Battery thermal management system control method based on multi-mode switching and predictive optimization

Abstract

The invention provides a battery thermal management system control method based on multi-mode switching and predictive optimization, and the method comprises the following steps: building a discrete time state space model based on the physical structure of a battery module; performing first-order Taylor expansion at a nominal working point for a nonlinear coupling term in the discrete time state space model so as to obtain a linear prediction model adapted to online optimization; designing a model prediction controller based on the linear prediction model; a dual-threshold trigger logic based on an instantaneous temperature tracking error and a battery output power demand is defined, so that a battery thermal management mode is adaptively switched. According to the method, the prediction controller based on the linear dynamic state space model is established, and a mode switching mechanism based on temperature error and battery power dual-threshold triggering is introduced, so that the model prediction controller can realize accurate tracking and quick response of the battery temperature, the health state of the battery is improved, and the service life of the battery is prolonged.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of battery thermal management, in particular to a battery thermal management system control method based on multi-mode switching and predictive optimization. BACKGROUND

[0002] With the rapid development of electric vehicles and large-scale energy storage applications, the thermal safety and efficiency of battery systems have become increasingly prominent. Lithium-ion batteries generate nonlinear and fluctuating heat during charging and discharging, which can lead to performance degradation, shortened life, and even safety hazards if not properly cooled or regulated. Existing thermal management methods rely on single proportional-integral-derivative controllers or empirical rule-based control strategies, which can maintain basic temperatures in static conditions but often have limited control bandwidth, regulation lag, and low energy efficiency in dynamic loads and complex environments.

[0003] In recent years, some studies have begun to use predictive optimization control frameworks to improve the accuracy and real-time performance of battery temperature regulation. This method uses rolling prediction of system state and optimization calculation to obtain control input, which can improve dynamic performance to some extent. However, traditional predictive control mostly uses a single mode, which cannot flexibly switch and schedule logic when the battery is in different power output or environmental conditions, making it difficult to balance energy efficiency and stability. Therefore, how to achieve adaptive control mode adjustment in diverse conditions has become a key challenge to improve battery thermal management performance. SUMMARY

[0004] To solve the problems in the prior art, the present application provides a battery thermal management system control method based on multi-mode switching and predictive optimization, aiming to improve the safety and operational economy of the battery system as a whole.

[0005] The battery thermal management system control method based on multi-mode switching and predictive optimization includes the following steps:

[0006] Step 1: Based on the physical structure of the battery module, a discrete-time state-space model is constructed, which considers each row of batteries and their air ducts as a thermal node. The temperature state vector includes the battery temperature and air temperature at the thermal node, and the control input vector includes the temperature difference between the inlet air temperature and the ambient temperature and the reciprocal of the inlet air flow rate.

[0007] Step 2: For the nonlinear coupling term in the discrete-time state-space model, which is the product of the reciprocal of the inlet air flow rate and the temperature state vector and the heat generation disturbance term, a first-order Taylor expansion is performed at the nominal operating point to convert it into a linear affine function of the reciprocal of the inlet air flow rate and the heat generation disturbance term, thereby obtaining a linear prediction model suitable for online optimization.

[0008] Step 3: Design a model predictive controller based on the linear prediction model; the model predictive controller performs the following rolling optimization procedure at each control period: predict the future temperature sequence according to the current system state, obtain the optimal control increment sequence in the future control time domain by solving a quadratic programming problem that optimizes the temperature tracking error and the control input variation at the same time, and only apply the first control increment in the sequence to update the actual control input, while introducing a feedback correction mechanism to compensate for model errors and unknown disturbances;

[0009] Step 4: Define a double-threshold trigger logic based on the instantaneous temperature tracking error and the battery output power demand, so that the drive system can adaptively switch between the shutdown mode, the MPC low-speed control mode and the MPC high-speed control mode; wherein the MPC low-speed control mode and the MPC high-speed control mode are distinguished by setting different control input constraint ranges.

[0010] Further, in step 1, the discrete-time state-space model is:

[0011]

[0012] wherein, is the temperature state vector, is the control input vector, is the output vector, is the disturbance term;

[0013] For the discrete-time state-space model, the temperature state vector is augmented to a new vector containing the original state increment and system output, and the control input vector is converted to a control input increment vector, so that the discrete-time state-space model is converted to a standard increment model form suitable for model predictive control, i.e.

[0014] ;

[0015] ;

[0016] wherein, is the augmented temperature state vector, represents the control input increment; is an order identity matrix, is a zero matrix of appropriate dimension; the matrix represents the coupling relationship between the air nodes and the battery nodes; represents the influence of the control input vector on the discrete-time state-space model; represents the extraction of the measurable output temperature from the temperature state vector; is the system period.

