A multi-objective thermal management control method for lithium-ion batteries
By constructing an SOH adaptive thermal model and a multi-rate execution architecture, the problems of thermal safety and real-time adaptability in lithium-ion battery thermal management are solved, achieving coordinated optimization of hot spot temperature, temperature difference and energy consumption, extending battery life and reducing energy consumption.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- EAST CHINA JIAOTONG UNIVERSITY
- Filing Date
- 2026-06-18
- Publication Date
- 2026-07-21
AI Technical Summary
Existing thermal management control methods for lithium-ion batteries fail to effectively adapt to the degradation of battery health status, resulting in decreased thermal safety. The large computational load affects real-time adaptability and makes it difficult to achieve comprehensive optimization of temperature control accuracy, cooling energy consumption, and life-related thermal behavior.
An adaptive dual-state lumped parameter thermal model with SOH is constructed. A dynamic safety domain is built by combining SOH, SOC, charge/discharge rate and ambient temperature. Online optimization is performed using a multi-rate execution architecture and an improved starfish optimization algorithm. A multi-objective comprehensive fitness function is constructed to achieve synergistic optimization of hot spot temperature, temperature difference suppression, cooling energy consumption and lifetime-related thermal behaviors.
It improves the thermal safety and real-time adaptability of lithium-ion batteries throughout their entire life cycle, extends battery life, reduces cooling energy consumption, and improves driving comfort, thus realizing the comprehensive benefits of the thermal management system.
Smart Images
Figure CN122436619A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power battery technology for new energy vehicles, and specifically to a multi-objective thermal management control method for lithium-ion batteries. Background Technology
[0002] As a core power component of new energy vehicles, the operating temperature of lithium-ion batteries directly determines their charge-discharge performance, cycle life, and safety. The optimal operating temperature range for lithium-ion batteries is typically 25~40℃, and the maximum temperature difference within the battery pack must be controlled within 5℃. Excessive temperature accelerates irreversible aging of the battery's SEI film, degrades active materials, and may even trigger thermal runaway safety accidents; excessively low temperature leads to a sharp decline in battery charge-discharge capacity, directly reducing the vehicle's driving range; while excessive temperature differences within the battery pack cause deterioration in the consistency of individual cells, further shortening the overall lifespan of the battery pack and increasing the total lifespan cost of the vehicle.
[0003] Model predictive control (MPC) has become the mainstream control method for current vehicle battery thermal management systems due to its inherent advantages such as multivariable collaborative control, constrained rolling optimization, and the ability to predict operating conditions. However, existing thermal management MPC solutions still face the following serious challenges in practical applications: First, existing control frameworks generally adopt fixed safety constraints and fixed parameter thermal models, which do not take into account the decrease in thermal stability and changes in heat generation characteristics caused by the degradation of battery state of health (SOH), making it difficult to adapt to the thermal safety requirements of aging batteries at high SOC, high rate and high ambient temperature. Secondly, most existing online optimization schemes for MPC weights are direct applications of general swarm intelligence algorithms, which require a large number of global iterative calculations to be performed at high frequency in each control cycle. This results in a large amount of computation and affects the real-time adaptability of complex predictive control under limited computing power. Third, existing control methods mostly take "minimizing temperature tracking error" as the single core objective, without fully considering temperature difference suppression, cooling energy consumption, and life-related thermal behaviors such as temperature fluctuations and high-temperature residence. This makes it difficult for thermal management systems to achieve comprehensive optimization between temperature control accuracy, cooling energy consumption, and long-term life. Summary of the Invention
[0004] In view of this, the present invention provides a multi-objective thermal management control method for lithium-ion batteries, which improves thermal safety and the real-time adaptability of complex predictive control under limited computing power conditions, and achieves multi-objective collaborative optimization of hot spot temperature, internal temperature difference, cooling energy consumption and life-related thermal behavior while meeting dynamic safety constraints.
[0005] A multi-objective thermal management control method for lithium-ion batteries includes: Step S1: Obtain the real-time state information of the lithium-ion battery and establish an SOH adaptive dual-state lumped parameter thermal model. The model uses the hot spot temperature and edge temperature of the battery pack as state variables and the coolant flow rate and coolant inlet temperature as control variables to obtain the thermal model prediction results. The thermal model prediction results include the predicted hot spot temperature and edge temperature. Step S2: Based on the thermal model prediction results obtained in step S1, and combined with SOH, SOC, charge / discharge rate, ambient temperature and predicted heat load, construct a dynamic safety domain coupled with operating conditions and health. The dynamic safety domain includes at least a dynamic temperature safety upper limit, a dynamic temperature difference upper limit and a temperature rise rate constraint. Step S3: Construct a safety penalty term based on the dynamic safety domain obtained in step S2, and simultaneously construct a safety boundary crossing risk penalty term based on the absolute safety boundary, thereby constructing a multi-objective comprehensive fitness function. The fitness function includes a temperature tracking term, a temperature difference suppression term, a cooling energy consumption term, a control smoothing term, a safety penalty term, a safety boundary crossing risk penalty term, and a lifetime-related thermal behavior penalty term. Step S4: Based on the fitness function constructed in step S3, online optimization is performed using a multi-rate execution architecture. When the low-frequency optimization cycle or risk triggering condition is met, the improved starfish optimization algorithm is used to update the control weights and prediction time-domain parameters. In the high-frequency control cycle, the updated control weights and prediction time-domain parameters are loaded through the model predictive controller, and the optimal control quantity is obtained by rolling solution in combination with the dynamic safety domain in step S2. Step S5: Map the optimal control quantity obtained in step S4 into control commands for the thermal management actuator, and complete closed-loop correction based on battery temperature feedback to achieve multi-objective coordinated control of hot spot temperature, internal temperature difference, cooling energy consumption and life-related thermal behavior.
