 0 to 10MPa and an accuracy of ±0.5%; the vibration sensor is a piezoelectric accelerometer with a measurement range of ±50g and a frequency response of 0.5Hz to 10kHz; the water pressure sensor is a silicon piezoresistive pressure sensor with a measurement range of 0 to 2MPa and an accuracy of ±0.3%; the flow sensor is a turbine flow meter with a measurement range of 0.5 to 100L / min and an accuracy of ±1%; and the humidity sensor is a capacitive humidity sensor with a measurement range of 0% to 100%RH and an accuracy of ±2%RH.

[0025] The infrared thermal imaging preprocessing module 2 is communicatively connected to the multi-sensor acquisition module 1, and is used to process the two-dimensional planar infrared temperature data acquired in real time by the infrared thermal imager to filter out background noise. In one embodiment of the present invention, the spatial resolution of the infrared thermal imager is 640×480 pixels, the temperature resolution is 0.05℃, the measurement range is -20℃ to 150℃, and the sampling frequency is 10Hz. The infrared thermal imaging preprocessing module 2 uses a neural network to filter and temperature-filter the image, and to remove imaging noise and bad pixels. Preferably, the neural network adopts a U-Net structure, including a 4-layer encoder and a 4-layer decoder, using mean squared error (MSE) as the loss function, and is trained using an Adam optimizer with a learning rate of 0.001 and 1000 iterations.

[0026] The battery pack thermal diffusion mechanism model module 3 is communicatively connected to the infrared thermal imaging preprocessing module 2 and the multi-sensor acquisition module 1. It is used to construct a dynamic model of the battery pack's electro-thermal-fluid multiphysics field based on differential geometry theory. The surface temperature field of the battery pack is regarded as a Riemannian manifold embedded in three-dimensional space. The geometric representation of the thermal flow field is constructed, the critical points and topological structures in the thermal flow field are identified, and thermal field characteristic data are generated.

[0027] The data fusion model module 4 is connected to the battery pack thermal diffusion mechanism model module 3 to receive thermal field characteristic data and generate thermal diffusion imbalance criteria and thresholds based on multi-source data fusion.

[0028] The cooling decision module 5 is connected to the data fusion model module 4 and is used to optimize cooling parameters and generate and execute targeted cooling strategies based on the heat diffusion imbalance criterion and threshold using reinforcement learning algorithms.

[0029] As mentioned above, the multi-sensor acquisition module 1 is used to collect data from various sensors. In one embodiment of the present invention, the sensor data in the multi-sensor acquisition module 1 is transmitted to a host computer via an IoT transmitter or wired transmission. The infrared thermal imaging preprocessing module 2 also transmits the preprocessed thermal diffusion data to the host computer. The host computer controls the infrared thermal imager and the IoT transmitter. The IoT transmitter preferably uses a 4G / 5G communication module with a data transmission rate of not less than 10Mbps to ensure real-time data transmission. The wired transmission method preferably uses an RS485 bus or Ethernet interface with a data transmission rate of not less than 100Mbps.

[0030] The data processing of the multi-sensor acquisition module 1 includes real-time filtering, amplification, and analog-to-digital conversion of the acquired signals using industrial signal instruments. Preferably, a Butterworth low-pass filter is used for filtering, with the cutoff frequency set to 1 / 5 of the sampling frequency to remove high-frequency noise; an instrumentation amplifier is used for signal amplification, with the gain adjusted from 10 to 100 times according to different sensor signals; a 16-bit ADC is used for analog-to-digital conversion, with the sampling frequency set to 10Hz to ensure the real-time performance and accuracy of data acquisition.

[0031] To ensure consistency of data from different sensors, the multi-sensor acquisition module 1 performs time synchronization and normalization on the acquired data. Time synchronization is achieved through a timestamp mechanism, with a time resolution better than 10ms; data normalization uses a minimum-maximum normalization method to unify data of different dimensions into the [0,1] interval, and the calculation formula is as follows: , in, These are the normalized data values. The original data values, The minimum value of the data. This represents the maximum value of the data. This normalization process ensures that data from different types of sensors have the same dimensions and scale range in subsequent processing.

[0032] The infrared thermal imaging preprocessing module 2 uses a neural network to filter and temperature-filter the image, and removes imaging noise and bad pixels. It obtains three-dimensional thermal diffusion data through interaction with the battery pack thermal diffusion mechanism model module 3.