[0017] Further, in step 2, the nominal operating point includes the nominal airflow reciprocal, nominal battery temperature, nominal air temperature, and nominal battery heat generation rate. These nominal values ​​are dynamically determined based on the battery's historical operating data or real-time estimates; the first-order Taylor expansion is:

[0018] ;

[0019] in, It is the reciprocal of the nominal airflow volumetric flow rate at the inlet. and They are the first The nominal air temperature and nominal battery temperature of the outlet. The nominal battery heat generation rate; For the first The first in the cycle The temperature of the exhaust air, The convective heat transfer coefficient is... For effective heat exchange area, For the first The first in the cycle The battery temperature of the outlet, It is the reciprocal of the inlet air volumetric flow rate. For battery heat generation rate, , These are the specific heat capacity and density of air, respectively. , These represent the battery's specific heat capacity and mass, respectively.

[0020] Further, in step 3, the feedback correction mechanism specifically involves treating the heat generated inside the battery as an unobservable disturbance term, and using the deviation between the actual measured value of the battery temperature and the model prediction output value in each control cycle to dynamically adjust the internal state of the model prediction controller or compensate for future predictions.

[0021] The dynamically adjusted model prediction controller is:

[0022]

[0023] This refers to the temperature difference between the inlet air temperature and the ambient temperature.

[0024] Furthermore, the objective function of the quadratic programming problem is specifically:

[0025] ;

[0026] in, To predict the output, For reference temperature, To control the input increment, a weight factor; a prediction horizon, a control horizon; the optimization is performed subject to linear inequality constraints on the control inputs and their increments.

[0027] Further, the specific decision rule of the double threshold trigger logic is:

[0028] when the system is in the OFF mode, if the temperature tracking error exceeds a first error threshold, or the battery output power demand exceeds a first power threshold, switch to the MPC high-speed control mode;

[0029] when the system is in the MPC high-speed control mode, if the temperature tracking error is below a second error threshold, and the battery output power demand is below a second power threshold, switch to the MPC low-speed control mode;

[0030] when the system is in the MPC low-speed control mode, if the temperature tracking error is below a third error threshold, and the battery output power demand is below a third power threshold, switch to the OFF mode.

[0031] Further, the first error threshold is 0.3 K, the second error threshold is 0.1 K, and the third error threshold is 0.05 K; the first power threshold and the second power threshold are both 40 kW, and the third power threshold is 20 kW.

[0032] Further, the control input constraint range corresponding to the MPC high-speed control mode is: inlet air flow [0, 20] CFM, and inlet air temperature [283, 303] K; the control input constraint range corresponding to the MPC low-speed control mode is: inlet air flow [0, 10] CFM, and inlet air temperature [293, 303] K.

[0033] A battery thermal management system, comprising a memory, a processor, and a computer program stored on the memory, wherein the processor implements the method as described above when executing the computer program.

[0034] An electric vehicle comprising the battery thermal management system as described above.

[0035] The beneficial effects of the present application: the present application establishes a predictive controller based on a linearized dynamic state space model, and introduces a mode switching mechanism based on temperature error and battery power double threshold triggering, so that the model predictive controller (MPC) can realize accurate tracking (average absolute error <0.5K) and fast response of battery temperature; and provides reliable performance guarantee for the introduction of the double threshold mode switching mechanism, so that the system has the courage to switch to the non-MPC mode or the low speed mode when the error and the power meet the conditions, thereby directly achieving the effect of significantly reducing the system energy consumption (experimental verification energy consumption reduction 8.6%); and further improves the battery state of health (SOH) and prolongs the service life. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 Flowchart of the present application;

[0037] Figure 2 Battery module and heat dissipation structure schematic diagram of the embodiment of the present application;

[0038] Figure 3 Model predictive control prediction time domain and control time domain relationship schematic diagram;

[0039] Figure 4 Temperature following effect diagram under constant reference temperature;

[0040] Figure 5 Temperature following effect comparison diagram before and after adding mode switching mechanism;

[0041] Figure 6 MPC control effect diagram with optimized path as dynamic reference temperature;

[0042] Figure 7 Improved MPC-MS strategy output control path diagram;

[0043] Figure 8 Comparison diagram of MPC and other control strategies on BTMS control effect. DETAILED DESCRIPTION

[0044] The present application will be described in detail below with reference to the accompanying drawings. The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the 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 with reference to the drawings are exemplary and are only used to explain the present application, and cannot be interpreted as a limitation on the present application. The left, middle, right, top, bottom and other orientation terms in the embodiments of the present application are only relative concepts or are with reference to the normal use state of the product, and should not be considered as limiting.