[0006] The multi-objective thermal management control method for lithium-ion batteries provided by the present invention has the following beneficial effects: 1. This invention constructs a dynamic safety domain coupled with operating conditions and health. By introducing a safety boundary that dynamically changes with SOH, SOC, charge / discharge rate, ambient temperature and predicted heat load in thermal management control, it can dynamically tighten temperature and temperature difference constraints under complex operating conditions of aging batteries, thereby improving thermal safety throughout the entire life cycle.
[0007] 2. This invention adopts a multi-rate execution architecture and a risk-triggered online update mechanism to reduce unnecessary online global iterative calculations while ensuring the optimization effect, thereby improving the real-time adaptability of complex predictive control under limited computing power conditions.
[0008] 3. This invention introduces temperature difference suppression, cooling energy consumption, control smoothing, and lifespan-related thermal behavior constraints into the control optimization closed loop, expanding the control objective from single temperature tracking to the synergistic optimization of hot spot temperature, intra-pack temperature difference, and energy consumption. By suppressing temperature fluctuations and high-temperature residence, battery capacity decay can be effectively delayed. By penalizing drastic changes in control variables, actuator wear can be reduced and driving comfort improved. At the same time, the control focus can be automatically adjusted according to the degree of battery aging and the level of operating risk. Ultimately, under the premise of ensuring thermal safety, the invention achieves a balance between extended battery life, reduced cooling energy consumption, and smooth operation of actuators, significantly improving the overall benefits of the thermal management system throughout its entire life cycle.
[0009] 4. This invention employs a dual control mechanism that coordinates the coolant flow rate and coolant inlet temperature, and executes this control through an electronic water pump and a refrigeration unit compressor, achieving strong engineering compatibility and system adaptability. Attached Figure Description
[0010] Figure 1 This is a flowchart illustrating the multi-objective thermal management control method for lithium-ion batteries provided in an embodiment of the present invention. Figure 2 The graph shows a comparison of the battery pack hotspot temperature changes over time under three control methods. Figure 3 The graph shows a comparison of the battery pack temperature difference control effects under three control methods. Detailed Implementation
[0011] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain embodiments of the present invention, and should not be construed as limiting the present invention.
[0012] Please see Figure 1 The multi-objective thermal management control method for lithium-ion batteries provided by the present invention includes steps S1 to S5: Step S1: Obtain the real-time state information of the lithium-ion battery and establish an SOH adaptive dual-state lumped parameter thermal model. The model uses the battery pack hot spot temperature and edge temperature as state variables and the coolant flow rate and coolant inlet temperature as control variables to obtain the thermal model prediction results, which include the predicted hot spot temperature and edge temperature.
[0013] The continuous-time heat balance equation of the SOH adaptive two-state lumped-parameter thermal model is as follows:
[0014]
[0015] in, for The temperature of hot spots at all times for The rate of temperature change at hotspot nodes at any given time for Temperature of the edge node at any given time for The rate of temperature change at the edge nodes at any given time. and They are respectively Real-time hotspot and edge node heat capacity. This represents the equivalent thermal resistance between hotspot nodes and edge nodes. for The ambient temperature at any given time The equivalent thermal resistance between the hotspot node and the environment. The equivalent heat transfer coefficient between the hot spot node and the coolant. for The mass flow rate of the coolant at all times. for The inlet temperature of the coolant at all times. for The total heat generation power of the battery at any given time. This represents the equivalent thermal resistance between the edge node and the environment. This is the equivalent heat transfer coefficient between the edge node and the coolant. and They are respectively The heat generation distribution coefficients of hotspot nodes and edge nodes at any given time.
[0016] The nonlinear heat transfer function of the coolant flow rate is:
[0017]
[0018] in, The nonlinear heat transfer function of the coolant flow rate can be obtained through bench test calibration, function fitting, or table lookup. , These are the heat transfer correction coefficients for hotspot nodes and edge nodes, respectively.