[0033] In practice, the infrared thermal imaging preprocessing workflow includes: 1) Image noise removal: A combination of bilateral filtering and adaptive median filtering is used to effectively preserve edge information while filtering out noise. The mathematical expression for bilateral filtering is: , in, For the filtered image at position Temperature value at that location, For the original image at position Temperature value at that location, In pixels The central window area, This is a range kernel function used to measure the similarity of temperature values. This is a spatial kernel function used to measure the impact of spatial distance. Indicates position and The difference in temperature values, Indicates position and Spatial distance, This is a normalization factor to ensure that the filtered temperature value remains within a reasonable range. In this embodiment, the window size is set to 5×5 pixels. and All use the Gaussian kernel function. The standard deviation is set to 10℃. The standard deviation is set to 2 pixels.

[0034] 2) Defect Pixel Detection and Repair: A local outlier detection method is used to identify defective pixels in the thermal imager, and repair is performed through neighborhood interpolation. The criteria for outlier detection are as follows: , in, For pixels Temperature value at, in units of , For The local area centered on (usually) The average temperature of the pixel window, in °C. This represents the standard deviation of temperature in this local area, in °C, and k is the threshold coefficient, typically set to 3. This represents absolute value operations. When a bad pixel is detected, it is repaired using bilinear interpolation, which replaces the bad pixel value with the weighted average of the surrounding normal pixels.

[0035] 3) Temperature Field Reconstruction: A deep learning network maps the two-dimensional temperature field to three-dimensional space. An encoder-decoder network is used, with the processed infrared thermal image as input and a three-dimensional temperature field representation as output. The encoder employs a residual network structure containing 5 residual blocks; the decoder uses a transposed convolutional structure containing 5 transposed convolutional layers. During training, temperature reconstruction error and gradient consistency error are used as loss functions. , in, The total loss function value is dimensionless. The temperature reconstruction error represents the root mean square difference between the predicted and actual temperatures; it is dimensionless. Gradient consistency error represents the difference between the predicted temperature gradient and the actual temperature gradient; it is dimensionless. and These are weighting coefficients, set to 0.8 and 0.2 respectively, dimensionless, used to balance the importance of the two types of errors. The formula for calculating the temperature reconstruction error is the mean square error: Where n is the number of samples, and Let be the predicted temperature and the actual temperature of the i-th sample, respectively. The gradient consistency error is calculated using the following formula: ,in Represents the gradient of the temperature field. express Norm.

[0036] Through the above processing, the infrared thermal imaging preprocessing module 2 can reconstruct a high-quality three-dimensional temperature field distribution from the two-dimensional infrared image, providing basic data for subsequent thermal diffusion analysis.

[0037] like Figure 2 As shown, the battery pack thermal diffusion mechanism model module 3 includes a multi-level thermal manifold building unit 31, a manifold metric building unit 32, and a thermal flow topology analysis unit 33.

[0038] The multi-level thermal manifold building unit 31 is used to map the surface temperature distribution of a single cell to a two-dimensional curved surface, the internal temperature field to a three-dimensional manifold, combine multiple single cell manifolds into a module-level manifold through continuous mapping, and integrate multiple module manifolds into a battery pack-level manifold.

[0039] In the specific implementation, the multi-level thermal manifold building unit 31 first divides the surface of each battery cell into a triangular mesh to construct a discrete manifold representation. For a typical 18650 lithium battery, the surface mesh is divided into 1000–2000 triangular elements; for a prismatic battery, the surface mesh is divided into 2000–4000 triangular elements. Based on the surface temperature data, the internal temperature distribution is derived by solving the boundary value problem of the heat conduction equation, thus constructing a three-dimensional temperature field manifold.

[0040] The general form of the heat conduction equation is: , in, This refers to the material density, expressed in kg / m³. Specific heat capacity is expressed in J / (kg·K), temperature is expressed in K, time is expressed in s, and thermal conductivity is expressed in W / (m·K). For gradient operators, The term represents thermal diffusivity, measured in W / m³, while Q represents internal heat source, also measured in W / m³. Typical parameter values ​​for lithium-ion batteries are: , , This equation describes the change in internal temperature of a battery over time. The left side represents the rate of change of heat stored per unit volume, the first term on the right side represents the heat flow caused by heat conduction, and the second term represents the heat generated by the internal heat source.

