A method and system for dynamic optimization of thermal management parameters in liquid-cooled lithium iron phosphate batteries

Abstract

This invention discloses a method and system for dynamically optimizing thermal management parameters of liquid-cooled lithium iron phosphate batteries, relating to the field of dynamic optimization of battery thermal management parameters. The method includes: constructing an offline training database for the thermal state of the carrier based on carrier state, thermal state, cooling system, and environmental parameters, and performing data cleaning and alignment; constructing a general thermal state prediction model for the carrier based on the LightGBM model; verifying the reliability of the predicted carrier thermal state by combining actual monitoring results; dynamically optimizing the general thermal state prediction model for the carrier based on an online gradient descent algorithm; dynamically optimizing the coolant flow rate based on the thermal balance equation; and setting up a sensor group for real-time acquisition of current state parameters and actual thermal state data. The advantages of this invention are: by integrating data-driven and physical models, and utilizing online learning to dynamically optimize thermal management parameters, it significantly improves the thermal management accuracy, adaptability, and safety of liquid-cooled lithium iron phosphate batteries.

Description

Technical Field

[0001] This invention relates to the field of dynamic optimization of battery thermal management parameters, specifically to a method and system for dynamic optimization of thermal management parameters for liquid-cooled lithium iron phosphate batteries. Background Technology

[0002] With the rapid development of the new energy vehicle industry, the thermal safety of power batteries has become increasingly prominent. Lithium iron phosphate batteries have become one of the mainstream choices due to their high safety and long cycle life. However, their thermal management still faces severe challenges. Batteries generate a lot of heat during charging and discharging. If the heat is not dissipated in time, it will lead to an increase in battery pack temperature, a decrease in battery life, and even safety accidents such as thermal runaway. Liquid cooling systems have become the mainstream thermal management solution for large-capacity battery packs due to their efficient and uniform cooling performance. The core of the system is to dynamically adjust parameters such as coolant flow rate according to the thermal state of the battery in order to achieve precise temperature control.

[0003] Traditional thermal management strategies are typically based on preset thresholds or simple PID controllers, relying on extensive bench testing and engineer experience. This is a static and passive control approach. However, the actual operating conditions of vehicles are complex and varied, including different ambient temperatures, driving habits, and road conditions. These factors dynamically affect the balance between battery heat generation and dissipation. Static strategies cannot adapt to these dynamic changes, often exhibiting control lag or over-adjustment, leading to low energy efficiency and insufficient temperature control accuracy. Therefore, developing an intelligent thermal management method that can sense the status in real time, predict trends, and dynamically optimize parameters is of paramount importance for improving battery performance, ensuring safety, and extending battery life. It is also an urgent need for current industry technological development.

[0004] Existing prediction methods based on large-scale machine learning models such as neural networks tend to rely heavily on the scale and comprehensiveness of training data. For vehicles with low usage frequency or unique behavioral patterns, the difficulty in obtaining sufficient sample data for training leads to a significant decrease in model prediction accuracy, insufficient generalization ability, and an inability to effectively capture diverse driving habits. Furthermore, they are prone to the "black box" effect and lack physical constraints, resulting in a lack of interpretability in the prediction process. More importantly, the prediction results may violate basic physical laws (such as the law of conservation of energy and the laws of thermodynamics), leading to outputs that contradict reality. This makes it difficult for existing technologies to effectively integrate data-driven learning capabilities with the interpretability and safety of physical models, ultimately resulting in a trade-off between prediction accuracy and system reliability. Summary of the Invention

[0005] To address the aforementioned technical problems, this paper provides a method and system for dynamic optimization of thermal control parameters for liquid-cooled lithium iron phosphate batteries. This technical solution solves the problem mentioned in the background that the model is prone to relying on the scale and comprehensiveness of training data. For carriers with low usage frequency or special behavior patterns, it is difficult to obtain sufficient sample data for training, resulting in a significant decrease in model prediction accuracy, insufficient generalization ability, and inability to effectively capture diverse driving habits. Secondly, it is prone to the "black box" effect and lacks physical constraints, and its prediction process lacks interpretability. More importantly, the prediction results may violate basic physical laws (such as energy conservation and thermodynamic laws), leading to outputs that contradict the actual situation. This makes it difficult for existing technologies to effectively integrate data-driven learning capabilities with the interpretability and safety of physical models, ultimately making it difficult to balance prediction accuracy and system reliability.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A method for dynamically optimizing thermal management parameters for liquid-cooled lithium iron phosphate batteries includes:

[0008] Based on carrier state, thermal state, cooling system and environmental parameters, an offline training database for carrier thermal state was constructed, and the data was cleaned and aligned.

[0009] Based on the offline training database of carrier thermal state, a general thermal state prediction model for carriers is constructed using the LightGBM model to predict the thermal state of carriers.

[0010] Based on the general thermal state prediction model for carriers, and combined with actual monitoring results, the reliability of the predicted thermal state of carriers was verified.

[0011] Based on the online gradient descent algorithm and combined with the vehicle's driving data, the general thermal state prediction model of the vehicle is dynamically optimized so that it can automatically learn driving habits.

[0012] Based on the output of the general thermal state prediction model of the carrier, the coolant flow rate is dynamically optimized according to the thermal balance equation.

[0013] A sensor array is set up to collect in real time the current state parameters required for model prediction and the actual thermal state data at future moments to verify the accuracy of the prediction.

[0014] Preferably, the step of constructing a general thermal state prediction model for the carrier based on the LightGBM model and an offline training database of the carrier's thermal state specifically includes:

[0015] The carrier state parameters, cooling system parameters, and environmental parameters are used as inputs, and the carrier thermal state parameters over a future period of time are used as outputs.

[0016] Among them, environmental parameters are used to characterize the dynamic constraints on the coolant inlet temperature and the system heat dissipation boundary, thereby improving the prediction accuracy of the model under different operating conditions.

[0017] Based on the carrier thermal state offline training database, it is divided into training set, validation set and test set in a ratio of 7:2:1;

[0018] Based on the LightGBM framework, a general thermal state prediction model for the carrier is constructed using its built-in gradient boosting decision tree algorithm. This model consists of multiple regression trees added together.

[0019] Each regression tree splits nodes with the goal of minimizing the weighted squared error. During splitting, the optimal feature and its splitting threshold are selected by maximizing the gain.

[0020] The initial learning rate is set to A. During training, the loss on the validation set is continuously monitored. If the loss does not decrease for five consecutive iterations, the current learning rate is halved.

[0021] With the goal of minimizing the average temperature deviation of the battery, a custom loss function with safety penalty is constructed based on the mean square error formula.

[0022] The gradient boosting algorithm is used to optimize the model parameters. In each iteration, the first and second gradients of the current model with respect to the custom loss function are calculated, and a new regression tree is added to fit the negative gradient direction to reduce the loss.

