A connected electric vehicle cabin and battery thermal management cooling optimization method

Abstract

The application provides a connected electric vehicle cabin and battery thermal management cooling optimization method, which is used for real-time prediction optimization combined with connected road information, seeks to quickly reduce the temperature of the battery pack while ensuring the comfort requirements of the cabin, and minimizes the cooling energy consumption; the connected communication technology is used to obtain historical vehicle speed information, the future driving speed of the vehicle is predicted based on the extreme learning machine algorithm, and the thermal management optimization strategy is more effectively adjusted; in view of the high coupling characteristics of the cabin and the battery thermal management system, a thermal model containing the cabin loop, the battery pack liquid cooling loop and the air conditioning loop is established; considering the slow dynamic characteristics of temperature and the high energy consumption characteristics of the air conditioning system under high temperature environment, a double-layer model prediction framework is proposed, and the energy optimization and temperature control tracking are integrated into the two-layer framework; due to the difference of the double-layer prediction view and the sampling time, the calculation complexity is reduced, so that the system control precision is better improved and the energy consumption is reduced.

Description

Technical Field

[0001] This invention belongs to the field of new energy vehicle power battery technology, and provides a cooling optimization method for cabin and battery thermal management of connected electric vehicles. More specifically, it is a method for optimizing cabin temperature and lithium-ion battery pack cooling of intelligent connected electric vehicles using a two-layer architecture model predictive control. Background Technology

[0002] Thermal management is a decisive factor in optimizing the overall energy consumption of electric vehicles. For electric vehicles, the battery is the only power source providing the necessary traction and thermal management capabilities. During periods of high acceleration and deceleration, the demand for traction increases dramatically, generating significant heat within the battery. When the battery temperature is too high or too low, its efficiency, safety, durability, and thermal runaway resistance all deteriorate. Therefore, the thermal management system for regulating battery temperature consumes a large amount of electrical energy, especially for cooling the battery in high-temperature environments. Furthermore, the power required to cool the battery pack is provided by the battery itself, resulting in a direct feedback loop between battery thermal management and the vehicle's power load. Therefore, efficient thermal management of the electric vehicle battery pack is crucial for improving the overall energy efficiency of the vehicle.

[0003] The two main objectives of efficient thermal management are to reduce the energy consumption of cooling the battery pack while ensuring the required thermal state of the system. Therefore, to enhance the cooling capacity of the battery pack, this design employs a combination of liquid cooling and active air cooling for the lithium-ion battery pack. Liquid cooling of the battery pack dissipates the heat discharged from the battery through a heat exchanger to the refrigerant in the HVAC circuit via the coolant in the liquid cooling circuit. Active air cooling uses a variable-speed fan to blow cool air from the passenger compartment into the battery pack to lower its temperature. However, the air in the passenger compartment is cooled by the compressor in the air conditioning system; therefore, the active air cooling of the battery pack is also coupled to the air conditioning system. This means that the cooling capacity of the battery thermal management system comes from the air conditioning system, and it indicates that the passenger compartment circuit, the air conditioning circuit, and the battery pack liquid cooling circuit are directly interacting and coupled. As the most important load in all thermal systems, the air conditioning consumes a significant amount of energy to maintain the operating temperature of the passenger compartment and the battery pack, which affects the overall energy efficiency of the vehicle and significantly reduces the driving range. In hot weather, to maintain passenger comfort, the driver has to consume a large amount of air conditioning energy, which undoubtedly limits the vehicle's energy-saving potential. Therefore, to ensure the thermal requirements of electric vehicle battery packs and cabins, it is crucial to integrate high-value connected information into the optimization process, and to reduce energy consumption through real-time prediction and multi-layer optimization strategies.

[0004] Traditional battery thermal management optimization assumes the entire driving cycle is known and performs offline global optimization to minimize cooling power and maintain the battery pack temperature within its optimal operating range, without considering future road information. However, due to the slow dynamic characteristics of thermodynamics, optimizing the battery pack's thermal response requires a long operating range. Predictive control offers significant advantages for this characteristic of batteries because it can fully utilize the favorable connected driving environment to obtain more comprehensive road information and improve overall vehicle energy efficiency through real-time prediction and optimization. However, previous predictive control methods have a trade-off: pursuing high temperature accuracy in the short field of view sacrifices energy consumption, while achieving higher energy-saving potential in the long field of view only yields a wider control range. Therefore, to address the uncertainty of future vehicle speed changes, this invention first considers the impact of this uncertainty on the thermal management of electric vehicle systems. It utilizes intelligent connected vehicle-to-everything (V2X) communication technology to allow real-time communication between vehicles and other vehicles (V2V), infrastructure (V2I), pedestrians (V2P), and networks (V2N) to acquire and analyze historical traffic data. Furthermore, it samples data based on an extreme learning machine algorithm to predict future vehicle speeds. Incorporating connected information into speed prediction can improve the efficiency of the cooling system and minimize energy waste. Secondly, addressing the coupling issue between the battery pack liquid cooling circuit, air conditioning circuit, and cabin circuit, a control-oriented temperature model for the cabin and battery pack is established, and a Layered Architecture Model Predictive Control (LA-MPC) framework is proposed. In this framework, the upper layer uses long-field-of-view traffic condition information obtained from connected information to calculate the minimum global energy demand within the field of view, thereby optimizing the temperature thermal trajectory. The lower layer employs a model predictive controller with short prediction cycles and fewer optimization variables and constraints to track the thermal trajectory of the upper layer input, thereby reducing the total system power consumption, ensuring the battery pack's operating temperature, and meeting the constraints of passenger comfort requirements. Because the prediction and control time scales differ at each layer, computational complexity is reduced, and battery pack energy optimization can be integrated into the two-layer control framework to improve the thermal efficiency of the thermal management system and minimize energy consumption. Summary of the Invention

