Lithium ion battery thermal management control method based on joint state estimation

Abstract

A lithium ion battery thermal management control method based on joint state estimation comprises the following steps: establishing a lithium ion battery second-order RC equivalent electric model, and performing electric model parameter identification by adopting an improved robust adaptive multi-forgetting factor recursive least square method; establishing a dual-state lumped parameter thermal model of the lithium ion battery, and performing thermal model parameter identification by adopting an improved goat optimization algorithm; constructing an electrothermal coupling model based on the identified electrical model and the parameters of the dual-state lumped parameter thermal model, then constructing a minimum entropy adaptive Kalman filter based on a near-end strategy optimization algorithm, and then carrying out joint estimation on the SOC and SOT of the battery; and based on the joint estimation result, constructing a deep reinforcement learning optimized adaptive model prediction control algorithm, and performing thermal management control on the lithium ion battery through the deep reinforcement learning optimized adaptive model prediction control algorithm. According to the invention, more accurate and effective thermal management control of the lithium ion battery can be realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of lithium ion batteries, in particular to a lithium ion battery thermal management control method based on joint state estimation. BACKGROUND

[0002] With the wide application of high energy density lithium ion batteries in electric vehicles, its thermal safety and state estimation accuracy have become the core bottleneck restricting the performance of the whole vehicle. Especially in the distributed electric drive scene such as wheel hub motor drive, the battery frequently withstands large current impact and complex working condition switching, which leads to three challenges for traditional single state estimation method: first, the electric model parameters are strongly time-varying, and the conventional recursive least squares method is difficult to track dynamic characteristics due to the fixed forgetting factor; second, the thermal model parameter identification accuracy is insufficient, and the existing optimization algorithm is easy to fall into local optimum, affecting the reliability of temperature field prediction; third, the strong coupling characteristics of SOC (state of charge) and SOT (battery temperature) are not fully modeled, and independent estimation will introduce cross error.

[0003] In view of the above problems, it is urgent to build an electric-thermal joint estimation and cooperative control system to realize more accurate and effective lithium ion battery thermal management control. SUMMARY

[0004] In view of the above problems, the present application provides a lithium ion battery thermal management control method based on joint state estimation to build an electric-thermal joint estimation and cooperative control system to realize more accurate and effective lithium ion battery thermal management control.

[0005] A lithium ion battery thermal management control method based on joint state estimation, comprising: Step S1, establishing a second-order RC equivalent electric model of lithium ion battery, and using an improved robust adaptive multi-forgetting factor recursive least squares method to identify the electric model parameters; Step S2, establishing a double-state lumped parameter thermal model of lithium ion battery, and using an improved goat optimization algorithm to identify the thermal model parameters; Step S3, based on the electric model and double-state lumped parameter thermal model parameters identified in steps S1 and S2, constructing an electric-thermal coupling model, and based on the electric-thermal coupling model, constructing a minimum entropy adaptive Kalman filter based on a proximal policy optimization algorithm, and then jointly estimating the SOC and SOT of the battery; Step S4, based on the joint estimation result in step S3, constructing a deep reinforcement learning optimized adaptive model predictive control algorithm, and using the deep reinforcement learning optimized adaptive model predictive control algorithm to perform thermal management control on the lithium ion battery.

[0006] According to the lithium ion battery thermal management control method based on joint state estimation provided by the present application, the following beneficial effects are achieved: 1、The application comprehensively considers the lithium ion battery electrical model and the two-state lumped parameter thermal model, the dynamic characteristics of the battery are represented by the second-order RC equivalent electrical model and the two-state lumped parameter thermal model respectively, and the improved robust adaptive multi-forgetting factor recursive least square method and the improved goat optimization algorithm are used to improve the parameter identification accuracy, thereby providing the parameter and model basis for the joint estimation of SOT and SOC; 2、Based on the electrical model and the thermal model of the lithium ion battery, a new electrical-thermal coupling model is constructed, and based on the identified electrical model and two-state lumped parameter thermal model parameters, a minimum entropy adaptive Kalman filter based on the proximal policy optimization algorithm (PPO) is constructed to jointly estimate SOC and SOT, and solve the state coupling interference; 3、The SOC and the battery temperature estimated by the minimum entropy adaptive Kalman filter based on the proximal policy optimization algorithm of the application are used to construct an adaptive model predictive control algorithm optimized by deep reinforcement learning to perform thermal management control on the lithium ion battery, so that accurate and effective thermal management control of the lithium ion battery is realized, and the whole chain innovation of parameter identification-joint estimation-dynamic control is used to break through the defects of traditional thermal management response lag and excessive conservatism, thereby providing effective protection for the safety of the battery under high dynamic conditions. BRIEF DESCRIPTION OF DRAWINGS

