Lithium battery module charging and discharging process temperature control system and method

Abstract

The invention relates to the technical field of battery thermal management, and particularly discloses a lithium battery module charging and discharging process temperature control system and method, and the system comprises a thermal field dynamic sensing module, a heat flow path prediction module, a self-adaptive regulation and control decision module and an actuator array cooperation module. By constructing a global temperature field digital twinborn model, predicting a future heat flow path and a hot spot, and driving an actuator array to perform spatial directional cooperative regulation and control, closed-loop, dynamic and accurate control of the internal temperature of the lithium battery module is realized, and the thermal safety and the temperature uniformity are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of battery thermal management, and particularly relates to a lithium battery module charging and discharging process temperature control system and method. BACKGROUND

[0002] As a high-efficiency energy storage device, lithium batteries are widely used in electric vehicles, energy storage power stations, portable electronic devices and other fields. Their performance and safety directly affect the reliable operation of the entire system. A large amount of heat is generated during the charging and discharging process of the lithium battery module. If the heat cannot be dissipated in time and evenly, the internal temperature distribution of the module will be uneven, and local overheating may cause thermal runaway, which seriously affects the battery life and brings serious safety hazards. Therefore, accurate monitoring and effective control of the working temperature of the lithium battery module is one of the core technologies to ensure its safe and efficient operation.

[0003] Temperature control during the charging and discharging process of the lithium battery module is a key technology direction to improve the safety and reliability of the battery system. This technology aims to monitor the internal temperature field distribution of the module in real time, and dynamically adjust the thermal management strategy based on the monitoring data, so as to maintain the battery temperature within a safe and efficient working range, prevent local overheating and optimize battery performance.

[0004] The prior art usually uses a limited number of temperature sensors arranged at fixed positions inside the module for temperature monitoring. This static sensor layout scheme has significant defects: since the sensor positions are fixed, its monitoring range is limited and it is difficult to cover the dynamic heat migration path caused by uneven current distribution, internal resistance difference and changes in heat dissipation conditions during the charging and discharging process, especially in the key areas such as cell connection sheets and busbars where concentrated heat is easily generated, forming a monitoring blind area.

[0005] This results in the system being unable to capture the formation and evolution of local hot spots in real time and accurately, causing a serious delay in overheating warning and failing to provide timely and accurate control basis for the thermal management system, greatly increasing the risk of battery thermal runaway. SUMMARY

[0006] The purpose of the present application is to provide a lithium battery module charging and discharging process temperature control system and method to solve the problem of monitoring blind area caused by fixed position temperature sensor layout in the prior art, and the resulting technical contradiction of not capturing dynamic heat migration path and local hot spot in time and accurately, which leads to delay and failure of thermal management control.

[0007] To achieve the above purpose, the present application provides a lithium battery module charging and discharging process temperature control system. The system includes a thermal field dynamic perception module, a heat flow path prediction module, an adaptive control decision module and an actuator array coordination module.

[0008] The thermal field dynamic perception module is used to construct and update the global temperature field digital twin model of the lithium battery module in real time. The module includes a high-density distributed temperature sensing network and a data fusion and reconstruction unit. The high-density distributed temperature sensing network is composed of multiple miniature temperature sensor nodes deployed in key heat-sensitive areas inside the lithium battery module, including but not limited to the tab connection of each cell, the surface of the bus bar inside the module, the center point of the gap between cells, and the inner surface of the module shell.

[0009] The data fusion and reconstruction unit receives real-time temperature data from all temperature sensor nodes, and based on the preset three-dimensional geometric model of the lithium battery module and the material thermal physical property parameter database, uses the Kriging spatial interpolation algorithm to perform spatial interpolation calculation on the sparse sensor measurement point data, to generate a three-dimensional temperature field distribution map covering the entire module internal space with a preset spatial resolution, i.e. the global temperature field digital twin model. The model is updated at a preset refresh frequency.

[0010] The thermal flow path prediction module is used to predict the dynamic migration path of heat and potential hot spot locations within a future time window based on current and historical temperature field states. The module includes a heat source calculation unit and a computational fluid dynamics simulation unit. The heat source calculation unit receives real-time module total current, total voltage, and cell voltage data provided by the battery management system, and calculates the instantaneous heat generation rate of each calculation grid cell in the module at the current location based on the Joule law and the electrochemical heat generation model.

[0011] The computational fluid dynamics simulation unit takes the current three-dimensional temperature field distribution map output by the thermal field dynamic perception module as the initial boundary condition, and takes the instantaneous heat generation rate distribution output by the heat source calculation unit as the volume heat source, calls the preset fluid control equation and heat transfer equation for numerical solution, simulates the coupling process of air or cooling medium flow and heat conduction, convection inside the module within a preset time length, predicts the evolution trend of the temperature field, and identifies the area where the temperature gradient exceeds the preset safety threshold, marking it as a potential hot spot area.

[0012] The adaptive regulation and control decision module is used to generate targeted and spatially differentiated thermal management regulation and control instructions based on the output of the thermal flow path prediction module. The module includes a strategy mapping unit and an instruction generation unit. The strategy mapping unit has a multi-dimensional strategy query table built-in, which takes the spatial coordinates of the predicted potential hot spot area, the temperature exceeding amplitude, and the temperature rise rate as input indexes, and maps the corresponding optimal composite control strategy.

