A thermal management system and control method for a lithium-ion battery pack
By acquiring the state parameters of individual battery cells, identifying aging states and predicting future trends, a dynamic thermal prediction model is constructed, and differentiated thermal management control commands are generated. This solves the problems of safety and energy efficiency optimization in existing lithium-ion battery thermal management strategies, and achieves precise thermal management and energy consumption optimization throughout the entire life cycle.
Patent Information
- Application Number
- CN202611246721.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-17
- Publication Date
- 2026-11-17
AI Technical Summary
Existing lithium-ion battery thermal management strategies cannot differentiate and set safety boundaries, making it difficult to balance thermal safety and energy efficiency optimization throughout the entire life cycle. Furthermore, traditional methods have low computational efficiency and are difficult to reconstruct the continuous spatiotemporal temperature field inside the battery pack in real time.
By acquiring the state parameters of individual battery cells, identifying aging states and predicting future trends, a dynamic thermal prediction model is constructed, generating differentiated thermal management control commands, and using physical information neural networks and digital twins for accurate thermal field prediction and safety constraint optimization.
It achieves thermal safety assurance throughout the entire life cycle, improves the spatiotemporal resolution of thermal field sensing, reduces sensor deployment costs, and optimizes temperature uniformity and energy consumption, transforming into proactive look-ahead control.
Smart Images

Figure CN123120979A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of battery thermal management, and more particularly to a thermal management system and control method for a lithium-ion battery pack. Background Technology
[0002] Thermal management of lithium-ion batteries is crucial for their safe and efficient operation. Existing thermal management strategies are mostly based on fixed temperature thresholds or simple feedback logic, failing to consider the varying aging states of individual battery cells due to manufacturing and usage differences. This makes it impossible to differentiate safety boundaries, leaving severely aged batteries vulnerable to thermal runaway. Furthermore, traditional thermal field prediction relies on lumped parameter models or finite element simulations, which are computationally inefficient and struggle to reconstruct the continuous spatiotemporal temperature field within the battery pack in real time. In addition, existing strategies are mostly reactive controls, lacking forward-looking consideration of future battery aging trends, making it difficult to continuously balance thermal safety and energy consumption optimization throughout the entire battery lifecycle.
[0003] Therefore, there is an urgent need for an intelligent thermal management method that integrates aging state identification, accurate thermal field prediction, and safety constraint optimization. Summary of the Invention
[0004] This application provides a thermal management system and control method for a lithium-ion battery pack, which solves the technical problems of existing thermal management strategies that are difficult to balance safety and energy efficiency optimization throughout the entire life cycle and cannot achieve differentiated and precise thermal regulation.
[0005] To achieve the above objectives, this application adopts the following technical solution: In a first aspect, a control method for a thermal management system of a lithium-ion battery pack is provided, comprising: Obtain the state parameters of each battery cell in the lithium-ion battery pack, wherein the state parameters include at least voltage, current, surface temperature and state of charge; Based on the state parameters of each battery cell, the current aging state parameters of each battery cell are identified, and the future aging trend of each battery cell is predicted based on the historical aging trajectory of each battery cell. Based on the current aging state parameters and future aging trends of each battery cell, the thermal safety constraint boundaries of each battery cell are determined in a differentiated manner. Based on the state parameters of each battery cell, a dynamic thermal prediction model for the lithium-ion battery pack is constructed, and the spatiotemporal distribution of the thermal field of the lithium-ion battery pack in the future is predicted. Based on the predicted spatiotemporal distribution of the thermal field and the thermal safety constraint boundary, thermal management control instructions are generated and thermal management operations are executed. The thermal management control instructions include at least differentiated execution parameters for adjusting the cooling or heating capabilities of different spatial regions of the lithium-ion battery pack.
[0006] Based on the above technical solutions, the control method for a thermal management system of a lithium-ion battery pack provided in this application achieves differentiated and precise thermal management by integrating online identification of aging states, thermal field prediction using physical information neural networks, and safety constraint optimization control. This method sets thermal safety constraint boundaries based on the differentiated health states of each battery cell, enabling severely aged cells to obtain more conservative safety thresholds and ensuring thermal safety throughout their entire lifespan. By embedding the thermal conduction partial differential equation into a loss function using a physical information spatiotemporal graph neural network, it achieves accurate reconstruction and multi-step prediction of the continuous spatiotemporal thermal field across the entire battery pack using limited measurement point data. Through digital twin advanced extrapolation and a CBF-MPC hybrid framework, thermal safety is embedded as a hard constraint into the optimization process, simultaneously optimizing temperature uniformity and energy consumption while theoretically ensuring safety.
[0007] In conjunction with the first aspect above, in one possible implementation, the dynamic heat prediction model is a dynamic heat prediction model based on a physical information neural network, comprising: The partial differential equation of heat conduction is embedded as the loss function of the physical information neural network as a physical constraint, and the voltage, current and surface temperature of each battery cell are used as inputs to reconstruct the global continuous spatiotemporal thermal field distribution of the lithium-ion battery pack in real time, and predict the spatiotemporal evolution trend of the thermal field in the future preset time domain. The physical information neural network is a physical information spatiotemporal graph neural network. The physical information spatiotemporal graph neural network uses each battery cell as a graph node and the heat conduction path between battery cells as a graph edge. The heat conduction partial differential equation is used as a physical constraint and embedded as a loss function in the graph neural network. The network training is driven by measured data from temperature sensors to reconstruct and predict the continuous spatiotemporal temperature field within the lithium-ion battery pack in multiple steps.
