Parameter simulation estimation method and device for battery heat effect model

Abstract

The invention provides a parameter simulation estimation method and device for a battery heat effect model, and the method comprises the steps: testing battery heat effect data, and dividing the battery heat effect data into a training set, a verification set and a test set; a Gaussian process regression model is constructed, and a training set is adopted for training; defining a genetic algorithm, performing hyper-parameter tuning by adopting a verification set, and optimizing the trained Gaussian process regression model according to the optimal parameter to obtain a GPR-GA model; verifying the GPR-GA model by using the data of the test set, and evaluating the precision of the model; adjusting model parameters according to an evaluation result until the model precision meets a set requirement; and performing parameter simulation estimation on the battery heat effect model by adopting the adjusted model. By means of the Gaussian process regression model, the genetic algorithm is introduced to optimize the hyper-parameters and the thermal effect physical parameters of the model, the battery thermal effect data can be more accurately fitted, the hyper-parameters can be adaptively adjusted, and the prediction precision of the model and the applicability under the actual working condition are further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of batteries, and particularly to a method and device for parameter simulation estimation of a battery thermal effect model. Background Art

[0002] Battery thermal effect models are widely used in electric vehicles, aerospace, and renewable energy storage systems. By accurately simulating the thermal effects of batteries, the battery management system and thermal management scheme can be optimized to ensure the stability and efficiency of the battery under various working conditions.

[0003] Currently, the parameter estimation of battery thermal effect models generally uses direct measurement or curve fitting methods. The direct measurement method directly measures parameters such as temperature, current, and voltage of the battery during charge and discharge through experiments, and then calculates the parameters of the thermal effect model based on these parameters. This method lacks the generality to widely adapt to different battery types and working conditions. Measuring each battery consumes manpower, material resources, and financial resources, and has low engineering application value. The curve fitting method is to establish the current-voltage characteristic curve of the battery and obtain the parameters of the thermal effect model through experimental data fitting. However, the parameters of the battery thermal effect model are complex non-linear relationships. Using the curve fitting method requires a large amount of experimental data, and the experimental process may be relatively complex, and it is also difficult to guarantee the accuracy.

[0004] In order to improve generality, although in related technologies, equivalent circuit models or finite element simulation methods are used for simulation and parameter simulation estimation, the finite element simulation method needs to consider the electrochemical process, heat conduction process, etc. inside the battery. Therefore, a large number of physical property parameters are required, and the calculation amount is large, and the requirements for computing resources are high. While the equivalent circuit model method has a smaller calculation amount, it cannot accurately describe the complex processes inside the battery, and the accuracy is not high. Summary of the Invention

[0005] To solve the above technical problems, the first object of the present invention is to propose a method for parameter simulation estimation of a battery thermal effect model.

[0006] The second object of the present invention is to propose a device for parameter simulation estimation of a battery thermal effect model.

[0007] The technical solution adopted by the present invention is as follows:

[0008] An embodiment of the first aspect of the present invention provides a method for parameter simulation estimation of a battery thermal effect model, including the following steps: testing battery thermal effect data, preprocessing the battery thermal effect data, and dividing the preprocessed data into a training set, a validation set, and a test set; constructing a Gaussian process regression model based on the physical laws of battery thermal effects and the parameters of the battery thermal effect model, and training the Gaussian process regression model GPR (Gaussian Process Regression) using the training set; defining a genetic algorithm, and using the validation set to perform hyperparameter tuning on the genetic algorithm GA (Genetic Algorithm) to iteratively search for the optimal parameters of the battery thermal effect model, and optimizing the trained Gaussian process regression model GPR according to the optimal parameters to obtain an optimized GPR-GA model; using the test set data to verify the optimized GPR-GA model and evaluating the model accuracy of the GPR-GA model; adjusting the parameters of the GPR-GA model according to the evaluation results until the model accuracy meets the set requirements; and using the adjusted GPR-GA model to perform parameter simulation estimation of the battery thermal effect model.