[0023] When the absolute value of the change in the value of the validation set loss function over K consecutive iterations is less than the preset threshold a, the model is determined to have converged, and the training process is terminated early to avoid overfitting.

[0024] The gain expression is:

[0025]

[0026] In the formula, Gain is the gain value used to evaluate the quality of the split point, G L G R These are the sum of the first-order gradients of all samples in the left and right child nodes after the split, and H, respectively. L H R denoted as the sum of the second-order gradients of all samples in the left and right child nodes after the split, respectively; λ is the L2 regularization coefficient; and γ is the minimum gain threshold required for the split. Splitting will only occur when Gain > γ;

[0027] The expression for the custom loss function with safety penalty is:

[0028]

[0029] In the formula, L is a custom loss value with a safety penalty, M is the number of samples participating in training, and Y... true.i Y represents the true average battery temperature of the i-th sample. pred.i The predicted average battery temperature for the i-th sample, Y safe α is the preset safe temperature threshold, and α is the safety penalty factor.

[0030] Preferably, the step of verifying the reliability of the predicted thermal state of the carrier based on the general thermal state prediction model and actual monitoring results specifically includes:

[0031] Based on the trained general thermal state prediction model of the carrier, and based on the actual monitored carrier state parameters, cooling system parameters and environmental parameters, the thermal state parameter data of the carrier for a future period of time is obtained.

[0032] Acquire the predicted thermal state parameters of the model under different carrier states and the corresponding actual monitoring data, and perform data preprocessing;

[0033] Based on the predicted data of thermal state parameters and the corresponding actual monitoring data, the rolling average error of the predicted values ​​of each thermal state parameter is calculated.

[0034] The rolling average error of the predicted value is calculated for each thermal state parameter, and the consistency of each thermal state parameter is judged to screen the credibility of the predicted data and the corresponding actual monitoring data. If the rolling average error of a thermal state parameter exceeds the corresponding preset threshold for three consecutive times, an alarm is triggered, and the parameter data will depend on the monitoring data rather than the prediction. Otherwise, the predicted parameter is considered reliable.

[0035] The expression for the rolling average error value of the two data points is:

[0036] MAE_r j (k)=δ·MAE_r j (k-1)+(1-δ)·|Y Measurement,j (k)-Y pred.j (k|k-Δt)|

[0037] In the formula, MAE_r j (k) represents the rolling average error of the correction for the j-th thermal state parameter at the k-th calculation time, where k is the index of the error calculation time, δ is the smoothing factor, and MAE_r j (k-1) represents the rolling average error of the correction for the j-th thermal state parameter at the (k-1)-th calculation time, Y Measurement，j (k) represents the actual monitoring data of the j-th thermal state parameter at the k-th calculation time, Y pred，j(k|k-Δt) represents the predicted data of the j-th thermal state parameter at the k-th calculation time, and Δt represents the prediction time span of the model.

[0038] Preferably, the step of dynamically optimizing the general thermal state prediction model of the vehicle based on the online gradient descent algorithm and combined with the vehicle's driving data, so that it can automatically learn driving habits, specifically includes:

[0039] Set up a fixed-size first-in-first-out (FIFO) buffer to store the N most recent driving habit data pairs;

[0040] An incremental learning mechanism is set up to automatically trigger incremental learning of the basic model for predicting the general thermal state of the vehicle based on the rolling average error of the data and the number of driving habit data pairs in the FIFO buffer.

[0041] Based on the incremental learning mechanism, the general thermal state prediction model of the carrier is dynamically optimized, enabling it to automatically learn driving habits.

[0042] Preferably, the step of dynamically optimizing the coolant flow rate based on the output of the carrier's general thermal state prediction model and the thermal balance equation specifically includes:

[0043] The output of the general thermal state prediction model of the carrier is used as the input value, and the optimized value of the coolant flow rate is used as the output value.

[0044] Based on the thermal balance formula, a physical mapping model between the thermal state parameters of the carrier and the flow rate of the coolant is constructed, so that the parameter changes strictly follow physical rules.

[0045] The output of the general thermal state prediction model of the carrier is input into the physical mapping model of the thermal state parameters of the carrier and the flow rate of the coolant to calculate the predicted value of the thermal power accumulated by the battery pack in the future.

[0046] Based on the top-level temperature control target, dynamically generate the target value of the heat power accumulated by the battery pack itself;

[0047] To minimize the error between the accumulated heat power of the battery pack and the target value, and to ensure a smooth change in the coolant flow rate, an optimal flow rate control objective function under thermal balance constraints is constructed.

[0048] In each control cycle, the objective function is minimized based on the gradient descent algorithm to obtain the optimal coolant flow rate sequence for a future period.

[0049] The first control variable in the optimal coolant flow rate sequence for a future period is output as the actual control command to the coolant pump actuator, and the above rolling optimization steps are repeated until the next cycle.

[0050] The physical mapping model expression between the thermal state parameters of the carrier and the flow rate of the coolant is as follows:

[0051]

[0052] In the formula, P 电池 For battery heat generation power, P 冷却液 P is the heat power carried away by the coolant circulation. 电池包 For the heat power accumulated by the battery pack itself, c p ρ is the specific heat capacity at constant pressure of the coolant, v is the density of the coolant, and T is the flow rate of the coolant. out T is the outlet temperature of the battery pack coolant. in Here, m is the inlet temperature of the battery pack coolant, and c is the total mass of the battery pack. 电池 For the specific heat capacity of the battery, dT 电池 / dt is the rate of change of the average temperature of the battery;

[0053] The objective function for optimal flow velocity control under thermal balance constraints is expressed as follows:

[0054]

[0055] In the formula, minJ is the objective function value of the flow velocity optimal control under thermal balance constraints, and P 电池包，pred (k) represents the predicted heat power accumulated by the battery pack itself at the k-th calculation time, P 电池包，set (k) represents the accumulated thermal power of the target battery pack at the k-th calculation time, θ is the equilibrium constant, dimensionless, and is adjusted through experiments or simulations to balance temperature accuracy and flow rate stability, v(k) is the predicted value of coolant flow rate at the k-th calculation time, and v(k-1) is the predicted value of coolant flow rate at the (k-1)-th calculation time.

[0056] Furthermore, this solution proposes a dynamic optimization system for thermal management parameters of liquid-cooled lithium iron phosphate batteries, used to implement the aforementioned dynamic optimization method for thermal management parameters of liquid-cooled lithium iron phosphate batteries, including:

[0057] The data acquisition and database module is used to construct an offline training database for the thermal state of the carrier based on the carrier state, thermal state, cooling system and environmental parameters, and to perform data cleaning and alignment; the data acquisition module is used to set up a sensor group to collect in real time the current state parameters required for model prediction and the actual thermal state data at future moments for verifying the accuracy of the prediction.