[0008] In conjunction with the first aspect above, in one possible implementation, predicting the spatiotemporal distribution of the thermal field of the lithium-ion battery pack over a future period includes: Advanced simulations are performed using a pre-constructed digital twin of the lithium-ion battery pack. The digital twin takes the real-time state parameters of each battery cell as input and continuously injects measured data into the physical information neural network through a data assimilation algorithm, so that the state of the digital twin is synchronized with that of the physical battery pack in real time. Multiple candidate control strategies are run in parallel in the digital twin, and the thermal field evolution results under each strategy are simulated. Based on the simulation results, the optimal candidate control strategy is selected, and the thermal field evolution result corresponding to the optimal candidate control strategy is determined as the predicted thermal field spatiotemporal distribution.
[0009] In conjunction with the first aspect above, in one possible implementation, the method for determining the thermal safety constraint boundary includes: The health status value of each battery cell is determined based on the current aging state parameters of each individual battery cell. A safety threshold is then set for each battery cell based on the differences in its health status value. A lower health status value corresponds to a more conservative safety threshold. Using the safety threshold as the boundary parameter of the safety state set, a control barrier function is used to construct the thermal safety constraint boundary. This control barrier function satisfies the following condition: for any given time, if the current thermal state of each battery cell is within the safety state set defined by the control barrier function, then the control input generated under the constraint of the control barrier function ensures that the thermal state of each battery cell remains within the safety state set at all subsequent times.
[0010] In conjunction with the first aspect above, in one possible implementation, the process of generating thermal management control commands includes: A hybrid framework of control barrier function and model predictive control is adopted to solve the multi-objective optimization problem. The control barrier function serves as a hard constraint to ensure the thermal safety of the lithium-ion battery pack throughout its entire life cycle, while model predictive control acts as an optimizer to optimize the energy consumption and temperature uniformity objectives of the thermal management system within the feasible region defined by the hard constraint.
[0011] In conjunction with the first aspect above, in one possible implementation, the constraints of the multi-objective optimization problem include: s Under the low-temperature preheating mode, the lithium plating risk coefficient of each battery cell is lower than the preset lithium plating risk threshold. The lithium plating risk coefficient of the i-th battery cell... Calculated using the following formula: ; in, Based on the current health status value of the i-th battery cell Determined lithium plating sensitivity coefficient; ν is the molar activation energy of the lithium deposition reaction; Q is the ideal gas constant; Let be the current temperature of the i-th battery cell; Let be the current charge / discharge current of the i-th battery cell; Reference current; The current influence index; This represents the current state of charge of the i-th battery cell; For reference state of charge; The state of charge influence index.
[0012] In conjunction with the first aspect above, in one possible implementation, predicting the future aging trend of each battery cell based on its historical aging trajectory includes: The internal resistance growth curve and capacity decay curve of each battery cell are obtained throughout its entire life cycle. A time-series prediction model is used to predict the expected internal resistance and capacity of each battery cell after a preset number of cycles based on the internal resistance growth curve and capacity decay curve. When generating the thermal management control command, the cooling resource distribution is increased in advance for battery cells that are predicted to experience accelerated aging, based on the expected internal resistance and capacity values.
[0013] In conjunction with the first aspect above, in one possible implementation, the differentiated execution parameters include a reconfigurable flow channel topology switching command for the cooling system and a cooling medium flow distribution command for each cooling zone; the cooling system includes multiple cooling medium flow channels whose on / off state and flow direction can be independently controlled, and each cooling medium flow channel forms a grid-like segmented network inside the lithium-ion battery pack.
[0014] In conjunction with the first aspect above, in one possible implementation, the step of generating the reconfigurable flow channel topology switching instruction includes: Using the current flow channel topology as the initial state, and at the cost of the operating energy consumption of each flow channel switching valve, the target flow channel topology is solved under the constraint of the thermal safety constraint boundary to achieve the optimal combination of temperature uniformity, cooling medium transport energy consumption, and topology switching energy consumption. Based on the target flow channel topology, the on / off state and flow direction of each cooling medium flow channel are determined so that the spatial topology of the cooling medium flow channel matches the spatial distribution of heat generation intensity in the spatiotemporal distribution of the thermal field.
[0015] Secondly, this application provides a thermal management system for a lithium-ion battery pack, comprising: a state parameter acquisition module, an aging state identification and prediction module, a thermal safety constraint boundary determination module, a dynamic thermal prediction module, a control command generation module, and a thermal management execution module; wherein, the state parameter acquisition module is used to acquire the state parameters of each battery cell in the lithium-ion battery pack; the aging state identification and prediction module is used to identify the current aging state parameters of each battery cell based on the state parameters of each battery cell, and predict the future aging trend of each battery cell based on the historical aging trajectory of each battery cell; the thermal safety constraint boundary determination module is used to differentiately determine the thermal safety constraint boundary of each battery cell based on the current aging state parameters and the future aging trend of each battery cell; the dynamic thermal prediction module is used to construct a dynamic thermal prediction model of the lithium-ion battery pack based on the state parameters of each battery cell, and predict the spatiotemporal distribution of the thermal field of the lithium-ion battery pack in a future period; the control command generation module is used to generate thermal management control commands based on the predicted spatiotemporal distribution of the thermal field and the thermal safety constraint boundaries; and the thermal management execution module is used to execute thermal management operations according to the thermal management control commands.
[0016] This application provides a thermal management system and control method for a lithium-ion battery pack, which can achieve the following beneficial effects: First, by identifying the current aging state of each battery cell online and predicting its future aging trend, the thermal safety constraint boundary of each battery cell is determined differently, so that battery cells with lower health state values obtain a more conservative safety threshold, thereby providing thermal safety assurance for each battery cell adapted to its aging degree throughout its entire life cycle; Second, by constructing a dynamic thermal prediction model based on a physical information neural network, the partial differential equation of heat conduction is embedded as a physical constraint loss function, and the battery is controlled using limited measurement point data. The precise reconstruction and multi-step prediction of the continuous spatiotemporal thermal field across the entire domain significantly improves the spatiotemporal resolution of thermal field perception and reduces sensor deployment costs. Furthermore, by using digital twins for advanced extrapolation and parallel strategy selection, a shift from passive response control to proactive look-ahead control is achieved, completing multi-strategy effect evaluation and optimization before control commands are executed. Finally, through a hybrid framework of control barrier function-model predictive control, thermal safety constraints are embedded as hard constraints into the optimization process, theoretically ensuring safety while simultaneously optimizing temperature uniformity and system energy consumption, effectively balancing thermal safety, comfort, and economy.