[0045] Battery thermal management system control method based on multi-mode switching and predictive optimization, as shown in Figure 1 comprises the following steps:

[0046] Step 1: Based on the physical structure of the battery module, a discrete-time state space model is constructed, which regards each row of battery and its air duct as a thermal node, and the temperature state vector includes the battery temperature and air temperature at the thermal node, and the control input vector includes the temperature difference of the inlet air temperature relative to the ambient temperature and the reciprocal of the inlet air flow volume; Specifically:

[0047] Step 1.1: On the battery test platform, the application collects thermal characteristic data such as battery surface temperature, current, voltage and cooling air volume by performing constant current charge and discharge experiments on power batteries under different state of charge and ambient temperature; These data help identify key thermal parameters such as battery heat capacity, heat transfer coefficient, flow passage cross-sectional area and fluid density, providing accurate inputs for subsequent modeling and control;

[0048] Step 1.2: Based on the collected data, the application designs a dynamic state space model to describe the temperature change of the battery; The state variable of the model is the internal temperature of the battery, the control input is the cooling air volume and the inlet air temperature, and the output is the battery surface temperature; The internal heat generation of the battery changes dynamically with the battery operating conditions (such as driving conditions), and a prediction model needs to be established according to the thermal characteristics of the battery. In order to capture these dynamic characteristics, the application uses a lumped parameter thermal network model; The discrete-time state space model is:

[0049] (1)

[0050] Where, is the temperature state vector, is the control input vector, is the output vector, is the disturbance term;

[0051] Where, on the basis of model simplification and linearization, the system described by the application is a discrete controlled object of the battery pack thermal process. In order to describe the dynamic behavior of the system, a discrete-time state space model is established, and the state matrix , the input matrix and the output matrix are extracted, and the state increment form is introduced in the control implementation to ensure that the system has good controllability and observability, laying a foundation for the design of the subsequent rolling optimization controller.

[0052] Under the above assumptions and simplifications, the discrete-time model of the system can be further represented in the form of state space; According to the above content, the system matrix can be represented as:

[0053] (2)

[0054] (3)

[0055] (4)

[0056] where, is the convective heat transfer coefficient, is the effective heat transfer area; , , are the specific heat capacity, density and inlet volumetric flow rate of air, respectively; , are the specific heat capacity and mass of the battery, respectively; is the nominal air flow rate reciprocal of the inlet air flow volumetric flow rate, and are the nominal air temperature and nominal battery temperature of the first row, respectively, is the nominal battery heat generation rate;

[0057] In order to introduce the input increment term in MPC modeling , the discrete-time state-space model is reconfigured by augmenting the temperature state vector into a new vector containing the original state increment and system output, and transforming the control input vector into a control input increment vector, so as to convert the discrete-time state-space model into a standard increment model form suitable for model predictive control, i.e.

[0058] (5)

[0059] (6)

[0060] where, is the augmented temperature state vector, denotes the control input increment; is the order identity matrix, is a zero matrix of appropriate dimension; the matrix denotes the coupling relationship between the air nodes and the battery nodes; denotes the influence of the control input vector on the discrete-time state-space model; denotes the extraction of the measurable output temperature from the temperature state vector; is the system period;

[0061] Step 2: For the nonlinear coupling term formed by multiplying the inverse of the inlet airflow volumetric flow rate with the temperature state vector and the heat generation perturbation term in the discrete-time state-space model, a first-order Taylor expansion is performed at the nominal operating point to transform it into a linear affine function of the inverse of the inlet airflow volumetric flow rate and the heat generation perturbation term, thereby obtaining a linear prediction model suitable for online optimization; the nominal operating point includes the nominal airflow inverse, nominal battery temperature, nominal air temperature, and nominal battery heat generation rate, which are dynamically determined based on the battery's historical operating data or real-time estimates;

[0062] Among them, the battery module and heat dissipation structure, such as Figure 2 As shown, the battery module consists of 32 cylindrical cells arranged in 8 rows, with 4 cells in each row. The spacing between battery cells and between battery cells and module walls is 2mm; the cooling air entering the module ( The cells are uniformly distributed in the flow channel, with each row of cells coupled to its corresponding cooling airflow channel to achieve heat exchange. Over 90% of the heat generated by the battery is carried away by convection, while the contributions of conduction and radiation are negligible. Therefore, in the modeling process of this invention, only the convective heat transfer between the battery and the airflow is considered. This invention introduces air temperature... The thermal coupling relationship between the battery cell temperature and the internal temperature establishes the interaction between airflow and heat generation in the battery cell. Figure 2 middle The inlet volumetric flow rate is denoted as . Heat generation is caused by the battery's internal resistance, while heat dissipation is achieved through conduction with external fluids via cooling ducts. Based on this, the present invention constructs the following coupling equation, which details the heat exchange process between air and the battery cell.