[0019] It should be noted that, as a specific example, the SOH adaptive dual-state lumped parameter thermal model also includes a high-flow-rate enhanced heat transfer correction term and an equivalent temperature uniformity correction term between nodes. When the coolant flow rate increases or the operating condition risk level increases, the equivalent heat transfer coefficient between hot spots and edge nodes increases accordingly, and the equivalent thermal resistance between nodes decreases accordingly, in order to characterize the enhanced cooling and temperature uniformity capabilities.
[0020] After applying forward Euler discretization, the discretized state update equation of the thermal model is:
[0021]
[0022]
[0023]
[0024] in, for The state variable at time t, for Control variables at time, for The perturbation variable at any given time. for The state variable at time t, To control the sampling period, It is a nonlinear state change function.
[0025] The heat generation distribution coefficient of the SOH adaptive two-state lumped parameter thermal model is dynamically corrected based on the charge / discharge rate, SOC, and SOH, while the heat capacity parameter is dynamically updated based on SOH. The expressions for the heat generation distribution coefficient and the updated heat capacity are as follows:
[0026]
[0027]
[0028]
[0029] in, This serves as the baseline value for the heat generation distribution coefficient of hotspot nodes. , , For correction factor, This is the charge / discharge rate correction function. This is a state-of-charge correction function. For health status correction function, for Battery charge / discharge rate at any given time for The state of charge of the battery at all times. for Monitor the battery's health status at all times; and The reference thermal capacities for hotspot nodes and edge nodes are obtained through cell thermal characteristic tests. and These are the heat capacity correction functions for hotspot nodes and edge nodes, respectively.
[0030] Step S2: Based on the thermal model prediction results obtained in step S1, and combined with SOH, SOC, charge / discharge rate, ambient temperature and predicted heat load, construct a dynamic safety domain coupled with operating conditions and health. The dynamic safety domain includes at least a dynamic temperature safety upper limit, a dynamic temperature difference upper limit and a temperature rise rate constraint.
[0031] Among them, the dynamic safety upper limit of hot spot temperature and the upper limit of temperature difference are based on the tightened expression of SOH, and are corrected by combining SOC, charge and discharge rate, ambient temperature and predicted heat load.
[0032] Specifically, the expression for the dynamic temperature safety upper limit is:
[0033] in, for The safe upper limit of hotspot temperature at any time. This represents the upper limit of the safe temperature for hot spots under baseline operating conditions. , , , , This is a correction factor for the dynamic temperature safety upper limit. This is a reference value for the state of charge. This is a reference value for the magnification. This is a reference value for ambient temperature. for Predicted heat load at any given time This indicates the operation of taking the positive part; The expression for the upper limit of dynamic temperature difference is:
[0034] in, for The upper limit of dynamic temperature difference at any given time. As the upper limit of the baseline temperature difference, , , This is the correction factor for the upper limit of dynamic temperature difference.
[0035] The aforementioned dynamic temperature safety upper limit and dynamic temperature difference upper limit tighten as the battery SOH decays, high SOC, high rate, high ambient temperature and high heat load conditions, and are used to characterize the condition-health coupling safety domain of aged batteries under complex conditions.
[0036] The expression for the temperature rise rate constraint is:
[0037] in, for The hotspot temperature is predicted in real time. for The hotspot temperature is predicted in real time. for The upper limit of the rate of temperature rise at any given time.
[0038] The expression for the upper limit of the temperature rise rate is:
[0039] in, for The upper limit of the rate of temperature rise at any given time. As the upper limit of the reference temperature rise rate, , , For calibration correction factors.
[0040] The upper limit of the temperature rise rate is based on the tightened expression for SOH, and is modified in conjunction with SOC, charge / discharge rate, ambient temperature, and predicted heat load. This embodiment sets hard constraints on the control quantity to ensure that the actuator operates within a safe operating range; the expression is:
[0041]
[0042]
[0043]
[0044] in, and These are the lower and upper limits of coolant flow rate, respectively. and These are the lower and upper limits of the coolant inlet temperature, respectively. for The mass flow rate of the coolant at all times. This represents the maximum allowable change in coolant flow rate between adjacent control times. for The inlet temperature of the coolant at all times. This represents the maximum allowable change in coolant inlet temperature between adjacent control moments.
[0045] Step S3: Based on the dynamic safety domain obtained in step S2, a safety penalty term is constructed, and a safety boundary crossing risk penalty term is constructed based on the absolute safety boundary. Then, a multi-objective comprehensive fitness function is constructed. The fitness function includes a temperature tracking term, a temperature difference suppression term, a cooling energy consumption term, a control smoothing term, a safety penalty term, a safety boundary crossing risk penalty term, and a lifetime-related thermal behavior penalty term.
[0046] In step S3, the expression for the safety penalty item is:
[0047] in, For safety penalties, To predict the length of the time domain, This is the soft constraint penalty coefficient for dynamic temperature out-of-bounds conditions. This indicates taking the maximum value. for The corresponding dynamic temperature safety limit at any given time. This is the soft constraint penalty coefficient for dynamic temperature difference exceeding the limit. for The battery pack temperature difference is predicted in real time. for The upper limit of dynamic temperature difference at any given time. This is the soft constraint penalty coefficient for exceeding the temperature rise rate limit.