[0017] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description
[0018] Figure 1 A system architecture diagram of a thermal management system for a lithium-ion battery pack provided in this application embodiment; Figure 2 A flowchart illustrating a control method for a thermal management system of a lithium-ion battery pack provided in an embodiment of this application; Figure 3 A flowchart illustrating a method for thermal field prediction and strategy selection based on digital twins, provided in an embodiment of this application; Figure 4 A flowchart illustrating a future aging trend prediction and forward-looking thermal management method provided in an embodiment of this application; Figure 5 This is a flowchart illustrating a reconfigurable cooling channel topology optimization method provided in an embodiment of this application. Detailed Implementation
[0019] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0020] The control method for a thermal management system of a lithium-ion battery pack provided in this application embodiment can be applied to, for example... Figure 1 The thermal management system of the lithium-ion battery pack shown includes: a state parameter acquisition module 101, an aging state identification and prediction module 102, a thermal safety constraint boundary determination module 103, a dynamic thermal prediction module 104, a control command generation module 105, and a thermal management execution module 106.
[0021] The state parameter acquisition module 101 is used to acquire the state parameters of each battery cell in the lithium-ion battery pack. The state parameters include at least voltage, current, surface temperature and state of charge.
[0022] The aging state identification and prediction module 102 is used to identify the current aging state parameters of each battery cell based on the state parameters of each battery cell, and to predict the future aging trend of each battery cell based on the historical aging trajectory of each battery cell.
[0023] The thermal safety constraint boundary determination module 103 is used to differentiate the thermal safety constraint boundary of each battery cell based on the current aging state parameters and the future aging trend of each battery cell.
[0024] The dynamic thermal prediction module 104 is used to construct a dynamic thermal prediction model of the lithium-ion battery pack based on the state parameters of each battery cell, and to predict the spatiotemporal distribution of the thermal field of the lithium-ion battery pack in the future.
[0025] The control command generation module 105 is used to generate thermal management control commands based on the predicted spatiotemporal distribution of the thermal field and the thermal safety constraint boundary.
[0026] The thermal management execution module 106 is used to perform thermal management operations according to the thermal management control instructions, the thermal management control instructions including at least differentiated execution parameters for adjusting the cooling or heating capacity of different spatial regions of the lithium-ion battery pack.
[0027] To address the technical problems of existing thermal management strategies failing to simultaneously consider safety and energy efficiency optimization throughout the entire lifecycle and lacking differentiated and precise thermal control, this application provides a control method for a thermal management system of a lithium-ion battery pack. This method includes: acquiring the state parameters of each individual battery cell in the lithium-ion battery pack; identifying the current aging state and predicting future aging trends based on the state parameters of each individual battery cell; differentially determining the thermal safety constraint boundaries of each individual battery cell; constructing a dynamic thermal prediction model to predict the spatiotemporal distribution of the thermal field in future time periods; and generating thermal management control commands based on the thermal field distribution and thermal safety constraints, and executing thermal management operations. Based on this, an integrated closed-loop management system is achieved for the precise perception, forward prediction, and safety optimization control of the thermal state of the lithium-ion battery pack throughout its entire lifecycle.
[0028] like Figure 2 As shown in the embodiment of this application, a control method for a thermal management system of a lithium-ion battery pack includes: S201. Obtain the state parameters of each individual cell in the lithium-ion battery pack.
[0029] The state parameters include at least voltage, current, surface temperature, and state of charge (SOC). These state parameters are acquired in real time using temperature sensors located on the surface of each battery cell, current sensors connected in series in the circuit, and voltage sensors connected in parallel across each battery cell. The sampling frequency can be dynamically adjusted according to actual operating conditions; for example, the sampling frequency can be increased under high-heat-generating conditions such as fast charging or high-rate discharging.
[0030] It should be noted that the state of charge (SOC) can be estimated by combining the ampere-hour integration method with the open-circuit voltage correction method to ensure the accuracy of the SOC estimate under different operating conditions.
[0031] S202. Based on the state parameters of each battery cell, identify the current aging state parameters of each battery cell, and predict the future aging trend of each battery cell based on the historical aging trajectory of each battery cell.
[0032] The current aging parameters include at least the State of Health (SOH) and internal resistance of each battery cell. SOH can be determined by comparing the ratio of current usable capacity to nominal capacity, or by combining this with an assessment of internal resistance growth. Specifically, an equivalent circuit model (ECM) can be used to identify the parameters of each battery cell. Using voltage, current, and SOC as model inputs, algorithms such as recursive least squares or Kalman filtering are used to estimate the ohmic and polarization internal resistance of the battery online. Then, the estimated SOH value is calculated based on the mapping relationship between internal resistance growth and capacity decay.
[0033] It should be noted that, for acquiring historical aging trajectories, the system continuously records the internal resistance and capacity data of each battery cell throughout the entire lifecycle of the battery pack, forming an aging trajectory database in time-series format. Based on this, a time-series prediction model (such as a Long Short-Term Memory (LSTM) network or a gated recurrent unit (GRU)) is used to predict the expected internal resistance and capacity values of each battery cell after a preset number of cycles, according to the internal resistance growth curve and capacity decay curve. This time-series prediction model is trained offline using a large amount of battery aging experimental data to learn the nonlinear evolution of internal resistance and capacity with the number of cycles, thereby achieving accurate prediction of future aging trends online.
[0034] S203. Based on the current aging state parameters and future aging trends of each battery cell, determine the thermal safety constraint boundary of each battery cell in a differentiated manner.