[0041] The multi-level thermal manifold construction adopts a bottom-up strategy, first constructing the individual-level manifold, and then constructing the module-level and package-level manifolds through mapping relationships. Thermal coupling between individual units is characterized by interfacial thermal resistance, typically ranging from 0.01 to 0.05 K·m² / W. Thermal coupling between modules considers the influence of cooling channels and structural components, and is described using an equivalent thermal resistance network model.

[0042] The manifold metric building unit 32 is used to construct a Riemannian metric that reflects thermal anisotropy based on the spatial distribution of the material's thermal conductivity, and to establish an affine connection that describes the change of thermal gradient.

[0043] In Riemannian geometry, the metric tensor of a manifold defines a distance measurement in space. For a thermal flux field, the metric tensor can be expressed as the inverse of the thermal conductivity tensor: , in, To measure the components of a tensor, dimensionless, representing a distance metric on a manifold. This is the inverse of the thermal conductivity tensor, with units of (W / (m·K)). -1, where i and j are tensor indices, ranging from 1 to 3, corresponding to the three coordinate directions in three-dimensional space. For anisotropic materials, the thermal conductivity tensor is a 3×3 symmetric positive definite matrix; for isotropic materials, the thermal conductivity tensor simplifies to a scalar multiplied by the identity matrix. The physical meaning of the metric tensor is to describe the ease of heat diffusion in different directions. Directions with high thermal conductivity correspond to small values ​​of the metric tensor, indicating that heat flow is easy to pass through; directions with low thermal conductivity correspond to large values ​​of the metric tensor, indicating that heat flow is difficult to pass through.

[0044] Based on the Riemannian metric, the gradient operator on a manifold can be expressed as: , in, Let f be the i-th component of the gradient of the function f. The inverse of the tensor is a dimensionless metric. Let be the partial derivative of function f with respect to the j-th coordinate, and i and j be indices. The Einstein summation convention is used, i.e., summing over repeated subscripts and superscripts. This formula represents a method for calculating gradients in non-Euclidean space, taking into account the influence of material anisotropy on heat flux.

[0045] Affine connections describe parallel transport on a manifold. For a heat flow field, the Christopher notation for an affine connection can be represented as: , in, The Christopher notation represents the rate of change of the coordinate basis as a point moves to a neighboring point on a manifold; it is dimensionless. The inverse of the tensor is a dimensionless metric. To measure the components of a tensor, dimensionless. The partial derivative of a tensor with respect to coordinates is measured in units of m. -1 i, j, k and The value ranges from 1 to 3. The physical meaning of the Christopher sign is to describe the degree of curvature of the heat flow path in a non-uniform material. When the material properties are uniformly distributed in space, the Christopher sign is zero, and the heat flow propagates in a straight line; when the material properties are non-uniform, the Christopher sign is non-zero, and the heat flow path will be curved.

[0046] The heat flow topology analysis unit 33 is used to perform vector field decomposition on the heat flow field, identify critical points such as source points, sink points and saddle points in the heat flow field, calculate the topology index of each critical point, and identify the branch lines connecting the critical points.

[0047] The vector field of the heat flow field can be represented as a temperature gradient field: , Where v is the heat flow vector field, representing the direction and magnitude of heat flow, with units of W / m², and k is the thermal conductivity, with units of W / (m·K). The temperature gradient is expressed in K / m, and the negative sign indicates that heat flows from a high-temperature region to a low-temperature region. This formula is an expression of Fourier's law of heat conduction, describing the relationship between heat flow and temperature gradient.

[0048] According to the Helmholtz decomposition theorem, a vector field can be decomposed into the sum of an irrotational field and a divergence-free field: , Where v is the heat flux vector field, with units of W / m². A is a scalar potential function representing thermal potential, with units of W / m, and A is a vector potential function, with units of W. The gradient of the potential function is represented by the irrotational field component, and its unit is W / m². The curl of the vector potential, representing the divergence-free part, is expressed in W / m². Irrotational field. This represents pure diffusion of heat, without a scattering field. This represents the circulation of heat. This decomposition helps in analyzing the structural characteristics of the heat flow field and identifying the main modes of heat propagation.