[0009] The method for parameter simulation estimation of the battery thermal effect model proposed above in the present invention may further have the following additional technical features:

[0010] According to an embodiment of the present invention, preprocessing the battery thermal effect data specifically includes: removing noise and outliers from the battery thermal effect data for cleaning; performing time alignment on the battery thermal effect data; extracting data feature values required for model training from the battery thermal effect data; and performing normalization or standardization on the data feature values.

[0011] According to an embodiment of the present invention, the parameters of the battery thermal effect model include: battery internal resistance R, specific heat capacity Cp, heat convection coefficient h, thermal conductivity k, battery surface area A, charge-discharge efficiency η charge and η discharge , and heat radiation rate ∈.

[0012] According to an embodiment of the present invention, the physical laws of battery thermal effects are obtained according to the following formula:

[0013]

[0014] where m is the mass of the battery, Cp is the specific heat capacity of the battery, dT / dt is the change rate of the battery temperature, Qin is the heat generated inside the battery, Qconduct is the heat dissipated through heat conduction, Qconv is the heat dissipated through heat convection, Qradiation is the heat dissipated through heat radiation, A is the battery surface area, T is the battery temperature value, T ambient is the external environmental temperature, and h is the heat convection coefficient.

[0015] According to an embodiment of the present invention, a Gaussian process regression model is constructed based on the physical laws of battery thermal effects and the parameters of the battery thermal effect model, and the Gaussian process regression model GPR is trained using the training set. Specifically, it includes: setting the input variable X and output variable Y of the Gaussian process regression model, where X = {I, V, Tenv, R, Cp, h, k, ∈, A, η charge , η discharge}, Y is the predicted battery temperature value T, I is the current, V is the working voltage, and Tenv is the ambient temperature; selecting a kernel function and a mean function, and establishing the relationship between the input variable X and the output variable Y using the kernel function and the mean function; calculating the covariance matrix according to the kernel function; calculating the log-likelihood function according to the covariance matrix; training the Gaussian process regression model GPR by maximizing the log-likelihood function using the training set.

[0016] According to an embodiment of the present invention, the kernel function includes: RBF (Radial Basis Function) kernel and white noise kernel.

[0017] According to an embodiment of the present invention, a genetic algorithm is defined, and the hyperparameters of the genetic algorithm GA are tuned using the validation set to iteratively search for the optimal parameters of the battery thermal effect model. Specifically, it includes: defining an objective function using the minimization problem of the error between the predicted value of the Gaussian process regression model GPR and the test value of the validation set; randomly generating an initial population, where each individual is a set of parameters of the battery thermal effect model; calculating the fitness value of the genetic algorithm objective function for each individual; selecting individuals with high fitness from the current population to enter the next generation; performing parameter recombination on the selected individuals to generate offspring; randomly perturbing the parameters of the individuals; calculating the fitness value of each individual and selecting the individual with the highest fitness value as the core of the new population; when the objective function value converges to a set threshold or the genetic algorithm reaches the maximum number of iterations, outputting the optimal parameters of the battery thermal effect model.

[0018] According to an embodiment of the present invention, the optimized GPR-GA model is verified using the test set data, and the model accuracy of the GPR-GA model is evaluated. Specifically, it includes: loading the optimized GPR-GA model; inputting the input variables in the test set into the GPR-GA model to obtain the predicted values of the GPR-GA model; comparing the predicted value curve of the GPR-GA model with the test curve of the test set data, and evaluating the model accuracy of the GPR-GA model by analyzing the similarity of the curves.

[0019] An embodiment of the second aspect of the present invention provides an apparatus for parameter simulation estimation of a battery thermal effect model, including: a test module configured to test battery thermal effect data, preprocess the battery thermal effect data, and divide the preprocessed data into a training set, a validation set, and a test set; a training module configured to construct a Gaussian process regression model based on the physical laws of battery thermal effects and the parameters of the battery thermal effect model, and train the Gaussian process regression model GPR using the training set; an optimization module configured to define a genetic algorithm, perform hyperparameter tuning on the genetic algorithm GA using the validation set to iteratively search for the optimal parameters of the battery thermal effect model, and optimize the trained Gaussian process regression model GPR according to the optimal parameters to obtain an optimized GPR-GA model; an evaluation module configured to verify the optimized GPR-GA model using the test set data and evaluate the model accuracy of the GPR-GA model; an adjustment module configured to adjust the parameters of the GPR-GA model according to the evaluation results until the model accuracy meets the set requirements; and a simulation estimation module configured to perform parameter simulation estimation of the battery thermal effect model using the adjusted GPR-GA model.