[0063] Step 2.1: When the inlet airflow rate When ≠0, the heat generation and dissipation model of the battery can be described in differential form as follows:

[0064] (7)

[0065] in, Indicates the first Temperature of each airflow channel Indicates the first The temperature of the battery pack; The convective heat transfer coefficient is... For effective heat exchange area; , , These are the specific heat capacity, density, and inlet volumetric flow rate of air, respectively. , These are the battery's specific heat capacity and mass, respectively. The internal heat generation rate of the battery;

[0066] After the model is discretized by Euler formula, the discrete-time expression of the system can be obtained:

[0067] (8)

[0068] where, is the reciprocal of the inlet air volume flow; the reciprocal term is introduced to simplify the expression related to the air flow and facilitate the subsequent modeling and solution of the model predictive controller (MPC);

[0069] In the implementation process of the controller, the input vector is defined as follows:

[0070] (9)

[0071] where, is the temperature difference of the inlet air temperature relative to the ambient temperature, ;

[0072] The state vector is defined as follows:

[0073] (10)

[0074] Step 2.2: Linearization processing:

[0075] In order to solve the nonlinear coupling problem between the cooling air volume and the heat production, the first-order Taylor expansion method is adopted to linearize the state space model; this processing method effectively improves the convergence and real-time executability of the MPC algorithm, and ensures the feasibility in actual control;

[0076] For the coupling terms , and , the system state and control input are coupled, resulting in nonlinear dynamics of the system; in order to make the optimization problem solvable, the first-order Taylor expansion is adopted. Assuming that the values of , , and at time are , , and respectively, the first-order Taylor expansion is:

[0077]

[0078] where, is the nominal air flow reciprocal of the inlet air volume flow, and are the nominal air temperature and the nominal battery temperature of the th row, is the nominal battery heat generation rate; are the battery temperature and the air temperature of the th row in the th period, is the convective heat transfer coefficient, is the effective heat transfer area, is the battery temperature of the th row in the th period, is the inverse of the inlet air flow volume flow rate, is the battery heat generation rate, , are the specific heat capacity and the density of air, respectively, , are the specific heat capacity and the mass of the battery, respectively;

[0079] This linearization form decouples the original bilinear term into an affine function with respect to the air flow related inputs and the disturbance term ; therefore, while maintaining sufficient accuracy around the selected operating point, the computational burden of solving the MPC optimization problem is greatly reduced;

[0080] To improve the computational feasibility of the model, several simplifying assumptions are made during the establishment of the thermal model:

[0081] Air flow distribution assumption: it is assumed that the air flow inside the module is uniformly distributed in each parallel cooling channel;

[0082] Battery node assumption: it is assumed that the internal heat conduction speed of each row of batteries is fast, so the entire row of batteries can be regarded as a single concentrated heat node;

[0083] Heat exchange mechanism assumption: since convective heat transfer dominates under the studied operating conditions, the influence of radiative heat transfer is ignored;

[0084] Disturbance processing assumption: the internal heat generation term of the battery is difficult to measure accurately in real time, and its change is strongly dependent on uncertain factors such as load characteristics and aging state;

[0085] Based on this, the disturbance dynamic term is not explicitly introduced in the nominal prediction model, and the error caused by the mismatch of the model is compensated by the subsequent closed-loop feedback correction mechanism; these assumptions, while simplifying the model implementation and ensuring computational feasibility, inevitably introduce deviations between the model prediction and the real system response, which is the research motivation for the MPC and mode switching strategy based on feedback correction proposed later;

[0086] Step 3: Design a model predictive controller based on the linear prediction model; the model predictive controller performs the following rolling optimization process at each control period: predict the future temperature sequence according to the current system state, obtain the optimal control increment sequence in the future control time domain by solving a quadratic programming problem that optimizes the temperature tracking error and the control input variation at the same time, and only apply the first control increment in the sequence to update the actual control input, while introducing a feedback correction mechanism to compensate for model errors and unknown disturbances; specifically:

[0087] Step 3.1: Based on the constructed linear prediction model, the present application proposes a model predictive controller based on rolling optimization and feedback correction mechanism, which is used to realize dynamic and accurate control of battery temperature; the controller can automatically adjust the control input under different loads and working conditions to ensure temperature accuracy and energy efficiency optimization;

[0088] After the thermal model is established, the MPC framework is applied to the regulation of battery temperature, and dynamic control is achieved through recursive prediction; at each sampling time, the controller will solve a finite time domain optimization problem, but only implement the first control amount of the calculated control sequence; then, the optimization is re-executed according to the latest state measurement at the next time, thereby realizing closed-loop adaptive control; specifically:

[0089] After the linear prediction model is established, the output variable, battery temperature, needs to be optimized under the recursive prediction framework; let the control time domain be , and the prediction time domain be ; then the control vector and the prediction vector are defined as follows:

[0090] (12)

[0091] Where, represents the input increment at time step , corresponding to the changes of the inlet air flow reciprocal and the inlet air temperature .