[0007] Figure 1 A flowchart of the lithium ion battery thermal management control method based on joint state estimation provided by the embodiments of the application is shown in the figure. Figure 2 A comparison chart of the estimation of SOC by the method proposed in the application and the traditional extended Kalman algorithm and the real value is shown in the figure. Figure 3 A comparison chart of the estimation of SOT by the method proposed in the application and the traditional extended Kalman algorithm and the real value is shown in the figure. Figure 4 A comparison chart of the temperature control by the method proposed in the application and the traditional PID algorithm and the ideal value is shown in the figure. DETAILED DESCRIPTION

[0008] The embodiments of the application will be described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary and are intended to explain the embodiments of the application, and cannot be understood as a limitation of the application.

[0009] Please refer to Figure 1 The embodiments of the application provide a lithium ion battery thermal management control method based on joint state estimation, which comprises steps S1-S4: Step S1, a lithium-ion battery second-order RC equivalent circuit model is established, and an improved robust adaptive multi-forgetting factor recursive least square method is used for parameter identification of the equivalent circuit model.

[0010] wherein, according to the Kirchhoff voltage and current law in circuit theory, and letting the discharging direction of the battery be positive, the relationship between the voltage and the current of the lithium-ion battery second-order RC equivalent circuit is expressed as:

[0011]

[0012]

[0013] wherein, is the working current, and are the electrochemical polarization voltage and the concentration polarization voltage of the lithium-ion battery, and are the electrochemical polarization resistance and the concentration polarization resistance of the lithium-ion battery, and are the electrochemical polarization capacitance and the concentration polarization capacitance of the lithium-ion battery, is the differential of , is the differential of , is the differential of time , is the open-circuit voltage, is the resistance parameter, is the output voltage; the SOC of the lithium-ion battery is expressed as:

[0014] wherein, is the SOC value of the battery at time , is the initial SOC value of the battery at time , is the coulomb efficiency, is the maximum available capacity, is the time between and , is the current at time step , is the differential of ; after discretization, we obtain:

[0015]

[0016]

[0017] in, and They are The electrochemical polarization voltage and concentration polarization voltage of the lithium-ion battery at the moment, and They are The electrochemical polarization voltage and concentration polarization voltage of the lithium-ion battery at the moment, for The current at the moment, for The current at the moment, for The output voltage at the moment, for The open circuit voltage at the moment is the sampling time, is the electrochemical polarization time constant of lithium-ion batteries, Concentration polarization time constant of lithium-ion batteries; The state space is represented as:

[0018]

[0019] in, and for The two internal state variables at time , for The SOC value of the battery at the moment, and for The two internal state variables at time , and is the resistance parameter; set up , , then the parameters to be identified in the electrical model are for: .

[0020] Specifically, the following steps are sampled to obtain the improved robust adaptive multi-forgetting factor recursive least squares method: Data preprocessing, the expression is:

[0021]

[0022] in, and are the normalized input data and output data, respectively, are the original input variables at time t, are the original output variables at time t, and are the mean of input data and output data, respectively, and are the standard deviation of input data and output data, respectively; Collect experimental data, fit SOC-OCV curve, and obtain the open circuit voltage at time t under target temperature through the relationship between SOC-OCV curve and sampling time. are The robust adaptive multiple forgetting factor recursive least squares filter parameter update is:

[0023]

[0024] where, are the gain vector at time t, are the error covariance matrix at time t, is the adaptive forgetting factor, denotes the transpose, are the parameter estimation vector at time t, are the parameter estimation vector at time t, is the gain matrix; The covariance matrix update is:

[0025] where, are the error covariance matrix at time t, is the regularization coefficient, is the identity matrix; The adaptive forgetting factor update is:

[0026] where, and are the minimum and maximum values of the forgetting factor, respectively, is the adjustment coefficient, is the square of the Euclidean norm.