[0049] The critical point is the point in the vector field where the velocity is zero, that is: , in, Let be the heat flow vector at point p, and 0 be the zero vector. Critical points represent special locations in the heat flow field, such as heat sources, heat sinks, or heat flow junctions.

[0050] Critical points can be classified using the eigenvalues ​​of the Jacobian matrix: If the real parts of all eigenvalues ​​are negative, then it is a sink (cooling zone). If the real part of all eigenvalues ​​is positive, then it is a source point (heat source region). If some eigenvalues ​​have positive real parts and some have negative real parts, then it is a saddle point (heat flow convergence zone). The topological index of the critical point is defined as: , in, Let p be the topological index, which is dimensionless. Let p be the Jacobian matrix of the vector field at point p, and its elements are... The unit is W / m³, det represents determinant operation, and sign represents the sign function. It returns +1 when the parameter is positive, -1 when it is negative, and 0 when it is zero. The topological exponents for source and sink points are +1, and the topological exponent for saddle points is -1. The physical meaning of the topological exponent is to describe the convergence or divergence characteristics of heat flow around the critical point. A positive topological exponent indicates convergence or divergence of heat flow, while a negative topological exponent indicates saddle point properties.

[0051] The heat flow topology analysis unit 33 is also used to calculate the Betti number under different thresholds of the heat flow field, track the generation and disappearance of heat field topology features, calculate the degree of distortion of heat flow lines, and identify abnormal heat flow patterns.

[0052] The Betti number is a topological invariant representing the number of connected structures in different dimensions. For a three-dimensional heat flux field, three Betti numbers can be calculated: Indicates the number of connected components. Indicates the number of rings (or channels). The number of cavities represents the number of dimensions, all of which are dimensionless integers. By tracking the changes in the Betty number at different temperature thresholds, changes in the topological structure of heat flux can be identified. For example, An increase indicates the emergence of a new heat source or cold source area. An increase in [something] indicates the formation of a heat flow loop, which is often an early sign of thermal runaway.

[0053] The distortion of thermal streamlines can be characterized by Frenet curvature and deflection: , , in, Frenet curvature, representing the degree of curvature of thermal flow lines, is measured in meters (m). -1 , Torque, representing the degree of distortion of thermal flow lines, is measured in meters (m). -1 v is the heat flux vector, in W / m², a is the spatial derivative (acceleration) of the heat flux vector, in W / m³, and j is the spatial derivative (rapid change rate) of the acceleration, in W / m². 4 , Represents the cross product of vectors. Represents the dot product of vectors. This represents the vector magnitude. Regions with high curvature and torsion typically correspond to regions of abnormal heat flow and are potential sources of thermal runaway. For example, when the curvature of a certain region... More than 0.5m -1 And torsion More than 0.3m -1 At that time, the area was marked as an area of ​​abnormal heat flow and required close monitoring.

[0054] Using the methods described above, the battery pack thermal diffusion mechanism model module 3 can accurately characterize the thermal flow field structure and dynamic characteristics of the battery pack, providing a theoretical basis for thermal runaway early warning and cooling decisions.

[0055] like Figure 3 As shown, the data fusion model module 4 includes a feature mapping unit 41, a multi-source data integration unit 42, and a risk assessment unit 43.

[0056] The feature mapping unit 41 is used to map the thermal field feature data output by the battery pack thermal diffusion mechanism model module 3 to a standard feature space. In the implementation process, the feature mapping adopts a manifold learning method to reduce the dimensionality of the high-dimensional thermal field features to a low-dimensional representation space. Specifically, the isomap algorithm is used to map the thermal field features to a 10-20 dimensional feature space by preserving the geodesic distance relationship.

[0057] The main steps of the isometric mapping algorithm include: 1) Construct an adjacency graph and determine the neighbors of each data point using the K-nearest neighbor or ε-neighborhood method; 2) Calculate the shortest path distance between any two points on the map, as an approximation of the geodesic distance; 3) Apply multidimensional scaling analysis (MDS) to reduce dimensionality based on the geodesic distance matrix.

[0058] For thermal field characteristics, the K-nearest neighbor method with K=10 is preferred for constructing the adjacency graph to ensure local connectivity and global consistency.

[0059] The multi-source data integration unit 42 is used to perform correlation analysis between the data collected by the multi-sensor acquisition module 1 and the thermal field feature data. The data integration adopts an attention mechanism-based Transformer architecture, which can adaptively focus on key information in different modal data.