[0058] The thermal management optimization module is used to train an offline database of the vehicle's thermal state. Based on the LightGBM model, it constructs a general thermal state prediction model for the vehicle to predict its thermal state. Based on this model and actual monitoring results, it verifies the reliability of the predicted thermal state. Using an online gradient descent algorithm and combined with the vehicle's driving data, it dynamically optimizes the general thermal state prediction model, enabling it to automatically learn driving habits. Based on the output of the model, and using the thermal balance equation, it dynamically optimizes the coolant flow rate.

[0059] Preferably, the data acquisition and database module includes:

[0060] The database unit is used to construct an offline training database for the thermal state of the carrier based on the carrier state, thermal state, cooling system and environmental parameters, and to perform data cleaning and alignment.

[0061] The data acquisition unit is used to set up a sensor group to collect in real time the current state parameters required for model prediction and the actual thermal state data at future moments for verifying the accuracy of the prediction.

[0062] Preferably, the thermal management optimization module includes:

[0063] The model prediction unit is used to train the carrier thermal state offline database. Based on the LightGBM model, it constructs a general thermal state prediction basic model for the carrier to predict the thermal state of the carrier.

[0064] The model verification unit is used to verify whether the predicted thermal state of the carrier is reliable based on the general thermal state prediction model of the carrier and the actual monitoring results.

[0065] The incremental learning unit is used to dynamically optimize the general thermal state prediction basic model of the vehicle based on the online gradient descent algorithm and combined with the vehicle's driving data, so that it can automatically learn driving habits.

[0066] A thermal management optimization unit is used to dynamically optimize the coolant flow rate based on the output results of the carrier's general thermal state prediction basic model and the thermal balance equation.

[0067] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0068] This invention provides a dynamic optimization method for thermal control parameters of liquid-cooled lithium iron phosphate batteries. Based on a general thermal state prediction model constructed using LightGBM, it ensures initial predictive capability even in the absence of individual samples. Simultaneously, it introduces a thermal balance equation as a physical constraint to dynamically optimize the coolant flow rate, making the control strategy both interpretable and strictly adhering to physical laws. This fundamentally avoids unreasonable results that might be output by a "black box" model. Furthermore, by utilizing rolling average error for model verification and monitoring, and triggering incremental learning based on real-time driving data, the model can automatically learn the habitual characteristics of different drivers. This allows it to rapidly evolve into a personalized prediction model based on a small number of samples, achieving self-evolution and personalized adaptation. This effectively solves the core defects of existing technologies, such as heavy reliance on large amounts of data, weak generalization ability, "black box" effect, and lack of physical constraints. It achieves a unification of high-precision prediction and personalized adaptive control under conditions of scarce samples. Attached Figure Description

[0069] Figure 1 This is a flowchart of the dynamic optimization method for thermal control parameters of liquid-cooled lithium iron phosphate batteries according to the present invention.

[0070] Figure 2 The present invention provides a flowchart for predicting the thermal state of a carrier by constructing a general thermal state prediction model based on the LightGBM model and an offline training database of the carrier thermal state.

[0071] Figure 3 The flowchart of the present invention is as follows: Based on the online gradient descent algorithm, combined with the vehicle's driving data, the general thermal state prediction basic model of the vehicle is dynamically optimized so that it can automatically learn driving habits.

[0072] Figure 4 Based on the output of the general thermal state prediction model of the carrier, the flow chart of coolant flow rate is dynamically optimized according to the thermal balance equation. Detailed Implementation

[0073] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0074] Reference Figure 1 As shown, a method for dynamic optimization of thermal management parameters for liquid-cooled lithium iron phosphate batteries includes:

[0075] Based on carrier state, thermal state, cooling system and environmental parameters, an offline training database for carrier thermal state was constructed, and the data was cleaned and aligned.

[0076] Based on the offline training database of carrier thermal state, a general thermal state prediction model for carriers is constructed using the LightGBM model to predict the thermal state of carriers.

[0077] Based on the general thermal state prediction model for carriers, and combined with actual monitoring results, the reliability of the predicted thermal state of carriers was verified.

[0078] Based on the online gradient descent algorithm and combined with the vehicle's driving data, the general thermal state prediction model of the vehicle is dynamically optimized so that it can automatically learn driving habits.

[0079] Based on the output of the general thermal state prediction model of the carrier, the coolant flow rate is dynamically optimized according to the thermal balance equation.

[0080] A sensor array is set up to collect in real time the current state parameters required for model prediction and the actual thermal state data at future moments to verify the accuracy of the prediction.

[0081] It can be explained that the thermal management performance of liquid-cooled lithium iron phosphate batteries is mainly affected by factors such as coolant type, cell arrangement, battery structure, liquid cooling plate design, heat transfer medium, and coolant flow rate. Among these, hardware parameters such as coolant type and cell arrangement are inherent design parameters that cannot be dynamically adjusted. However, coolant flow rate and the vehicle's motion state (strongly correlated with driving habits) are key dynamic variables. This invention achieves advanced prediction of the thermal state by fitting the mapping relationship between the vehicle's motion state and thermal state, and then dynamically adjusts the coolant flow rate accordingly to achieve the goal of advanced thermal management. Therefore, this solution forms a complete dynamic optimization technology solution through a closed-loop process from offline database construction → general model establishment → prediction reliability verification → online incremental learning → flow rate optimization under physical constraints. This effectively solves the core defects of existing technologies such as strong dependence on data volume, insufficient generalization ability, model "black box" nature, and lack of physical constraints, and achieves the unity of high-precision prediction and personalized adaptive control under limited sample conditions.

[0082] The process of constructing an offline training database for the thermal state of the carrier based on carrier state, thermal state, cooling system, and environmental parameters, and then performing data cleaning and alignment, specifically includes:

[0083] Based on carrier state parameters, thermal state parameters, cooling system parameters, and environmental parameters, the types of thermal control parameters for liquid-cooled lithium iron phosphate batteries are determined.

[0084] By selecting different carriers, battery capacities and cooling system architectures, and simulating the environment and carrier movement, test experiments were conducted to obtain thermal control parameter samples of liquid-cooled lithium iron phosphate batteries.

[0085] Based on the median filtering algorithm, the collected parameter sample data is filtered to eliminate the influence of abnormal peaks;

[0086] Based on the min-max normalization formula, the collected parameter sample data is normalized to map the data to the [0, 1] interval, thus eliminating the influence of data units.

[0087] Based on cubic spline interpolation, the data is interpolated to ensure that the data of all channels are strictly aligned to a unified time axis;

[0088] An offline training database for the thermal state of the carrier was constructed to store and record sample data of thermal control parameters of liquid-cooled lithium iron phosphate batteries, facilitating subsequent parameter adjustments and updates.

