Battery thermal management method and device, vehicle and equipment

Abstract

The invention provides a battery thermal management method and device, a vehicle and equipment, relates to the technical field of automobiles, and can solve the technical problems of poor battery thermal management strategy adaptability, low energy efficiency, insufficient performance output and dynamic working condition response lag in the prior art on the premise of ensuring safe operation of a battery. The method is applied to a controller, the controller performs thermal management on a battery through a battery thermal management model, and the method comprises the following steps: acquiring first working condition data of the battery, the first working condition data being used for describing the state of the battery and the operating environment of the battery, and performing a thermal management decision based on the first working condition data through the battery thermal management model, the control instruction is used for adjusting the temperature of the battery, parameter updating is carried out on the battery thermal management model based on the first working condition data and the second working condition data, the second working condition data is working condition data corresponding to the battery after the temperature of the battery is adjusted, and thermal management is carried out on the battery based on the updated battery thermal management model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automobiles, in particular to a battery thermal management method and device, a vehicle and equipment. BACKGROUND

[0002] With the rapid development of new energy fields such as electric vehicles and energy storage systems, the safety, energy efficiency and long-term performance output of battery systems as core power and energy storage components have become key factors restricting product competitiveness. Temperature is a variable that directly affects the charging and discharging efficiency, cycle life and safety and stability of the battery, so high-precision and adaptive battery thermal management technology has become the focus of industry research and development.

[0003] The existing battery thermal management control scheme maintains the battery temperature between fixed maximum and minimum safety thresholds, which can ensure basic safety, but the overly conservative regulation and control method leads to insufficient release of battery energy efficiency and peak performance, and such a scheme lacks timeliness and is difficult to adapt to complex and variable actual operating scenarios. SUMMARY

[0004] The present application provides a battery thermal management method, device, vehicle and equipment, which can solve the technical problems of poor adaptability, low energy efficiency, insufficient performance output and dynamic condition response lag of the battery thermal management strategy in the prior art while ensuring safe operation of the battery.

[0005] To achieve the above-mentioned purpose, the technical scheme adopted by the present application is as follows: In a first aspect, a battery thermal management method is provided. The method is applied to a controller, and the controller performs thermal management on a battery through a battery thermal management model. The method comprises: obtaining first working condition data of the battery, the first working condition data being used to describe a state of the battery and an operating environment in which the battery is located; making a thermal management decision based on the first working condition data through the battery thermal management model to obtain a control instruction, the control instruction being used to adjust a temperature of the battery; updating parameters of the battery thermal management model based on the first working condition data and second working condition data, the second working condition data being working condition data corresponding to the battery after the temperature of the battery is adjusted; and performing thermal management on the battery based on the updated battery thermal management model.

[0006] According to the above technical means, by constructing a dynamically updated battery thermal management model, accurate control instructions are generated in combination with real-time working condition data, and model parameters are continuously optimized based on adjusted data, so that the strategy is adapted to changes in the state of the battery and the environment, the temperature adjustment precision and the cell consistency are improved, the energy consumption is reduced, the battery life is prolonged, and the operation safety is ensured.

[0007] In a possible embodiment, the thermal management model comprises a prediction layer network and a decision layer network, and the battery thermal management model is used to make a thermal management decision based on the first working condition data to obtain a control instruction, which can be specifically implemented as follows: the prediction layer network is used to make a prediction based on the first working condition data to obtain battery temperature feature data in a first time period; the prediction layer network refers to a time series prediction layer network constructed based on a long short-term memory (LSTM) network; and the decision layer network is used to make a decision based on the battery temperature feature data and the first working condition data to obtain the control instruction, and the decision layer network refers to a network constructed based on a reinforcement learning algorithm.

[0008] According to the above technical means, the temperature change trend is predicted in advance through the LSTM network, and the PPO reinforcement learning dynamic decision is combined, so that the thermal management strategy has predictability and adaptability. The battery state and environmental changes can be accurately responded to, and the temperature stability, cell consistency and energy consumption can be balanced, so that the thermal management precision and reliability are improved.

[0009] In an embodiment, the decision layer network is used to make a decision based on the battery temperature feature data and the first working condition data to obtain the control instruction, which can be specifically implemented as follows: the battery temperature feature data and the battery working condition data are fused to obtain a state vector, and the state vector is input into the decision layer network for processing to obtain the control instruction.

[0010] According to the above technical means, the battery temperature feature data and the working condition data are integrated into a state vector through data fusion, so that the decision layer network can obtain the temperature change trend and real-time running state information at the same time, effectively avoiding the one-sidedness of single data decision. The control instruction that adapts to the current working condition is accurately generated, the temperature regulation is improved in pertinence and timeliness, the cell consistency is ensured, and the thermal management energy consumption is reduced.

[0011] In another embodiment, the decision layer network comprises a hidden layer and an output layer, and the state vector is input into the decision network to obtain the control instruction, which can be specifically implemented as follows: the hidden layer of the decision layer network is used to perform nonlinear transformation on the state vector to obtain high-order features, the high-order features refer to features extracted from the state vector for describing the battery temperature regulation effect, the output layer of the decision layer network is used to generate a first probability distribution of the control instruction based on the high-order features, the first probability distribution takes an effective regulation range of the control instruction as a definition domain and contains probability density information of different regulation intensities, a candidate control instruction is obtained by sampling from the first probability distribution, and the candidate control instruction is subjected to physical constraint verification to obtain the control instruction.

[0012] According to the above technical means, high-order characteristics reflecting the temperature regulation effect are extracted through the hidden layer, and the probability distribution generated by the output layer covers different regulation intensities. After sampling and physical constraint verification, it is ensured that the control command adapts to the working condition demand and meets the hardware limit of the actuator, improving the accuracy and feasibility of the command, optimizing the temperature regulation effect and system safety.

[0013] In another embodiment, sampling is performed from the first probability distribution to obtain a candidate control command, which can be specifically implemented as: sampling in the first probability distribution based on the mean and standard deviation of the first probability distribution, and determining the candidate control command based on the sampling result.

[0014] According to the above technical means, by sampling based on the mean and standard deviation of the first probability distribution, the optimal regulation direction is locked based on the mean, and the sampling flexibility is given based on the standard deviation, avoiding rigid command generation. The optimality and explorability of the control command are effectively balanced to improve the adaptability to dynamic working conditions, make the temperature regulation more accurate and flexible, and ensure the stability of the battery thermal management effect.

[0015] In another embodiment, the candidate control command is subjected to physical constraint verification to obtain a control command, which can be specifically implemented as: in the case where the regulation range of the candidate control command is greater than the regulation range threshold, determining the regulation range of the candidate control command based on the regulation range threshold to obtain the control command.