[0060] The core of Transformer is its self-attention mechanism, calculated using the following formula: , in, The attention calculation result is a matrix with the same dimensions as V, where Q is the query matrix representing the features to be focused on, K is the key matrix representing the features that can be focused, and V is the value matrix representing the actual content of the features. This means that multiplying the query matrix by the transpose of the key matrix results in an attention score matrix. is the dimension of the key vector, used as a normalization factor to prevent excessively large attention scores. `softmax` is the softmax function, which converts the attention scores into a probability distribution. The physical meaning of the self-attention mechanism is to adaptively allocate attention weights based on the correlation between features, enabling the model to focus on key information in different modalities of data.

[0061] In the specific implementation, sensor data such as temperature, pressure, and vibration, as well as thermal field feature data, are converted into 128-dimensional feature vectors by dedicated encoders, and then fused by a 6-layer Transformer encoder to finally generate a 256-dimensional fused feature vector.

[0062] The risk assessment unit 43 is used to construct a thermal runaway risk assessment model based on the output of the multi-source data integration unit 42, and to generate thermal diffusion imbalance criteria and thresholds. The risk assessment model adopts an ensemble learning method based on gradient boosting trees, comprehensively considering factors such as temperature gradient, temperature change rate, and thermal flux topology characteristics, and outputs a thermal runaway risk score.

[0063] Criteria for thermal diffusion imbalance include: 1) Temperature gradient anomaly index (TGI): Characterizes the degree of anomaly in the distribution of temperature gradient; 2) Heat flux topological change rate (TTCR): Characterizes the rate of change of the topological structure of the heat flux field; 3) Heat source density index (HSD): Characterizes the density and intensity of local heat sources.

[0064] The formulas for calculating these criteria are as follows: , Wherein, TGI stands for Temperature Gradient Anomaly Index, which is dimensionless. The modulus of the temperature gradient is expressed in K / m. This represents the average historical temperature gradient modulus, expressed in K / m. The standard deviation of the historical temperature gradient modulus is expressed in K / m. A TGI greater than 3 is considered an anomaly in the temperature gradient, indicating that the current temperature gradient deviates significantly from the normal range, which is a possible sign of thermal runaway.

[0065] , Where TTCR is the heat flux topological change rate, in units per second. Let be the set of critical points at time t. For a moment The set of critical points, This represents a set difference measure, which calculates the number of distinct elements in two sets. The time interval is in seconds. A TTCR greater than 0.5 times / second is considered a rapid change in topology, indicating that the thermal flow field structure is changing rapidly, which may be a precursor to thermal runaway.

[0066] , Where HSD is the heat source density index, measured in W / m², and S is the set of heat sources, containing all identified heat sources. The intensity of the i-th heat source is expressed in W. The distance to the nearest battery boundary, in meters. This represents the summation of all elements in set S. An HSD greater than 100 W / m² is considered an anomaly in heat source density, indicating a concentrated and intense local heat source that may lead to rapid heat accumulation.

[0067] The risk assessment model comprehensively evaluates the risk of thermal runaway based on these criteria, outputting a risk score between 0 and 1. A risk score greater than 0.7 is considered high risk, requiring immediate cooling measures; a risk score between 0.4 and 0.7 is considered medium risk, requiring enhanced monitoring; and a risk score less than 0.4 is considered low risk, requiring routine monitoring.

[0068] like Figure 4 As shown, the cooling decision module 5 includes a thermal runaway risk classification unit 51, a cooling strategy generation unit 52, an execution control unit 53, and a feedback optimization unit 54.

[0069] Thermal runaway risk classification unit 51 is used to classify the risk of thermal runaway based on thermal diffusion imbalance criteria and thresholds. The risk classification adopts a three-level system: Level 1 Risk (Low): Risk score < 0.4, continue monitoring; Level 2 Risk (Medium): Risk score between 0.4 and 0.7, strengthen monitoring and prepare for cooling down; Level 3 Risk (High): Risk score > 0.7, immediately initiate cooling.

[0070] Furthermore, depending on the location of thermal runaway, it can be further divided into four cases: Single-cell thermal runaway: The thermal runaway occurs inside a single battery cell; Termination thermal runaway: The thermal runaway occurs in the terminal region of the battery module; Boundary thermal runaway: The thermal runaway occurs in the boundary region between modules; External thermal runaway: The thermal runaway occurs outside the battery pack.