[0089] The following parameters can be explained: Carrier state parameters include: accelerator pedal opening, brake pedal opening, motor requested torque, motor actual power, carrier speed, longitudinal acceleration, battery current, battery total voltage, and battery SOC; cooling system parameters include: coolant actual flow rate, electric water pump power, electric water pump actual speed, radiator fan actual speed, fan power, PTC heater actual power, and PTC heater outlet water temperature; environmental parameters include: ambient temperature, ambient humidity, and atmospheric pressure; thermal state parameters include: battery average temperature, coolant inlet temperature, and coolant outlet temperature. Since these parameters (such as throttle, power, and current) directly determine the inflow and outflow of energy, and the generation and exchange of heat are essential for energy conversion... However, this follows the laws of physics (Joule's law, the laws of thermodynamics). By conducting comprehensive environmental chamber simulations, standard cycles of chassis dynamometers, and extreme tests on multiple real-world roads, thermal management parameter samples under different operating conditions are systematically collected. The test content includes: high and low temperature environment adaptability assessment, steady-state and transient load response, and the impact of different driving styles and road conditions on thermal behavior. Key data such as battery heat generation, cooling system efficiency, and temperature distribution are comprehensively obtained. By cleaning, aligning, and fusing these multi-dimensional and multi-boundary condition data, a high-quality, high-coverage offline training database is constructed, providing sufficient and reliable sample support for subsequent data-driven and physical model-integrated thermal state prediction.

[0090] Reference Figure 2 As shown, the step of constructing a general thermal state prediction model for the carrier based on the LightGBM model and an offline training database of the carrier's thermal state specifically includes:

[0091] The carrier state parameters, cooling system parameters, and environmental parameters are used as inputs, and the carrier thermal state parameters over a future period of time are used as outputs.

[0092] Among them, environmental parameters are used to characterize the dynamic constraints on the coolant inlet temperature and the system heat dissipation boundary, thereby improving the prediction accuracy of the model under different operating conditions.

[0093] Based on the carrier thermal state offline training database, it is divided into training set, validation set and test set in a ratio of 7:2:1;

[0094] Based on the LightGBM framework, a general thermal state prediction model for the carrier is constructed using its built-in gradient boosting decision tree algorithm. This model consists of multiple regression trees added together.

[0095] Each regression tree splits nodes with the goal of minimizing the weighted squared error. During splitting, the optimal feature and its splitting threshold are selected by maximizing the gain.

[0096] The initial learning rate is set to A. During training, the loss on the validation set is continuously monitored. If the loss does not decrease for five consecutive iterations, the current learning rate is halved.

[0097] With the goal of minimizing the average temperature deviation of the battery, a custom loss function with safety penalty is constructed based on the mean square error formula.

[0098] The gradient boosting algorithm is used to optimize the model parameters. In each iteration, the first and second gradients of the current model with respect to the custom loss function are calculated, and a new regression tree is added to fit the negative gradient direction to reduce the loss.

[0099] When the absolute value of the change in the value of the validation set loss function over K consecutive iterations is less than the preset threshold a, the model is determined to have converged, and the training process is terminated early to avoid overfitting.

[0100] The gain expression is:

[0101]

[0102] In the formula, Gain is the gain value used to evaluate the quality of the split point, G L G R These are the sum of the first-order gradients of all samples in the left and right child nodes after the split, and H, respectively. L H R denoted as the sum of the second-order gradients of all samples in the left and right child nodes after the split, respectively; λ is the L2 regularization coefficient; and γ is the minimum gain threshold required for the split. Splitting will only occur when Gain > γ;

[0103] The expression for the custom loss function with safety penalty is:

[0104]

[0105] In the formula, L is a custom loss value with a safety penalty, M is the number of samples participating in training, and Y... true.iY represents the true average battery temperature of the i-th sample. pred.i The predicted average battery temperature for the i-th sample, Y safe α is the preset safe temperature threshold, and α is the safety penalty factor.

[0106] This can be explained by the fact that the LightGBM model, as a lightweight gradient boosting machine learning framework, can handle the high-dimensionality features and complex nonlinear interactions between carrier state parameters, cooling system parameters, and environmental parameters in battery thermal state prediction inputs. Its decision tree-based ensemble algorithm is naturally adept at handling tabular data and can automatically capture complex nonlinear relationships and interactions between features, without requiring complex feature engineering or dimensionality transformations like neural networks. Secondly, its built-in gradient boosting decision tree algorithm, through iterative error correction, typically achieves higher accuracy than single models (such as SVR) or traditional neural networks on small samples. Furthermore, its built-in L2 regularization effectively controls model complexity, prevents overfitting, and ensures generalization performance. Therefore, this solution uses the LightGBM framework as a foundation and its built-in gradient boosting decision tree algorithm to construct a general thermal state prediction model for carriers. This model is used to fit the mapping relationship between carrier state parameters, cooling system parameters, environmental parameters, and carrier thermal state parameters, thereby overcoming the lag of traditional feedback control based on the current state and achieving advanced prediction of future battery temperature changes. This approach facilitates incremental learning and optimization using small sample data of driving habits, enhancing the personalization of the general thermal state prediction model for the vehicle. In this scheme, environmental parameters (especially ambient temperature) are used to characterize the dynamic constraints on the cooling system's heat dissipation boundary. Specifically, ambient temperature directly limits the minimum theoretical outlet temperature of the radiator, thus determining the lower limit of the coolant inlet temperature. In high-temperature environments, the system's heat dissipation capacity decreases, and the coolant's ability to remove heat weakens, leading to more rapid battery temperature rise. Therefore, using environmental parameters as model input is crucial. It enables the model to perceive changes in the external thermal environment and dynamically adjust its prediction baseline, significantly improving the model's prediction accuracy and generalization ability under complex operating conditions such as different regions and seasons, and avoiding prediction errors that violate thermodynamic principles. Parameters λ and γ, as hyperparameters, are determined by defining a search range, using cross-validation, or evaluating the model's performance under different λ values ​​on the validation set, selecting the λ value that performs best on the validation set. Parameter α is tuned based on the validation set, evaluating the model under different α values ​​on the validation set, and selecting the value that significantly reduces the predicted temperature from exceeding the safety threshold Y. safe The number of times the parameter Y is calculated without significantly worsening the overall prediction error, and the value of α is the parameter Y. safe It is a non-adjustable, rigid parameter determined by the nature of the battery itself. You can find the battery cell's operating temperature range and the temperature threshold for irreversible damage by referring to the cell manufacturer's specifications.