[0016] According to the above technical means, by performing physical constraint verification on the candidate control command and correcting based on the threshold when the regulation range exceeds the threshold, it can be ensured that the command does not exceed the hardware limit of the actuator, avoiding equipment damage or regulation failure due to overrange. While ensuring system safety, the effective regulation ability of the control command is maintained, and the reliability and stability of thermal management are improved.

[0017] In another embodiment, the battery thermal management model is updated based on the first working condition data and the second working condition data, which can be specifically implemented as: performing decision-making based on the second working condition data through the battery thermal management model to obtain a second probability distribution, obtaining a probability ratio based on the first probability distribution and the second probability distribution, calculating a reward value based on the first working condition data, the second working condition data and a reward function, the reward function being used to evaluate the temperature stability, the cell consistency and the energy consumption level of the battery, the reward value being used to reflect the performance of the battery in temperature stability, cell consistency and energy consumption level after execution of the control command, calculating an advantage function value based on the reward value, the first working condition data and the second working condition data, constructing a target function based on the probability ratio and the advantage function value, obtaining adjustment parameters of the battery thermal management model based on the back propagation algorithm and the target function, and adjusting the battery thermal management model based on the adjustment parameters.

[0018] According to the above technical means, by combining the front and rear working condition data, the strategy change is quantified in a probability ratio, the reward function and the advantage function are used to evaluate the adjustment effect, and then the target function and the back propagation are used to optimize the model parameters, so that the thermal management model can dynamically adapt to the battery state and environmental changes, accurately balance the temperature stability, cell consistency and energy consumption, and significantly improve the strategy adaptability and adjustment reliability.

[0019] In another embodiment, based on the back propagation algorithm and the target function, the adjustment parameters of the battery thermal management model can be specifically implemented as follows: the gradient of the target function with respect to the parameters of the battery thermal management model is calculated based on the back propagation algorithm, the gradient is optimized to obtain the adjustment amount of the parameters of the battery thermal management model, and the adjustment amount is the adjustment parameters of the battery thermal management model.

[0020] According to the above technical means, the gradient of the target function with respect to the model parameters is accurately calculated by the back propagation algorithm, the optimization direction of the parameters is determined, and then the adjustment parameters are obtained through gradient optimization, so that the model parameter updating is more targeted and efficient, blind adjustment is avoided, the temperature adjustment precision, cell consistency control and energy consumption optimization capability are improved, and the thermal management effect is steadily improved.

[0021] In another embodiment, the battery thermal management method provided by the present application can also be specifically implemented as follows: the updated battery thermal management model is simulated and verified in the working condition database, and in the case that the simulation verification result meets the termination condition, the adjusted parameters are rolled back to the parameters before adjustment.

[0022] According to the above technical means, the performance of the updated model in various scenarios can be detected in advance by simulating and verifying the model in the working condition database; if the verification is not up to standard, the parameters are rolled back, which can effectively avoid the risk of model performance degradation, ensure that the thermal management strategy is always reliable, avoid problems such as temperature out of control and energy consumption surge in actual application, and improve the robustness and system safety of model iteration.

[0023] In a second aspect, a battery thermal management device is provided, which includes an acquisition module, a decision module, an update module and a management module.

[0024] The acquisition module is used to acquire first working condition data of the battery, the first working condition data being used to describe the state of the battery and the running environment in which the battery is located; the decision module is used to make a thermal management decision based on the first working condition data through a battery thermal management model to obtain a control instruction, the control instruction being used to adjust the temperature of the battery; the update module is used to update the parameters of the battery thermal management model based on the first working condition data and second working condition data, the second working condition data being the working condition data corresponding to the battery after the temperature of the battery is adjusted; and the management module is used to perform thermal management on the battery based on the updated battery thermal management model.

[0025] The decision module is further configured to obtain battery temperature feature data in the first time period based on the first working condition data by a prediction layer network, the prediction layer network being a time series prediction layer network constructed based on a long short-term memory (LSTM) network, and obtain the control instruction based on the battery temperature feature data and the first working condition data by a decision layer network, the decision layer network being a network constructed based on a reinforcement learning algorithm.

[0026] The decision module is further configured to perform data fusion on the battery temperature feature data and the battery working condition data to obtain a state vector, input the state vector to the decision layer network for processing, and obtain the control instruction.

[0027] The decision module is further configured to perform nonlinear transformation on the state vector by a hidden layer of the decision layer network to obtain high-order features, the high-order features being features extracted from the state vector and used to describe the battery temperature regulation effect, generate a first probability distribution of the control instruction based on the high-order features by an output layer of the decision layer network, the first probability distribution having an effective regulation range of the control instruction as a definition domain and containing probability density information of different regulation intensities, sample the first probability distribution to obtain a candidate control instruction, and perform physical constraint verification on the candidate control instruction to obtain the control instruction.

[0028] The decision module is further configured to sample the first probability distribution based on a mean value and a standard deviation of the first probability distribution, and determine the candidate control instruction based on a result of the sampling.

[0029] The decision module is further configured to, in a case where a regulation range of the candidate control instruction is greater than a regulation range threshold, determine the regulation range of the candidate control instruction based on the regulation range threshold to obtain the control instruction.

[0030] The update module is further configured to obtain a second probability distribution based on the second working condition data by the battery thermal management model, obtain a probability ratio based on the first probability distribution and the second probability distribution, calculate a reward value based on the first working condition data, the second working condition data, and a reward function, the reward function being used to evaluate temperature stability, cell consistency, and energy consumption level of the battery, the reward value being used to reflect performance of the battery in the temperature stability, the cell consistency, and the energy consumption level after execution of the control instruction, calculate an advantage function value based on the reward value, the first working condition data, and the second working condition data, construct a target function based on the probability ratio and the advantage function value, obtain an adjustment parameter of the battery thermal management model based on a back propagation algorithm and the target function, and adjust the battery thermal management model based on the adjustment parameter.

[0031] The updating module is further configured to calculate a gradient of the objective function with respect to the battery thermal management model parameter based on a back propagation algorithm, optimize the gradient, and obtain an adjustment amount of the battery thermal management model parameter, the adjustment amount being an adjustment parameter of the battery thermal management model.

[0032] The updating module is further configured to simulate and verify the updated battery thermal management model in the working condition database, and roll back the adjusted parameter to the pre-adjustment parameter in a case where a simulation verification result meets a termination condition.

[0033] In a third aspect, an embodiment of the present application provides a vehicle, and the battery thermal management of the vehicle applies the method of any one of the first aspect.

[0034] In a fourth aspect, an embodiment of the present application provides an electronic device, including a processor, a memory storing a computer program, and the computer program is executed by the processor to implement the method of any one of the first aspect.