[0020] Advantages of the present invention:

[0021] With the non-parametric characteristics and powerful uncertainty quantification ability of the Gaussian process regression model, the present invention can efficiently capture the complex non-linear relationship of temperature changes during battery operation, comprehensively simulate the battery thermal effect behavior. At the same time, by introducing a genetic algorithm to optimize the hyperparameters and thermal effect physical parameters of the model, it can more accurately fit the battery thermal effect data, adaptively adjust the hyperparameters, further improve the prediction accuracy and applicability under actual working conditions of the model, reduce the overfitting risk of the model, and improve the generalization ability of the model in a changing environment. Description of the Drawings

[0022] Figure 1 is a flowchart of a method for parameter simulation estimation of a battery thermal effect model according to an embodiment of the present invention;

[0023] Figure 2 is a block diagram of an apparatus for parameter simulation estimation of a battery thermal effect model according to an embodiment of the present invention. Detailed Embodiments

[0024] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0025] Figure 1 is a flowchart of a method for parametric simulation estimation of a battery thermal effect model according to an embodiment of the present invention. As Figure 1 shown, the method for parametric simulation estimation of the battery thermal effect model includes the following steps:

[0026] S1, test battery thermal effect data, preprocess the battery thermal effect data, and divide the preprocessed data into a training set, a validation set, and a test set.

[0027] Specifically, various data of the battery thermal effect can be measured through HPPC (Hybrid Pulse Power Characterization) experiments. The HPPC experiment is a battery performance test method. By applying a standardized pulse charge and discharge load, the voltage response, internal impedance, and thermal effect of the battery are measured. This method can accurately capture battery thermal effect data at different SOCs (State Of Charge), including temperature changes and thermal accumulation characteristics, and reflect the thermal performance of the battery under actual working conditions.

[0028] The specific test experimental operation conditions for each battery are as follows: stand still for 1 min; perform constant power charging at a power of 300 W until the cell voltage reaches 3.55 V; set the constant temperature box to T = 5 °C and let the cell stand still for 4 h; perform constant current charging at 28 A until the cut-off voltage, and conduct 12 HPPC experiments (stand still for 1 h, discharge at a constant current of 120 A for 0.2 h, stand still for 1 h, charge at a constant current of 30 A for 0.2 h)

[0029] After obtaining the battery thermal effect data from the HPPC test, the test data is preprocessed. In an embodiment of the present invention, the preprocessing of the battery thermal effect data specifically includes: removing noise and outliers from the battery thermal effect data for cleaning; aligning the time of the battery thermal effect data; extracting data feature values required for model training from the battery thermal effect data; and normalizing or standardizing the data feature values.

[0030] Specifically, noise removal mainly uses filtering methods (such as low-pass filters) to remove high-frequency noise in temperature, voltage, and current signals. Outlier processing generally involves identifying and removing abnormal data points (such as sudden temperature changes or voltage jumps) during the experiment, which can be accomplished through statistical methods or anomaly detection algorithms. Time alignment is to synchronize the time of data such as temperature, voltage, and current to ensure that the sampling times of different sensors are consistent. If missing data is encountered, interpolation methods (such as linear interpolation) are used to fill it. Feature extraction is to extract key features from the original data of the HPPC experiment to construct the dataset required for the model. Data standardization is to normalize or standardize the data feature values to avoid the influence of features with different dimensions on the modeling process.

[0031] After the preprocessing is completed, the preprocessed data is divided into a training set, a validation set, and a test set. The training set generally accounts for 70% of the data, the validation set accounts for 15%, and the test set accounts for 15%.