[0071] The cooling strategy generation unit 52 is used to generate targeted cooling strategies based on the location, risk level, and thermal flow topology of thermal runaway. The cooling strategy generation employs deep reinforcement learning methods, particularly the Deep Q-Network (DQN) algorithm.

[0072] The core of DQN is the estimation of the state-action value function. , representing the long-term cumulative reward of taking action a in state s. The Q-function is updated using temporal difference learning: , in, For state Take action below The Q value represents the expected cumulative return, and is dimensionless. The learning rate controls the update step size; it is dimensionless and typically set to 0.01 or 0.1. Instant reward refers to the reward obtained immediately after taking an action; it is dimensionless. This is a discount factor that controls the importance of future returns; it is dimensionless and typically set to 0.9 or 0.99. For the next state The maximum Q value represents the maximum possible future return; it is dimensionless. This indicates an update operation. The formula describes the iterative update process of the Q-value, continuously optimizing the Q-value estimate by combining the current immediate reward and the maximum possible future reward.

[0073] In the cooling strategy generation, state s includes the current thermal flow field characteristics, thermal runaway risk score, and historical cooling effect; action a includes cooling water flow rate adjustment, spray system control, and fan power adjustment; reward r is designed based on cooling effect and energy consumption balance, and the calculation formula is: , Where r is the reward value, which is dimensionless. The value is the reduction in maximum temperature, expressed in Kelvin (K). A negative sign indicates a larger temperature reduction and a greater reward. The value is E, representing cooling energy consumption, expressed in jJ. A negative sign indicates a larger reward for lower energy consumption. This represents the reduction in risk score, is dimensionless, and a positive sign indicates that the greater the risk reduction, the greater the reward. , , The weighting coefficients, dimensionless, are set to 0.5, 0.3, and 0.2 respectively, to balance the three objectives of temperature control, energy consumption, and risk management. This reward function encourages the algorithm to find cooling strategies that are both effective in reducing temperature and conserving energy.

[0074] The execution control unit 53 is used to convert the cooling strategy into control commands for the cooling equipment, controlling the cooling water flow rate, spray system, and fan power. The control commands are sent to the execution device via CAN bus or Modbus protocol, with control accuracy of: cooling water flow rate ±0.5L / min, spray system start / stop time ±10ms, and fan power ±5%.

[0075] Different cooling strategies are implemented for different types of thermal runaway: Individual thermal runaway: The cooling device at the thermal runaway location is cooled down with maximum cooling power, and the adjacent modules are moderately cooled. The cooling water flow rate is set to 8-12 L / min, the spray system pulse frequency is set to 5 Hz, and the duty cycle is set to 60%. End-stage thermal runaway: Prioritize cooling the thermal runaway module board, dynamically adjust cooling parameters according to the thermal diffusion rate, set cooling water flow rate to 6-10 L / min, spray system pulse frequency to 3 Hz, and duty cycle to 50%; Boundary thermal runaway: Directional cooling is implemented at the thermal runaway location by controlling the corresponding cooling valves, with the cooling water flow rate set to 10-15 L / min, the spray system pulse frequency set to 8 Hz, and the duty cycle set to 80%. External thermal runaway: The external area of ​​thermal runaway is cooled by a spray system. The pulse frequency of the spray system is set to 10Hz, the duty cycle is set to 100%, and the duration is 30 to 60 seconds.

[0076] The feedback optimization unit 54 continuously optimizes the cooling parameters based on the cooling effect, improving the cooling strategy through iterative learning. Optimization employs an online learning method, adjusting the Q-network parameters in real time based on the cooling effect. The learning rate is dynamically adjusted over time, initially set to 0.01 and gradually reduced to 0.001 to ensure algorithm convergence and adaptability.

[0077] The system of the present invention also includes a thermal runaway feature library and a visualization monitoring module.

[0078] The thermal runaway feature library is communicatively connected to the battery pack thermal diffusion mechanism model module 3 and the data fusion model module 4. It stores thermal flow topology features, typical thermal runaway evolution paths, and corresponding risk ratings under various thermal runaway scenarios. The thermal runaway feature library adopts a hierarchical storage structure, including: Basic feature layer: Stores basic thermal runaway feature templates, such as temperature gradient, thermal flow topology, etc. Evolutionary path layer: stores the temporal evolution sequences of different types of thermal runaway; Risk rating layer: Stores risk ratings corresponding to different combinations of features.