[0027] Based on the above improved robust adaptive multiple forgetting factor recursive least squares method, the internal resistance, capacitance and other parameters of the second-order RC equivalent circuit model can be identified.

[0028] Step S2, a double-state lumped parameter thermal model of the lithium ion battery is established, and the improved goat optimization algorithm is used for thermal model parameter identification.

[0029] The expression of the double-state lumped parameter thermal model of the lithium ion battery is:

[0030]

[0031] wherein, and are the core thermal capacity and the surface thermal capacity of the battery respectively, is the differential of the output temperature , , is the core temperature of the battery, is the ambient temperature, is the differential of the temperature difference , , is the surface temperature of the battery, is the differential of the time , is the heat generation rate of the battery, is the core thermal resistance of the battery, is the surface thermal resistance of the battery. The Laplace transform is used to obtain:

[0032]

[0033] wherein, is a complex frequency variable in the Laplace transform; according to the above formula, when , , , and are determined, the heat generation rate is taken as the input of the model, and and can be outputted, and the to-be-identified parameter in the thermal model is: .

[0034] The improved goat optimization algorithm is used for thermal model parameter identification, and the specific steps include: The difference equations of the discrete are obtained by differentiating and obtained by the Laplace transform:

[0035]

[0036]

[0037]

[0038]

[0039] in, for Output temperature at the moment, for Output temperature at the moment, for Output temperature at the moment, 、 、 、 is the unknown number to be identified related to the thermophysical parameters, for The heat flow at the moment, for The heat flow at the moment, For the adoption time interval; Initialize the position of the goat population, the expression is:

[0040] in, For the The initial value of the individual, and are the lower and upper bounds of the search space, respectively. is a random number; Calculate the fitness value of each goat and determine the initial global optimal position ; For each iteration, the adaptive weight and dynamic search radius are calculated as follows:

[0041]

[0042] in, For the The adaptive weight of the iteration, and are the maximum and minimum values ​​of adaptive weight respectively, is the maximum number of iterations, For the The dynamic search radius of the iteration, is the maximum search radius, is the decay constant; The elite guide position is:

[0043] wherein, is the position of the elite individual in the i th iteration, is the total number of elite individuals, is the position of the j th elite solution in the i th iteration, is the position of the j th elite solution in the i th iteration, is the position of the j th elite solution in the i th iteration, is the position of the j th elite solution in the i th iteration, The position of each goat is updated, and the expression is:

[0044] wherein, is the position of the j th goat in the i th iteration, is the position of the j th goat in the i th iteration, is the position of the j th goat in the i th iteration, is the position of the j th goat in the i th iteration, is the position of the j th goat in the i th iteration, , is a random number, is the global optimal position in the i th iteration, is the position of a randomly selected goat in the i th iteration, is a random number of standard normal distribution, is the elite guide coefficient; When the maximum number of iterations is reached or the convergence condition is met, the algorithm is terminated, the global optimal solution is output, the improved goat optimization algorithm is constructed, and then the improved goat optimization algorithm is used for thermal model parameter identification. Specifically, the voltage, current and temperature data of the lithium ion battery discharge condition are used as identification data input, combined with the heat generation formula and the identified entropy heat coefficient, the thermal physical parameters are identified. Step S3, based on the electrical model and the dual-state lumped parameter thermal model parameters identified in steps S1 and S2, an electro-thermal coupling model is constructed, and based on the electro-thermal coupling model, a minimum entropy adaptive Kalman filter based on the proximal policy optimization algorithm is constructed, and then the SOC and SOT of the battery are jointly estimated. Wherein, step S3 specifically includes:

[0045] The state space equation of the second-order RC equivalent electrical model of the lithium ion battery is constructed and discretized, and the expression is:

[0046]

[0047]

[0048] ​​​

[0049]

[0050]

[0051]

[0052]

[0053] in, for The SOC value of the battery at the moment, is the rated capacity of the battery, for The process noise at the moment, and for The two internal state variables at time , 、 is the polarized capacitance, is the voltage at the observation terminal, for The open circuit voltage that changes with temperature at all times, for The observation noise at time for The state variables of the electrical model at time t, for The input vector of the electrical model at time t, for The observed variables of the electrical model at each moment; The state space equation of the two-state lumped parameter thermal model is constructed and discretized, and the expression is:

[0054]

[0055]

[0056]

[0057]

[0058]

[0059]

[0060] in, for The state variables of the thermal model at each moment, for The state variables of the thermal model at time is the state transition matrix, is the input matrix, is the observation matrix, is the input vector of the thermal model at time t, is the observation variable of the thermal model at time t; By the bidirectional coupling relationship between the electrical model and the thermal model, an electro-thermal coupling model is constructed, and the temperature-dependent parameters of the electrical model are updated synchronously, expressed as:

[0061] wherein, is the total internal heat generation power of the battery at time t, is the ohmic resistance varying with the battery temperature, is the current absolute temperature of the battery, is the partial derivative of the open-circuit voltage with respect to temperature; Then a minimum entropy adaptive Kalman filter based on the proximal policy optimization algorithm is constructed, the flow is: The entropy of the filter is defined as:

[0062]

[0063] wherein, is the entropy at time t, is the logarithmic function with the natural logarithm as the base, is the covariance matrix of the filter, is the observation matrix of the filter, is the error covariance matrix of the state estimation at time t, is the covariance matrix of the measurement noise; The optimization objective is to minimize the cumulative entropy, expressed as:

[0064] wherein, is the optimization objective function, is the total time; The proximal policy optimization algorithm is used to optimize the minimum entropy adaptive Kalman filter, and the adjustment of the hyperparameters of the filter is modeled as a policy , is the probability of selecting an action under the state , including the filter and the observation information, is the parameter adjustment amount, and reward The expression is:

[0065] where, is the smoothing coefficient, and are the state estimation values at time and time ; Objective function of the proximal policy optimization algorithm is:

[0066]

[0067]

[0068] where, is the expected function, represents taking the minimum value, is the importance sampling ratio, is the advantage function, is the clipping function, is a hyperparameter, is the old policy, is the discount factor, is the generalized advantage estimation (GAE) parameter, is the entropy at time Loss of the value function is:

[0069] where, is the value function, is the discounted cumulative return from time to time ; Adding an entropy regularization term in policy optimization , the expression is:

[0070] where, is the entropy coefficient, is the entropy of the policy in state ; Initialize the filter parameters, the policy network and the value network of the proximal policy optimization algorithm, then perform prediction and update of the filter, the expression is:

[0071]

[0072]

[0073]

[0074]

[0075] in, for The state prediction value at the moment, is the state transition matrix, is the forecast error covariance matrix, for The error covariance matrix of the state estimate at the moment, is the process noise covariance matrix, is the Kalman gain, for The actual measured value at the moment, is the identity matrix; Then calculate entropy and reward, and then according to the strategy Select parameter adjustment action ,renew and , then store the data, and finally execute the proximal strategy optimization algorithm to update the objective function and the loss of the value function , thereby constructing a minimum entropy adaptive Kalman filter based on the proximal strategy optimization algorithm, and then jointly estimating the SOC and SOT of the battery.

[0076] It should be noted that the joint estimation proposed in this invention is mainly divided into two parts: SOC estimation and SOT estimation. The SOC estimation part uses the identified open-circuit voltage and electrical model parameters that vary with temperature and SOC. The two state estimation modules are coupled by transferring external characteristic parameters and estimation results. Combined with the electrical model of a second-order RC equivalent circuit, the two-state lumped parameter thermal model state equation, and a minimum entropy adaptive Kalman filter based on a proximal strategy optimization algorithm, the joint estimation of SOC and SOT is achieved.

[0077] Step S4: Based on the joint estimation result in step S3, a deep reinforcement learning optimized adaptive model predictive control algorithm is constructed, and thermal management control of the lithium-ion battery is performed through the deep reinforcement learning optimized adaptive model predictive control algorithm.

[0078] Wherein, step S4 specifically includes: Define the state variables of the electrothermal coupling model , the input vector of the electrothermal coupling model , for Battery temperature at the moment, for Cooling power at all times; Objective function of adaptive model predictive control algorithm for:

[0079] in, is Time-predicted Output at any moment, is the reference value of the output, , is the reference SOC value of the battery, is the battery temperature reference value, is the total prediction step length, is Time-predicted The control input at the moment, is Predicted at all times The control state at all times, is the state reference value, 、 、 are the cost weight matrices of output, control input, and terminal state respectively; The goal of the adaptive model predictive control algorithm is to minimize the objective function , using quadratic programming to minimize the solution, in which deep reinforcement learning is used to dynamically adjust 、 、 and 、 , in order to optimize the control performance, thereby constructing an adaptive model predictive control algorithm optimized by deep reinforcement learning, and then performing thermal management control on lithium-ion batteries, realizing a closed-loop thermal management strategy of real-time state estimation and adaptive optimization control.