[0035] The second aspect to the fourth aspect provide solutions for implementing the method of the first aspect, and the specific implementation is not described again. The technical effects corresponding to any one of the implementation manners of the second aspect to the fourth aspect can be referred to the technical effects corresponding to any one of the implementation manners of the first aspect, and the description is not repeated here.

[0036] It should be noted that the various possible implementation manners of any one of the aspects can be combined on the premise that the solutions are not contradictory. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0038] Figure 1 A schematic diagram of the architecture of a controller provided by an embodiment of the present application is shown in the figure. Figure 2 A flowchart of a battery thermal management method provided by an embodiment of the present application is shown in the figure. Figure 3 A flowchart of another battery thermal management method provided by an embodiment of the present application is shown in the figure. Figure 4 A logic diagram of a battery thermal management method provided by an embodiment of the present application is shown in the figure. Figure 5 A structural diagram of a battery thermal management device provided by an embodiment of the present application is shown in the figure. Figure 6 FIG. 1 is a structural schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0039] In the embodiments of the present application, in order to clearly describe the technical solutions of the embodiments of the present application, the same items or similar items with basically the same functions and effects are distinguished by using "first", "second", and the like. Those skilled in the art can understand that "first", "second", and the like do not limit the quantity and execution order, and "first", "second", and the like do not necessarily mean different. The technical features described by "first" and "second" do not have a prior or size order.

[0040] In the embodiments of the present application, the words "exemplarily" or "for example" are used to represent an example, illustration, or description. Any embodiment or design scheme described as "exemplarily" or "for example" in the embodiments of the present application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the words "exemplarily" or "for example" are used to present the relevant concept in a specific manner, for the purpose of understanding.

[0041] In the embodiments of the present application, at least one can also be described as one or more, and the plurality can be two, three, four, or more, which is not limited in the present application.

[0042] In addition, the specific implementation manners of the technical solutions described in the embodiments of the present application are for more clearly illustrating the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art can know that, with the emergence of new vehicle control technology and sensor technology, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems. Exemplarily, in the prior art, battery thermal management control is mostly passive strategies such as fixed threshold triggering or preset working condition lookup table. For example, the cooling system is started only when the battery monomer temperature reaches 45℃, and the heating system is started when the temperature is lower than 0℃. This way cannot predict the temperature change trend in advance, and is easy to cause the active attenuation of the battery cell due to the sudden rise and fall of the temperature. Or a control instruction is generated through a preset mapping table, but the mapping table is calibrated based on the initial state of the new battery, without considering the influence of battery aging and dynamic working condition on temperature response, resulting in poor adaptability of the control instruction, and it is difficult to solve the local overheating problem caused by the accumulation of temperature difference between the battery cells, and finally affect the battery cycle life and safety performance.

[0043] Based on this, the application provides a battery thermal management method, device, vehicle and equipment, solves the technical problems that the existing thermal management strategy lacks foresight, cannot dynamically adapt to the battery state and complex working conditions, and the temperature regulation precision and energy consumption optimization are difficult to balance, generates a control instruction through a thermal management model, and updates the thermal management model to realize foresight, self-adaptation and high-precision thermal management of the battery temperature, maximizes the reduction of thermal management energy consumption on the premise of ensuring the safety and full life cycle of the battery.

[0044] The embodiments of the application will be described in detail below with reference to the drawings.

[0045] The scheme provided by the application can be applied to Figure 1 The controller 100 shown, Figure 1 The controller 100 provided by the embodiments of the application includes a data processing module 101, an operation module 102 and an execution control module 103.

[0046] Exemplarily, Figure 1 The controller 100 shown can be a vehicle controller, a battery management system controller, or an automobile thermal management domain controller.

[0047] Among them, the data processing module 101, the operation module 102 and the execution control module 103 are connected through an internal bus.

[0048] The data processing module 101 receives the raw data collected by the battery sensor and the vehicle-mounted environment sensor, performs outlier rejection, time alignment, dimension standardization and other preprocessing, and outputs clean first working condition data to the operation module 102, while receiving the second working condition data fed back by the execution control module 103.

[0049] The operation module 102 completes temperature feature prediction, control instruction probability distribution generation and sampling based on the preprocessed first working condition data, and completes model parameter update through back propagation optimization.

[0050] The execution control module 103 sends the control instruction output by the operation module 102 to the thermal management actuator through the bus, collects the battery working condition data after execution, and feeds back to the data processing module 101.

[0051] Figure 2 The flowchart of the battery thermal management method provided by the embodiments of the application. The method can be executed by a controller, which can be Figure 1controller 100 in the figure. The battery thermal management method provided in the embodiments of the present application can be applicable to various scenarios requiring accurate and adaptive regulation of battery temperature, including but not limited to: under different driving conditions such as urban congestion, high-speed driving, mountain climbing, and under extreme environmental temperatures such as extreme cold and extreme heat, the temperature of power batteries such as ternary lithium batteries and lithium iron phosphate batteries is regulated in advance, which not only guarantees the charging and discharging performance and cycle life of the battery, but also avoids the risk of thermal runaway caused by temperature out of control; in the scenarios of grid peak shaving and new energy consumption, the temperature change of large-scale lithium ion battery packs under high-rate charging and discharging and long-term standing is controlled, and the energy consumption is optimized to ensure the safety and economy of the energy storage system; adapt to the state change of the battery from new battery to aged battery, continuously update the model parameters, and always maintain the optimal thermal management strategy to solve the problem of decreased regulation accuracy of the traditional fixed strategy after the battery is aged.

[0052] As shown in Figure 2 The battery thermal management method provided in the embodiments of the present application can include: S201: Obtain first working condition data of the battery.

[0053] The state of the battery refers to the current electrochemical characteristics and operating state parameters of the battery, including battery remaining capacity, health state, single cell temperature / temperature field distribution, charging and discharging current, terminal voltage, maximum temperature difference of the cell, temperature rise rate, etc., which directly reflects the energy storage capacity and thermal safety state of the battery.

[0054] The operating environment of the battery refers to external working condition conditions that affect the temperature change of the battery, including but not limited to environmental temperature, environmental humidity, vehicle speed, driving mode, road slope, vehicle load demand, etc.

[0055] Optionally, the first working condition data can include: a single cell temperature set, a battery pack total voltage, a charging and discharging current, a battery remaining capacity value (State of Charge, SOC), a battery health state value (State of Health, SOH), an environmental temperature, a vehicle speed, a driving mode, and a cell temperature difference.