[0013] The optimal composite control strategy is a combination of specific parameters, including the number of actuators to be started, the target power level of each actuator, and the start-stop timing relationship between different actuators. The instruction generation unit generates a digital control instruction sequence containing specific actuator address code, power control parameters, and timestamps according to the optimal composite control strategy output by the strategy mapping unit.

[0014] The actuator array coordination module is used to receive and execute the control instructions issued by the adaptive control decision module to implement precise spatial directional thermal management of the module. This module includes an actuator array composed of multiple independently controllable micro actuators and an array drive controller. The micro actuators in the actuator array are divided into two types according to their physical effects: The first type is active heat dissipation actuators, including micro fans and semiconductor cooling plates. The second type is heat flow guiding actuators, including micro guide vanes and variable thermal resistance material sheets.

[0015] These micro actuators are embedded in the inner wall of the heat dissipation duct of the lithium battery module or installed on the surface of the key cells in a matrix form.

[0016] The array drive controller parses the received digital control instruction sequence and accurately controls the operating state and power output of each addressed micro actuator through independent pulse width modulation channels or digital analog conversion channels, so that it works according to the required time and intensity of the instructions, thereby forming a dynamic adjustable forced convection field or local cooling field matching the predicted heat flow path inside the module, achieving directional suppression and heat dissipation of the predicted hot spots.

[0017] Further, the Kriging spatial interpolation algorithm used by the data fusion and reconstruction unit in the thermal field dynamic perception module has a variogram model that is adaptively selected according to the material composition of different regions inside the lithium battery module. For the cell body region, a Gaussian variogram model is used to reflect its continuous and smooth temperature change characteristics; for the metal connecting piece and busbar region, an exponential variogram model is used to better fit the possible temperature change characteristics. The data fusion and reconstruction unit will dynamically adjust the main parameters of the variogram, including the range and base value, to ensure that the interpolation model matches the current actual thermal field physical characteristics, before each interpolation calculation.

[0018] Furthermore, the computational fluid dynamics simulation unit in the heat flow path prediction module operates within a simplified real-time simulation framework. This framework performs model order reduction processing on the complete fluid control equations and heat transfer equations. Specifically, it obtains the intrinsic modes of the flow field and temperature field under different typical operating conditions of the module through high-fidelity offline simulation in advance. During online prediction, the current temperature field and heat source distribution are projected onto a low-dimensional subspace composed of these preset intrinsic modes for solution, which greatly reduces the amount of computation. This allows the temperature field prediction within the next 5 to 30 seconds time window to be completed within 100 milliseconds, meeting the requirements of real-time control.

[0019] Furthermore, the policy mapping unit in the adaptive control decision module generates its multi-dimensional policy lookup table offline through a deep reinforcement learning algorithm. The training process takes place in a virtual environment containing a high-fidelity thermofluid simulation model and a battery electrothermal coupling model. The reinforcement learning agent uses the predicted hotspot region state as the observation value, the action space as the action combination of the actuator array, and the comprehensive reward function as maintaining module temperature uniformity, minimizing control energy consumption, and avoiding frequent actuator start-stop. Through millions of scenario trials and policy iterations, it finally converges to obtain a lookup table that maps the observed state to the optimal action policy. After system deployment, this lookup table can be periodically updated incrementally based on actual operating data.

[0020] Furthermore, the micro-guide blades in the actuator array coordination module have blade angles driven by micro-stepping motors, which can be continuously adjusted within the range of 0 to 90 degrees. The variable thermal resistance material sheet is made of doped polymer composite material, and its thermal conductivity can be linearly varied within the range of 0.5 W / m Kelvin to 5 W / m Kelvin under the action of an externally applied control voltage of 0 to 5 V. The array drive controller dynamically changes the heat conduction path resistance in its local area by adjusting the voltage applied across the variable thermal resistance material sheet, thereby achieving active guidance of the heat flow direction.

[0021] The present invention also provides a method for temperature control during the charging and discharging process of a lithium battery module applied to the above-mentioned system. This method is automatically executed by the system and includes the following steps:

[0022] Step 1: System initialization, loading the 3D model of the lithium battery module, material thermal property parameters, preset variogram model set, reduced-order simulation intrinsic mode database, and multi-dimensional strategy lookup table.

[0023] Step 2: The thermal field dynamic sensing module continuously collects temperature data from each monitoring point inside the module through a high-density distributed temperature sensing network, and uses the data fusion and reconstruction unit to build and refresh the digital twin model of the global temperature field in real time based on the Kriging space interpolation algorithm and the adaptive variogram model.

[0024] Step 3: The heat flow path prediction module simultaneously acquires the current temperature field data and the electrical parameters of the battery management system. The instantaneous heat generation rate distribution is calculated by the heat source calculation unit, and the computational fluid dynamics simulation unit performs rapid numerical simulation of the temperature field evolution within a preset future time window based on a reduced-order model to predict the heat flow path and identify potential hotspot areas.