[0032] S2. Based on the physical laws of battery thermal effects and the parameters of the battery thermal effect model, a Gaussian process regression model is constructed, and the training set is used to train the Gaussian process regression model GPR.

[0033] In an embodiment of the present invention, the parameters of the battery thermal effect model include: battery internal resistance R (5 - 50 mΩ), specific heat capacity Cp (800–1400 J / (kg·K)), heat convection coefficient h (5–50 W / (m 2 ·K)), thermal conductivity k (0.1–2.0 W / (m·K)), battery surface area A (0.005–0.1 m 2 ), charge and discharge efficiency ηcharge and ηdischarge (85%–98%), and thermal radiation rate ∈ (0.8–0.98).

[0034] The physical laws of battery thermal effects mainly include the following (1)-(7):

[0035] (1) Ohmic heat Q ohmic (the heat generated by the current passing through the internal resistance), which is generated by the current I passing through the battery internal resistance R: Q ohmic = I 2 R; where I is the battery charge and discharge current (A), and R is the battery internal resistance (Ω).

[0036] (2) Electrochemical reaction heat Q chem (the reaction heat during charge and discharge), during the battery charge and discharge process, the chemical reaction of the battery (such as the insertion / extraction of lithium ions in a lithium battery) will release or absorb heat. The electrochemical reaction heat Q chem can be expressed as:

[0037] Q chem = η charge IVcharge +η discharge IV discharge

[0038] where: η charge , η discharge are the charging and discharging efficiencies respectively; V charge , V discharge are the charging and discharging voltages respectively.

[0039] (3) Total internal heat source Q in : Q in = Q ohmic + Q chem

[0040] (4) Heat loss Q conduct by heat conduction. The heat inside the battery is transferred through the thermal conductivity k. For a solid battery, the heat loss Q conduct by heat conduction is given by the formula:

[0041]

[0042] where: k is the thermal conductivity of the material (W / (m·°C)), A is the heat conduction surface area (m 2 ), T is the current temperature of the battery (°C), T ambient is the external ambient temperature (°C), and d is the thickness of the heat conduction path (m).

[0043] (5) Heat loss Qconv by heat convection. The heat exchange between the battery surface and the external environment usually occurs through heat convection. Newton's law of cooling can be used to represent the heat transfer by heat convection:

[0044] Q conv = hA(T - T ambient )

[0045] where the heat loss Qconv by heat convection, h is the heat convection coefficient (W / (m 2 ·°C)), A is the battery surface area (m 2 ), the battery temperature value (°C), and T ambient is the ambient temperature (°C).

[0046] (6) Heat loss Qradiation by heat radiation. The heat transfer by heat radiation of the battery is described by the Stefan-Boltzmann law:

[0047]

[0048] where ∈ is the emissivity of the battery surface (dimensionless, 0 ≤ ∈ ≤ 1), σ is the Stefan-Boltzmann constant, and A is the battery surface area (m 2), where T is the battery temperature value (the temperature on the battery surface), T ambient is the external environmental temperature.

[0049] (7) Considering all factors of the internal heat source and heat transfer of the battery comprehensively, the heat balance equation of the battery describes the change of the battery temperature over time. The heat balance equation of the battery is:

[0050]

[0051] Among them, m is the mass of the battery, Cp is the specific heat capacity of the battery, dT / dt is the change rate of the battery temperature, Qin is the heat generated inside the battery, Qconduct is the heat dissipated through heat conduction, Qconv is the heat dissipated through heat convection, and Qradiation is the heat dissipated through heat radiation.

[0052] Substituting each heat transfer process into the heat balance equation, we get:

[0053]

[0054] Among them, A is the surface area of the battery, T is the battery temperature value, T ambient is the external environmental temperature, h is the heat convection coefficient, k is the thermal conductivity, A is the surface area of the battery, ∈ is the heat radiation rate, and σ is the Stefan-Boltzmann constant.