[0080] Figure 2 and Figure 3 The proposed method for the joint estimation of SOC and SOT is compared with the traditional extended Kalman algorithm and the true value. The software-in-the-loop test is built by AMESIM software. Figure 2 It can be seen that the method proposed in the present invention significantly improves the estimation accuracy of SOC, with an accuracy increase of about 10%. Figure 3 It can be seen that the method proposed in the present invention significantly improves the estimation accuracy of SOT, with the accuracy increased by about 10%. Figure 4 To compare the temperature of the control method of the present invention with that of the traditional PID control, Figure 4Compared with the traditional PID control algorithm, the temperature control stability of the application is higher.

[0081] In summary, the lithium ion battery thermal management control method based on joint state estimation according to the above embodiment has the following beneficial effects: 1、The present application comprehensively considers the lithium ion battery electrical model and the dual-state lumped parameter thermal model, and the dynamic characteristics of the battery are represented by the second-order RC equivalent electrical model and the dual-state lumped parameter thermal model respectively, and the improved robust adaptive multi-forgetting factor recursive least squares method and the improved goat optimization algorithm are used to improve the parameter identification accuracy, thereby providing a parameter and model basis for SOT and SOC joint estimation. 2、Based on the electrical model and the thermal model of the lithium ion battery, a new electrical-thermal coupling model is constructed, and based on the identified electrical model and dual-state lumped parameter thermal model parameters, a minimum entropy adaptive Kalman filter based on the proximal policy optimization algorithm (PPO) is constructed to jointly estimate SOC and SOT, and solve the state coupling interference. 3、The minimum entropy adaptive Kalman filter based on the proximal policy optimization algorithm of the present application estimates the SOC and the battery temperature in real time, constructs an adaptive model predictive control algorithm optimized by deep reinforcement learning, and performs thermal management control on the lithium ion battery, thereby realizing accurate and effective thermal management control of the lithium ion battery.

[0082] The above-mentioned embodiments only express several embodiments of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent. It should be noted that for ordinary skilled persons in the art, without departing from the concept of the present application, some modifications and improvements can be made, which are all within the scope of the present application. Therefore, the protection scope of the present application patent should be subject to the appended claims.

Claims

1. A lithium-ion battery thermal management control method based on joint state estimation, characterized in that: include: Step S1, establishing a second-order RC equivalent electrical model of a lithium-ion battery, and using an improved robust adaptive multi-forgetting factor recursive least squares method to identify electrical model parameters; Step S2, establishing a dual-state lumped parameter thermal model of the lithium-ion battery, and using an improved goat optimization algorithm to identify the thermal model parameters; Step S3: Based on the electrical model and the two-state lumped parameter thermal model parameters identified in steps S1 and S2, an electrothermal coupling model is constructed, and based on the electrothermal coupling model, a minimum entropy adaptive Kalman filter based on a proximal strategy optimization algorithm is constructed to jointly estimate the SOC and SOT of the battery; Step S4: Based on the joint estimation result in step S3, a deep reinforcement learning optimized adaptive model predictive control algorithm is constructed, and thermal management control of the lithium-ion battery is performed through the deep reinforcement learning optimized adaptive model predictive control algorithm.