[0092] In the prediction time domain, the output prediction sequence of the discrete-time state-space model can be represented as:

[0093] (13)

[0094] Where, is the stacked vector of predicted outputs, is the free response matrix related to the system dynamics, and is the dynamic matrix, which describes the relationship between input increments and future outputs;

[0095] The definitions of these matrices are as follows:

[0096] (14)

[0097] (15)

[0098] Step 3.2: Prediction horizon The selection of the control horizon has a significant impact on both computational complexity and control performance; a longer prediction horizon helps to better predict the system dynamics, thereby reducing the risk of constraint violation and control lag; however, it also significantly increases the computational burden;

[0099] Since only the first component of the control sequence is implemented at each sampling time, the control horizon should be kept within a moderate range to avoid unnecessary computational overhead; in general, the control horizon is selected to satisfy , as shown in Figure 3 .

[0100] Step 3.3: Optimization objective function and solving method:

[0101] The objective function set by the present application considers two aspects:

[0102] (1) Reducing the deviation between the battery temperature and the desired reference trajectory;

[0103] (2) Suppressing rapid fluctuations in control input and reducing fan and actuator energy consumption;

[0104] The objective function adopts a quadratic programming (QP) form, which balances temperature control accuracy and energy consumption through weighting coefficients , and calculates the control input online by a real-time optimization solver;

[0105] In the feedback correction phase, the quadratic programming problem is solved with the control increment as the decision variable; the general form of quadratic programming can be represented as:

[0106] (16)

[0107] (17)

[0108] where is the Hessian matrix, is the linear cost vector, and the constraint condition defines the feasible region of the system;

[0109] Unlike traditional on-off control, MPC requires a temperature reference trajectory to be set in advance. This trajectory is planned in advance according to traffic conditions, environmental factors and other related variables; the optimization objective of MPC has two, one is to make the output in the prediction time domain as close as possible to the reference trajectory ; the second is to make the increment of the control as small as possible ; the objective function (penalty function) of the quadratic programming problem is as follows:

[0110] (18)

[0111] where, is the predicted output, is the reference temperature, is the control input increment, is the weight factor; the optimization is carried out under the condition of satisfying the linear inequality constraint of the control input and its increment; formula (18) is the "arithmetic expression" of the objective function, which shows its physical meaning (penalty on error and control increment at each time point), and formula (19) is the "compact matrix form" of the objective function, which is the mathematical form used for controller design and actual solution;

[0112] The corresponding objective function can be expressed in matrix form as follows:

[0113] (19)

[0114] where, and are the stacked prediction vector and the reference vector respectively, is the stacked control increment sequence; by substituting the prediction model (formula (13)-(15)) into the cost function (formula (19)), the optimization problem can be rewritten into the standard quadratic programming form:

[0115] (20)

[0116] where:

[0117] (21)

[0118] (22)

[0119] where, is a unit matrix with appropriate dimensions;

[0120] Step 3.4: Modeling of control input constraints:

[0121] To ensure the physical feasibility of system operation, the present application sets input constraints and incremental constraints for cooling air volume and inlet air temperature respectively; the constraint forms are: (1) input boundary constraint: ensure that the air volume and air temperature are within the allowed range; (2) incremental boundary constraint: limit the change rate of air volume and air temperature to avoid excessive fluctuation. The double constraint design ensures the stability and executability of MPC optimization;

[0122] The following constraint conditions are imposed on the input variables:

[0123] The input variables of the system and the control increments all need to be subjected to constraint conditions to ensure that the control behavior meets the physical and safety limits:

[0124] (23)

[0125] These constraint conditions can be rewritten as the following linear inequality form:

[0126] (24)

[0127] Wherein, is a lower triangular matrix used for cumulative summation of control input increments; is an identity matrix used to represent independent constraints on each input increment;

[0128] Therefore, the constraint conditions can be further written in the form of standard quadratic programming:

[0129] (25)

[0130] At each control step, solving the quadratic programming problem composed of equations (20-22) and (25) can obtain the optimal control increment sequence ;

[0131] Step 3.5: Rolling solution and feedback update:

[0132] Within the control period, the present application solves the optimization problem in real time, and only the first control input is used to update the system, and the remaining solution vectors are recalculated in the next period; preferably, a feedback correction mechanism is introduced in each period to correct the difference between prediction and reality; in this way, the present application forms a closed-loop rolling optimization structure to ensure the dual optimization of temperature control accuracy and energy consumption control;

[0133] By solving the optimization problem in real time, the control input is updated, and the next control period is entered. The whole process forms a closed-loop control mechanism to ensure the optimization of temperature accuracy and energy efficiency;

[0134] After converting the above optimization problem into a standard quadratic programming problem, a quadratic programming solver is called to obtain the control increment sequence in real time ; only the first control increment is kept for current control input update:

[0135] (26)

[0136] In actual implementation, only the first element of the sequence is taken to update the current control input:

[0137] (27)

[0138] where, denotes the optimal control increment sequence obtained by solving the quadratic programming problem, only the first component of which is implemented in each control period ; this receding predictive control strategy can ensure the closed-loop stability and adaptivity of the system under dynamic operating conditions;

[0139] Then the system enters the next control period, and the above prediction, optimization and update process is repeated to form a complete rolling closed-loop control mechanism;