[0025] Step 4: The strategy mapping unit of the adaptive control decision module queries the multi-dimensional strategy lookup table based on the predicted potential hotspot area information to obtain the corresponding optimal composite control strategy, and the instruction generation unit converts it into a specific digital control instruction sequence.

[0026] Step 5: The array drive controller of the actuator array coordination module receives the control command sequence, parses and drives the corresponding micro fans, semiconductor cooling chips, guide vanes and variable thermal resistance material sheets in the actuator array to coordinate their actions according to the specified power, angle or thermal resistance value, and generate the target control field inside the module.

[0027] Step 6: The system executes steps 2 to 5 cyclically at a frequency of not less than 10 Hz to achieve closed-loop, dynamic, and spatially oriented temperature control of the entire lithium battery module charging and discharging process.

[0028] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention constructs a digital twin model of the global temperature field by deploying a high-density distributed temperature sensing network and integrating advanced spatial interpolation algorithms, fundamentally eliminating the monitoring blind spots inherent in fixed sensor layouts. This system can present the three-dimensional temperature distribution inside the module in real time with high spatial resolution, accurately capturing subtle temperature changes in key areas such as cell connectors and busbars, providing an unprecedented panoramic and high-precision data foundation for thermal management, and achieving comprehensive perception of the thermal state.

[0029] 2. This invention innovatively introduces a real-time heat flow path prediction mechanism based on computational fluid dynamics. This system not only monitors the current temperature but also proactively simulates the migration and accumulation of heat in the near future, identifying potential hotspots in advance. This paradigm shift from "passive monitoring" to "active prediction" enables the thermal management system to intervene and regulate before local overheating actually occurs, nipping overheating risks in the bud and significantly improving the timeliness of early warning and the proactiveness of control.

[0030] 3. This invention employs an adaptive decision-making and actuator array collaborative control mechanism based on deep reinforcement learning optimization. The system can automatically match and execute a composite control strategy that is spatially highly oriented and precisely matched in intensity, based on predicted hotspot characteristics. By collaboratively controlling various types of micro-actuators, a complex, dynamic, and controllable forced convection and thermal conductivity regulation field can be formed within the module, achieving precise "targeted" cooling and heat flow dissipation of predicted hotspots. Compared to traditional global uniform heat dissipation, this control method is more efficient, consumes less energy, and effectively avoids the problems of local overcooling or insufficient heat dissipation.

[0031] 4. This invention establishes a high-speed closed loop integrating sensing, prediction, decision-making, and execution, achieving intelligent temperature control across the entire chain, from macroscopic modules to microscopic hotspots, and from current state to future trends. This system significantly improves the thermal safety and temperature uniformity of lithium battery modules under complex charge and discharge conditions, effectively extends battery cycle life, and provides reliable technical assurance for the safe application of high-power, high-energy-density battery systems. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the overall technical architecture of the temperature control system for the charging and discharging process of a lithium battery module proposed in this invention. Figure 2 This is a schematic diagram of the core principle framework for constructing a digital twin model of the global temperature field and predicting heat flow paths in this invention; Figure 3 This is a logical flow diagram of the adaptive control decision and actuator array coordination in this invention; Figure 4 This is a schematic diagram of the multi-level interaction relationship and data flow between the thermal field dynamic sensing module and the heat flow path prediction module in this invention. Detailed Implementation

[0033] Please refer to Figures 1 to 4 This invention provides a specific embodiment of a temperature control system and method for the charging and discharging process of a lithium battery module. Please refer to the appendix. Figure 1 This system is a closed-loop control system integrating four major functions: high-precision sensing, real-time prediction, intelligent decision-making, and precise execution. Using a lithium battery module as the physical entity, the system achieves dynamic monitoring of the three-dimensional temperature distribution within the module, prediction of future thermal trends, and spatially differentiated active thermal management through a series of tightly coupled hardware modules and software algorithms. During operation, the modules interact with each other according to a preset logical sequence and data interface, forming a high-speed, adaptive control loop.

[0034] The system's physical carrier is a standardized lithium battery module, which contains multiple lithium-ion cells connected in series or parallel. The cells are electrically connected through metal busbars, and the entire module is encapsulated within a casing with an internal airflow structure. The core of this invention lies in the deep integration of a sensor network, computing unit, and actuators into this physical module, making it an intelligent thermal management entity.

[0035] First, the specific implementation details of the dynamic thermal field sensing module are explained. Please refer to the appendix. Figure 2 With appendix Figure 4 The core task of the dynamic thermal field sensing module is to construct and maintain a digital twin model of the global temperature field that is updated synchronously with the physical module. This module consists of a high-density distributed temperature sensing network at the hardware layer and a data fusion and reconstruction unit at the software layer.

[0036] The deployment of the high-density distributed temperature sensing network follows a strict set of engineering guidelines. The selected miniature temperature sensor nodes are digital output surface-mount thermocouples or digital temperature sensors with integrated circuit bus interfaces, with a size less than 2 mm × 2 mm × 1 mm to ensure minimal interference with the internal space and airflow of the module. Deployment locations are based on a map of "critical heat-sensitive areas" determined by pre-performed finite element thermal simulations. Specifically, one sensor node is deployed at the solder joint between the aluminum or copper tabs of each battery cell and the busbar to monitor Joule heating that may be generated due to contact resistance. Inside the module, one sensor node is deployed every 50 mm on the surface of the copper or aluminum busbars connecting the battery cells to monitor the temperature distribution on the busbars. A miniature probe-type sensor is vertically inserted at the center point of the gap between adjacent battery cells to monitor heat dissipation from the sides of the battery cells and the air temperature in the gap.