[0048] The expression for the penalty item for safety boundary violations is:
[0049] in, Penalties for safety boundary crossing risks, and These are the penalty weighting coefficients corresponding to the risks of hotspot temperature exceeding the limit and the risks of temperature difference exceeding the limit, respectively. This is the upper limit of the allowable temperature for hotspots. This represents the upper limit of the allowable temperature difference for the battery pack.
[0050] It is mainly used for dynamic soft constraint processing in the MPC rolling optimization process, specifically to penalize situations where the hot spot temperature, temperature difference within the packet, and temperature rise rate exceed the dynamic safety domain constraint in the prediction time domain. It is used to evaluate the safety risk of candidate weight vectors in the comprehensive fitness function, specifically to suppress temperature fluctuations and high-temperature residence in the prediction time domain.
[0051] In step S3, the expression for the multi-objective integrated fitness function is:
[0052]
[0053]
[0054] in, For multi-objective integrated fitness function, , , , These are the weighting coefficients for the temperature tracking term, temperature difference suppression term, cooling energy consumption term, and control smoothing term, respectively. The optimal operating reference temperature for the battery. express The observable temperature difference surrogate quantity at any given time, express Real-time power consumption of the cooling system express Increment of control quantity at any given time, For lifespan-related thermal behavior penalties, The penalty term is for deviation from the benchmark weight. To predict the amplitude of temperature difference fluctuations within the time domain, To predict the root mean square value of hotspot temperature in the time domain, To predict the cumulative time during which the temperature of hotspots exceeds a preset high-temperature threshold within the time domain, , , For calibration coefficients, This is the current candidate weight vector. This is the baseline weight vector corresponding to the current operating condition. To be the L2 norm, by introducing This can prevent the weight vector from deviating excessively from the current operating condition benchmark value, reduce the risk of control quantity oscillation, and improve the stability of online optimization.
[0055] In this embodiment, the temperature difference within the package is represented by an observable temperature difference surrogate quantity. This observable temperature difference surrogate quantity is obtained by smoothing the difference between the hot spot temperature and the edge temperature. It is used to avoid the appearance of an unrealistic zero temperature difference phenomenon when the equivalent node temperatures of two states cross. Specifically, the observable temperature difference surrogate quantity is calculated by the following formula:
[0056] in, express The observable temperature difference surrogate quantity at any given time, This is a temperature difference smoothing correction coefficient, used to avoid an unrealistic zero temperature difference phenomenon when the temperatures of hot nodes and edge nodes are close to or cross each other.
[0057] Step S4: Based on the fitness function constructed in step S3, online optimization is performed using a multi-rate execution architecture. When the low-frequency optimization cycle or risk triggering condition is met, the improved starfish optimization algorithm is used to update the control weights and prediction time-domain parameters. In the high-frequency control cycle, the updated control weights and prediction time-domain parameters are loaded through the model predictive controller, and the optimal control quantity is obtained by rolling solution in combination with the dynamic safety domain in step S2.
[0058] Among them, the improved starfish optimization algorithm is based on the multi-rate execution architecture and the vehicle scenario. The multi-rate execution architecture is set as follows: the high-frequency control cycle is 1 control cycle, the low-frequency optimization cycle is greater than 1 control cycle, and the weight iterative optimization operation is performed by combining fixed backup update and risk-triggered online update.
[0059] In this embodiment, the low-frequency optimization cycle is set to 12 control cycles, and the high-frequency control cycle is set to 1 control cycle; in one embodiment, the prediction time-domain parameters include the prediction time-domain length. and control time domain length When the risk level of the working condition When the temperature rises, the control time domain is shortened and the safety constraint weights are increased to improve the thermal safety response speed; when When reducing the load, the predicted time-domain parameters are maintained or appropriately extended to balance cooling energy consumption optimization and control stability. Weighted iterative optimization is performed when a fixed backup update cycle is reached or risk triggering conditions are met. In other high-frequency control cycles, the underlying MPC controller directly loads the optimal weight vector or predicted time-domain parameters output from the previous optimization cycle to complete the rolling solution, reducing the amount of online optimization computation without sacrificing control performance and adapting to the computing power limitations of the vehicle controller.
[0060] In this embodiment, the iterative optimization steps of the improved starfish optimization algorithm are as follows: Step S401, Initialization of basic algorithm parameters: Determine the weight vector to be optimized as follows Set the core parameters of the algorithm: upper and lower bounds of the weight vector. , Population size Maximum number of iterations Adaptive compression coefficient upper and lower bounds , random disturbance trigger probability Early termination threshold for iteration .