[0056] Specifically, the battery state parameters are collected through a sensor array of a battery management system, the environmental and driving condition data sent by a vehicle controller are received through a vehicle bus or Ethernet, the collected raw data are subjected to pretreatment operations such as filtering and denoising, unifying data collection period, and mapping parameters to the [0, 1] interval, and the first working condition data are output.

[0057] S202: Make a thermal management decision based on the first working condition data through a battery thermal management model to obtain a control instruction.

[0058] The control instruction refers to a specific parameter instruction for driving the battery thermal management actuator to act, including but not limited to the fan speed of the cooling system, the water pump flow, the electronic expansion valve opening degree, the heater power of the heating system, or the valve switching state of the thermal management loop, directly determining the strength and mode of battery temperature regulation.

[0059] In some embodiments, the battery temperature feature data in the first time period is obtained by predicting based on the first working condition data through a prediction layer network, and the control instruction is obtained by deciding based on the battery temperature feature data and the first working condition data through a decision layer network.

[0060] Specifically, the preprocessed first working condition data is input into the prediction layer network, the prediction layer network learns the time sequence dependence of the first working condition data, and outputs the battery temperature feature data in the future first time period; the temperature feature data and the original first working condition data are fused to generate a state vector and input into the decision layer network, and the decision layer network outputs the executable control instruction through the probability modeling and constraint checking of the PPO algorithm.

[0061] The battery temperature feature data refers to the time sequence evolution characteristics of the battery temperature in the first time period, including the predicted temperature value, the maximum temperature difference prediction value, the temperature rise rate prediction value, the temperature peak value and the occurrence time of each time in the future, reflecting the dynamic change trend of the battery temperature.

[0062] The prediction layer network refers to a time sequence prediction model constructed with LSTM as the core, the input of which is the time sequence of the first working condition data, and the output of which is the battery temperature feature data in the first time period, which is used to capture the battery temperature change rule in advance.

[0063] The decision layer network refers to a network constructed based on a reinforcement learning algorithm, the input of which is the fused state vector, and the output of which is the probability distribution of the control instruction, which is optimized through policy iteration to achieve the optimal balance among temperature stability, cell consistency and energy consumption of the generated control instruction.

[0064] Specifically, the decision layer network refers to a decision network constructed based on a reinforcement learning (RL) algorithm, which includes core components such as policy network and value network, can fuse the battery temperature feature data and the first working condition data, and dynamically generate the control instruction adapted to the current battery state and operating environment through continuous interaction with the battery thermal management environment, i.e. obtaining feedback after executing the control instruction, learning the optimal decision strategy, to achieve the comprehensive goal of battery temperature stability, cell consistency and energy consumption optimization.

[0065] The long short-term memory network LSTM refers to a special recurrent neural network, which solves the gradient disappearance or gradient explosion problem of traditional RNN in long-time sequence data processing through the gating mechanism of the forgetting gate, the input gate and the output gate, learns the long-term dependence relationship between the working condition data and the temperature change, and adapts to the time sequence prediction demand of the battery temperature. Exemplarily, the battery temperature feature data is data-fused with the battery working condition data to obtain a state vector, and the state vector is input to a decision layer network for processing to obtain a control instruction.

[0066] Specifically, the battery temperature feature data and the first working condition data are dimensionally matched, for example, the time sequence dimension of the temperature feature data is flattened into a vector, and the information of the two types of data is fused through feature splicing or attention mechanism to generate a dimensionally unified state vector.

[0067] Data fusion refers to the process of integrating feature data of different sources and different dimensions into a unified feature vector, for example, splicing "future 5-second predicted temperature sequence, dimension 5", "current SOC, dimension 1", "environmental temperature, dimension 1", and "cell temperature difference, dimension 1" into a state vector of dimension 8, so that the decision layer network can simultaneously capture the time sequence prediction information and real-time working condition information.

[0068] Exemplarily, the state vector is nonlinearly transformed by the hidden layer of the decision layer network to obtain high-order features. Based on the high-order features, the output layer of the decision layer network generates a first probability distribution of the control instruction, samples from the first probability distribution to obtain a candidate control instruction, and performs physical constraint verification on the candidate control instruction to obtain the control instruction.

[0069] The decision layer network includes a hidden layer and an output layer.

[0070] The hidden layer refers to the neural network layer between the input layer and the output layer in the decision layer network, which is composed of a fully connected layer and an activation function, and is responsible for nonlinear transformation and feature abstraction of the input state vector.

[0071] The output layer refers to the final output layer of the decision layer network, which is usually a fully connected layer, and outputs the parameters of the control instruction probability distribution based on the high-order features output by the hidden layer.

[0072] The high-order feature refers to a feature extracted from the state vector for describing the battery temperature regulation effect.

[0073] For example, the coupling relationship feature of the predicted temperature rise rate and the cooling power, the mapping feature of the sensitivity of the SOC level to the temperature regulation, and the trade-off feature of the environmental temperature and the energy consumption optimization are the basis for generating a reasonable control instruction probability distribution.

[0074] Specifically, the full connection layer of the input hidden layer performs linear transformation, such as mapping the dimension from 8 to 64, and then performs non-linear mapping through an activation function to filter irrelevant information and strengthen key features, and finally outputs a high-order feature vector with a dimension of 64.

[0075] The first probability distribution takes the effective adjustment range of the control instruction as the domain, and contains probability density information of different adjustment intensities.

[0076] Specifically, the first probability distribution is a continuous Gaussian distribution, and the domain is the physical adjustment range of the thermal management actuator, such as the effective range of the cooling fan speed being 0 to 1000 rpm and the effective range of the heater power being 0 to 5 kW. The higher the probability density of the adjustment intensity, the greater the possibility that the control instruction can make the battery temperature adjustment more optimal at this intensity.

[0077] The candidate control instruction refers to a preliminary adjustment instruction randomly sampled from the first probability distribution without hardware constraint verification, and the value of the candidate control instruction may exceed the actual operating limit of the actuator.

[0078] The physical constraint verification refers to the process of checking and correcting the candidate control instruction for hardware feasibility, to ensure that the final output control instruction is within the physical performance limit of the thermal management actuator, and to avoid damage to the actuator or failure of temperature adjustment due to excessive instruction range.

[0079] For example, based on the mean and standard deviation of the first probability distribution, a candidate control instruction is determined based on the sampling result.

[0080] The mean and standard deviation of the first probability distribution refer to the core parameters of the Gaussian distribution, where the mean represents the control instruction intensity that is most likely to optimize the adjustment effect, and the standard deviation represents the fluctuation range of the control instruction intensity.

[0081] The sampling result refers to a specific control instruction value randomly extracted based on the probability density of the Gaussian distribution.