[0035] The method for determining the thermal safety constraint boundary is as follows: First, the health state value of each battery cell is determined based on its current aging state parameters. Then, a safety threshold is set for each battery cell based on the differences in its health state value, with a more conservative safety threshold corresponding to a lower health state value. Specifically, for battery cells with a state of equilibrium (SOH) below 80%, the upper limit of their safe temperature is reduced by 5°C to 10°C compared to battery cells with an SOH above 90%. Using the safety threshold as the boundary parameter of the safety state set, a control barrier function (CBF) is used to construct the thermal safety constraint boundary. The control barrier function satisfies the following condition: for any given time, if the current thermal state of each battery cell is within the safety state set defined by the control barrier function, then the control input generated under the constraint of the control barrier function ensures that the thermal state of each battery cell remains within the safety state set at all subsequent times.
[0036] The construction method for the control barrier function is as follows: For the i-th cell in a lithium-ion battery pack, its thermal dynamics can be expressed as the following lumped-parameter thermal model: ; in, Let be the heat capacity of the i-th battery cell; Let be the rate of change of temperature of the i-th cell over time; Let be the heat generation power of the i-th battery cell; The heat power carried away by the cooling system from the i-th battery cell; Let represent the heat power lost by the i-th battery cell to the surrounding environment. This thermal model describes the physical law of temperature change in a battery cell: the product of heat capacity and temperature rise rate equals the heat generation power minus the heat dissipation power, that is, the rate of change of the battery's internal energy is determined by the net effect of heat generation and heat dissipation.
[0037] Define the safety state set of the i-th battery cell as ;in, Let be the instantaneous temperature of the i-th battery cell. To determine the health status value of the i-th battery cell Differentiated safety temperature thresholds. With The reduction, The corresponding reduction reflects the safety requirements of aging batteries being more sensitive to temperature.
[0038] Constructing control barrier functions as follows: ; According to the theory of control barrier functions, if an extended class exists... function This makes the following equation true: Then the set This is a forward-invariant set, meaning that under the constraint of the control barrier function, as long as the current battery temperature is within the safe state set, applying a control input that satisfies the above inequality constraint can guarantee that the battery temperature will always remain within the safe state set in all future moments. This is the first derivative with respect to time. In practical implementation, It can be taken as a linear function. The larger its value, the faster the system response and the stronger the conservatism of the safety boundary.
[0039] because Substituting into the above formula, we get: ,Right now ; Combining the aforementioned lumped-parameter thermal model, the effect of cooling power can be obtained. Constraints: ; This constraint is embedded as a hard constraint in the subsequent control command generation process to ensure that the cooling power output by the control system at any time is not lower than the lower limit value, thereby theoretically ensuring that the battery temperature is always within the safe state set.
[0040] S204. Based on the state parameters of each battery cell, construct a dynamic thermal prediction model for the lithium-ion battery pack and predict the spatiotemporal distribution of the thermal field of the lithium-ion battery pack in the future.
[0041] The dynamic thermal prediction model is based on a Physics-Informed Neural Network (PINN). The PINN takes the voltage, current, and surface temperature of each battery cell as input, embeds the partial differential equation of heat conduction as a physical constraint into the loss function of the PINN, and reconstructs the global continuous spatiotemporal thermal field distribution of the lithium-ion battery pack in real time, predicting the spatiotemporal evolution trend of the thermal field within a preset time domain.
[0042] The specific structure and training method of the physical information spatiotemporal graph neural network are explained in detail below.
[0043] The physical information neural network is preferably a Physics-Informed Spatial-Temporal Graph Neural Network. This network constructs a graph structure with each battery cell as a graph node and the heat conduction paths between battery cells as graph edges. For a battery pack containing N battery cells, a graph is defined... Among them, the set of nodes For N individual battery cells, the edge set This represents the thermal conduction path relationship between battery cells. If two battery cells are physically adjacent or connected through a thermally conductive structure, there is a graph edge between them. The weight of the edge is determined based on the thermal conductivity coefficient and geometric distance between the battery cells.
[0044] The network's input features include the voltage, current, and surface temperature of each node (cell). In the time dimension, a sliding window approach is used to input time-series data from past periods. The network captures the spatial thermal coupling relationships between cells through graph convolutional layers and captures the temporal evolution of the thermal field through temporal processing layers (such as gated recurrent units (GRUs) or temporal convolutional networks (TCNs). To obtain temperature predictions for continuous spatial locations between nodes, this application introduces a spatial coordinate encoding layer on top of the spatiotemporal graph neural network: the spatial coordinates of each cell are input as additional node features to the network, and a fully connected decoder is added after the graph convolution operation. This decoder takes the coordinates and temporal features of the target spatial location as input, outputs the temperature value at that location at the corresponding time, and finally outputs the future temperature value. The predicted temperature values at each node and at continuous spatial locations between nodes within a time period are used to reconstruct and predict the continuous spatiotemporal thermal field across the entire domain.
[0045] The specific method of embedding physical information is as follows: the three-dimensional heat conduction partial differential equation is added as a physical constraint term to the network's loss function. The three-dimensional heat conduction equation is: ; in, For the density of battery materials, For isobaric specific heat capacity, Let x be the temperature at spatial location x and time t. Thermal conductivity, Let x be the volumetric heat production rate at spatial location x and time t.
[0046] It should be noted that in the above formula... Let be the Hamiltonian operator, representing the partial derivative operation with respect to three-dimensional spatial coordinates. Wherein, It represents the temperature gradient, pointing in the direction of the fastest spatial increase in temperature. The divergence of the thermal flux density represents the net amount of heat conducted out through the surface of a point element. Thus, this partial differential equation actually reveals the physical process of energy conservation: "The change in energy within the battery element (left side) equals the internal heat generation (second term on the right side) minus the net heat dissipated to the outside (negative value of the first term on the right side)".