[0079] The thermal runaway feature library supports incremental learning, enabling it to extract features from new thermal runaway cases and update the library content. The initial library capacity is 100 typical thermal runaway scenarios, which can be expanded to 500+ scenarios through actual operation.

[0080] The visualization monitoring module communicates with the data fusion model module 4 and the cooling decision module 5 to present the heat flow field distribution, thermal runaway risk rating, and cooling strategy in a visual manner. The visualization interface includes: Real-time display of three-dimensional heat flow field, supporting multiple display modes such as temperature, temperature gradient, and heat flow vector; The thermal runaway risk rating dashboard visually displays the current risk level. The cooling strategy execution status panel displays the current operating status and parameters of the cooling equipment; Historical data trend charts show the historical trends of key parameters.

[0081] The visual monitoring module supports web access and mobile terminal access, making it convenient for administrators to remotely monitor the system status.

[0082] Corresponding to the system embodiment, the present invention also provides a method for multimodal thermal flow field coupling early warning and cooling decision-making of energy storage battery modules, including: 1) Collect temperature, pressure and vibration data of the battery pack, module and individual cells of the energy storage battery module, as well as water pressure data of the cooling pipes in the battery pack, flow data of the inlet and outlet of the cooling pipes and humidity data of the battery pack bottom plate; 2) Process the two-dimensional planar infrared temperature data acquired in real time by the infrared thermal imager to filter out background noise; 3) Based on differential geometry theory, construct a multi-physics dynamic model of the battery pack, which is based on the electric-thermal-fluid fields. Treat the surface temperature field of the battery pack as a Riemannian manifold embedded in three-dimensional space, construct a geometric representation of the thermal flow field, identify the critical points and topological structures in the thermal flow field, and generate thermal field characteristic data. 4) Based on thermal field characteristic data, a thermal diffusion imbalance criterion and threshold are generated through multi-source data fusion; 5) Based on the thermal diffusion imbalance criterion and threshold, a reinforcement learning algorithm is used to optimize the cooling parameters, generate and execute targeted cooling strategies.

[0083] The specific implementation of the method corresponds to the implementation of each module in the system, and will not be elaborated here.

[0084] The multimodal thermal flow field coupling early warning and cooling decision system for energy storage battery modules of this invention has achieved significant results in a test at a large-scale energy storage power station. This energy storage power station uses lithium iron phosphate batteries with a total capacity of 100MWh, comprising 5000 battery modules, each containing 16 individual battery cells.

[0085] During a six-month test, the system successfully issued warnings for 27 potential thermal runaway events, providing an average of 7.5 minutes in advance and allowing ample time for the cooling system to respond. Compared to traditional thermal management systems, cooling energy consumption was reduced by 38%, and the incidence of thermal runaway events was reduced by 92%.

[0086] In particular, in a thermal runaway event caused by an internal short circuit in a single unit, the system identified early signs of thermal runaway when the temperature rose by only 2°C by analyzing changes in the thermal flow field topology. It issued an early warning 9 minutes earlier than traditional systems and implemented precise cooling, successfully preventing the spread of thermal runaway.

[0087] The embodiments described above are merely illustrative of specific implementations of the present invention, and while the descriptions are detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A multi-modal thermal flow field coupled early warning and cooling decision system for energy storage battery modules, characterized in that, The application relates to a battery pack thermal diffusion mechanism model module, which is in communication connection with the infrared thermal imaging pretreatment module and the multi-sensor acquisition module, and is used for constructing an electric-thermal-flow multi-physical field dynamic model of a battery pack based on a differential geometry theory, taking a surface temperature field of the battery pack as a Riemann manifold embedded in a three-dimensional space, constructing geometric characterization of a heat flow field, identifying critical points and topological structures in the heat flow field, and generating heat field characteristic data. The battery pack thermal diffusion mechanism model module comprises a multi-level heat flow manifold construction unit, a manifold metric construction unit, and a heat flow topological analysis unit. The infrared thermal imaging pretreatment module adopts a neural network to filter and filter temperature process images, filters out imaging noise and bad points, and obtains three-dimensional thermal diffusion data through interaction with the battery pack thermal diffusion mechanism model module. The multi-sensor acquisition module is in communication connection with the infrared thermal imaging pretreatment module and the multi-sensor acquisition module, and is used for constructing an electric-thermal-flow multi-physical field dynamic model of a battery pack based on a differential geometry theory, taking a surface temperature field of the battery pack as a Riemann manifold embedded in a three-dimensional space, constructing geometric characterization of a heat flow field, identifying critical points and topological structures in the heat flow field, and generating heat field characteristic data. The data fusion model module is in communication connection with the battery pack thermal diffusion mechanism model module, is used for receiving the heat field characteristic data, and generates a thermal diffusion imbalance criterion and a threshold value based on multi-source data fusion. The cooling decision module is in communication connection with the data fusion model module, is used for optimizing cooling parameters by using a reinforcement learning algorithm according to the thermal diffusion imbalance criterion and the threshold value, and generates and executes a targeted cooling strategy.