2. The lithium-ion battery thermal management control method based on joint state estimation according to claim 1, characterized in that: In step S1, a second-order RC equivalent electrical model of a lithium-ion battery is established, specifically including: The relationship between the voltage and current of the second-order RC equivalent circuit of a lithium-ion battery is expressed as: in, is the operating current, and are the electrochemical polarization voltage and concentration polarization voltage of lithium-ion batteries, and are the electrochemical polarization resistance and concentration polarization resistance of lithium-ion batteries, and are the electrochemical polarization capacitance and concentration polarization capacitance of lithium-ion batteries, for The differential of for The differential of For time The differential of is the open circuit voltage, is the resistance parameter, is the output voltage; The SOC of a lithium-ion battery is expressed as: in, for The SOC value of the battery at the moment, For the initial The SOC value of the battery at the moment, is the Coulomb efficiency, is the maximum available capacity, for arrive The moment between For the time step The current when for The differential of After discretization, we get: in, and They are The electrochemical polarization voltage and concentration polarization voltage of the lithium-ion battery at the moment, and They are The electrochemical polarization voltage and concentration polarization voltage of the lithium-ion battery at the moment, for The current at the moment, for The current at the moment, for The output voltage at the moment, for The open circuit voltage at the moment is the sampling time, is the electrochemical polarization time constant of lithium-ion batteries, Concentration polarization time constant of lithium-ion batteries; The state space is represented as: in, and for The two internal state variables at time , for The SOC value of the battery at the moment, and for The two internal state variables at time , and is the resistance parameter; set up , , then the parameters to be identified in the electrical model are for: 。 3. The lithium-ion battery thermal management control method based on joint state estimation according to claim 2, characterized in that: In step S1, the following steps are sampled to obtain an improved robust adaptive multi-forgetting factor recursive least squares method: Data preprocessing, the expression is: in, and are the normalized input data and output data, for The original input variables at time t, for The original output variable at time t, and are the means of the input data and output data respectively, and are the standard deviations of input and output data respectively; The relationship between SOC-OCV curve and sampling time is used to obtain the target temperature The open circuit voltage at the moment; The robust adaptive forgetting factor recursive least squares filter parameter update is: in, for The gain vector at time t, for The error covariance matrix at time , is the adaptive forgetting factor, represents transpose, for The parameter estimation vector at time , for The parameter estimation vector at time , is the gain matrix; The covariance matrix is ​​updated as: in, for The error covariance matrix at time , is the regularization coefficient, is the identity matrix; The adaptive forgetting factor is updated as: in, and are the minimum and maximum values ​​of the forgetting factor, respectively. is the adjustment coefficient, is the square of the Euclidean norm.

4. The lithium-ion battery thermal management control method based on joint state estimation according to claim 3, characterized in that: In step S2, the expression of the dual-state lumped parameter thermal model of the lithium-ion battery is: in, and are the core heat capacity and surface heat capacity of the battery, Output temperature The differential of , is the core temperature of the battery, is the ambient temperature, Temperature difference The differential of , is the surface temperature of the battery, For time The differential of is the heat generation rate of the battery, is the battery core thermal resistance, is the battery surface thermal resistance; Using Laplace transform we get: in, is the complex frequency variable in Laplace transform; The parameters to be identified in the thermal model are for: 。 5. The lithium-ion battery thermal management control method based on joint state estimation according to claim 4, characterized in that: In step S2, the improved goat optimization algorithm is used to identify the thermal model parameters, which specifically includes: The Laplace transform of and Perform difference processing to obtain the discretized difference equation: in, for Output temperature at the moment, for Output temperature at the moment, for Output temperature at the moment, 、 、 、 is the unknown number to be identified related to the thermophysical parameters, for The heat flow at the moment, for The heat flow at the moment, For the adoption time interval; Initialize the position of the goat population, the expression is: in, For the The initial value of the individual, and are the lower and upper bounds of the search space, respectively. is a random number; Calculate the fitness value of each goat and determine the initial global optimal position ; For each iteration, the adaptive weight and dynamic search radius are calculated as follows: in, For the The adaptive weight of the iteration, and are the maximum and minimum values ​​of adaptive weight respectively, is the maximum number of iterations, For the The dynamic search radius of the iteration, is the maximum search radius, is the decay constant; Elite guide positions are: in, For the The elite individual position of the iteration, is the total number of elite individuals, For the In the iteration an elite solution; Update the position of each goat. The expression is: in, For the Goats in the The position in the iteration, For the Goats in the The position in the iteration, 、 is a random number, For the The global optimal position in the iteration, For the The position of a goat randomly selected in the iteration, is a random number from a standard normal distribution, is the elite guidance coefficient; When the maximum number of iterations is reached Or when the convergence conditions are met, the algorithm terminates and outputs the global optimal solution, completing the construction of the improved goat optimization algorithm, and then the improved goat optimization algorithm is used to identify the thermal model parameters.