[0107] The verification of the reliability of the predicted thermal state of the carrier based on the general thermal state prediction model and actual monitoring results specifically includes:

[0108] Based on the trained general thermal state prediction model of the carrier, and based on the actual monitored carrier state parameters, cooling system parameters and environmental parameters, the thermal state parameter data of the carrier for a future period of time is obtained.

[0109] Acquire the predicted thermal state parameters of the model under different carrier states and the corresponding actual monitoring data, and perform data preprocessing;

[0110] Based on the predicted data of thermal state parameters and the corresponding actual monitoring data, the rolling average error of the predicted values ​​of each thermal state parameter is calculated.

[0111] The rolling average error of the predicted value is calculated for each thermal state parameter, and the consistency of each thermal state parameter is judged to screen the credibility of the predicted data and the corresponding actual monitoring data. If the rolling average error of a thermal state parameter exceeds the corresponding preset threshold for three consecutive times, an alarm is triggered, and the parameter data will depend on the monitoring data rather than the prediction. Otherwise, the predicted parameter is considered reliable.

[0112] The expression for the rolling average error value of the two data points is:

[0113] MAE_r j (k)=δ·MAE_r j (k-1)+(1-δ)·|Y Measurement,j (k)-Y pred.j (k|k-Δt)|

[0114] In the formula, MAE_r j (k) represents the rolling average error of the correction for the j-th thermal state parameter at the k-th calculation time, where k is the index of the error calculation time, δ is the smoothing factor, and MAE_r j (k-1) represents the rolling average error of the correction for the j-th thermal state parameter at the (k-1)-th calculation time, Y Measurement，j (k) represents the actual monitoring data of the j-th thermal state parameter at the k-th calculation time, Y pred，j (k|k-Δt) represents the predicted data of the j-th thermal state parameter at the k-th calculation time, and Δt represents the prediction time span of the model.

[0115] This can be explained by the fact that the performance of any data-driven model will gradually decline after deployment due to changes in the external environment or encountering new operating conditions not seen during training (such as extreme driving behavior). This phenomenon is called "model drift." By continuously comparing predicted values ​​with actual values, the degradation trend of model performance can be monitored in real time. Once a systematic increase in prediction error is detected, an alarm can be triggered, prompting the system to undergo maintenance or retraining. Completely trusting model predictions is dangerous. If the model outputs an incorrect but seemingly reasonable prediction value for unknown reasons, such as predicting a temperature lower than the actual temperature, directly using it for control may lead to the risk of battery overheating. Therefore, without ensuring the accuracy of predictions... When the values ​​are accurate, monitoring data can be relied upon instead of prediction, thereby reducing the risk of model misjudgment. Secondly, it provides constraints for subsequent incremental learning. Incremental learning is not always happening; it needs to be triggered at the right time to avoid unnecessary calculations or learning noise. The rolling average error is a key indicator for determining whether the model needs to learn. When the rolling average error is higher than its historical baseline value multiple times but does not reach the fault threshold, this is used as a signal to trigger the incremental learning mechanism in subsequent steps. The model is fine-tuned using new driving data to adapt to changes. The smoothing factor δ is determined on the validation set through grid search, and its typical empirical value is 0.9.

[0116] Reference Figure 3 As shown, the method of dynamically optimizing the general thermal state prediction model of the vehicle based on the online gradient descent algorithm and combined with the vehicle's driving data, so that it can automatically learn driving habits, specifically includes:

[0117] Set up a fixed-size first-in-first-out (FIFO) buffer to store the N most recent driving habit data pairs;

[0118] An incremental learning mechanism is set up to automatically trigger incremental learning of the basic model for predicting the general thermal state of the vehicle based on the rolling average error of the data and the number of driving habit data pairs in the FIFO buffer.

[0119] Based on the incremental learning mechanism, the general thermal state prediction model of the carrier is dynamically optimized, enabling it to automatically learn driving habits.

[0120] This can be explained by the fact that, based on the general thermal state prediction model of the carrier, this scheme introduces an incremental learning mechanism, combined with driving habit data, to autonomously learn the driving habits of the carrier. As the amount of driving habit data increases, more sample data is used to participate in the incremental learning mechanism, thus making the dynamic optimization of the thermal control parameters of the carrier's liquid-cooled lithium iron phosphate battery more personalized and intelligent. Here, the driving habit data pair is a tuple [X(k), Y...]. Measurement [(k)], where X(k) is the carrier state input data at the k-th calculation time, Y Measurement(k) represents the actual monitoring data of the thermal state at the k-th calculation time. Continuously collecting this data can characterize driving habits and the resulting systemic thermal response patterns, such as the frequency of rapid acceleration and deceleration, the commonly used vehicle speed range, typical thermal loads, and other personalized characteristics.

[0121] The incremental learning mechanism includes an incremental learning triggering mechanism and an incremental learning updating mechanism. The incremental learning triggering mechanism refers to automatically triggering incremental learning when two conditions are met simultaneously: the amount of data accumulated in the FIFO buffer exceeds a preset quantity threshold and the rolling average error is higher than its historical baseline value multiple times but does not reach the fault threshold. The preset quantity threshold, historical baseline value, and fault threshold are formulated based on big data, normal distribution, and expert experience.

[0122] Incremental learning and update mechanisms include:

[0123] Fix all existing decision tree parameters in the pre-trained general thermal state prediction model of the carrier, use it as a fixed feature extractor, and add a linear fully connected layer after its output (i.e. the sum of all decision tree predictions).

[0124] Randomly sample a batch of driving habit data pairs from the FIFO buffer and calculate the loss value of the general thermal state prediction model for the vehicle on the current batch;

[0125] An online gradient descent algorithm is used to calculate the gradient of the loss function with respect to the weights and biases of the newly added linear layer. By backpropagating the error, only the values ​​of the weights and biases in the newly added linear layer are updated.

[0126] The updated parameters are used to replace the parameters in the original general thermal state prediction model for the carrier for subsequent predictions.

[0127] In this process, a batch of driving habit data pairs is randomly sampled from the FIFO buffer, the loss value of the general thermal state prediction model of the vehicle is calculated for the current batch, and a custom loss function with safety penalty is directly applied to ensure the consistency of the optimization objectives in the online incremental learning and offline training stages.

[0128] Reference Figure 4 As shown, the dynamic optimization of coolant flow rate based on the output of the carrier's general thermal state prediction model and the thermal balance equation specifically includes:

[0129] The output of the general thermal state prediction model of the carrier is used as the input value, and the optimized value of the coolant flow rate is used as the output value.

[0130] Based on the thermal balance formula, a physical mapping model between the thermal state parameters of the carrier and the flow rate of the coolant is constructed, so that the parameter changes strictly follow physical rules.

[0131] The output of the general thermal state prediction model of the carrier is input into the physical mapping model of the thermal state parameters of the carrier and the flow rate of the coolant to calculate the predicted value of the thermal power accumulated by the battery pack in the future.