[0082] For example, when the adjustment range of the candidate control instruction is greater than the adjustment range threshold, the adjustment range of the candidate control instruction is determined based on the adjustment range threshold to obtain the control instruction.

[0083] The adjustment range threshold refers to the maximum safe adjustment limit of the thermal management actuator, which is determined by the hardware performance of the actuator.

[0084] The adjustment range of the candidate control instruction being greater than the adjustment range threshold means that the value of the candidate control instruction exceeds the maximum safe adjustment limit of the actuator.

[0085] The adjustment range of the candidate control instruction based on the adjustment range threshold refers to truncating the candidate control instruction exceeding the threshold to the threshold range, such as correcting the candidate instruction of 1300 rpm to 1200 rpm, to ensure that the final output control instruction meets the hardware safety constraint.

[0086] S203: Perform parameter updating on the battery thermal management model based on the first working condition data and the second working condition data.

[0087] The second working condition data is the working condition data corresponding to the battery after the temperature of the battery is adjusted.

[0088] In some embodiments, by the battery thermal management model, a decision is made based on the second working condition data to obtain a second probability distribution, a probability ratio is obtained based on the first probability distribution and the second probability distribution, a reward value is calculated based on the first working condition data, the second working condition data, and a reward function, an advantage function value is calculated based on the reward value, the first working condition data, and the second working condition data, a target function is constructed based on the probability ratio and the advantage function value, adjustment parameters of the battery thermal management model are obtained based on a back propagation algorithm and the target function, and the battery thermal management model is adjusted based on the adjustment parameters.

[0089] The second probability distribution refers to the continuous probability distribution of the control instruction output by the battery thermal management model under the second working condition data, which has the same form as the first probability distribution and a definition domain of the effective adjustment range of the control instruction, and is used to quantify the preference of the model for different adjustment intensities under the battery state after executing the control instruction. The corresponding formula is: wherein, is the current model parameter, is the second working condition data.

[0090] The probability ratio refers to the ratio of the second probability distribution to the first probability distribution, also known as the strategy ratio, which is used to measure the difference in probability preference for the same control instruction before and after the model parameter updating, and the calculation formula is:

[0091] wherein, is the strategy ratio, representing the probability density ratio of the new strategy to the old strategy for the same control instruction at time t, and is used to measure the amplitude of the strategy updating, is the new strategy function based on the current model parameter to be updated The decision strategy constructed outputs the probability distribution of the control instruction, is the old strategy function based on the model parameter before updating The decision strategy constructed outputs the probability distribution of the control instruction, is the state at time t: a feature vector describing the current state of the battery and the operating environment, These are the model parameters for the new strategy, and the battery thermal management model parameters to be updated. Control commands in the second probability distribution The probability density, In the first probability distribution The probability density, This refers to control commands, which are actuator action signals used to regulate battery temperature. If the value is approximately equal to 1, then the new strategy is comparable to the old strategy. Their preferences are similar, and the update magnitude is small; if If the value is much greater than or much less than 1, the new strategy differs greatly from the old strategy, and the update magnitude is aggressive.

[0092] The reward function is used to evaluate the battery's temperature stability, cell consistency, and energy consumption level. Its mathematical expression is:

[0093] in, It's the reward value, a quantitative evaluation result after the execution of a single control command. A higher value indicates better control performance, while a lower value indicates worse control performance. It's the core feedback signal for model updates. The target temperature in the first operating condition data. This refers to the actual temperature in the second operating condition data. This refers to the maximum temperature difference of the battery cells in the second operating condition data. This refers to the energy consumption of the thermal management system in the second operating condition data. , and For example, negative weighting coefficients, =-0.1、 =-0.05、 =-0.02, which results in a better reward value when the temperature is more stable, the temperature difference is smaller, and the energy consumption is lower. This is a negative weighting coefficient for temperature stability, used to penalize deviations between the actual and target temperatures. This is a weighting coefficient for cell consistency, used to penalize the maximum temperature difference within the cell. It takes a negative value; the smaller the temperature difference, the higher the reward value for this contribution. This is a negative weighting coefficient for energy consumption levels, used to penalize the energy consumption of the thermal management system. The reward value reflects the battery's performance in temperature stability, cell consistency, and energy consumption level after the control command is executed. The specific calculation logic is as follows: extract from the data of the first operating condition. Extract from the second working condition data , , Substituting into the reward function formula, we obtain the single-step reward value. .

[0094] Specifically, according to the first working condition data, the second working condition data and the reward function, the process of calculating the reward value is to analyze the target temperature parameter from the first working condition data, extract the actual temperature sequence, the cell temperature difference sequence and the thermal management actuator energy consumption data from the second working condition data, perform numerical operation according to the preset reward function formula, output the quantitative reward value as the feedback signal of model updating.

[0095] The specific process of constructing the target function based on the probability ratio and the advantage function value is to use the clipping target function of the PPO algorithm, and the expression is:

[0096] Among them, is the target function of PPO, used to measure the advantages and disadvantages of the new strategy, and the target is to optimize to minimize or maximize, guide the policy to generate better battery temperature control instructions, [ ] is the mean value of the batch samples, is the clipping threshold, used to limit the policy update amplitude, when exceeds the range of [0.8, 1.2], it is forced to be calculated according to the boundary value, avoiding the instability of battery temperature regulation caused by policy mutation, clip( ) is a clipping function that limits the input value within a specified range to avoid too large or too small values, is the boundary of the clipping interval, is the preset clipping threshold, so the interval is [0.8, 1.2], and exceeding the range will be forced to be truncated to the boundary value.

[0097] Based on the back propagation algorithm and the target function, the adjustment parameters of the battery thermal management model are the gradients of the model parameters, the weights of the prediction layer LSTM and the hidden layer / output layer of the decision layer network calculated by the back propagation of the target function. After gradient clipping, the random gradient descent is used to update the parameters with a preset learning rate, such as 1e-4 for the decision layer and 5e-5 for the prediction layer, to obtain the adjustment amount of each parameter.

[0098] Illustratively, the gradient of the target function with respect to the parameters of the battery thermal management model is calculated based on the back propagation algorithm, and the adjustment amount of the battery thermal management model parameters is obtained by optimizing the gradient. The adjustment amount is the adjustment parameter of the battery thermal management model.

[0099] ​Specifically, the back propagation process starts from the objective function, and calculates the gradients of the output layer, hidden layer, and input layer parameters in turn: for the decision layer output layer, the gradient of the objective function on the Gaussian distribution mean / standard deviation parameters is calculated; for the hidden layer, the gradient of the objective function on the high-order feature mapping weight is calculated; for the prediction layer LSTM, the gradient of the objective function on the gating mechanism weight is calculated. Gradient clipping ensures smooth parameter updates by limiting the L2 norm of the gradient vector within a preset threshold.