6. The lithium-ion battery thermal management control method based on joint state estimation according to claim 5, characterized in that: Step S3 specifically includes: The state space equation of the second-order RC equivalent electrical model of the lithium-ion battery is constructed and discretized, and the expression is: in, for The SOC value of the battery at the moment, is the rated capacity of the battery, for The process noise at the moment, and for The two internal state variables at time , 、 is the polarized capacitance, is the voltage at the observation terminal, for The open circuit voltage that changes with temperature at all times, for The observation noise at time for The state variables of the electrical model at time t, for The input vector of the electrical model at time t, for The observed variables of the electrical model at each moment; The state space equation of the two-state lumped parameter thermal model is constructed and discretized, and the expression is: in, for The state variables of the thermal model at time for The state variables of the thermal model at time is the state transition matrix, is the input matrix, is the observation matrix, for The input vector of the moment thermal model, for Observed variables of the moment thermal model; Through the bidirectional coupling relationship between the electrical model and the thermal model, an electrical-thermal coupling model is constructed, and the temperature-dependent parameters of the electrical model are updated synchronously. The expression is: in, for The total internal heat generation power of the battery at all times, is the ohmic resistance that varies with battery temperature, is the current absolute temperature of the battery, is the partial derivative of the open circuit voltage with respect to temperature; Then, a minimum entropy adaptive Kalman filter based on the proximal policy optimization algorithm is constructed. The process is as follows: The entropy of the filter is defined as: in, for The entropy of the moment, The natural logarithm The logarithmic function with base , is the covariance matrix of the filter, is the measurement matrix of the filter, for The error covariance matrix of the state estimate at the moment, is the covariance matrix of the measurement noise; The optimization goal is to minimize the cumulative entropy, which is expressed as: in, is the optimization objective function, is the total moment; The minimum entropy adaptive Kalman filter is optimized by the proximal strategy optimization algorithm, and the adjustment of the filter's hyperparameters is modeled as a strategy , Is in state Select Action The probability of reward The expression is: in, is the smoothing coefficient, and They are Moment and The estimated value of the state at the moment; Objective function of the proximal policy optimization algorithm for: in, is the expected function, Indicates taking the minimum value, is the importance sampling ratio, is the advantage function, is the clipping function, is a hyperparameter, For the old strategy, is the discount factor, is the generalized advantage estimation parameter, for the entropy of the moment; Loss of value function for: in, is the value function, For Time has come The discounted cumulative return at each moment; Adding entropy regularization to policy optimization , the expression is: in, is the entropy coefficient, For strategy In state Entropy under Initialize the filter parameters, the policy network and the value network of the proximal policy optimization algorithm, and then predict and update the filter. The expression is: in, for The predicted state value at the moment, is the state transition matrix, is the forecast error covariance matrix, for The error covariance matrix of the state estimate at the moment, is the process noise covariance matrix, is the Kalman gain, for The actual measured value at the moment, is the identity matrix; Then calculate entropy and reward, and then according to the strategy Select parameter adjustment action ,renew and , then store the data, and finally execute the proximal strategy optimization algorithm to update the objective function and the loss of the value function , thereby constructing a minimum entropy adaptive Kalman filter based on the proximal strategy optimization algorithm, and then jointly estimating the SOC and SOT of the battery.

7. The lithium-ion battery thermal management control method based on joint state estimation according to claim 6, characterized in that: Step S4 specifically includes: Define the state variables of the electrothermal coupling model , the input vector for the electrothermal coupling model , for Battery temperature at the moment, for Cooling power at all times; Objective function of adaptive model predictive control algorithm for: in, is Predicted at all times Output at any moment, is the reference value of the output, , is the reference SOC value of the battery, is the battery temperature reference value, is the total prediction step length, is Time-predicted The control input at the moment, is Time-predicted The control state at all times, is the state reference value, 、 、 are the cost weight matrices of output, control input, and terminal state respectively; The goal of the adaptive model predictive control algorithm is to minimize the objective function , using quadratic programming to minimize the solution, in which deep reinforcement learning is used to dynamically adjust 、 、 and 、 , in order to optimize the control performance, thus constructing an adaptive model predictive control algorithm optimized by deep reinforcement learning, and then performing thermal management control of lithium-ion batteries.

Images