[0061] Step S402, Neighborhood population initialization for dual-sensor operation condition and health: Calculate the risk level of the current operation condition based on the current average heat load, temperature tracking deviation, battery pack temperature difference, and battery SOH. The value range is 0~1, and the expression is:
[0062] in, For the first Risk level of operating conditions in a low-frequency optimization cycle This is a limiting function that restricts the input value to the range of 0 to 1; This is a reference value for heat generation power. This is a reference value for temperature deviation. This is a reference value for temperature difference. , , , Calibrate the weighting coefficients. For the first The trigger control sampling time corresponding to each low-frequency optimization cycle. From The average heat generation power in the prediction time domain starting from time point [time]. for The hot spot temperature of the battery pack at any time. for The observable temperature difference surrogate quantity at any given time, for Real-time battery health status.
[0063] Based on the operating condition risk level and battery SOH, a benchmark weight vector corresponding to the operating condition is generated. Population initialization is completed within ±15% neighborhood of the benchmark weight vector, replacing the global random initialization of the traditional algorithm, thus narrowing the optimization range and improving the convergence speed.
[0064] Specifically, the initial position of the i-th individual in the population is
[0065] in, For the first Initial candidate weights, This is the current operating condition baseline weight vector. This indicates element-wise multiplication. The neighborhood coefficient is 0.15 in this embodiment. For the first The random perturbation vector corresponding to each candidate individual.
[0066] Step S403, Adaptive Compression Coefficient Iterative Update: The iterative process is divided into a global exploration phase and a local development phase. The adaptive compression coefficient decreases linearly with the number of iterations, balancing the algorithm's global exploration capability and local development capability. The expression is:
[0067] in, For the first The adaptive compression coefficient of the next iteration. and These are the upper and lower bounds of the compressibility coefficient, respectively. This represents the maximum number of iterations.
[0068] Early iteration A larger value ensures global exploration capability and avoids getting trapped in local optima; in the later stages of iteration... The value is relatively small, ensuring the accuracy of local development and improving the accuracy of optimization.
[0069] Step S404, Candidate Weight Vector Position Update: Based on the adaptive compression coefficient, update the position of each individual in the population to complete one iteration of optimization. The candidate weight position update formula is:
[0070]
[0071] in, For the first The adaptive compression coefficient of the next iteration. and These are the upper and lower bounds of the compressibility coefficient, respectively. This represents the maximum number of iterations. Indicates the first During the nth iteration, the 1st The positions of the candidate weight vectors; Indicates the first During the nth iteration, the 1st The positions of the candidate weight vectors; For the first The globally optimal individual in the next iteration. For random disturbance coefficients, For the first During the nth iteration candidate weight vectors, For the first During the nth iteration candidate weight vectors, For the first The globally optimal weight vector in the next iteration. , These are the algorithm coefficients.
[0072] After the update, the candidate weight vector is subjected to boundary processing to keep it within the preset upper and lower bounds of the weight.
[0073] Step S405, Adaptive Random Perturbation Mechanism: When the perturbation trigger condition of no decrease in global optimal fitness for two consecutive generations or a sudden change in risk level is met, a random perturbation update is performed with a 15% probability. This avoids the algorithm getting trapped in local optima under sudden changes in operating conditions and improves the stability of optimization. The adaptive random perturbation update expression is:
[0074] in, It follows a standard normal distribution. This represents the optimal individual guidance coefficient.
[0075] Step S406, Fitness Value Calculation: Substitute each candidate weight vector in the population into the multi-objective integrated fitness function constructed in step S3 to calculate the fitness value of each individual. The smaller the fitness value, the better the integrated control effect corresponding to the weight vector.
[0076] Step S407, Global Optimal Individual Update: Compare the current iteration's optimal fitness with the historical global optimal fitness. If the current iteration's optimal fitness is smaller, then update the global optimal individual and the optimal fitness.
[0077] Step S408, Early termination judgment: Detect the rate of change of the global optimal fitness for three consecutive generations. If the rate of change is less than the set early termination threshold, terminate the iteration, output the optimal weight vector, and reduce the consumption of ineffective computing power. If the termination condition is not met and the maximum number of iterations has not been reached, return to step S403 to continue iterating. When the hot spot temperature is close to the dynamic temperature safety limit, the temperature difference within the package is close to the dynamic temperature difference limit, or the actuator is close to saturation, the online update of the next low-frequency optimization cycle is allowed to be triggered early.
[0078] Step S409, Adaptive Weight Smoothing Processing: The optimal weight vector output by the algorithm is subjected to first-order low-pass smoothing to avoid control oscillations and actuator wear caused by sudden weight changes. The adaptive smoothing coefficient and weight smoothing processing expression are as follows:
[0079]
[0080] in, For the first The adaptive smoothing coefficient for each low-frequency optimization cycle. and These are the upper and lower bounds of the smoothing coefficient, respectively. For the first The optimal weight vector after smoothing a low-frequency optimization cycle. For the first The optimal weight vector after smoothing a low-frequency optimization cycle; For the first The original optimal weight vector is directly output by the improved starfish optimization algorithm in each low-frequency optimization cycle.