[0037] On the inner surface of the module housing, corresponding to the air duct inlet, outlet, and corners, surface-mount sensors are deployed with a grid density of 100 mm × 100 mm to monitor boundary temperature conditions. All sensor nodes are connected via a redundantly designed integrated circuit bus network or wireless sensor network. The bus controller is integrated into the data fusion and reconstruction unit, synchronously acquiring raw temperature data from all nodes at a sampling frequency of no less than 100 Hz. Each data packet contains a unique identifier for the sensor node, the temperature measurement value, a timestamp, and a data checksum.

[0038] The data fusion and reconstruction unit is an embedded industrial computer or a high-performance field-programmable gate array (FPGA). Internally, it stores a precise 3D computer-aided design geometric model of the lithium battery module, discretized into millions of volumetric element meshes. Simultaneously, the unit has a built-in database of material thermophysical parameters, defining precise density, specific heat capacity, and thermal conductivity parameters for each material, such as the cell casing, electrode material, electrolyte, busbar copper, and outer plastic casing. These parameters are functions of temperature. After receiving the raw temperature data stream from the sensor network, the data fusion and reconstruction unit first performs data preprocessing, including removing outliers caused by communication interference and applying a moving average filter to the historical data of each sensor node to suppress noise. The preprocessed data is a spatially discrete set of temperature points with three-dimensional coordinates.

[0039] To generate a three-dimensional temperature field covering the entire continuous space inside the module from these sparse point sets, the data fusion and reconstruction unit uses the Kriging spatial interpolation algorithm for calculation. The core of this algorithm is to construct a variogram model that reflects the correlation of spatial variables. In this implementation, the algorithm does not use a single global variogram, but rather adaptively selects one based on the material properties of the region where the interpolation point is located. The specific process is as follows: When it is necessary to estimate the temperature at a certain location within the module, the algorithm first queries the material type to which that location belongs in the three-dimensional geometric model. If the location is located inside or on the surface of the cell body, a Gaussian variogram model is used, whose mathematical form expresses the characteristic of a gradual and continuous temperature change within that region.

[0040] If the location is within a metal connector or busbar area, an exponential variogram model is used. This model can better fit the localized temperature fluctuations caused by the skin effect or poor contact. Before each interpolation calculation, the data fusion and reconstruction unit calculates the temperature statistical characteristics of different material regions in real time based on the latest batch of sensor data, including variance and spatial autocorrelation length. It then dynamically adjusts the key parameters of the selected variogram function, primarily the range and sill value. The range parameter determines the influence range of spatial correlation, while the sill value represents the overall temperature fluctuation amplitude of the region. Through this dynamic parameter adjustment, the interpolation model can adaptively match the actual thermophysical characteristics of the module under the current charging and discharging conditions, significantly improving interpolation accuracy.

[0041] The interpolation calculation process is accelerated on a graphics processing unit (GPU). The algorithm uses the 3D coordinates and temperature of all sensor nodes as known conditions, and the coordinates of all nodes in the entire module's 3D mesh as interpolation points. It solves the Kriging equations and finally outputs the temperature estimate for each mesh node. These temperature values ​​are organized into a 3D array with a preset spatial resolution, such as 1 mm, thus forming the global temperature field digital twin model. This model is updated at a refresh rate of 50 Hz to ensure high synchronization with the actual thermal state of the physical module. The generated temperature field data is sent to the heat flow path prediction module in real time via a high-speed communication interface.

[0042] The following details the specific implementation of the heat flow path prediction module. Please refer to the appendix. Figure 2 With appendix Figure 4 This module's function is to predict the dynamic migration process of heat over the next few seconds to tens of seconds based on the precise temperature field and real-time electrical parameters at the current moment. The module consists of a heat source calculation unit and a computational fluid dynamics simulation unit connected in series.

[0043] The heat source calculation unit continuously acquires real-time data streams from the battery management system of the vehicle or energy storage system. This data includes the total input and output current and total terminal voltage of the lithium battery module, as well as the terminal voltage of each cell obtained through the monitoring circuitry within the battery management system. Based on these electrical parameters, the heat source calculation unit calculates the instantaneous heat generation rate at each spatial location within the module. The heat generation model is a composite model of Joule heating and electrochemical reaction heat.

[0044] For purely conductive regions such as busbars and connectors, the heat generation rate is calculated based on Joule's law, using the following formula: ,in, For volumetric heat production rate, For current density, The resistivity of the material is given. The current density distribution is obtained by proportionally allocating the current using pre-calibrated finite element electromagnetic simulation results combined with the real-time total current.