[0047] The overall loss function of the network It consists of three parts: ; in, The data fitting loss measures the deviation between the network's predicted temperature and the actual temperature sensor measurement. To compensate for the residual loss of the partial differential equation, the residual of the above heat conduction equation is calculated at several collocation points and added as a penalty term to the loss function, forcing the network's prediction results to satisfy the physical laws of heat conduction. To compensate for boundary condition losses, ensure that the temperature field predicted by the network satisfies the thermal boundary conditions of the battery pack (such as adiabatic boundary or convective heat transfer boundary). , and These are the weighting coefficients for the data fitting loss, the partial differential equation residual loss, and the boundary condition loss, respectively, used to balance the relative importance of data-driven factors and physical constraints. By minimizing this loss function during network training, the network can output a global temperature field distribution that conforms to both the measured data and physical laws, driven by a small amount of measured temperature data.
[0048] S205. Based on the predicted spatiotemporal distribution of the thermal field and the thermal safety constraint boundary, generate thermal management control commands and execute thermal management operations.
[0049] The thermal management control command includes at least differentiated execution parameters for adjusting the cooling or heating capacity of different spatial regions of the lithium-ion battery pack.
[0050] In the process of generating thermal management control commands, a control barrier function-model predictive control hybrid framework (CBF-MPC Hybrid Framework) is used to solve the multi-objective optimization problem. The control barrier function serves as a hard constraint to ensure the thermal safety of the lithium-ion battery pack throughout its entire lifespan, while model predictive control (MPC) acts as an optimizer to optimize the energy consumption and temperature uniformity objectives of the thermal management system within the feasible region defined by the hard constraints.
[0051] Specifically, in each control cycle, MPC solves the following finite-time optimization problem: The thermal management control command includes at least differentiated execution parameters for adjusting the cooling or heating capacity of different spatial regions of the lithium-ion battery pack.
[0052] In the process of generating thermal management control commands, a control barrier function-model predictive control hybrid framework (CBF-MPC Hybrid Framework) is used to solve the multi-objective optimization problem. The control barrier function serves as a hard constraint to ensure the thermal safety of the lithium-ion battery pack throughout its entire lifespan, while model predictive control (MPC) acts as an optimizer to optimize the energy consumption and temperature uniformity objectives of the thermal management system within the feasible region defined by the hard constraints.
[0053] Specifically, in each control cycle, MPC solves the following finite-time optimization problem: ; The constraints are: , , , ; in, This represents the start time of the k-th control cycle. For at any time The predicted control input sequence for the j-th future time, j=0,1,... ; For predicting discrete-time indexes in the time domain; For prediction in the time domain; For the i-th battery cell at time... The predicted temperature; For the i-th battery cell at time... The predicted temperature The target reference temperature, For the cooling system at all times power consumption, The rated power of the cooling system, and These are the dimensionless weighting coefficients for the temperature uniformity target and the energy consumption target, respectively. The thermal dynamic model of the i-th battery cell is obtained by first-order forward Euler discretization of the aforementioned continuous-time lumped-parameter thermal model. The discretization time step is preset according to the sampling period of the temperature sensor, and this value is less than the control period of the thermal management system to ensure that at least one thermal state recursive prediction is completed within one control period. This indicates the control input sequence corresponding to future times. The single-step control input vector, The aforementioned control barrier function, and These are the lower and upper bound vectors of the control input, respectively. The core idea of this optimization problem is: under the thermal safety constraints guaranteed by the control barrier function... The goal is to find the optimal control sequence that makes the battery pack temperature as uniform as possible (first objective) and the cooling system can be depleted as little as possible (second objective).
[0054] Based on the above technical solutions, the control method for the thermal management system of the lithium-ion battery pack provided in this application introduces an aging state identification and prediction mechanism, enabling the thermal management strategy to adaptively adjust the safety constraint boundary as the battery aging degree deepens, overcoming the deficiency of insufficient safety in the later stages of the battery's entire life cycle by the traditional fixed threshold strategy; through the thermal field prediction driven by the physical information neural network, the function of reconstructing the continuous spatiotemporal thermal field of the entire battery pack can be achieved by relying only on temperature data from a limited number of measurement points, which significantly reduces the sensor deployment cost while improving the spatiotemporal resolution of thermal field perception; through the CBF-MPC hybrid framework, safety is incorporated into the optimization process as a hard constraint rather than a soft penalty, fundamentally ensuring the safety feasibility of the control strategy under any operating condition.
[0055] Furthermore, in this embodiment, the constraints for the multi-objective optimization problem involved in generating thermal management control commands also include that the lithium plating risk coefficient of each battery cell is lower than a preset lithium plating risk threshold under the low-temperature preheating mode. The lithium plating risk coefficient of the i-th battery cell... Calculated using the following formula: ; in, Based on the current health status value of the i-th battery cell A given lithium plating sensitivity coefficient, which varies with The decrease and increase reflect the physical fact that aged batteries are more prone to lithium plating due to the loss of negative electrode active material and the thickening of the SEI film. is the molar activation energy of the lithium deposition reaction, characterizing the energy barrier that the lithium deposition reaction needs to overcome; Q is the ideal gas constant (8.314 J / (mol·K)); The current temperature of the i-th cell is given; the lower the temperature, the higher the exponent. The larger the value, the higher the risk of lithium plating. This index term accurately describes the Arrhenius-type effect of temperature on the lithium plating reaction rate. The current charge / discharge current of the i-th battery cell is denoted as . The larger the charging current, the more severe the negative electrode polarization and the stronger the tendency for lithium plating. The reference current is used to dimensionless current variables; The current influence index is determined by fitting experimental data, and its value is usually taken as 1 to 2. The current state of charge of the i-th battery cell is as follows: under high SOC, the degree of lithium intercalation in the negative electrode is high, and the space for lithium plating is reduced. For reference state of charge; The state-of-charge (POC) influence index was determined through fitting experimental data. This formula, based on the Arrhenius and Butler-Volmer equations in electrochemical reaction kinetics, quantitatively characterizes the combined impact of four core factors—aging degree, temperature, current, and POC—on lithium plating risk. Under low-temperature preheating mode, by constraining this POC risk coefficient below a preset threshold, it ensures that the heating strategy will not induce lithium plating due to excessive pursuit of heating rate.