[0081] When in a transient high-risk operating condition Increase The weighting decreases, reducing the proportion of the weight vector from the previous optimization cycle, allowing the current weights to respond quickly to changes in operating conditions; when in a steady-state, low-risk operating condition, Decrease Increasing the weight vector increases the proportion of the previous optimization cycle, thereby ensuring a smooth control process and reducing wear on the actuator.
[0082] In this embodiment, the risk triggering conditions include at least one of the following: risk level mutation, dynamic security domain near activation, and actuator near saturation.
[0083] Step S5: Map the optimal control quantity obtained in step S4 into control commands for the thermal management actuator, and complete closed-loop correction based on battery temperature feedback to achieve multi-objective coordinated control of hot spot temperature, internal temperature difference, cooling energy consumption and life-related thermal behavior.
[0084] The optimal control variables include the optimal coolant flow rate and the optimal coolant inlet temperature. Through the mapping relationship calibrated on the bench, the optimal coolant flow rate is mapped to the electric water pump speed command, and the optimal coolant inlet temperature is mapped to the refrigeration unit compressor control command. The electric water pump speed command and the refrigeration unit compressor control command satisfy the following mapping relationships:
[0085]
[0086] in, This is the speed command for the electronic water pump; For the compressor control commands of the refrigeration unit; To achieve the optimal coolant flow rate; To achieve the optimal coolant inlet temperature; The ambient temperature; , , , , , This is the test bench calibration coefficient.
[0087] Specifically, the control commands are encapsulated according to the vehicle CAN bus communication protocol and sent to the vehicle controller. At the same time, based on the battery temperature data collected in real time by the BMS, the deviation between the temperature predicted by the thermal model and the actual temperature is calculated, and the thermal model parameters and MPC prediction output are corrected online to complete the closed-loop control of thermal management.
[0088] The method provided by this invention will now be tested.
[0089] This embodiment sets up three comparison schemes: traditional incremental PID control (PID for short), traditional fixed-weight MPC control (fixed-weight MPC for short), and the method proposed in this invention. Under the thermal load condition set in this embodiment, the simulation results are as follows: Figure 2 and Figure 3 As shown.
[0090] Depend on Figure 2As can be seen, in the initial stage of simulation, the hot spot temperature of the battery pack is about 36℃. All three control methods can gradually reduce the hot spot temperature to approach the ideal cooling target temperature of 25℃. In the medium-high heat load stage, the hot spot temperature under PID control and fixed weight MPC control shows a significant rebound, with its peak value reaching more than 28℃. However, the method of the present invention can suppress the peak hot spot temperature to about 27℃, with a smaller increase in hot spot temperature, and the entire process is below the dynamic temperature safety limit. This shows that the method of the present invention can adjust the cooling intensity in advance under heat load change conditions, and improve the safety margin and stability of hot spot temperature control.
[0091] Depend on Figure 3 As can be seen, during the initial cooling phase, the internal temperature difference of all three control methods showed a short-term increase, with a peak value of approximately 2.3℃ to 2.4℃, but none exceeded the dynamic temperature difference upper limit. During the later stages of increased heat load, the internal temperature difference under PID control and fixed-weight MPC control increased again, approaching the dynamic temperature difference upper limit. However, the method of this invention, through dynamic safety domain constraints and a risk-triggered online update mechanism, significantly reduced the internal temperature difference compared to the comparative methods, keeping the peak temperature difference within approximately 1℃ in the later stages, with smaller temperature fluctuations. Therefore, the method of this invention not only suppresses the rise in hot spot temperature but also ensures temperature consistency within the battery pack, achieving coordinated optimization control of hot spot temperature and internal temperature difference while meeting both the dynamic temperature safety upper limit and the dynamic temperature difference upper limit.
[0092] In summary, the multi-objective thermal management control method for lithium-ion batteries according to the above embodiments has the following beneficial effects: 1. This invention constructs a dynamic safety domain coupled with operating conditions and health. By introducing a safety boundary that dynamically changes with SOH, SOC, charge / discharge rate, ambient temperature and predicted heat load in thermal management control, it can dynamically tighten temperature and temperature difference constraints under complex operating conditions of aging batteries, thereby improving thermal safety throughout the entire life cycle.
[0093] 2. This invention adopts a multi-rate execution architecture and a risk-triggered online update mechanism to reduce unnecessary online global iterative calculations while ensuring the optimization effect, thereby improving the real-time adaptability of complex predictive control under limited computing power conditions.
[0094] 3. This invention introduces temperature difference suppression, cooling energy consumption, control smoothing, and lifespan-related thermal behavior constraints into the control optimization closed loop, expanding the control objective from single temperature tracking to the synergistic optimization of hot spot temperature, intra-pack temperature difference, and energy consumption. By suppressing temperature fluctuations and high-temperature residence, battery capacity decay can be effectively delayed. By penalizing drastic changes in control variables, actuator wear can be reduced and driving comfort improved. At the same time, the control focus can be automatically adjusted according to the degree of battery aging and the level of operating risk. Ultimately, under the premise of ensuring thermal safety, the invention achieves a balance between extended battery life, reduced cooling energy consumption, and smooth operation of actuators, significantly improving the overall benefits of the thermal management system throughout its entire life cycle.