[0055] In an embodiment of the present invention, based on the physical laws of the battery thermal effect and the parameters of the battery thermal effect model, a Gaussian process regression model is constructed, and the Gaussian process regression model GPR is trained using a training set, specifically including the following steps S21 - S25:

[0056] S21, set the input variable X and the output variable Y of the Gaussian process regression model. Among them, X = {I, V, Tenv, R, Cp, h, k, ∈, A, η charge , η discharge}, Y is the predicted battery temperature value T, I is the current, V is the working voltage, and Tenv is the environmental temperature.

[0057] S22, select the kernel function and the mean function, and establish the relationship between the input variable X and the output variable Y using the kernel function and the mean function.

[0058] Specifically, Gaussian process regression uses the kernel function to establish the relationship between the input variable X and the output variable Y, that is, T(x)~GP(m(x),k(x,x′)); where m(x) is the mean function; k(x,x′) is the kernel function, which describes the correlation between the input variables, x is obtained from the training set, and x′ is obtained from the test data. The mean function m(x) is usually set to zero, indicating no deviation. If there is prior knowledge, the mean function can also be selected as a constant function or a linear function.

[0059] In a specific embodiment of the present invention, the kernel function of the selected model includes: an RBF kernel and a white noise kernel, which can be specifically a combination of the RBF kernel and the white noise kernel:

[0060]

[0061] Where x and x′ are input variables; is the amplitude hyperparameter, l is the length scale parameter, is the noise term, and δ represents the δ function. In the present invention, the RBF kernel is used to capture complex data patterns, while the white noise kernel is used to describe the measurement noise in the data.

[0062] Therefore, based on the physical laws of battery thermal effect and the parameters of the battery thermal effect model, a Gaussian process regression model is constructed, and the physical thermal effect parameter optimization kernel function is introduced to improve the physical constraints and generalization ability of the model. GPR characterizes the high-dimensional nonlinear mapping relationship between input variables and output variables through kernel functions, and integrates key thermal parameters such as internal resistance, thermal conductivity, and thermal convection coefficient of the battery to construct a priori kernel functions, thereby ensuring that the model is not only driven by experimental data, but also conforms to the physical mechanism of battery thermal behavior. This method can ensure the robustness of the prediction results, and enhance the fit and interpretability of the model with the physical mechanism of battery thermal effect, thereby providing accurate data prediction for the simulation of battery thermal effect.

[0063] S23, calculating the covariance matrix according to the kernel function.

[0064] S24, calculating the log-likelihood function according to the covariance matrix.

[0065] S25, uses the training set to maximize the log-likelihood function to train the Gaussian process regression model GPR.

[0066] Specifically, the log-likelihood logp(y|X,θ) can be calculated according to the following formula:

[0067]

[0068] K(X,X;θ) is the covariance matrix of the training data, the calculation method depends on the selected kernel function and its hyperparameters; y is the target output vector; n is the number of training samples.

[0069] Maximize the log-likelihood function through numerical optimization (genetic algorithm) to obtain the hyperparameter θ * .

[0070] After training, GPR can be used to predict the battery temperature T and compare it with the actual test results, requiring it to conform to the physical laws of battery thermal effects and facilitating further iterative parameter optimization using the genetic algorithm.

[0071] S3. Define the genetic algorithm, and use the validation set to tune the hyperparameters of the genetic algorithm GA to iteratively search for the optimal parameters of the battery thermal effect model. Optimize the trained Gaussian process regression model GPR according to the optimal parameters to obtain the optimized GPR-GA model.

[0072] In an embodiment of the present invention, define the genetic algorithm, and use the validation set to tune the hyperparameters of the genetic algorithm GA to iteratively search for the optimal parameters of the battery thermal effect model, specifically including the following steps S31 - S38:

[0073] S31. Define the objective function using the minimization problem of the error between the predicted value of the Gaussian process regression model GPR and the test value of the validation set.

[0074] S32. Randomly generate the initial population, and each individual is a set of parameters of the battery thermal effect model.

[0075] Individual = {I, V, Tenv, R, Cp, h, k, ∈, A, η charge , η discharge}

[0076] S33. Calculate the fitness value of the objective function of each individual in the genetic algorithm.