[0100] According to the rule of parameter new value = parameter old value - learning rate x clipped gradient, the adjustment amount of each parameter is calculated, such as the adjustment amount of a certain hidden layer weight is -1e-4x0.02=-2e-6, which is the adjustment parameter used to update the model.

[0101] S204: performing thermal management on the battery based on the updated battery thermal management model.

[0102] Specifically, the second working condition data of the battery is collected, and clean data is output according to the pre-processing logic, the updated battery thermal management model is called, and thermal management decisions are made based on the second working condition data: the prediction layer network more accurately outputs the temperature feature data of the future first time period, the decision layer network fuses the feature data and the second working condition data to generate a control instruction probability distribution more adaptive to the current state, and after sampling and physical constraint verification, an optimized control instruction is obtained; the control instruction is sent to the thermal management actuator through the vehicle-mounted bus to drive the actuator to act to adjust the battery temperature; at the same time, the new working condition data after execution is collected as the input of the next round of model parameter update.

[0103] Figure 3 Another flowchart of a battery thermal management method provided by the embodiments of the present application is provided. The method can be executed by a controller, which can be the controller 100 in the vehicle thermal management system 1000. Figure 1

[0104] S301: obtaining first working condition data of a battery.

[0105] The introduction of this step can be referred to S201, which will not be described in detail here.

[0106] S302: making a thermal management decision based on the first working condition data by a battery thermal management model to obtain a control instruction.

[0107] The introduction of this step can be referred to S202, which will not be described in detail here.

[0108] S303: updating the parameters of the battery thermal management model based on the first working condition data and the second working condition data.

[0109] The introduction of this step can be referred to S203, which will not be described in detail here.

[0110] ​S304: Simulate and verify the updated battery thermal management model in the working condition database.

[0111] Specifically, historical data of multiple groups of typical battery working condition scenarios are extracted from the working condition database, which are used as inputs to drive the updated battery thermal management model to perform thermal management decisions and generate control instructions. At the same time, based on the ideal thermal management effect preset in the working condition database, such as the target temperature curve, the battery cell temperature difference threshold, and the energy consumption standard, the control instructions output by the model are quantitatively evaluated in terms of temperature stability, battery cell consistency, and energy consumption level, and a simulation verification report is generated to determine whether the model parameter update meets the actual application requirements.

[0112] The working condition database refers to a collection of data of various typical running scenarios of the battery throughout its life cycle, including working condition data under different environmental temperatures, different charge and discharge rates, different battery aging states, and different vehicle load demands, as well as corresponding ideal thermal management strategy effect labels, which are used to simulate and verify the robustness of the model in complex scenarios.

[0113] S305: If the simulation verification result meets the termination condition, the adjusted parameters are rolled back to the parameters before adjustment.

[0114] Specifically, if the simulation verification report shows that the updated battery thermal management model does not meet the preset standard in terms of key performance indicators, it is determined that the simulation verification result meets the termination condition. At this time, the parameters of the battery thermal management model are rolled back from the updated state to the state before parameter adjustment, to avoid risks such as temperature loss of control and excessive energy consumption in the actual battery thermal management process due to model performance degradation.

[0115] The simulation verification result refers to the quantitative evaluation results of the control instructions output by the updated model after simulation in the working condition database in terms of temperature regulation effect, battery cell temperature difference control, and energy consumption level, including but not limited to temperature deviation mean, maximum battery cell temperature difference statistical value, and thermal management energy consumption mean. The termination condition refers to the pre-set model performance degradation threshold.

[0116] For example, when any one of the indicators in the simulation verification result triggers the threshold, such as an increase in temperature deviation mean of more than 1°C, a proportion of times of exceeding the maximum battery cell temperature difference of more than 20%, and an increase in thermal management energy consumption of more than 15% compared to before adjustment, it is determined that the parameters need to be rolled back. The termination condition refers to the pre-set model performance degradation threshold, such as "temperature deviation mean increases by more than 1°C compared to before adjustment", "proportion of times of exceeding the maximum battery cell temperature difference is more than 20%", and "thermal management energy consumption increases by more than 15% compared to before adjustment". When any one of the indicators in the simulation verification result triggers the threshold, it is determined that the parameters need to be rolled back.

[0117] S306: Thermal management of the battery based on the updated battery thermal management model.

[0118] The introduction of this step can be referred to S204, which will not be described in detail here.

[0119] Figure 4 A logic diagram of a battery thermal management method provided by an embodiment of the present application. The method can be executed by a controller, which can be the controller 100 in the battery thermal management system 1000, and the steps are as follows: Figure 1 Input operating condition data 401: input the real-time state and operating environment data of the battery to provide a basis for subsequent analysis.

[0120] LSTM time series feature extraction 402: process the time series operating condition data through the long short-term memory network LSTM to extract the change trend of the battery temperature and other key features, and provide forward-looking information for decision-making.

[0121] RL decision 403: based on the RL reinforcement learning algorithm, combine the features extracted by LSTM and the real-time operating condition data to generate the optimal control instruction.

[0122] Execute the control instruction 404: send the control instruction output by the decision layer to the thermal management actuator to execute the temperature adjustment action.

[0123] Battery response 405: collect the battery operating condition data after the execution instruction to reflect the actual effect of the control instruction.

[0124] Reward feedback 406: according to the preset reward function, quantify the control effect and generate a reward signal as feedback for model optimization.

[0125] History data training 407: obtain historical operating condition data to continuously optimize the LSTM model and the decision model, and improve the prediction accuracy.

[0126] Specifically, in the history data training, 100,000 groups of historical operating condition data cover extreme temperatures, different driving modes, and battery full life cycle states, and the label of supervised training is the battery temperature sequence in the next 5 minutes under the corresponding time series. By minimizing the mean square error of the predicted temperature and the actual temperature, the LSTM learns the general temperature variation law, such as the polarization heat dominated temperature rise at low temperature and the heat dissipation efficiency decay characteristic at high temperature; the pre-trained LSTM and the decision model form an end-to-end framework, the parameters of the previous layer are frozen to retain the general time series feature extraction ability, and only the last layer weight is fine-tuned through the real-time feedback reward signal, so that the output temperature feature is more suitable for the individualized characteristics of the current battery. This reduces the training cost while taking into account the model generalization and scene adaptability, and improves the temperature prediction accuracy by more than 15%.