[0095] 4. This invention employs a dual control mechanism that coordinates the coolant flow rate and coolant inlet temperature, and executes this control through an electronic water pump and a refrigeration unit compressor, achieving strong engineering compatibility and system adaptability.
[0096] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.
Claims
1. A multi-objective thermal management control method for lithium-ion batteries, characterized in that, include: Step S1: Obtain the real-time state information of the lithium-ion battery and establish an SOH adaptive dual-state lumped parameter thermal model. The model uses the hot spot temperature and edge temperature of the battery pack as state variables and the coolant flow rate and coolant inlet temperature as control variables to obtain the thermal model prediction results. The thermal model prediction results include the predicted hot spot temperature and edge temperature. Step S2: Based on the thermal model prediction results obtained in step S1, and combined with SOH, SOC, charge / discharge rate, ambient temperature and predicted heat load, construct a dynamic safety domain coupled with operating conditions and health. The dynamic safety domain includes at least a dynamic temperature safety upper limit, a dynamic temperature difference upper limit and a temperature rise rate constraint. Step S3: Construct a safety penalty term based on the dynamic safety domain obtained in step S2, and simultaneously construct a safety boundary crossing risk penalty term based on the absolute safety boundary, thereby constructing a multi-objective comprehensive fitness function. The fitness function includes a temperature tracking term, a temperature difference suppression term, a cooling energy consumption term, a control smoothing term, a safety penalty term, a safety boundary crossing risk penalty term, and a lifetime-related thermal behavior penalty term. Step S4: Based on the fitness function constructed in step S3, online optimization is performed using a multi-rate execution architecture. When the low-frequency optimization cycle or risk triggering condition is met, the improved starfish optimization algorithm is used to update the control weights and prediction time-domain parameters. In the high-frequency control cycle, the updated control weights and prediction time-domain parameters are loaded through the model predictive controller, and the optimal control quantity is obtained by rolling solution in combination with the dynamic safety domain in step S2. Step S5: Map the optimal control quantity obtained in step S4 into control commands for the thermal management actuator, and complete closed-loop correction based on battery temperature feedback to achieve multi-objective coordinated control of hot spot temperature, internal temperature difference, cooling energy consumption and life-related thermal behavior.
2. The multi-objective thermal management control method for lithium-ion batteries according to claim 1, characterized in that, In step S1, the heat generation distribution coefficient of the SOH adaptive dual-state lumped parameter thermal model is dynamically corrected based on the charge / discharge rate, SOC, and SOH, and the heat capacity parameter is dynamically updated based on SOH. The expressions for the heat generation distribution coefficient and the updated heat capacity are as follows: in, and They are respectively The heat generation distribution coefficients of hotspot nodes and edge nodes at any given time. This serves as the baseline value for the heat generation distribution coefficient of hotspot nodes. , , For correction factor, This is the charge / discharge rate correction function. This is a state-of-charge correction function. For health status correction function, for Battery charge / discharge rate at any given time for The state of charge of the battery at all times. for Monitor the battery's health status at all times. and They are respectively Real-time hotspot and edge node heat capacity. and These are the baseline heat capacities for hotspot nodes and edge nodes, respectively. and These are the heat capacity correction functions for hotspot nodes and edge nodes, respectively; The continuous-time heat balance equation is: in, for The temperature of hot spots at all times for The rate of temperature change at hotspot nodes at any given time for Temperature of the edge node at any given time for The rate of temperature change at the edge nodes at any given time. This represents the equivalent thermal resistance between hotspot nodes and edge nodes. for The ambient temperature at any given time The equivalent thermal resistance between the hotspot node and the environment. The equivalent heat transfer coefficient between the hot spot node and the coolant. for The mass flow rate of the coolant at all times. for The inlet temperature of the coolant at all times. for The total heat generation power of the battery at any given time. This represents the equivalent thermal resistance between the edge node and the environment. This is the equivalent heat transfer coefficient between the edge node and the coolant; The discretized state update equation is: in, for The state variable at time t, for Control variables at time, for The perturbation variable at any given time. for The state variable at time t, To control the sampling period, It is a nonlinear state change function.
3. The multi-objective thermal management control method for lithium-ion batteries according to claim 2, characterized in that, In step S2, the expression for the dynamic temperature safety upper limit is: in, for The safe upper limit of hotspot temperature at any time. This represents the upper limit of the safe temperature for hot spots under baseline operating conditions. , , , , This is a correction factor for the dynamic temperature safety upper limit. This is a reference value for the state of charge. This is a reference value for the magnification. This is a reference value for ambient temperature. for Predicted heat load at any given time This indicates the operation of taking the positive part; The expression for the upper limit of dynamic temperature difference is: in, for The upper limit of dynamic temperature difference at any given time. As the upper limit of the baseline temperature difference, , , This is the correction factor for the upper limit of dynamic temperature difference; The expression for the temperature rise rate constraint is: in, for The hotspot temperature is predicted in real time. for The hotspot temperature is predicted in real time. for The upper limit of the rate of temperature rise at any given time.