[0056] In one possible implementation of the embodiments of this application, such as Figure 3 As shown, the above S204 (predicting the spatiotemporal distribution of the thermal field) can be specifically implemented through the following S301-S303, which are explained in detail below: S301. Perform advanced simulation using a pre-constructed digital twin of the lithium-ion battery pack.
[0057] The digital twin is a virtual mirror model of the physical battery pack, constructed in virtual space using the real-time state parameters of each individual battery cell as input. The construction of the digital twin includes the following steps: First, a high-fidelity multiphysics simulation model is established based on the battery pack's geometry, material properties, and thermophysical parameters. This model encompasses electrochemical heat generation, heat conduction, and convection heat dissipation models. Second, measured data is continuously injected into the physical information neural network using a data assimilation algorithm, ensuring real-time synchronization between the digital twin and the physical battery pack's state.
[0058] The specific data assimilation process is as follows: At each sampling moment, the system collects the measured values of voltage, current, and surface temperature of each battery cell, compares these measured values with the previous step prediction values of the physical information neural network, and calculates the residuals. Then, data assimilation algorithms such as ensemble Kalman filter (EnKF) or particle filter are used to correct the state variables of the digital twin (including the temperature field distribution and internal state variables of each node) based on the residuals, so that the state of the virtual model continuously converges to the real state of the physical entity. Through this continuous data assimilation process, the digital twin can track the state changes of the physical battery pack in real time, and can still maintain high fidelity even when there are parameter uncertainties in the model or sudden changes in operating conditions.
[0059] S302. Run multiple candidate control strategies in parallel in the digital twin and deduce the thermal field evolution results under each strategy.
[0060] The candidate control strategies are generated as follows: based on the current thermal field state and the predicted future thermal field trend, multiple different candidate control strategies are generated by random sampling within the feasible domain of the control input or by heuristic rules. Each candidate strategy corresponds to a set of cooling / heating execution parameters (such as the on / off state of each flow channel and the cooling medium flow distribution in each region). The digital twin utilizes its built-in physical information neural network thermal prediction model to rapidly extrapolate the thermal field evolution results of each candidate strategy in the future time domain, obtaining multiple sets of different thermal field evolution trajectories.
[0061] S303. Based on the deduction results, the optimal candidate control strategy is selected, and the thermal field evolution result corresponding to the optimal candidate control strategy is determined as the predicted thermal field spatiotemporal distribution.
[0062] The selection of the optimal candidate control strategy is based on a comprehensive evaluation of the temperature uniformity indicators (such as the standard deviation or maximum temperature difference of each battery cell) and energy consumption indicators (such as cooling medium pumping power consumption) of each candidate strategy in the prediction time domain. A weighted summation method or Pareto optimality method is used to determine the strategy with the best overall performance. The thermal field evolution trajectory corresponding to the optimal strategy is output as the final predicted spatiotemporal distribution of the thermal field. Simultaneously, the control parameters of this optimal strategy are passed to the control command generation module for execution of actual thermal management operations.
[0063] Based on the above technical solution, through advanced extrapolation using digital twins and parallel strategy selection, this application achieves a transformation from "passive response" thermal management to "proactive look-ahead" thermal management. Before control commands are actually applied to the physical battery pack, the system has already completed the effect evaluation and optimization of multiple possible control strategies in virtual space, thereby significantly reducing trial-and-error costs and improving the reliability and optimality of control decisions.
[0064] In one possible implementation of the embodiments of this application, such as Figure 4 As shown, the "predicting the future aging trend of each battery cell based on its historical aging trajectory" in S202 above can be specifically achieved through the following S401-S402, which are explained in detail below: S401. Obtain the internal resistance growth curve and capacity decay curve of each battery cell throughout its entire life cycle.
[0065] Among them, the internal resistance growth curve records the trajectory of the internal resistance value of each battery cell from its first use to the present moment as a function of the number of cycles or the usage time, while the capacity decay curve records the trajectory of the available capacity decay as a function of the number of cycles or the usage time. These two curves are the most critical indicators of battery aging. The system continuously collects and stores this data throughout the entire operation of the battery pack, forming a unique aging "fingerprint" profile for each battery cell.
[0066] S402. Using a time-series prediction model, predict the expected internal resistance and capacity of each battery cell after a preset number of cycles based on the internal resistance growth curve and capacity decay curve.
[0067] The preferred time-series prediction model is a Long Short-Term Memory (LSTM) network or a Gated Recurrent Unit (GRU), as these networks are particularly adept at handling long-term dependencies in time-series data. The model input is the internal resistance sequence over the past L cycles. and capacity sequence The output is the predicted internal resistance value after Z future cycles. and capacity forecast The model is trained offline using a large amount of battery aging experimental data. The training data covers diverse aging conditions such as different temperatures, different charge / discharge rates, and different cutoff voltages to ensure the model's generalization ability in different usage scenarios.
[0068] Based on the above technical solution, by predicting the future aging trend of the battery in advance through the time-series prediction model, the system can increase the distribution of cooling resources for battery cells that are predicted to age faster when the thermal management control command is generated, according to the predicted internal resistance and capacity values. This achieves proactive thermal management to prevent aging batteries from entering an unsafe state due to delayed thermal management.
[0069] In one possible implementation of the embodiments of this application, such as Figure 5 As shown, after S205, the control method of the thermal management system of the lithium-ion battery pack provided in this application embodiment further includes the following S501-S502: S501. Taking the current flow channel topology as the initial state, and at the cost of the operating energy consumption of each flow channel switching valve, under the constraints of the thermal safety constraint boundary, solve for the target flow channel topology that comprehensively optimizes temperature uniformity, cooling medium transport energy consumption, and topology switching energy consumption.