[0045] For the cell body region, the heat generation rate is calculated using the Bernham electrochemical heat generation model, which divides heat generation into reversible and irreversible heat. Irreversible heat is generated by ohmic internal resistance and polarization internal resistance, while reversible heat is related to the cell's entropy thermal coefficient and charge / discharge current. The heat source calculation unit stores tables of internal resistance and entropy thermal coefficient parameters for each cell at different states of charge and temperatures. By interpolating from these tables, combined with real-time current, voltage, and local cell temperature from the digital twin model, the heat generation rate of each computational grid cell within each cell is calculated. Finally, the heat source calculation unit outputs a three-dimensional distribution map of the instantaneous volumetric heat generation rate that is perfectly aligned with the three-dimensional temperature field grid.

[0046] The computational fluid dynamics simulation unit is responsible for performing fast forward simulation predictions. It receives two key inputs: First, the current three-dimensional temperature field distribution map provided by the thermal field dynamic sensing module serves as the initial temperature field for simulation. Second, the instantaneous heat generation rate distribution map provided by the heat source calculation unit is used as the volumetric heat source term in the simulation.

[0047] Traditional computational fluid dynamics (CFD) simulations are computationally intensive and cannot meet the requirements of real-time control. Therefore, this implementation adopts a real-time simulation framework based on model order reduction. The core of this framework is an "eigenmode" database. Before system deployment, high-fidelity, offline CFD simulations are required to simulate a large number of typical operating conditions for the target lithium battery module. These operating conditions cover different ambient temperatures, different charge and discharge rates, and different cooling airflow rates. From these high-fidelity simulation results, several eigenmodes that can characterize the main dynamic features of the module's flow and temperature fields are extracted using the intrinsic orthogonal decomposition method. Typically, the first 20 to 50 modes can capture more than 95% of the energy. These eigenmodes and their corresponding eigenvalues ​​are stored in the database of the CFD simulation unit.

[0048] During online prediction, the computational fluid dynamics simulation unit does not directly solve the complete Navier-Stokes equations and energy equations. Instead, it projects the current three-dimensional temperature field and three-dimensional heat production rate distribution onto a low-dimensional subspace spanned by the intrinsic modes, obtaining a set of low-dimensional modal coefficients. In this low-dimensional space, the fluid control equations and heat transfer equations are simplified to a set of ordinary differential equations (ODEs) concerning these modal coefficients. Solving these ODEs, whose dimensionality has been significantly reduced, allows for the rapid acquisition of the evolution trajectory of the modal coefficients over a future time period. Finally, the predicted modal coefficients for future moments are reconstructed in reverse to obtain the predicted three-dimensional temperature field for those future moments. The entire prediction process, targeting a time window of 5 to 30 seconds, can be completed within 100 milliseconds, fully meeting the timing requirements of real-time control. The simulation unit's output specifically identifies regions where the predicted temperature exceeds a preset safety threshold, such as 45 degrees Celsius, or where the temperature gradient exceeds 0.1 degrees Celsius per millimeter; these regions are marked as "potential hotspot regions." The predicted temperature field evolution sequence clearly reveals the main migration paths of heat flow.

[0049] Then, the specific implementation method of the adaptive control decision-making module is described. Please refer to the appendix. Figure 3 This module's function is to transform predicted, abstract thermal threats into specific, executable hardware control commands. The module includes a policy mapping unit and an instruction generation unit.

[0050] The core asset of the policy mapping unit is a multi-dimensional policy lookup table. This lookup table is not written using human-developed rules, but rather generated offline in a virtual environment through deep reinforcement learning algorithms. The training environment is a high-fidelity digital twin simulation platform that integrates a battery electrothermal coupling model, a computational fluid dynamics model, and a physical effect model of the actuator array. The observation space of the reinforcement learning agent is a feature vector of the predicted potential hotspot region. This vector includes the 3D bounding box coordinates of the hotspot region, the magnitude by which the highest temperature in the region exceeds the safety threshold, the average temperature rise rate of the region, and the relative position of the hotspot region within the module. The agent's action space is a combination of all possible actions of all micro-actuators, such as 10 speed settings for each micro-fan, 10 power levels for each thermoelectric cooler, 10 angles for each guide vane, and 10 voltage levels for each variable thermal resistance material sheet.

[0051] The agent's reward function is a multi-objective weighted sum, mainly consisting of three terms: a negative temperature uniformity exponent, which encourages uniform temperature distribution; a negative total energy consumption, which encourages energy conservation; and a negative penalty for actuator action changes, which encourages smooth control and avoidance of frequent switching. The agent learns how to select a set of actuator actions to maximize long-term cumulative rewards under various observed thermal threat states through millions of scenario trials and policy iterations in a virtual environment using a proximal policy optimization algorithm. The finally converged policy is discretized and stored as a multi-dimensional lookup table.

[0052] In the actual system, after receiving the feature vector of the potential hot spot area from the heat flow path prediction module, the strategy mapping unit uses it as an index to directly look up the table and output the corresponding "optimal composite control strategy" within milliseconds. This strategy is a structure that explicitly lists the unique identifiers of the micro actuators that need to be activated, the target power level, target angle value, or target voltage value corresponding to each actuator, and the start-stop timing relationship between different actuators. For example, it requires the guide vanes to rotate to a specified angle first, and then, after a delay of 200 milliseconds, start the micro fan in the corresponding area.