[0132] Based on the top-level temperature control target, dynamically generate the target value of the heat power accumulated by the battery pack itself;

[0133] To minimize the error between the accumulated heat power of the battery pack and the target value, and to ensure a smooth change in the coolant flow rate, an optimal flow rate control objective function under thermal balance constraints is constructed.

[0134] In each control cycle, the objective function is minimized based on the gradient descent algorithm to obtain the optimal coolant flow rate sequence for a future period.

[0135] The first control variable in the optimal coolant flow rate sequence for a future period is output as the actual control command to the coolant pump actuator, and the above rolling optimization steps are repeated until the next cycle.

[0136] The physical mapping model expression between the thermal state parameters of the carrier and the flow rate of the coolant is as follows:

[0137]

[0138] In the formula, P 电池 For battery heat generation power, P 冷却液 P is the heat power carried away by the coolant circulation. 电池包 For the heat power accumulated by the battery pack itself, c p ρ is the specific heat capacity at constant pressure of the coolant, v is the density of the coolant, and T is the flow rate of the coolant. out T is the outlet temperature of the battery pack coolant. in Here, m is the inlet temperature of the battery pack coolant, and c is the total mass of the battery pack. 电池 For the specific heat capacity of the battery, dT 电池 / dt is the rate of change of the average temperature of the battery;

[0139] The objective function for optimal flow velocity control under thermal balance constraints is expressed as follows:

[0140]

[0141] In the formula, minJ is the objective function value of the flow velocity optimal control under thermal balance constraints, and P 电池包，pred (k) represents the predicted heat power accumulated by the battery pack itself at the k-th calculation time, P 电池包，set(k) represents the accumulated thermal power of the target battery pack at the k-th calculation time, θ is the equilibrium constant, dimensionless, and is adjusted through experiments or simulations to balance temperature accuracy and flow rate stability, v(k) is the predicted value of coolant flow rate at the k-th calculation time, and v(k-1) is the predicted value of coolant flow rate at the (k-1)-th calculation time.

[0142] This can be explained by the fact that battery temperature changes are the result of accumulated heat power. Even with immediate and perfect cooling, the temperature will not drop instantly, but will change slowly according to the physical laws of heat capacity. If the target is temperature error, the controller is always chasing a delayed fact that has already occurred, resulting in a lag in the control response. The heat power accumulated by the battery pack itself directly determines the direction and rate of future temperature changes. If the accumulated heat power of the battery pack is greater than 0, it indicates a temperature rise; if it is less than 0, it indicates a temperature drop. By targeting the error of the accumulated heat power of the battery pack, controlling the current value of the accumulated heat power is equivalent to controlling the temperature trajectory over a future period, thus achieving true proactive temperature control. Secondly, the accumulated heat power of the battery pack has a clear physical mapping model. This model can effectively limit the optimal coolant flow rate sequence over a future period from exceeding physical limits, making it more accurate. Here, P... 电池包，set It is not a fixed value, but a setpoint dynamically generated based on the top-level temperature target. Its calculation formula is as follows:

[0143] P 电池包，set =K p ·(T set -T pred (k))

[0144] The controllable meaning of this formula is: by adjusting the accumulated thermal power of the target battery pack, the predicted average battery temperature is driven towards the set value T. set Convergence, where K p This is a proportionality coefficient, the value of which can be determined based on the battery pack's thermal capacity mc. 电池 Make initial settings.

[0145] The aforementioned sensor group, used to collect in real-time current state parameters required for model prediction and actual thermal state data for verifying prediction accuracy at future moments, specifically includes:

[0146] Sensor arrays are deployed at key locations on the carrier to collect thermal management parameter data of liquid-cooled lithium iron phosphate batteries, and the output data of the sensor arrays are divided into two categories.

[0147] The sensor group includes:

[0148] Temperature sensor array: used to monitor thermal status parameters, deployed in locations including the surface of the cells inside the battery pack, coolant inlet pipes, and coolant outlet pipes;

[0149] Carrier status sensor: used to monitor carrier motion parameters, acquired via the carrier CAN bus;

[0150] Cooling system sensors: used to monitor actuator status, including coolant flow rate sensors and electronic water pump speed sensors;

[0151] The output data of the sensor group is divided into two categories, specifically including:

[0152] The carrier state parameters, cooling system parameters, and environmental parameters are input into the carrier general thermal state prediction basic model in real time to drive the model to predict the future thermal state.

[0153] The actual monitoring data of thermal state parameters are compared with the prediction results of the model to calculate the rolling average error, verify the reliability of the prediction results, and provide feedback for optimized control.

[0154] By adopting a hardware synchronous acquisition method, all sensor data are ensured to have a unified and synchronized timestamp, thus guaranteeing data consistency in the time dimension.

[0155] The purpose of setting up a sensor array to collect thermal management parameters of liquid-cooled lithium iron phosphate batteries in real time is to provide input and verification data for the general thermal state prediction model of the carrier to achieve incremental model optimization. It also serves as a safety redundancy guarantee system to prevent thermal runaway in the event of model failure, provides necessary input to support thermal balance optimization control for the physical mapping model, and directly serves the calculation of the optimal coolant flow rate in the real-time control cycle, thus ensuring the feasibility, safety, and real-time optimization execution of thermal management in all dimensions.

[0156] Furthermore, based on the same inventive concept as the above-mentioned method for dynamic optimization of thermal management parameters for liquid-cooled lithium iron phosphate batteries, this solution proposes a system for dynamic optimization of thermal management parameters for liquid-cooled lithium iron phosphate batteries, comprising:

[0157] The data acquisition and database module is used to construct an offline training database for the thermal state of the carrier based on the carrier state, thermal state, cooling system and environmental parameters, and to perform data cleaning and alignment; the data acquisition module is used to set up a sensor group to collect in real time the current state parameters required for model prediction and the actual thermal state data at future moments for verifying the accuracy of the prediction.

[0158] The thermal management optimization module is used to train an offline database of the vehicle's thermal state. Based on the LightGBM model, it constructs a general thermal state prediction model for the vehicle to predict its thermal state. Based on the general thermal state prediction model and actual monitoring results, it verifies the reliability of the predicted thermal state. Based on an online gradient descent algorithm and the vehicle's driving data, it dynamically optimizes the general thermal state prediction model, enabling it to automatically learn driving habits. Based on the output of the general thermal state prediction model, it dynamically optimizes the coolant flow rate according to the thermal balance equation.

[0159] The data acquisition and database module includes:

[0160] The database unit is used to construct an offline training database for the thermal state of the carrier based on the carrier state, thermal state, cooling system and environmental parameters, and to perform data cleaning and alignment.