[0070] The cooling system includes multiple independently controllable cooling medium channels, each forming a grid-like segmented network within the lithium-ion battery pack. Each channel is equipped with an independent solenoid or electric valve to achieve independent control of its on / off state and flow direction. The state encoding method for each channel is as follows: on / off state is represented by 1 bit (1 for on, 0 for off), and flow direction state is represented by 1 bit (1 for forward flow, 0 for reverse flow), requiring a total of 2 bits for each channel. (Complete state vector) Where M is the total number of cooling medium channels, and each component can be configured independently.
[0071] The topology optimization problem can be formulated as follows: given the current flow channel topology state (0 / 1 state vectors of each flow channel's on / off state and flow direction) to find the target topology state. This minimizes the following combined cost function: ; Simultaneously satisfying thermal safety constraint boundaries: ; in, Topological state The standard deviation of the temperature of each individual cell in the battery pack characterizes the temperature uniformity. The reference temperature standard deviation is used for dimensionless temperature uniformity index. The pump for delivering cooling medium is in topology state The instantaneous power is used to characterize the energy consumption level of the cooling medium transport. To start from the current topology Switch to target topology The required valve operating energy consumption includes the product of the number of switching operations of each switching valve and the energy consumption of a single operation; For reference switching energy consumption, used to dimensionless topology switching energy consumption; , , These are the dimensionless weighting coefficients for the targets of temperature uniformity, transmission energy consumption, and switching energy consumption, respectively.
[0072] S502. Determine the on / off state and flow direction of each cooling medium channel according to the target flow channel topology, so that the spatial topology of the cooling medium channel matches the spatial distribution of heat generation intensity in the spatiotemporal distribution of the thermal field.
[0073] In this context, "matching" means that in areas with high heat generation intensity (i.e., spatial locations with higher temperatures or faster temperature rise rates in the spatial distribution of the thermal field), the corresponding cooling channels remain open with a larger distribution of cooling medium flow rate; in areas with low heat generation intensity, the corresponding cooling channels can be appropriately shut off or have their flow rate reduced, thereby achieving "on-demand allocation" of cooling resources. Through the reconfigurable design of the channel topology, the spatial layout of the cooling system is no longer fixed but can be adjusted in real time according to the dynamic changes in the thermal field, realizing dynamic optimization of the spatial distribution of cooling capacity.
[0074] Based on the above technical solutions, the reconfigurable flow channel topology allows the spatial cooling capacity distribution of the cooling system to be dynamically adjusted with the spatiotemporal evolution of the thermal field, which greatly improves the flexibility and energy efficiency of cooling resource utilization compared to a fixed flow channel structure.
[0075] In one possible implementation, after S203 and before S205, the control method of the thermal management system of the lithium-ion battery pack provided in this application embodiment further includes the following S601-S602: S601. Identify hot and cold areas in the lithium-ion battery pack based on the spatiotemporal distribution of the thermal field.
[0076] Hotspot regions are defined as continuous spatial areas where the predicted temperature is higher than the average battery pack temperature plus a preset temperature difference, while coldspot regions are defined as continuous spatial areas where the predicted temperature is lower than the average battery pack temperature minus a preset temperature difference. Identifying hotspot regions is crucial for thermal management decisions because hotspots are often potential sources of thermal runaway.
[0077] S602. Based on the location and severity of the hot spot and cold spot areas, the cooling medium flow rate of each cooling area is configured differently when generating the thermal management control command.
[0078] Specifically, for hot spots, the cooling medium flow rate is increased to enhance heat dissipation; for cold spots, in low-temperature environments, the cooling medium flow rate can be appropriately reduced or heating can be activated to avoid performance degradation due to excessive cooling. The specific values for differentiated flow configurations are determined uniformly through the aforementioned CBF-MPC optimization framework.
[0079] Based on the above technical solution, this application achieves spatial fine-grained allocation of thermal management resources by accurately identifying hot and cold spots and configuring differentiated flow rates, thereby minimizing unnecessary cooling energy consumption while ensuring thermal safety.
[0080] Some of the data in the above formula are calculated by removing dimensions and taking their numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.
Claims
1. A control method for a thermal management system of a lithium-ion battery pack, characterized in that, include: Obtain the state parameters of each battery cell in the lithium-ion battery pack, wherein the state parameters include at least voltage, current, surface temperature and state of charge; Based on the state parameters of each battery cell, the current aging state parameters of each battery cell are identified, and the future aging trend of each battery cell is predicted based on the historical aging trajectory of each battery cell. Based on the current aging state parameters and future aging trends of each battery cell, the thermal safety constraint boundaries of each battery cell are determined in a differentiated manner. Based on the state parameters of each battery cell, a dynamic thermal prediction model for the lithium-ion battery pack is constructed, and the spatiotemporal distribution of the thermal field of the lithium-ion battery pack in the future is predicted. Based on the predicted spatiotemporal distribution of the thermal field and the thermal safety constraint boundary, thermal management control instructions are generated and thermal management operations are executed. The thermal management control instructions include at least differentiated execution parameters for adjusting the cooling or heating capabilities of different spatial regions of the lithium-ion battery pack.
2. The control method for a thermal management system of a lithium-ion battery pack according to claim 1, characterized in that, The dynamic heat prediction model is a dynamic heat prediction model based on a physical information neural network, including: The partial differential equation of heat conduction is embedded as the loss function of the physical information neural network as a physical constraint, and the voltage, current and surface temperature of each battery cell are used as inputs to reconstruct the global continuous spatiotemporal thermal field distribution of the lithium-ion battery pack in real time, and predict the spatiotemporal evolution trend of the thermal field in the future preset time domain. The physical information neural network is a physical information spatiotemporal graph neural network. The physical information spatiotemporal graph neural network uses each battery cell as a graph node and the heat conduction path between battery cells as a graph edge. The heat conduction partial differential equation is used as a physical constraint and embedded as a loss function in the graph neural network. The network training is driven by measured data from temperature sensors to reconstruct and predict the continuous spatiotemporal temperature field within the lithium-ion battery pack in multiple steps.