4. The multi-objective thermal management control method for lithium-ion batteries according to claim 3, characterized in that, In step S3, the expression for the safety penalty item is: in, For safety penalties, To predict the length of the time domain, This is the soft constraint penalty coefficient for dynamic temperature out-of-bounds conditions. This indicates taking the maximum value. for The corresponding dynamic temperature safety limit at any given time. This is the soft constraint penalty coefficient for dynamic temperature difference exceeding the limit. for The battery pack temperature difference is predicted in real time. for The upper limit of dynamic temperature difference at any given time. This is the soft constraint penalty coefficient for exceeding the temperature rise rate limit; The expression for the penalty item for safety boundary violations is: in, Penalties for safety boundary crossing risks, and These are the penalty weighting coefficients corresponding to the risks of hotspot temperature exceeding the limit and the risks of temperature difference exceeding the limit, respectively. This is the upper limit of the allowable temperature for hotspots. This represents the upper limit of the allowable temperature difference for the battery pack.
5. The multi-objective thermal management control method for lithium-ion batteries according to claim 4, characterized in that, In step S3, the expression for the multi-objective integrated fitness function is: in, For multi-objective integrated fitness function, , , , These are the weighting coefficients for the temperature tracking term, temperature difference suppression term, cooling energy consumption term, and control smoothing term, respectively. The optimal operating reference temperature for the battery. express The observable temperature difference surrogate quantity at any given time, express Real-time power consumption of the cooling system express Increment of control quantity at any given time, For lifespan-related thermal behavior penalties, The penalty term is for deviation from the benchmark weight. To predict the amplitude of temperature difference fluctuations within the time domain, To predict the root mean square value of hotspot temperature in the time domain, To predict the cumulative time during which the temperature of hotspots exceeds a preset high-temperature threshold within the time domain, , , These are calibration coefficients.
6. The multi-objective thermal management control method for lithium-ion batteries according to claim 5, characterized in that, In step S3, the observable temperature difference surcharge is calculated using the following formula: in, express The observable temperature difference surrogate quantity at any given time, This is the temperature difference smoothing correction coefficient.
7. The multi-objective thermal management control method for lithium-ion batteries according to claim 6, characterized in that, In step S4, the improved starfish optimization algorithm is obtained based on the multi-rate execution architecture and the vehicle scenario. The multi-rate execution architecture is set as follows: the high-frequency control cycle is 1 control cycle, the low-frequency optimization cycle is greater than 1 control cycle, and the weight iterative optimization operation is performed by combining fixed backup update and risk-triggered online update. The adaptive compression coefficient and candidate weight position update expressions of the improved starfish optimization algorithm are as follows: in, For the first Initial candidate weights, This is the current operating condition baseline weight vector. This indicates element-wise multiplication. Neighborhood coefficient, For the first The random perturbation vector corresponding to each candidate individual. For the first The adaptive compression coefficient of the next iteration. and These are the upper and lower bounds of the compressibility coefficient, respectively. This represents the maximum number of iterations. Indicates the first During the nth iteration, the 1st The positions of the candidate weight vectors; Indicates the first During the nth iteration, the 1st The positions of the candidate weight vectors; For the first The globally optimal individual in the next iteration. For random disturbance coefficients, For the first During the nth iteration candidate weight vectors, For the first During the nth iteration candidate weight vectors, For the first The globally optimal weight vector in the next iteration. , These are the algorithm coefficients; The adaptive random perturbation update expression is: in, It follows a standard normal distribution. The optimal individual guidance coefficient; The expressions for the adaptive smoothing coefficient and weighted smoothing processing are as follows: in, For the first The adaptive smoothing coefficient for each low-frequency optimization cycle. and These are the upper and lower bounds of the smoothing coefficient, respectively. For the first Risk level of operating conditions in a low-frequency optimization cycle For the first The optimal weight vector after smoothing a low-frequency optimization cycle. For the first The optimal weight vector after smoothing a low-frequency optimization cycle; For the first The original optimal weight vector is directly output by the improved starfish optimization algorithm in each low-frequency optimization cycle.
8. The multi-objective thermal management control method for lithium-ion batteries according to claim 7, characterized in that, In step S5, the optimal control variables include the optimal coolant flow rate and the optimal coolant inlet temperature. Through the mapping relationship calibrated on the bench, the optimal coolant flow rate is mapped to the electric water pump speed command, and the optimal coolant inlet temperature is mapped to the refrigeration unit compressor control command. The electric water pump speed command and the refrigeration unit compressor control command satisfy the following mapping relationships: in, This is the speed command for the electronic water pump; For the compressor control commands of the refrigeration unit; To achieve the optimal coolant flow rate; To achieve the optimal coolant inlet temperature; The ambient temperature; , , , , , This is the test bench calibration coefficient.