[0140] Through the rolling optimization controller and feedback correction mechanism constructed by this step, the temperature tracking accuracy can be guaranteed while the control input constraints and energy consumption can be effectively controlled, laying a precise and efficient control foundation for the subsequent introduction of the mode switching mechanism;

[0141] The feedback correction mechanism is specifically: the heat generation inside the battery is regarded as an unobservable disturbance term, and the deviation between the actual measured value and the model predicted output value of the battery temperature in each control period is used to dynamically adjust the internal state of the model predictive controller or compensate for future prediction;

[0142] The dynamically adjusted model predictive controller is:

[0143] (28)

[0144] where, is the temperature difference between the inlet air temperature and the relative ambient temperature; in this model, the nonlinear disturbance related term has been removed, and the deviation correction caused by modeling error is completed by the closed-loop MPC framework;

[0145] Step 4: define a double-threshold trigger logic based on instantaneous temperature tracking error and battery output power demand, so that the drive system can adaptively switch between the off mode, MPC low-speed control mode and MPC high-speed control mode; wherein the MPC low-speed control mode and the MPC high-speed control mode are distinguished by setting different control input constraint ranges; specifically, the specific determination rule of the double-threshold trigger logic is:

[0146] ​When the system is in the off mode, if the temperature tracking error exceeds a first error threshold, or the battery output power demand exceeds a first power threshold, switch to the MPC high-speed control mode;

[0147] When the system is in the MPC high-speed control mode, if the temperature tracking error is lower than a second error threshold, and the battery output power demand is lower than a second power threshold, switch to the MPC low-speed control mode;

[0148] When the system is in the MPC low-speed control mode, if the temperature tracking error is lower than a third error threshold, and the battery output power demand is lower than a third power threshold, switch to the off mode;

[0149] The first error threshold is 0.3 K, the second error threshold is 0.1 K, and the third error threshold is 0.05 K; the first power threshold and the second power threshold are both 40 kW, and the third power threshold is 20 kW;

[0150] The control input constraint range corresponding to the MPC high-speed control mode is: inlet air volume [0, 20] CFM, and inlet air temperature [283, 303] K; the control input constraint range corresponding to the MPC low-speed control mode is: inlet air volume [0, 10] CFM, and inlet air temperature [293, 303] K.

[0151] A battery thermal management system includes a memory, a processor, and a computer program stored on the memory, wherein the processor executes the computer program to implement the method described above.

[0152] An electric vehicle includes the battery thermal management system described above.

[0153] Simulation verification and MPC strategy performance analysis:

[0154] Due to the unpredictability of future operating conditions and the absence of future vehicle speed prediction in the model, there will inevitably be some deviation between the system output and the reference value. Although the MPC algorithm has a feedback correction mechanism, it can respond to reference value changes by adjusting the control input increment, but this feedback process usually has a certain time lag. To evaluate the control effect of the MPC strategy in actual operation, this section further analyzes the dynamic tracking ability of the algorithm output to the reference path, focusing on the response and timeliness and accuracy of the system under different operating conditions. This analysis aims to verify the online control performance and practical application value of the proposed control strategy.

[0155] Selection and definition of error evaluation indicators:

[0156] To quantitatively evaluate the temperature tracking performance of MPC, the present application selects three commonly used error evaluation indexes: mean absolute error (MAE, Mean Absolute Error), root mean square error (RMSE, Root Mean Square Error), and mean relative error (MRE, Mean Relative Error) to measure the deviation between the output and the reference value; the calculation formula is as follows:

[0157] (29)

[0158] (30)

[0159] (31)

[0160] wherein, is the output value of the MPC, is the reference value, is the current time step, is the total number of steps for the entire working condition.

[0161] Constant reference temperature verification:

[0162] Under the condition of constant reference temperature, the present application verifies the MPC strategy and compares it with the traditional control method. In this verification, the initial inlet gas flow of the model is set to . The simulation is simulated at 298 K, 303 K and 308 K as the highest temperature reference path. The temperature tracking performance of the MPC is shown in Figure 4 , and the corresponding temperature tracking error is shown in Table 1;

[0163] Table 1 Tracking error under constant reference temperature

[0164]

[0165] According to the calculation results in Table 1, it can be seen that after the temperature first reaches the target temperature, the system following the three targets can achieve temperature following effect, the average absolute error MAE of the following error is less than 0.6 K, the root mean square error RMSE is less than 1.6 K, and the following effect of 303 K and 298 K is better than that of 308 K;

[0166] Compared with the case of following the target temperature of 298 K, although there is a temperature difference of 5 K between the initial temperature and the target temperature, since ABTMS only cools the inlet air and cannot heat it, the reference temperature of 298 K is cooled by quickly opening the active BTMS to cool the battery, so that the target temperature is reached faster. However, in the case of 308 K, the system can only heat up to the target temperature by itself;