2. The energy storage battery module multi-modal thermal flow field coupled warning and cooling decision system of claim 1, wherein, The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor.

3. The energy storage battery module multi-modal thermal flow field coupled warning and cooling decision system of claim 1, wherein, The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor.

4. The energy storage battery module multi-modal thermal flow field coupled warning and cooling decision system of claim 1, wherein, The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor.

5. The energy storage battery module multi-modal thermal flow field coupled warning and cooling decision system of claim 1, wherein, The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor.

6. The energy storage battery module multi-modal thermal flow field coupled warning and cooling decision system of claim 1, wherein, The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow sensor and a humidity sensor. The multi-sensor acquisition module comprises a temperature sensor, a pressure sensor, a vibration sensor, a water pressure sensor, a flow 7. The energy storage battery module multi-modal thermal flow field coupled warning and cooling decision system of claim 2, wherein, The heat flow topology analysis unit is also used to calculate the Betti number of the heat flow field under different thresholds, track the generation and disappearance of the heat field topology features, calculate the degree of distortion of the heat flow lines, and identify abnormal heat flow patterns.

8. The energy storage battery module multi-modal thermal flow field coupled warning and cooling decision system of claim 1, wherein, The cooling decision module includes: a thermal runaway risk grading unit for grading the thermal runaway risk according to the heat diffusion imbalance criterion and threshold; a cooling strategy generation unit for generating a targeted cooling strategy according to the location of thermal runaway, risk level, and heat flow topology structure; an execution control unit for converting the cooling strategy into control commands of the cooling equipment to control the cooling water flow, spray system, and fan power; a feedback optimization unit for continuously optimizing the cooling parameters according to the cooling effect and improving the cooling strategy through iterative learning.

9. The energy storage battery module multi-modal thermal flow field coupled warning and cooling decision system of claim 1, wherein, The system also includes: a thermal runaway feature library in communication connection with the battery pack heat diffusion mechanism model module and the data fusion model module, used to store the heat flow topology features, typical evolution paths of thermal runaway, and corresponding risk ratings under multiple thermal runaway scenarios; a visual monitoring module in communication connection with the data fusion model module and the cooling decision module, used to present the heat flow field distribution, thermal runaway risk rating, and cooling strategy in a visual manner.

10. The method of multi-modal thermal flow field coupled warning and cooling decision for energy storage battery module, using the multi-modal thermal flow field coupled warning and cooling decision system for energy storage battery module according to any one of claims 1-9, characterized in that, It includes: collecting temperature, pressure, and vibration data on the battery pack, module, and single cell of the energy storage battery module, as well as water pressure data of the cooling pipeline in the battery pack, flow data of the cooling pipeline inlet and outlet, and humidity data of the battery pack bottom plate; processing the two-dimensional plane infrared temperature data collected by the infrared thermal imager in real time to filter out background noise; based on the theory of differential geometry, constructing an electric-thermal-flow multi-physical field dynamic model of the battery pack, regarding the surface temperature field of the battery pack as a Riemannian manifold embedded in three-dimensional space, constructing a geometric representation of the heat flow field, identifying critical points and topology structures in the heat flow field, and generating heat field feature data; based on the heat field feature data, generating a heat diffusion imbalance criterion and threshold through multi-source data fusion; according to the heat diffusion imbalance criterion and threshold, optimizing the cooling parameters using a reinforcement learning algorithm to generate and execute a targeted cooling strategy.

Images