[0053] The instruction generation unit is responsible for translating the structured strategy output by the strategy mapping unit into low-level control instructions that the actuator array coordination module can directly recognize. Based on the actuator type and communication protocol, the instruction generation unit generates specific digital control instruction sequences. For micro fans and thermoelectric coolers, the instructions include the actuator address code, the target pulse width modulation duty cycle value, and the effective timestamp.

[0054] For miniature guide vanes, the instructions include the actuator address code, target angle value, rotational speed, and effective timestamp. For variable thermal resistance material sheets, the instructions include the actuator address code, target control voltage value, voltage ramp-up rate, and effective timestamp. All these instructions are encapsulated in transactional instruction packets, ensuring that either all are executed correctly or all are rolled back in case of errors. The instruction packets are sent to the actuator array coordination module via real-time Ethernet or the controller area network bus.

[0055] Finally, the specific implementation method of the actuator array coordination module is described in detail. Please refer to the appendix. Figure 1 With appendix Figure 3 This module is the final actuator that influences the physical world, and it consists of an actuator array and an array drive controller.

[0056] The actuator array is a collection of miniature actuators custom-embedded according to the module structure. In the active cooling actuator, the miniature fan, with a diameter between 10 mm and 20 mm, is driven by a brushless DC motor and embedded in the sidewall or partition of the module's cooling duct. Its airflow direction can be perpendicular to the main airflow or at a certain angle. The thermoelectric cooler, measuring 10 mm × 10 mm × 3 mm, is tightly attached to specific locations on the pre-identified cell casing prone to hot spots using thermal grease. Its cold side faces the cell, and its hot side connects to the miniature heat sink fins and is positioned within the airflow duct.

[0057] In the heat flow guiding actuator, the miniature guide vanes are made of a 0.5 mm thick aluminum alloy sheet, approximately 5 mm wide, and their length covers the height of the air duct cross-section. The vane shaft is driven by a miniature stepper motor with a step angle of 1.8 degrees. Through a reduction gear set, continuous and precise positioning of the vanes can be achieved within the range of 0 to 90 degrees. At 0 degrees, the vanes are parallel to the airflow direction; at 90 degrees, they are completely perpendicular to change or block the airflow. The variable thermal resistance material sheet is a functional material device with a polymer matrix doped with carbon nanotubes or graphene. This material sheet is made into a 1 mm thick sheet and inserted between the battery cell and the heat sink, or attached to the busbar surface. Its core characteristic is that its thermal conductivity changes linearly with the DC control voltage applied to its two electrodes. At 0 volts, the internal filler is dispersed, with a thermal conductivity of approximately 0.5 W / m Kelvin, exhibiting a high thermal insulation state; at 5 volts, the filler forms a thermally conductive chain under the action of an electric field, and the thermal conductivity can rise to 5 W / m Kelvin, exhibiting a high thermal conductivity state.

[0058] The array drive controller is a multi-channel motor servo drive and power amplifier circuit board. It receives a sequence of digital control commands from the adaptive control decision module, which is then parsed by the built-in microcontroller. For pulse-width modulation (PWM) control actuators, the controller generates a PWM waveform with the corresponding frequency and precision based on the duty cycle value in the command, driving the motor or a thermoelectric cooler via a power metal-oxide-semiconductor field-effect transistor.

[0059] For angle-controlled actuators, the controller converts the target angle value into the number of pulses and direction signals required by the stepper motor, driving the stepper motor for precise positioning. For voltage-controlled actuators, the controller outputs an analog voltage of 0 to 5 volts through a high-precision digital-to-analog converter, applying it across the variable thermal resistance material sheet, and monitors the voltage value in real time through a feedback circuit to ensure precise control. The array drive controller ensures that all addressed actuators can start synchronously or asynchronously in sequence according to the command requirements, and operate at the specified intensity, thereby constructing a complex and dynamic control field within the module's internal space. This control field may create a forced cooling airflow upstream of a predicted hot spot, enhance the heat conduction path in the hot spot area, or precisely guide cool air to the heat accumulation area through guide vanes.

[0060] The entire system operates on a strict time-series cycle. During system initialization, all necessary parameters are loaded from non-volatile memory: the 3D geometric model, material database, variogram model set, reduced-order simulation intrinsic mode database, and a trained multi-dimensional strategy lookup table. After entering the main loop, the thermal field dynamic sensing module updates the digital twin model at a frequency of 50 Hz.

[0061] The heat flow path prediction module acquires data synchronously and completes the next prediction within 100 milliseconds. The adaptive control decision module, upon receiving the prediction results, completes strategy query and instruction generation within 10 milliseconds. The actuator array coordination module has an execution latency of less than 50 milliseconds. Therefore, the entire closed-loop cycle from sensing to execution is less than 200 milliseconds, and the control frequency is higher than 5 Hz, typically set to 10 Hz in practice to ensure the system can quickly respond to rapidly changing thermal states during charging and discharging.

[0062] Through the coordinated operation of the above-described specific implementation methods, the system of this invention achieves panoramic perception, advanced prediction, and precise control of the internal temperature field of lithium battery modules. Unlike existing technologies that can only respond with a lag and globally after the temperature exceeds the limit at a fixed point, this system can intervene at specific locations before hotspots actually form, thereby suppressing temperature fluctuations within a smaller and safer range, significantly improving the proactiveness and safety of thermal management. The entire system is an adaptive intelligent agent, and its strategy lookup table can be incrementally updated periodically based on data accumulated during actual operation, thereby continuously optimizing the control effect and reducing overall energy consumption.