[0167] For the case of following the reference temperature of 308 K, during the slow heating of the system to reach 308 K for the first time, since the model needs to control the increment of the variable and the error between the output and the reference value to be as small as possible, the MPC algorithm always keeps the inlet air flow

[0168] On this basis, after adding the mode switching mechanism, the temperature following effect is further optimized; as shown in Figure 5 Since the target temperature to be followed is higher than the actual output temperature and the input power required by the motor is less than the set value, the control mode remains in the BTMS off state, and the battery naturally heats up to reach 308 K for the first time. After the system is basically stable at 308 K, the maximum deviation is less than 0.2 K. The improved MPC algorithm reduces the average absolute error MAE, root mean square error RMSE and average relative error MRE of temperature following by 74.64%, 73.51% and 74.6% respectively. Through mode switching, MPC-MS (MPC with mode switching) can maintain the stability of the system at high temperature, reduce energy consumption, and improve cooling efficiency at low temperature.

[0169] Dynamic temperature tracking verification steps:

[0170] Further, the present application introduces dynamic reference temperature verification; in this verification, hybrid dynamic programming and genetic algorithm (DP-GA) are used for multi-objective optimization to determine the dynamic reference temperature as the control target; during simulation, MPC adjusts the control input according to the change of battery temperature and the reference trajectory, and performs real-time prediction and optimization; the MPC dynamic temperature tracking performance and the corresponding control output trajectory are shown in Figure 6

[0171] As shown in Figure 6 ​​As shown, since the model needs to control the increment of the variable and the error between the output and the reference value as small as possible, the result of the algorithm control is that the tracking error is less than 0.1 K during 0~750 s , which shows that the selection of the initial value has an impact on the control effect; during 3000~4000 s, the maximum error of the battery temperature is not more than 0.1 K, and the maximum error is not more than 0.1 K; at this time, the power required by the motor is at a relatively low level, and it can be seen that the global optimization control strategy is to close the air conditioner, and the fan is mostly in the closed state, and is intermittently opened at a low wind speed, which shows that the cooling capacity required by the battery system at this time is small; the MPC maintains the fan at a medium wind speed, which provides a larger cooling capacity and increases the heat generation of the battery, and the two are balanced, so that the control temperature and the reference temperature remain basically the same, which obviously consumes more energy;

[0172] The control effect of MPC-MS after adding the mode switching module is shown in Figure 7 ; the simulation starts in the ABTMS closed state, and the tracking error is greater than the set value 0.1 K at 749 s, but the output power required by the battery is less than the set value 40 kW, and the MPC low speed mode is switched in; the error is further increased to 0.3 K at 900 s, and the MPC high speed mode is switched in; at 1350 s, the tracking error is less than the set value 0.1 K and the output power required by the battery is less than the set value 40 kW, which shows that the heat generation of the system will not increase significantly in a short time, and the MPC high speed mode is switched to the MPC low speed mode; then enter the US06 supplementary federal test procedure (a kind of aggressive driving cycle simulating high speed and high acceleration) working condition, the output power required by the battery exceeds the set value 40 kW, the system switches back to the high speed mode, and the BTMS closed model is switched from the MPC low speed mode when the tracking error is less than a certain value and the required output power in the front and rear time is not large.

[0173] Table 2 tracking error under dynamic reference temperature

[0174]

[0175] Table 2 compares the temperature tracking error of the original MPC strategy and the improved MPC-MS strategy, and it can be seen that the addition of the mode switching module has little effect on the tracking error; but during 3150~4000 s, the system is switched out of the MPC control mode; compared with the MPC algorithm, the MPC-MS strategy has certain improvement in the energy consumption of the BTMS in a single working condition, as shown in Figure 8As shown, the energy consumption of the active BTMS single working condition is reduced by 1.08 Wh, about 8.6%, and the minimum SOH of the battery is increased from 0.8104 to 0.8118, the maximum capacity loss of the battery is reduced by 1.1%, and the consistency change is small.

[0176] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above examples, and the above examples and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A battery thermal management system control method based on multi-mode switching and predictive optimization, characterized in that, The method comprises the following steps: Step 1: based on the physical structure of the battery module, a discrete-time state-space model is constructed, in which each row of batteries and its air duct is regarded as a thermal node, the temperature state vector of the thermal node comprises the battery temperature and the air temperature at the thermal node, and the control input vector comprises the temperature difference between the inlet air temperature and the ambient temperature and the reciprocal of the inlet air flow volume; Step 2: for the nonlinear coupling term in the discrete-time state-space model, which is formed by the multiplication of the reciprocal of the inlet air flow volume, the temperature state vector and the heat generation disturbance term, a first-order Taylor expansion is performed at the nominal operating point to convert it into a linear affine function about the reciprocal of the inlet air flow volume and the heat generation disturbance term, thereby obtaining a linear prediction model suitable for online optimization; Step 3: based on the linear prediction model, a model predictive controller is designed; the model predictive controller performs the following rolling optimization process at each control period: predicting the future temperature sequence according to the current system state, obtaining the optimal control increment sequence in the future control time domain by solving a quadratic programming problem that simultaneously optimizes the temperature tracking error and the control input change, and only applying the first control increment in the sequence to update the actual control input, while introducing a feedback correction mechanism to compensate for model errors and unknown disturbances; Step 4: a double-threshold trigger logic based on the instantaneous temperature tracking error and the battery output power requirement is defined to enable the driving system to adaptively switch between the shutdown mode, the MPC low-speed control mode and the MPC high-speed control mode; wherein the MPC low-speed control mode and the MPC high-speed control mode are distinguished by setting different control input constraint ranges.