[0077] S34. Select individuals with high fitness from the current population to enter the next generation according to the fitness value.

[0078] S35. Recombine the parameters of the selected individuals to generate offspring.

[0079] S36. Randomly perturb the parameters of the individuals.

[0080] S37. Calculate the fitness value of each individual and select the individual with the highest fitness value as the core of the new population.

[0081] S38. When the objective function value converges to the set threshold or the genetic algorithm reaches the maximum number of iterations, output the optimal parameters of the battery thermal effect model.

[0082] S4. Use the test set data to verify the optimized GPR-GA model and evaluate the model accuracy of the GPR-GA model.

[0083] In an embodiment of the present invention, the optimized GPR-GA model is verified using test set data to evaluate the model accuracy of the GPR-GA model, specifically including the following steps S41 - S43:

[0084] S41, load the optimized GPR-GA model.

[0085] S42, input the input variables in the test set into the GPR-GA model to obtain the predicted values of the GPR-GA model;

[0086] S43, compare the predicted value curve of the GPR-GA model with the test curve of the test set data, and evaluate the model accuracy of the GPR-GA model by analyzing the similarity of the curves.

[0087] S5, adjust the parameters of the GPR-GA model according to the evaluation results until the model accuracy meets the set requirements.

[0088] Specifically, the model prediction accuracy is evaluated through visual analysis. Compare the battery temperature curve output by the GPR-GA model with the temperature data obtained from the test. If the two data are close, it indicates that the model has strong dynamic prediction ability. If there are areas with significant deviations in the data, it is necessary to analyze the error sources and adjust the model.

[0089] The methods for adjusting the model include hyperparameter tuning, data enhancement and preprocessing, kernel function selection and optimization, optimization algorithm iteration, model integration, etc. First, the key parameters affecting the accuracy (such as battery internal resistance, conductivity, heat convection coefficient, etc.) should be identified by analyzing the error between the model prediction and experimental data, and the hyperparameters of the kernel function (such as signal variance and length scale) should be adjusted to improve the model fitting. Thus, the genetic algorithm is combined to optimize the thermal effect parameters, and a deviation correction factor is introduced to dynamically compensate the predicted values, ultimately improving the physical consistency of the model and the prediction accuracy under complex working conditions.

[0090] S6, use the GPR-GA model to perform parameter simulation estimation of the battery thermal effect model.

[0091] In summary, according to the parameter simulation estimation method of the battery thermal effect model in the embodiment of the present invention, by virtue of the non-parametric characteristics and powerful uncertainty quantification ability of the Gaussian process regression model, the complex nonlinear relationship of temperature changes during battery operation can be efficiently captured, and the battery thermal effect behavior can be comprehensively simulated. At the same time, by introducing the genetic algorithm to optimize the hyperparameters and thermal effect physical parameters of the model, the battery thermal effect data can be more accurately fitted, and the hyperparameters can be adaptively adjusted, further improving the prediction accuracy of the model and its applicability under actual working conditions, reducing the overfitting risk of the model, and improving the generalization ability of the model in a changing environment.

[0092] Corresponding to the above parameter simulation estimation method of the battery thermal effect model, the present invention also proposes a parameter simulation estimation device for the battery thermal effect model. Since the device embodiment of the present invention corresponds to the above method embodiment, for the details not disclosed in the device embodiment, reference can be made to the above method embodiment, and no further elaboration will be made in the present invention.

[0093] Figure 2 is a block diagram of a parameter simulation estimation device for a battery thermal effect model according to an embodiment of the present invention. As Figure 2 shown, the parameter simulation estimation device includes: a test module 10, a training module 20, an optimization module 30, an evaluation module 40, an adjustment module 50, and a simulation estimation module 60.