[0063] Example 2: This example provides a specific implementation variation based on the system architecture described in the above examples. It mainly focuses on the differentiated design of the computational fluid dynamics simulation unit of the heat flow path prediction module and the heat flow guiding actuator in the actuator array coordination module, in order to adapt to a lithium battery module structure using a liquid cooling plate.

[0064] In this embodiment, the lithium battery module is cooled by liquid cooling. A liquid cooling plate with internal flow channels is integrated on the bottom or side of the module, and the coolant circulates within the channels to remove heat. Therefore, the deployment of the high-density distributed temperature sensing network in the thermal field dynamic sensing module needs to be adjusted accordingly. In addition to deploying sensors at key locations such as the cell tabs and busbars, multiple thin-film pressure-temperature integrated sensors need to be deployed in a matrix at the contact interface between the liquid cooling plate and the cell to monitor contact thermal resistance and interface temperature. Simultaneously, high-precision platinum resistance temperature sensors are deployed at the liquid inlet, outlet, and key flow channel bends of the liquid cooling plate to monitor the temperature changes of the coolant itself.

[0065] The computational fluid dynamics simulation unit in the heat flow path prediction module changes its physical model from being dominated by air convection to being dominated by the coupling of solid heat conduction and liquid convection. Its reduced-order model's intrinsic mode database is obtained by offline simulation of the three-dimensional temperature field of the liquid cooling system under different coolant flow rates, inlet temperatures, and module heat generation powers. During online prediction, the calculation method for volumetric heat sources is the same as in the aforementioned embodiment. However, the simulation's governing equations focus more on solving the coupled heat conduction equation in the solid region and the convection heat transfer equation within the coolant flow channel. The prediction module needs to obtain the current flow rate and inlet temperature of the coolant circulation system in real time as boundary conditions for the simulation. In addition to the temperature field and potential hot spots in the module's solid region, the prediction output also includes the predicted coolant outlet temperature and the temperature distribution on the surface of the liquid cooling plate, used to evaluate the overall performance of the cooling system.

[0066] The multi-dimensional strategy lookup table of the adaptive control decision module needs to be retrained, resulting in significant changes to its action space. Because the cooling medium is liquid, the actuator array configuration differs from the embodiments described above. The actuator array coordination module in this embodiment includes two types of novel actuators: The first category is distributed micro piezoelectric pumps and valve units. The second category is materials with variable thermal conductivity interfaces.

[0067] Distributed micro piezoelectric pumps and valve units are integrated into the flow channel network of the liquid cooling plate. The main flow channels of the liquid cooling plate are designed with a tree-like branching structure, and a micro piezoelectric pump and a micro proportional valve are installed at the inlet of each branch flow channel. The piezoelectric pumps can provide additional local driving force, and the proportional valves can precisely regulate the coolant flow rate of that branch. Variable thermal conductivity interface material is filled in the assembly gap between the battery cell and the liquid cooling plate. Its function is similar to the variable thermal resistance material sheet in the above embodiment, but its form is paste or gasket-like. By applying an electric field, its thermal conductivity is changed, thereby dynamically adjusting the thermal resistance of heat transfer from the battery cell to the liquid cooling plate.

Claims

1. A temperature control system for the charging and discharging process of a lithium battery module, characterized in that, include: The thermal field dynamic sensing module is used to build and update the full-domain temperature field digital twin model of the lithium battery module in real time; The heat flow path prediction module is used to predict the dynamic migration path of heat and potential hotspot locations within a future time window based on the current and historical temperature field conditions. The adaptive control decision module is used to generate targeted, spatially differentiated thermal management control commands based on the output of the heat flow path prediction module. The actuator array coordination module is used to receive and execute the control commands issued by the adaptive control decision module to implement precise space-oriented thermal management of the module; The dynamic thermal field sensing module includes a high-density distributed temperature sensing network and a data fusion and reconstruction unit.

2. The temperature control system for the charging and discharging process of a lithium battery module according to claim 1, characterized in that, The heat flow path prediction module includes a heat source calculation unit and a computational fluid dynamics simulation unit. The heat source calculation unit receives the total current and voltage of the module and the voltage data of each cell provided by the battery management system in real time, and calculates the instantaneous heat generation rate of each calculation grid unit in the module based on Joule's law and the electrochemical heat generation model. The computational fluid dynamics simulation unit uses the current three-dimensional temperature field distribution map output by the thermal field dynamic sensing module as the initial boundary condition, and the instantaneous heat generation rate distribution output by the heat source calculation unit as the volume heat source. It calls the preset fluid control equation and heat transfer equation to perform numerical solution, simulates the coupling process of air or cooling medium flow and heat conduction and convection inside the module within a preset time period in the future, thereby predicting the evolution trend of the temperature field and identifying areas where the temperature gradient exceeds the preset safety threshold.