2. The battery thermal management system control method based on multi-mode switching and predictive optimization of claim 1, wherein: In step 1, the discrete-time state-space model is: ； wherein is a temperature state vector, is a control input vector, is an output vector, is a disturbance term; The discrete-time state-space model is reconstructed by dimension expansion to augment the temperature state vector into a new vector containing the original state increment and system output, and to convert the control input vector into a control input increment vector, thereby converting the discrete-time state-space model into a standard incremental model form suitable for model predictive control, i.e. ； ； wherein, is the extended temperature state vector, denotes the control input increment; is the is the identity matrix of order is a zero matrix of appropriate dimensions; the matrix denotes the coupling between the air nodes and the battery nodes; denotes the influence of the control input vector on the discrete-time state-space model; denotes the extraction of the measurable output temperatures from the temperature state vector; is the system period.

3. The battery thermal management system control method based on multi-mode switching and predictive optimization of claim 1, wherein: In step 2, the nominal operating point includes the nominal air flow reciprocal, the nominal battery temperature, the nominal air temperature and the nominal battery heat generation rate, which are dynamically determined based on historical operation data or real-time estimated values of the battery; the first-order Taylor expansion is: ； in, It is the reciprocal of the nominal airflow volumetric flow rate at the inlet. and They are the first The nominal air temperature and nominal battery temperature of the outlet. The nominal battery heat generation rate; For the first The first in the cycle The temperature of the exhaust air, The convective heat transfer coefficient is... For effective heat exchange area, For the first The first in the cycle The battery temperature of the outlet, It is the reciprocal of the inlet air volumetric flow rate. For battery heat generation rate, , These are the specific heat capacity and density of air, respectively. , These represent the battery's specific heat capacity and mass, respectively.

4. The battery thermal management system control method based on multi-mode switching and predictive optimization of claim 3, wherein: In step 3, the feedback correction mechanism specifically regards the heat generation inside the battery as an unobservable disturbance term, and dynamically adjusts the internal state of the model predictive controller or compensates for future predictions based on the deviation between the actual measured value and the model predicted output value of the battery temperature in each control period; The dynamically adjusted model predictive controller is: ； The temperature difference is the temperature difference between the inlet air temperature and the ambient temperature.

5. The battery thermal management system control method based on multi-mode switching and predictive optimization of claim 1, wherein: The objective function of the quadratic programming problem is specifically: ； wherein, is a prediction output, is a reference temperature, is a control input increment, is a weight factor; is a prediction horizon, is a control horizon; the optimization being performed subject to linear inequality constraints on the control input and its increment.

6. The battery thermal management system control method based on multi-mode switching and predictive optimization of claim 1, wherein: The specific determination rule of the double-threshold trigger logic is: When the system is in the shutdown mode, if the temperature tracking error exceeds the first error threshold or the battery output power requirement exceeds the first power threshold, the system switches to the MPC high-speed control mode. when the system is in the MPC high-speed control mode, if the temperature tracking error is lower than a second error threshold and the battery output power demand is lower than a second power threshold, switching to the MPC low-speed control mode; when the system is in the MPC low-speed control mode, if the temperature tracking error is lower than a third error threshold and the battery output power demand is lower than a third power threshold, switching to the shutdown mode.

7. The battery thermal management system control method based on multi-mode switching and predictive optimization of claim 6, wherein: The first error threshold is 0.3 K, the second error threshold is 0.1 K, and the third error threshold is 0.05 K; the first power threshold and the second power threshold are both 40 kW, and the third power threshold is 20 kW.

8. The battery thermal management system control method based on multi-mode switching and predictive optimization of claim 1, wherein: The control input constraint range corresponding to the MPC high-speed control mode is: inlet air volume [0, 20] CFM, and inlet air temperature [283, 303] K; the control input constraint range corresponding to the MPC low-speed control mode is: inlet air volume [0, 10] CFM, and inlet air temperature [293, 303] K.

9. A battery thermal management system comprising a memory, a processor, and a computer program stored on the memory, wherein, The processor implements the method of any one of claims 1 to 8 when executing the computer program.

10. An electric vehicle, characterized by A battery thermal management system comprising the battery thermal management system of claim 9.

Images