[0094] Among them, the test module 10 is used to test battery thermal effect data, preprocess the battery thermal effect data, and divide the preprocessed data into a training set, a validation set, and a test set; the training module 20 is used to construct a Gaussian process regression model based on the physical laws of battery thermal effects and the parameters of the battery thermal effect model, and use the training set to train the Gaussian process regression model GPR; the optimization module 30 is used to define a genetic algorithm, use the validation set to perform hyperparameter tuning on the genetic algorithm GA, iteratively search for the optimal parameters of the battery thermal effect model, optimize the trained Gaussian process regression model GPR according to the optimal parameters, and obtain an optimized GPR-GA model; the evaluation module 40 is used to verify the optimized GPR-GA model using the test set data and evaluate the model accuracy of the GPR-GA model; the adjustment module 50 is used to adjust the parameters of the GPR-GA model according to the evaluation results until the model accuracy meets the set requirements; the simulation estimation module 60 is used to perform parameter simulation estimation of the battery thermal effect model using the adjusted GPR-GA model.

[0095] The parameter simulation estimation device for the battery thermal effect model according to the embodiment of the present invention can efficiently capture the complex non-linear relationship of temperature changes during battery operation and comprehensively simulate the battery thermal effect behavior by virtue of the non-parametric characteristics and powerful uncertainty quantification ability of the Gaussian process regression model. At the same time, by introducing a genetic algorithm to optimize the hyperparameters and thermal effect physical parameters of the model, it can more accurately fit the battery thermal effect data, can adaptively adjust the hyperparameters, further improve the prediction accuracy and applicability under actual working conditions of the model, reduce the overfitting risk of the model, and improve the generalization ability of the model in a changing environment.

[0096] In the description of the present invention, the descriptions referring to terms such as "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" etc. mean that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Any process or method description shown in the flowchart or described in other ways herein can be understood as representing a module, segment or part of code including one or more executable instructions for implementing a customized logical function or process, and the scope of the preferred embodiments of the present invention includes additional implementations, where the functions can be executed in a substantially simultaneous manner or in a reverse order according to the functions involved, rather than in the order shown or discussed, which should be understood by those skilled in the art to which the embodiments of the present invention pertain.

[0097] Those of ordinary skill in the art in this technical field can understand that all or part of the steps carried by the methods of the above embodiments can be completed by instructing relevant hardware through a program, and the said program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0098] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A parameter simulation and estimation method for a battery thermal effect model, characterized in that: The following steps are involved: Testing battery thermal effect data, preprocessing the battery thermal effect data, and dividing the preprocessed data into a training set, a validation set, and a test set; Based on the physical law of battery thermal effect and the parameters of the battery thermal effect model, a Gaussian process regression model is constructed, and the Gaussian process regression model GPR is trained using the training set; A genetic algorithm is defined, and hyperparameter tuning of the genetic algorithm GA is performed using the validation set to iteratively search for optimal parameters of the battery thermal effect model, and the trained Gaussian process regression model GPR is optimized according to the optimal parameters to obtain an optimized GPR-GA model; Using the test set data to verify the optimized GPR-GA model and evaluate the model accuracy of the GPR-GA model; Adjust the GPR-GA model parameters according to the evaluation results until the model accuracy meets the set requirements; The adjusted GPR-GA model is used to simulate and estimate the parameters of the battery thermal effect model.

2. The parameter simulation and estimation method of the battery thermal effect model according to claim 1, characterized in that: Preprocessing the battery thermal effect data specifically includes: Remove noise and outliers from battery thermal effect data for cleaning; Time alignment of battery thermal effect data; Extract data feature values ​​required for model training from battery thermal effect data; The data feature values ​​are normalized or standardized.

3. The parameter simulation and estimation method of the battery thermal effect model according to claim 1, characterized in that: The parameters of the battery thermal effect model include: battery internal resistance R, specific heat capacity Cp, thermal convection coefficient h, thermal conductivity k, battery surface area A, charge and discharge efficiency η charge and η discharge and thermal emissivity ∈.