3. The temperature control system for the charging and discharging process of a lithium battery module according to claim 2, characterized in that, The adaptive control decision module includes a strategy mapping unit and an instruction generation unit; The strategy mapping unit has a built-in multi-dimensional strategy lookup table. The multi-dimensional strategy lookup table uses the spatial coordinates of the predicted potential hotspot area, the temperature exceedance range, and the temperature rise rate as input indexes to map out the corresponding optimal composite control strategy. The instruction generation unit generates a digital control instruction sequence containing specific actuator address codes, power control parameters, and timestamps based on the optimal composite control strategy output by the strategy mapping unit.

4. The temperature control system for the charging and discharging process of a lithium battery module according to claim 3, characterized in that, The actuator array coordination module includes an actuator array consisting of multiple independently controllable micro actuators and an array drive controller; The micro actuators in the actuator array include active heat dissipation actuators and heat flow directed actuators; The array drive controller parses the received digital control command sequence and precisely controls the operating status and power output of each addressed micro actuator through an independent pulse width modulation channel or digital-to-analog conversion channel.

5. The temperature control system for the charging and discharging process of a lithium battery module according to claim 4, characterized in that, The high-density distributed temperature sensing network consists of multiple miniature temperature sensor nodes deployed in key heat-sensitive areas inside the lithium battery module. The data fusion and reconstruction unit receives raw temperature data from all temperature sensor nodes in real time, and based on the preset three-dimensional geometric model of the lithium battery module and the material thermophysical parameter database, it uses the Kriging spatial interpolation algorithm to perform spatial interpolation calculations on the sparse sensor measurement point data to generate a three-dimensional temperature field distribution map that covers the entire internal space of the module and has a preset spatial resolution.

6. A temperature control system for the charging and discharging process of a lithium battery module according to claim 5, characterized in that, The data fusion and reconstruction unit uses the Kriging space interpolation algorithm, whose variogram model is adaptively selected based on the material composition of different regions inside the lithium battery module. For the cell body region, a Gaussian variogram model is adopted; For the metal connector and busbar area, an exponential variogram model is used; Before each interpolation calculation, the data fusion and reconstruction unit dynamically adjusts the main parameters of the variogram function based on the statistical characteristics of regional temperature changes calculated in real time from sensor data. The main parameters include the range and the sill value.

7. A temperature control system for the charging and discharging process of a lithium battery module according to claim 6, characterized in that, The computational fluid dynamics simulation unit operates within a simplified real-time simulation framework, which performs model order reduction processing on the complete fluid control equations and heat transfer equations; the specific method of model order reduction processing is as follows: The intrinsic modes of the flow field and temperature field of the module under different typical working conditions are obtained in advance through high-fidelity offline simulation; during online prediction, the current temperature field and heat source distribution are projected onto the low-dimensional subspace composed of the preset intrinsic modes for solution.

8. A temperature control system for the charging and discharging process of a lithium battery module according to claim 7, characterized in that, The multi-dimensional policy lookup table of the policy mapping unit is generated offline through deep reinforcement learning algorithm; The training process takes place in a virtual environment that includes a high-fidelity thermal fluid simulation model and a battery electrothermal coupling model; The reinforcement learning agent uses the predicted hotspot region state as the observation value, the action space as the action combination of the actuator array, and the comprehensive reward function as maintaining module temperature uniformity, minimizing control energy consumption, and avoiding frequent start-stop of actuators. Through scenario trial and error and policy iteration, it finally converges to obtain a lookup table that can map the observed state to the optimal action policy.

9. A temperature control system for the charging and discharging process of a lithium battery module according to claim 8, characterized in that, After system deployment, the multi-dimensional strategy query table is periodically updated through incremental learning based on actual operational data.

10. A method for temperature control during the charging and discharging process of a lithium battery module applied to the system described in any one of claims 1 to 9, characterized in that, The system executes automatically, including the following steps: Step 1: System initialization, loading the 3D model of the lithium battery module, material thermal property parameters, preset variogram model set, reduced-order simulation intrinsic mode database, and multi-dimensional strategy lookup table; Step 2: The thermal field dynamic sensing module continuously collects temperature data from each monitoring point inside the module through a high-density distributed temperature sensing network, and uses the data fusion and reconstruction unit to build and refresh the digital twin model of the global temperature field in real time based on the Kriging space interpolation algorithm and the adaptive variogram model. Step 3: The heat flow path prediction module simultaneously acquires the current temperature field data and the electrical parameters of the battery management system. The instantaneous heat generation rate distribution is calculated by the heat source calculation unit, and the computational fluid dynamics simulation unit performs a rapid numerical simulation of the temperature field evolution within a preset future time window based on a reduced-order model to predict the heat flow path and identify potential hotspot areas. Step 4: The strategy mapping unit of the adaptive control decision module queries the multi-dimensional strategy lookup table based on the predicted potential hotspot area information to obtain the corresponding optimal composite control strategy, and the instruction generation unit converts it into a specific digital control instruction sequence. Step 5: The array drive controller of the actuator array coordination module receives the control command sequence, parses and drives the corresponding micro fans, semiconductor cooling chips, guide vanes and variable thermal resistance material sheets in the actuator array to coordinate their actions according to the specified power, angle or thermal resistance value, and generate the target control field inside the module. Step 6: The system executes steps 2 to 5 in a loop at a frequency of not less than 10 Hz.

Images