[0161] A data acquisition unit is used to set up a sensor group to collect in real time the current state parameters required for model prediction and the actual thermal state data at future moments for verifying the accuracy of the prediction.

[0162] The thermal management optimization module includes:

[0163] The model prediction unit is used to train the carrier thermal state offline database. Based on the LightGBM model, it constructs a general thermal state prediction basic model for the carrier to predict the thermal state of the carrier.

[0164] The model verification unit is used to verify whether the predicted thermal state of the carrier is reliable based on the general thermal state prediction model of the carrier and the actual monitoring results.

[0165] The incremental learning unit is used to dynamically optimize the general thermal state prediction basic model of the vehicle based on the online gradient descent algorithm and combined with the vehicle's driving data, so that it can automatically learn driving habits.

[0166] A thermal management optimization unit is used to dynamically optimize the coolant flow rate based on the output results of the carrier's general thermal state prediction basic model and the thermal balance equation.

[0167] In summary, the advantages of this invention are: by integrating data-driven and physical models, and utilizing online learning to dynamically optimize thermal management parameters, the thermal management accuracy, adaptability, and safety of liquid-cooled lithium iron phosphate batteries are significantly improved.

[0168] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for dynamically optimizing thermal management parameters of liquid-cooled lithium iron phosphate batteries, characterized in that, include: Based on carrier state, thermal state, cooling system and environmental parameters, an offline training database for carrier thermal state was constructed, and the data was cleaned and aligned. Based on the offline training database of carrier thermal state, a general thermal state prediction model for carriers is constructed using the LightGBM model to predict the thermal state of carriers. Based on the general thermal state prediction model for carriers, and combined with actual monitoring results, the reliability of the predicted thermal state of carriers was verified. Based on the online gradient descent algorithm and combined with the vehicle's driving data, the general thermal state prediction model of the vehicle is dynamically optimized so that it can automatically learn driving habits. Based on the output of the general thermal state prediction model of the carrier, the coolant flow rate is dynamically optimized according to the thermal balance equation. A sensor array is set up to collect in real time the current state parameters required for model prediction and the actual thermal state data at future moments to verify the accuracy of the prediction.

2. The method for dynamic optimization of thermal control parameters for liquid-cooled lithium iron phosphate batteries according to claim 1, characterized in that, The process of constructing an offline training database for the thermal state of the carrier based on carrier state, thermal state, cooling system, and environmental parameters, and then performing data cleaning and alignment, specifically includes: Based on carrier state parameters, thermal state parameters, cooling system parameters, and environmental parameters, the types of thermal control parameters for liquid-cooled lithium iron phosphate batteries are determined. By selecting different carriers, battery capacities and cooling system architectures, and simulating the environment and carrier movement, test experiments were conducted to obtain thermal control parameter samples of liquid-cooled lithium iron phosphate batteries. Based on the median filtering algorithm, the collected parameter sample data is filtered to eliminate the influence of abnormal peaks; Based on the min-max normalization formula, the collected parameter sample data is normalized to map the data to the [0, 1] interval, thus eliminating the influence of data units. Based on cubic spline interpolation, the data is interpolated to ensure that the data of all channels are strictly aligned to a unified time axis; An offline training database for the thermal state of the carrier was constructed to store and record sample data of thermal control parameters of liquid-cooled lithium iron phosphate batteries, facilitating subsequent parameter adjustments and updates.

3. The method for dynamic optimization of thermal control parameters for liquid-cooled lithium iron phosphate batteries according to claim 2, characterized in that, The step of constructing a general thermal state prediction model for carriers based on the LightGBM model and an offline training database of carrier thermal state specifically includes: The carrier state parameters, cooling system parameters, and environmental parameters are used as inputs, and the carrier thermal state parameters over a future period of time are used as outputs. Among them, environmental parameters are used to characterize the dynamic constraints on the coolant inlet temperature and the system heat dissipation boundary, thereby improving the prediction accuracy of the model under different operating conditions. Based on the carrier thermal state offline training database, it is divided into training set, validation set and test set in a ratio of 7:2:1; Based on the LightGBM framework, a general thermal state prediction model for the carrier is constructed using its built-in gradient boosting decision tree algorithm. This model consists of multiple regression trees added together. Each regression tree splits nodes with the goal of minimizing the weighted squared error. During splitting, the optimal feature and its splitting threshold are selected by maximizing the gain. Set the initial learning rate to During training, the loss on the validation set is continuously monitored. If the loss does not decrease for five consecutive iterations, the current learning rate is halved. ; With the goal of minimizing the average temperature deviation of the battery, a custom loss function with safety penalty is constructed based on the mean square error formula. The gradient boosting algorithm is used to optimize the model parameters. In each iteration, the first and second gradients of the current model with respect to the custom loss function are calculated, and a new regression tree is added to fit the negative gradient direction, thereby reducing the loss. When the absolute value of the change in the value of the validation set loss function over K consecutive iterations is less than the preset threshold a, the model is determined to have converged, and the training process is terminated early to avoid overfitting. The gain expression is: In the formula, This is the gain value, used to evaluate the quality of the split point. , These are the sums of the first-order gradients of all samples in the left and right child nodes after the split, respectively. , These are the sums of the second-order gradients of all samples in the left and right child nodes after the split, respectively. The L2 regularization coefficient is... The minimum gain threshold required for splitting, where, , ,when It will only split when that happens; The expression for the custom loss function with safety penalty is: In the formula, For custom loss values ​​with safety penalties, This represents the number of samples used in the training process. For the first The true average battery temperature of each sample. No. Predicted average battery temperature for each sample. The preset safe temperature threshold, This is a safety penalty factor.

4. The method for dynamic optimization of thermal control parameters for liquid-cooled lithium iron phosphate batteries according to claim 3, characterized in that, The verification of the reliability of the predicted thermal state of the carrier based on the general thermal state prediction model and actual monitoring results specifically includes: Based on the trained general thermal state prediction model of the carrier, and based on the actual monitored carrier state parameters, cooling system parameters and environmental parameters, the thermal state parameter data of the carrier for a future period of time is obtained. Acquire the predicted thermal state parameters of the model under different carrier states and the corresponding actual monitoring data, and perform data preprocessing; Based on the predicted data of thermal state parameters and the corresponding actual monitoring data, the rolling average error of the predicted values ​​of each thermal state parameter is calculated. The rolling average error of the predicted value is calculated for each thermal state parameter, and the consistency of each thermal state parameter is judged to screen the credibility of the predicted data and the corresponding actual monitoring data. If the rolling average error of a thermal state parameter exceeds the corresponding preset threshold for three consecutive times, an alarm is triggered, and the parameter data will depend on the monitoring data rather than the prediction. Otherwise, the parameter is considered reliable. The rolling average error expression is as follows: In the formula, For the first The calculation time is... Rolling average error corrected for each thermal state parameter This is the index for the time of error calculation. As a smoothing factor, For the first The calculation time is... Rolling average error corrected for each thermal state parameter For the first The calculation time is... Actual monitoring data of each thermal state parameter For the first The calculation time is... Predicted data for each thermal state parameter, This represents the prediction time span of the model.