4. The parameter simulation and estimation method of the battery thermal effect model according to claim 3, characterized in that: The physical law of battery thermal effect is obtained according to the following formula: Where m is the mass of the battery, Cp is the specific heat capacity of the battery, dT / dt is the rate of change of the battery temperature, Qin is the heat generated inside the battery, Qconduct is the heat lost by thermal conduction, Qconv is the heat lost by thermal convection, Qradiation is the heat lost by thermal radiation, A is the battery surface area, T is the battery temperature value, T ambient is the external environment temperature, h is the thermal convection coefficient, k is the thermal conductivity, A is the battery surface area, ∈ is the thermal emissivity, and σ is the Stefan-Boltzmann constant.

5. The parameter simulation and estimation method of the battery thermal effect model according to claim 3, characterized in that: Based on the physical law of battery thermal effect and the parameters of the battery thermal effect model, a Gaussian process regression model is constructed, and the training set is used to train the Gaussian process regression model GPR, specifically including: Set the input variable X and output variable Y of the Stromberg process regression model, where X = {I, V, Tenv, R, Cp, h, k, ∈, A, η charge ,η discharge }, Y is the predicted battery temperature value T, I is the current, V is the operating voltage, and Tenv is the ambient temperature; Select the kernel function and the mean function, and use the kernel function and the mean function to establish the relationship between the input variable X and the output variable Y; Calculate the covariance matrix based on the kernel function; Calculating a log-likelihood function based on the covariance matrix; The training set is used to maximize the log-likelihood function to train the Gaussian process regression model GPR.

6. The parameter simulation and estimation method of the battery thermal effect model according to claim 5, characterized in that: The kernel function includes: RBF kernel and white noise kernel.

7. The parameter simulation and estimation method of the battery thermal effect model according to claim 1, characterized in that: A genetic algorithm is defined, and the validation set is used to perform hyperparameter tuning on the genetic algorithm GA to iteratively search for optimal parameters of the battery thermal effect model, specifically including: The objective function is defined by minimizing the error between the predicted value of the Gaussian process regression model GPR and the test value of the validation set; The initial population is randomly generated, and each individual is a set of parameters of the battery thermal effect model; Calculate the fitness value of the genetic algorithm objective function of each individual; According to the fitness value, individuals with high fitness are selected from the current population to enter the next generation; Reorganize the parameters of the selected individuals to generate offspring; Randomly perturb the parameters of individuals; Calculate the fitness value of each individual and select the individual with the highest fitness value as the core of the new population; When the objective function value converges to the set threshold or the genetic algorithm reaches the maximum number of iterations, the optimal parameters of the battery thermal effect model are output.

8. The parameter simulation and estimation method of the battery thermal effect model according to claim 1, characterized in that: The optimized GPR-GA model is verified using the test set data to evaluate the model accuracy of the GPR-GA model, specifically including: Load the optimized GPR-GA model; Input the input variables in the test set into the GPR-GA model to obtain the predicted values ​​of the GPR-GA model; The predicted value curve of the GPR-GA model is compared with the test curve of the test set data, and the model accuracy of the GPR-GA model is evaluated by analyzing the similarity of the curves.

9. A parameter simulation and estimation device for a battery thermal effect model, characterized in that: include: A test module, the test module is used to test battery thermal effect data, and preprocess the battery thermal effect data, and divide the preprocessed data into a training set, a validation set and a test set; A training module, wherein the training module is used to construct a Gaussian process regression model based on the physical law of battery thermal effect and the parameters of the battery thermal effect model, and to train the Gaussian process regression model GPR using the training set; An optimization module, wherein the optimization module is used to define a genetic algorithm, and use the verification set to perform hyperparameter tuning on the genetic algorithm GA to iteratively search for optimal parameters of the battery thermal effect model, and optimize the trained Gaussian process regression model GPR according to the optimal parameters to obtain an optimized GPR-GA model; An evaluation module, wherein the evaluation module is used to verify the optimized GPR-GA model using the test set data and evaluate the model accuracy of the GPR-GA model; An adjustment module, the adjustment module is used to adjust the GPR-GA model parameters according to the evaluation results until the model accuracy meets the set requirements; A simulation estimation module is used to use the adjusted GPR-GA model to perform parameter simulation estimation of the battery thermal effect model.

Images