[0127] Figure 5 ​A structural schematic diagram of a battery thermal management device is provided for an embodiment of the present application. The battery thermal management device can include an acquisition module 501, a decision module 502, an update module 503, and a management module 504. The acquisition module 501 is configured to perform Figure 2 the operation of S201 in the schematic method, and Figure 3 the operation of S301 in the schematic method; the decision module 502 is configured to perform Figure 2 the operation of S202 in the schematic method, and Figure 3 the operation of S302 in the schematic method; the update module 503 is configured to perform Figure 2 the operation of S203 in the schematic method, and Figure 3 the operations of S303, S304, and S305 in the schematic method; and the management module 504 is configured to perform Figure 2 the operation of S204 in the schematic method, and Figure 3 the operation of S306 in the schematic method.

[0128] The acquisition module 501 is configured to acquire first working condition data of a battery. The first working condition data is used to describe the state of the battery and the operating environment in which the battery is located. The decision module 502 is configured to perform thermal management decision-making based on the first working condition data by using a battery thermal management model, to obtain a control instruction. The control instruction is used to adjust the temperature of the battery. The update module 503 is configured to perform parameter updating on the battery thermal management model based on the first working condition data and second working condition data. The second working condition data is working condition data corresponding to the battery after the temperature of the battery is adjusted. The management module 504 is configured to perform thermal management on the battery based on the updated battery thermal management model.

[0129] The decision module 502 is further configured to perform prediction based on the first working condition data by using a prediction layer network, to obtain battery temperature feature data in a first time period. The prediction layer network is a time series prediction layer network constructed based on a long short-term memory network (LSTM). The decision module 502 is further configured to perform decision-making based on the battery temperature feature data and the first working condition data by using a decision layer network, to obtain the control instruction. The decision layer network is a reinforcement learning decision network constructed based on a proximal policy optimization (PPO) algorithm.

[0130] The decision module 502 is further configured to perform data fusion on the battery temperature feature data and the battery working condition data, to obtain a state vector. The decision module 502 is further configured to input the state vector into the decision layer network for processing, to obtain the control instruction.

[0131] The decision module 502 is further configured to perform a nonlinear transformation on the state vector through a hidden layer of the decision layer network to obtain high-order features, the high-order features being features extracted from the state vector and used to describe the temperature regulation effect of the battery, generate a first probability distribution of the control instruction based on the high-order features through an output layer of the decision layer network, the first probability distribution having an effective regulation range of the control instruction as a definition domain and containing probability density information of different regulation intensities, sample the first probability distribution to obtain a candidate control instruction, and perform physical constraint verification on the candidate control instruction to obtain the control instruction.

[0132] The decision module 502 is further configured to sample the first probability distribution based on a mean value and a standard deviation of the first probability distribution, and determine the candidate control instruction based on a result of the sampling.

[0133] The decision module 502 is further configured to, in a case where a regulation range of the candidate control instruction is greater than a regulation range threshold, determine the regulation range of the candidate control instruction based on the regulation range threshold to obtain the control instruction.

[0134] The update module 503 is further configured to perform decision-making based on the second working condition data through the battery thermal management model to obtain a second probability distribution, obtain a probability ratio based on the first probability distribution and the second probability distribution, calculate a reward value based on the first working condition data, the second working condition data and a reward function, the reward function being used to evaluate temperature stability, cell consistency and energy consumption level of the battery, the reward value being used to reflect performance of the battery in the temperature stability, the cell consistency and the energy consumption level after execution of the control instruction, calculate an advantage function value based on the reward value, the first working condition data and the second working condition data, construct a target function based on the probability ratio and the advantage function value, obtain an adjustment parameter of the battery thermal management model based on a back propagation algorithm and the target function, and adjust the battery thermal management model based on the adjustment parameter.

[0135] The update module 503 is further configured to calculate a gradient of the target function with respect to a parameter of the battery thermal management model based on the back propagation algorithm, optimize the gradient to obtain an adjustment amount of the parameter of the battery thermal management model, and the adjustment amount being the adjustment parameter of the battery thermal management model.

[0136] The update module 503 is further configured to simulate and verify the updated battery thermal management model in the working condition database, and roll back the adjusted parameter to a parameter before adjustment in a case where a simulation verification result meets a termination condition.

[0137] Please refer to Figure 6 , Figure 6A structural schematic diagram of an electronic device is provided in embodiments of the present application. The electronic device can include a processor 601, a memory 602. The processor 601 is in communication with the memory 602. The memory 602 is configured to store programs, and the processor 601 is configured to execute the programs, and specifically can execute the related steps in the above-described embodiments of the temperature control method.

[0138] It should be noted that those skilled in the art can understand that the electronic device structure shown in the above embodiments is not a limitation to the electronic device, and the electronic device can include more or fewer components than those shown in the above embodiments, or combine certain components, or different component arrangements. Figure 6 Figure 6 The electronic device structure shown in the above embodiments does not constitute a limitation to the electronic device, and the electronic device can include more or fewer components than those shown in the above embodiments, or combine certain components, or different component arrangements.

[0139] The processor 601 is the control center of the electronic device, connects various parts of the entire electronic device through various interfaces and lines, executes the software programs and / or modules stored in the memory 602 and calls the data stored in the memory 602, executes various functions of the electronic device and processes data, and thus monitors the entire electronic device. The processor 601 can include one or more processing units. Optionally, the processor 601 can integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, etc., and the modem processor mainly processes wireless communication. It can be understood that the above-mentioned modem processor can also not be integrated into the processor 601.

[0140] The memory 602 can be used to store software programs and various data. The memory 602 can mainly include a program storage area and a data storage area, wherein the program storage area can store the operating system, the application programs (such as determination unit, processing unit, etc.) required by at least one function module, etc. In addition, the memory 602 can include a high-speed random access memory, and can also include a non-volatile memory, for example, at least one magnetic disk storage device, a flash memory device, or other volatile solid-state memory device.