5. The method for dynamic optimization of thermal control parameters for liquid-cooled lithium iron phosphate batteries according to claim 4, characterized in that, The method of dynamically optimizing the general thermal state prediction model of the vehicle based on the online gradient descent algorithm and combined with the vehicle's driving data, so as to enable it to automatically learn driving habits, specifically includes: Set up a fixed-size first-in-first-out (FIFO) buffer to store the N most recent driving habit data pairs; An incremental learning mechanism is set up to automatically trigger incremental learning of the basic model for predicting the general thermal state of the vehicle based on the rolling average error of the data and the number of driving habit data pairs in the FIFO buffer. Based on the incremental learning mechanism, the general thermal state prediction model of the carrier is dynamically optimized, enabling it to automatically learn driving habits.

6. The method for dynamic optimization of thermal control parameters for liquid-cooled lithium iron phosphate batteries according to claim 5, characterized in that, The step of dynamically optimizing the coolant flow rate based on the output of the carrier's general thermal state prediction model and the thermal balance equation specifically includes: The output of the general thermal state prediction model of the carrier is used as the input value, and the optimized value of the coolant flow rate is used as the output value. Based on the thermal balance formula, a physical mapping model between the thermal state parameters of the carrier and the flow rate of the coolant is constructed, so that the parameter changes strictly follow physical rules. The output of the general thermal state prediction model of the carrier is input into the physical mapping model of the thermal state parameters of the carrier and the flow rate of the coolant to calculate the predicted value of the thermal power accumulated by the battery pack in the future. Based on the top-level temperature control target, dynamically generate the target value of the heat power accumulated by the battery pack itself; To minimize the error between the accumulated heat power of the battery pack and the target value, and to ensure a smooth change in the coolant flow rate, an optimal flow rate control objective function under thermal balance constraints is constructed. In each control cycle, the objective function is minimized based on the gradient descent algorithm to obtain the optimal coolant flow rate sequence for a future period. The first control variable in the optimal coolant flow rate sequence for a future period is output as the actual control command to the coolant pump actuator, and the process of obtaining the optimal coolant flow rate sequence is repeated until the next cycle. The physical mapping model expression between the thermal state parameters of the carrier and the flow rate of the coolant is as follows: In the formula, For battery heat generation power, The heat power carried away by the coolant circulation, For the heat power accumulated by the battery pack itself, The specific heat capacity at constant pressure of the coolant. For coolant density, This refers to the coolant flow rate. This refers to the outlet temperature of the battery pack coolant. This refers to the inlet temperature of the battery pack coolant. The total mass of the battery pack. For the specific heat capacity of the battery, The rate of change of the average battery temperature; The objective function for optimal flow velocity control under thermal balance constraints is expressed as follows: In the formula, The objective function value for flow velocity under thermal equilibrium constraints is... For the first The predicted value of the accumulated thermal power of the battery pack itself at the next calculation time. For the first The heat power accumulated by the target battery pack itself at the next calculation time. This is an equilibrium constant, dimensionless, which is adjusted through experiments or simulations to balance temperature accuracy and flow rate stability. For the first Predicted coolant flow rate at the next calculation time. For the first The predicted value of coolant flow rate at the next calculation time.

7. The method for dynamic optimization of thermal control parameters for liquid-cooled lithium iron phosphate batteries according to claim 6, characterized in that, The aforementioned sensor group, used to collect in real-time current state parameters required for model prediction and actual thermal state data for verifying prediction accuracy at future moments, specifically includes: Sensors are deployed at key locations on the carrier to collect thermal management parameter data of liquid-cooled lithium iron phosphate batteries, and the output data of the sensor group is divided into two categories. The sensor group includes: Temperature sensor array: used to monitor thermal status parameters, deployed in locations including the surface of the cells inside the battery pack, coolant inlet pipes, and coolant outlet pipes; Carrier status sensor: used to monitor carrier motion parameters, acquired via the carrier CAN bus; Cooling system sensors: used to monitor actuator status, including coolant flow rate sensors and electronic water pump speed sensors; The output data of the sensor group is divided into two categories, specifically including: The carrier state parameters, cooling system parameters, and environmental parameters are input into the carrier general thermal state prediction basic model in real time to drive the model to predict the future thermal state. The actual monitoring data of thermal state parameters are compared with the prediction results of the model to calculate the rolling average error, verify the reliability of the prediction results, and provide feedback for optimized control. By adopting a hardware synchronous acquisition method, all sensor data are ensured to have a unified and synchronized timestamp, thus guaranteeing data consistency in the time dimension.

8. A dynamic optimization system for thermal management parameters of liquid-cooled lithium iron phosphate batteries, characterized in that, A method for dynamically optimizing thermal management parameters for liquid-cooled lithium iron phosphate batteries as described in any one of claims 1-7, comprising: The database module is used to construct an offline training database for the thermal state of the carrier based on the carrier state, thermal state, cooling system and environmental parameters, and to perform data cleaning and alignment. The thermal management optimization module is used to train an offline database of the vehicle's thermal state. Based on the LightGBM model, it constructs a general thermal state prediction model for the vehicle to predict its thermal state. Based on the general thermal state prediction model and actual monitoring results, it verifies the reliability of the predicted thermal state. Based on an online gradient descent algorithm and the vehicle's driving data, it dynamically optimizes the general thermal state prediction model, enabling it to automatically learn driving habits. Based on the output of the general thermal state prediction model, it dynamically optimizes the coolant flow rate according to the thermal balance equation. The data acquisition module is used to set up a sensor group to collect in real time the current state parameters required for model prediction and the actual thermal state data at future moments for verifying the accuracy of the prediction.

9. A dynamic optimization system for thermal management parameters of liquid-cooled lithium iron phosphate batteries according to claim 8, characterized in that, The thermal management optimization module includes: The model prediction unit is used to train the carrier thermal state offline database. Based on the LightGBM model, it constructs a general thermal state prediction basic model for the carrier to predict the thermal state of the carrier. The model verification unit is used to verify whether the predicted thermal state of the carrier is reliable based on the general thermal state prediction model of the carrier and the actual monitoring results. The incremental learning unit is used to dynamically optimize the general thermal state prediction basic model of the vehicle based on the online gradient descent algorithm and combined with the vehicle's driving data, so that it can automatically learn driving habits. A thermal management optimization unit is used to dynamically optimize the coolant flow rate based on the output results of the carrier's general thermal state prediction basic model and the thermal balance equation.

Images