3. The control method for a thermal management system of a lithium-ion battery pack according to claim 2, characterized in that, The prediction of the spatiotemporal distribution of the thermal field of the lithium-ion battery pack in the future includes: Advanced simulations are performed using a pre-constructed digital twin of the lithium-ion battery pack. The digital twin takes the real-time state parameters of each battery cell as input and continuously injects measured data into the physical information neural network through a data assimilation algorithm. Multiple candidate control strategies are run in parallel within the digital twin, and the thermal field evolution results under each strategy are simulated. Based on the simulation results, the optimal candidate control strategy is selected, and the thermal field evolution result corresponding to the optimal candidate control strategy is determined as the predicted thermal field spatiotemporal distribution.
4. The control method for a thermal management system of a lithium-ion battery pack according to claim 1, characterized in that, The method for determining the thermal safety constraint boundary includes: The health status value of each battery cell is determined based on the current aging state parameters of each individual battery cell. A safety threshold is then set for each battery cell based on the differentiated health status values. Using the safety threshold as the boundary parameter of the safety state set, a control barrier function is used to construct the thermal safety constraint boundary. This control barrier function satisfies the following condition: for any given time, if the current thermal state of each battery cell is within the safety state set defined by the control barrier function, then the control input generated under the constraint of the control barrier function ensures that the thermal state of each battery cell remains within the safety state set at all subsequent times.
5. The control method for a thermal management system of a lithium-ion battery pack according to claim 4, characterized in that, The process of generating thermal management control commands includes: A hybrid framework of control barrier function and model predictive control is adopted to solve the multi-objective optimization problem. The control barrier function serves as a hard constraint to ensure the thermal safety of the lithium-ion battery pack throughout its entire life cycle, while model predictive control acts as an optimizer to optimize the energy consumption and temperature uniformity objectives of the thermal management system within the feasible region defined by the hard constraint.
6. The control method for a thermal management system of a lithium-ion battery pack according to claim 5, characterized in that, The constraints of the multi-objective optimization problem include: Under the low-temperature preheating mode, the lithium plating risk coefficient of each battery cell is lower than the preset lithium plating risk threshold. The lithium plating risk coefficient of the i-th battery cell... Calculated using the following formula: ; in, Based on the current health status value of the i-th battery cell Determined lithium plating sensitivity coefficient; ν is the molar activation energy of the lithium deposition reaction; Q is the ideal gas constant; Let be the current temperature of the i-th battery cell; Let be the current charge / discharge current of the i-th battery cell; Reference current; The current influence index; This represents the current state of charge of the i-th battery cell; For reference state of charge; The state of charge influence index.
7. The control method for a thermal management system of a lithium-ion battery pack according to claim 1, characterized in that, The method of predicting the future aging trend of each battery cell based on its historical aging trajectory includes: The internal resistance growth curve and capacity decay curve of each battery cell are obtained throughout its entire life cycle. A time-series prediction model is used to predict the expected internal resistance and capacity of each battery cell after a preset number of cycles based on the internal resistance growth curve and capacity decay curve. When generating the thermal management control command, the cooling resource distribution is increased in advance for battery cells that are predicted to experience accelerated aging, based on the expected internal resistance and capacity values.
8. The control method for a thermal management system of a lithium-ion battery pack according to claim 1, characterized in that, The differentiated execution parameters include the reconfigurable flow channel topology switching command of the cooling system and the cooling medium flow distribution command of each cooling zone; the cooling system includes multiple cooling medium flow channels that can be independently controlled in terms of on / off and flow direction, and each cooling medium flow channel forms a grid-like segmented network inside the lithium-ion battery pack.
9. The control method for a thermal management system of a lithium-ion battery pack according to claim 8, characterized in that, The steps for generating the reconfigurable flow channel topology switching command include: Using the current flow channel topology as the initial state, and at the cost of the operating energy consumption of each flow channel switching valve, the target flow channel topology is solved under the constraint of the thermal safety constraint boundary to achieve the optimal combination of temperature uniformity, cooling medium transport energy consumption, and topology switching energy consumption. Based on the target flow channel topology, the on / off state and flow direction of each cooling medium flow channel are determined so that the spatial topology of the cooling medium flow channel matches the spatial distribution of heat generation intensity in the spatiotemporal distribution of the thermal field.
10. A thermal management system for a lithium-ion battery pack, operating based on the control method for a thermal management system for a lithium-ion battery pack according to any one of claims 1-9, characterized in that, include: The module includes a status parameter acquisition module, an aging status identification and prediction module, a thermal safety constraint boundary determination module, a dynamic thermal prediction module, a control command generation module, and a thermal management execution module. The state parameter acquisition module is used to acquire the state parameters of each battery cell in the lithium-ion battery pack. The aging state identification and prediction module is used to identify the current aging state parameters of each battery cell based on the state parameters of each battery cell, and to predict the future aging trend of each battery cell based on the historical aging trajectory of each battery cell. The thermal safety constraint boundary determination module is used to differentiate the thermal safety constraint boundary of each battery cell based on the current aging state parameters and the future aging trend of each battery cell. The dynamic thermal prediction module is used to construct a dynamic thermal prediction model of the lithium-ion battery pack based on the state parameters of each battery cell, and to predict the spatiotemporal distribution of the thermal field of the lithium-ion battery pack in the future. The control command generation module is used to generate thermal management control commands based on the predicted spatiotemporal distribution of the thermal field and the thermal safety constraint boundary. The thermal management execution module is used to perform thermal management operations according to the thermal management control instructions.