[0141] From the above description of the embodiments, those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of functional modules is taken as an example for illustration, and in actual application, the above-mentioned functions can be completed by different functional modules according to needs, i.e., the internal structure of the module is divided into different functional modules to complete all or part of the functions described above. The specific working process of the above-described system, module and unit can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0142] ​The method steps in the embodiments can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, which can be stored in a random access memory (RAM), a flash memory, a read-only memory (ROM), a programmable ROM (PROM), an erasable PROM (EPROM), an electrically EPROM (EEPROM), a register, a hard disk, a mobile hard disk, a CD-ROM, or any other form of storage medium well known in the art. An exemplary storage medium is coupled to a processor, so that the processor can read information from, and write information to, the storage medium. Of course, the storage medium can be a component of the processor. The processor and the storage medium can be located in an ASIC. The ASIC can be located in a network device. Of course, the processor and the storage medium can also exist as discrete components in the network device. In the above embodiments, the entire or part of the flow or function can be implemented by software, hardware, firmware, or any combination thereof. When implemented by software, the entire or part of the flow or function can be implemented in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer programs or instructions are loaded on a computer, the entire or part of the flow or function of the embodiments of the present application is executed. The computer can be a general purpose computer, a special purpose computer, a computer network, a network device, a user equipment, or other programmable modules. The computer programs or instructions can be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another computer readable storage medium, for example, from a website site, a computer, a server, or a data center to another website site, a computer, a server, or a data center through a wired or wireless way. The computer readable storage medium can be any available medium or a data storage device integrated with one or more available media in a server, a data center, or the like, which can be accessed by a computer. The available medium can be a magnetic medium, such as a floppy disk, a hard disk, or a magnetic tape; an optical medium, such as a digital video disc (DVD); or a semiconductor medium, such as a solid state drive (SSD). The above are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any modification or replacement within the technical range disclosed in the present application can be easily thought of by those skilled in the art, and these modifications or replacements should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0143] Since the data processing apparatus in the embodiments of the present application can be applied to the above-mentioned method, the technical effects that can be obtained by the data processing apparatus can also be referred to the above-mentioned method embodiments, and the embodiments of the present application will not be described here. The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, any change or replacement within the technical scope disclosed in the present application should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0144] It should be understood that the application of the present application is not limited to the above examples, and those of ordinary skill in the art can improve or change it according to the above description, and all these improvements and changes should belong to the protection scope of the claims attached to the present application. Those of ordinary skill in the art can understand the implementation of all or part of the above-mentioned processes, and the equivalent changes made according to the claims of the present application still belong to the scope covered by the present application.

Claims

1. A battery thermal management method, characterized in that, The method is applied to a controller, which manages the battery's thermal performance using a battery thermal management model. The method includes: Acquire first operating condition data of the battery, the first operating condition data being used to describe the state of the battery and the operating environment in which the battery is located; Based on the first operating condition data, thermal management decisions are made using the battery thermal management model to obtain control commands, which are used to adjust the temperature of the battery. The battery thermal management model is updated based on the first operating condition data and the second operating condition data. The second operating condition data is the operating condition data of the battery after the temperature of the battery is adjusted. The battery is thermally managed based on the updated battery thermal management model.

2. The method according to claim 1, characterized in that, The thermal management model includes a prediction layer network and a decision layer network. The thermal management model, based on the first operating condition data, makes thermal management decisions and obtains control commands, including: The battery temperature characteristic data within the first time period is obtained by predicting based on the first operating condition data through the prediction layer network; the prediction layer network refers to a time-series prediction layer network constructed based on the Long Short-Term Memory (LSTM) network. The decision layer network makes decisions based on the battery temperature characteristic data and the first operating condition data to obtain control commands. The decision layer network is a network built based on a reinforcement learning algorithm.

3. The method according to claim 2, characterized in that, The decision-making process, based on the battery temperature characteristic data and the first operating condition data, through the decision-making layer network to obtain control commands includes: The battery temperature characteristic data is fused with the first operating condition data to obtain a state vector; The state vector is input into the decision layer network for processing to obtain the control command.

4. The method according to claim 3, characterized in that, The decision layer network includes a hidden layer and an output layer. The step of inputting the state vector into the decision network to obtain the control command includes: The state vector is nonlinearly transformed by the hidden layer of the decision layer network to obtain higher-order features, which are features extracted from the state vector to describe the battery temperature regulation effect. Based on the higher-order features, a first probability distribution of control commands is generated through the output layer of the decision layer network. The first probability distribution is defined by the effective adjustment range of the control commands and includes probability density information of different adjustment intensities. Candidate control commands are obtained by sampling from the first probability distribution; The candidate control commands are physically constrained to obtain the control commands.

5. The method according to claim 4, characterized in that, The step of sampling from the first probability distribution to obtain candidate control commands includes: Based on the mean and standard deviation of the first probability distribution, sampling is performed in the first probability distribution; Based on the sampling results, the candidate control commands are determined.

6. The method according to claim 4, characterized in that, The step of performing physical constraint verification on the candidate control commands to obtain control commands includes: If the adjustment range of the candidate control command is greater than the adjustment range threshold, the adjustment range of the candidate control command is determined based on the adjustment range threshold to obtain the control command.

7. The method according to claim 1, characterized in that, The step of updating the parameters of the battery thermal management model based on the first operating condition data and the second operating condition data includes: The battery thermal management model makes a decision based on the second operating condition data to obtain a second probability distribution; The probability proportions are obtained based on the first probability distribution and the second probability distribution; Based on the first operating condition data, the second operating condition data, and the reward function, a reward value is calculated. The reward function is used to evaluate the temperature stability, cell consistency, and energy consumption level of the battery. The reward value is used to reflect the battery's performance in temperature stability, cell consistency, and energy consumption level after the control command is executed. Based on the reward value, the first working condition data, and the second working condition data, the advantage function value is calculated; Construct an objective function based on the probability ratio and the advantage function value; Based on the backpropagation algorithm and the objective function, the adjustment parameters of the battery thermal management model are obtained; The battery thermal management model is adjusted based on the aforementioned adjustment parameters.

8. The method according to claim 7, characterized in that, The adjustment parameters of the battery thermal management model, obtained based on the backpropagation algorithm and the objective function, include: The gradient of the objective function with respect to the parameters of the battery thermal management model is calculated based on the backpropagation algorithm. The gradient is optimized to obtain the adjustment amount of the battery thermal management model parameters, and the adjustment amount is the adjustment parameter of the battery thermal management model.

9. The method according to any one of claims 1-8, characterized in that, The method further includes: The updated battery thermal management model was simulated and verified in the operating condition database. If the simulation verification results meet the termination conditions, the adjusted parameters will be reverted to the parameters before adjustment.

10. A battery thermal management device, characterized in that, The device includes: Acquisition module: used to acquire the first operating condition data of the battery, the first operating condition data being used to describe the state of the battery and the operating environment in which the battery is located; Decision module: used to make thermal management decisions based on the first operating condition data through the battery thermal management model, and obtain control commands, the control commands being used to adjust the temperature of the battery; Update module: used to update the parameters of the battery thermal management model based on the first operating condition data and the second operating condition data, wherein the second operating condition data is the operating condition data of the battery after adjusting the battery temperature; Management module: Used to perform thermal management on the battery based on the updated battery thermal management model.

11. A vehicle, characterized in that, The battery thermal management control device in the vehicle uses the method described in any one of claims 1-9.

12. An electronic device, characterized in that, include: processor; Memory, which stores computer programs; When the computer program is executed by the processor, it implements the method as described in any one of claims 1-9.

Images