[0005] This invention addresses the high energy consumption problem of rapidly reducing battery pack temperature while ensuring cabin comfort in high-temperature environments. It also considers the potential impact of uncertainties in future vehicle speed changes on the thermal management of electric vehicle systems and proposes an optimization method for cabin and battery thermal management cooling in connected electric vehicles.

[0006] This invention is achieved using the following technical solution:

[0007] A cooling optimization method for the thermal management of the passenger compartment and battery in a connected electric vehicle is disclosed. In this method, the passenger compartment and battery thermal management system includes an air conditioning circuit, a battery pack liquid cooling circuit, and a passenger compartment air circuit. Battery pack cooling is achieved simultaneously through non-contact liquid cooling and active air cooling. Non-contact liquid cooling uses a water pump to drive the circulation of coolant in the liquid cooling pipes, exchanging heat with the air conditioning refrigerant to reduce the battery pack temperature. Active air cooling uses a variable-speed fan to blow cool air from the passenger compartment into the battery pack to further reduce its temperature. The passenger compartment temperature is further reduced by a blower that delivers cooled air from the air conditioning system into the passenger compartment. Therefore, this method achieves rapid temperature reduction of the battery pack and passenger compartment by controlling the water pump speed, variable-speed fan speed, and blower air mass flow rate, thereby reducing energy consumption. The specific steps are as follows:

[0008] The first step is to establish a vehicle speed prediction model based on the Extreme Learning Machine algorithm.

[0009] Based on historical vehicle speed data, an extreme learning machine algorithm is constructed to establish a prediction model for future vehicle speed within the field of view. This method is a training algorithm for a single hidden layer feedforward neural network, which consists of three layers: an input layer, a hidden layer, and an output layer.

[0010] The first step is the vehicle speed training process based on the Extreme Learning Machine algorithm, given a known vehicle speed training set S = {(X...} j ,t j )|X j ∈R n ,t j ∈R m Let X = {j = 1, 2, ..., N}, where X = ... j =[x j1 ,x j2 ,...,x jn ] T Given an n-dimensional historical vehicle speed sequence, t j =[t j1 ,t j2 ,...,t jm ] T Let R be an m-dimensional target vehicle speed sequence, N be the number of historical vehicle speed samples in the vehicle speed training set, j be a natural number from 1 to N, and R be a variable. n Let R be the set of n-dimensional real numbers. m Given an m-dimensional set of real numbers, the output o of a single-hidden-layer feedforward neural network j Represented as:

[0011]

[0012] Where L is the number of hidden layer nodes, 0 < L < N, W i =[w i1 ,w i2 ,...,win ] T Let b be the input weight of the i-th hidden layer node. i β is the bias of the i-th hidden layer node. i Let g(·) be the output weight of the i-th hidden layer node, i = 1, 2, ..., L, and g(·) be the activation function of the hidden layer, a nonlinear function that provides nonlinear mapping for the system. Let g(·) be the Sigmoid function, in the form g(X) j )=1 / (1+e -Xj During training, the input weights W of the hidden layer nodes... i and the bias b of the hidden layer nodes i Since the output is randomly generated and requires no updating, the desired output o of a single-hidden-layer feedforward neural network is... j The output weights β of the hidden layer nodes need to be calculated. i ;

[0013] Furthermore, because when the number of hidden layer nodes L approaches the number of historical vehicle speed samples N in the vehicle speed training set infinitely, the output o of the feedforward neural network... j It can approach the target vehicle speed t in the training set infinitely. j That is, satisfying:

[0014]

[0015] From equations (1) and (2), we get:

[0016]

[0017] The above formula can be expressed in matrix form as follows:

[0018] Hβ=T (4)

[0019] Where H is the output matrix containing L hidden layer nodes, β is the output weight matrix between the hidden layer and the output layer, and T is the desired output matrix, then matrices H, β, and T are expressed as follows:

[0020]

[0021]

[0022] Since the number of historical vehicle speed samples N in the training set is greater than the number of hidden layer nodes L, the goal of training the neural network is to minimize the approximation error. Finding the output weight matrix β is equivalent to finding the least-squares solution for the linear system Hβ = T, i.e.:

[0023] ||Hβ * -T||=min β ||Hβ-T||,β∈R L×m (6)

[0024] The least squares solution β is then obtained. * This is the optimal solution to equation (4):

[0025] β * =H + T (7)

[0026] Among them, H + Let H be the Moore-Penrose generalized inverse matrix;

[0027] Secondly, the vehicle speed prediction process is based on the Extreme Learning Machine (ELM) algorithm. The previous step, the vehicle speed training process based on the ELM algorithm, determined the input weight matrix W = [W1, W2, ..., W...]. L ] T And the bias matrix b = [b1, b2, ..., b L ] T The output weight matrix β = β is calculated. * The test vehicle speed sequence X is now being presented. q =[x q1 ,x q2 ,...,x qm ] T Substituting q = 1, 2, ..., M into equation (1), we obtain the predicted vehicle speed output sequence o. q =βg(WX q +b), where o q To predict the vehicle speed output sequence, M is the number of test vehicle speed sequence samples, and q is a natural number from 1 to M. Therefore, the predicted vehicle speed V obtained by the Extreme Learning Machine algorithm is denoted as:

[0028] V=o q (8)

[0029] The second step is to establish a thermal model of the battery pack.

[0030] The thermal model of a battery pack must reflect both the electrical and thermal characteristics of the battery. Therefore, based on the law of conservation of energy, a thermal model for a lithium-ion battery pack is established:

[0031]

[0032] in, I represents the temperature change of the battery pack per unit time. bat R represents the current of the battery pack during vehicle operation. bat m is the internal resistance of the battery pack. bat For the mass of the battery pack, c bat Q is the specific heat capacity of the battery pack. bat_liq Q is used to cool and absorb heat from the battery pack in the liquid cooling circuit. bat_cabActive air cooling absorbs heat from the battery pack; battery pack voltage U bat Able to be determined by open-circuit voltage U oc and battery pack current I bat With the internal resistance R of the battery pack bat The product is represented as:

[0033] U bat =U oc -I bat R bat (10)

[0034] Battery pack current I bat The function to meet the success rate requirement is as follows:

[0035] I bat =(P tr +P comp +P bl +P pump +P fan ) / U bat (11)

[0036] Among them, P tr For the power consumption of vehicle traction, P comp For the power consumption of the air conditioner compressor, P bl For the power consumption of the cabin air intake blower, P pump For the power consumption of the liquid cooling circuit water pump, P fan The power consumption of the air-cooled fan is used; therefore, the battery pack current I is obtained from equations (10) and (11). bat for:

[0037]

[0038] Substituting equation (12) into equation (9), the battery thermal model is updated as follows:

[0039]

[0040] In equation (13), the traction power consumption P tr Represented as:

[0041]

[0042] Where V is the vehicle's speed. Let F be the acceleration of the car. Due to the uncertainty of the future speed, the car's speed is the predicted speed V obtained in the first step based on the extreme learning machine algorithm. m is the mass of the car. a It is the aerodynamic drag during driving, F roll It is the rolling resistance of the car, F gxt The slope resistance is expressed as follows:

[0043]

[0044] F roll =mgC r (16)

[0045] F gxt =mgsinφ (17)

[0046] Among them, C d ρ is the aerodynamic drag coefficient. a It is air density, A f It is the frontal area of ​​the vehicle when it is in motion, C r Here, φ is the rolling resistance coefficient, and φ is the slope angle. We assume φ = 0 here, so F... gxt =0;

[0047] The third step is to establish discrete temperature models for the air conditioning loop and cabin loop, oriented towards control.

[0048] Because air conditioning simulation models involve complex thermodynamics and fluid dynamics, they are unsuitable for controller design. Therefore, it is necessary to establish a discrete model of the air conditioning loop and cabin temperature for control purposes. The discrete model of cabin temperature can be represented as...

[0049]

[0050] Where k = 0, 1, 2..., when k = 0 is the initial time, T cab (k) represents the cabin temperature at time k, T cab (k+1) represents the cabin temperature at time k+1, T in (k) represents the temperature of the contents inside the cabin at time k, T shell (k) represents the cabin outer shell temperature at time k, W. bl (k) represents the blower inlet mass flow rate at time k, T ain (k) represents the cabin air inlet temperature at time k, where it is assumed that the temperature of the interior contents is T. in and cabin shell temperature T shell Since the changes in operating conditions are not significant, both are set as constants. γ1, γ2, γ3 and τ1 are identification parameters obtained by curve fitting using the least squares method based on the cabin temperature output of the electric vehicle cooling model in AMESim.

[0051] Since the discrete model of evaporator temperature can be expressed as:

[0052]

[0053] Among them, T evap (k) represents the evaporator temperature at time k, T evap(k+1) represents the evaporator temperature at time k+1. γ4, γ5, and τ2 are the target temperature of the evaporator, and the identification parameters are obtained by curve fitting using the least squares method through the evaporator output of the electric vehicle cooling model in AMESim.

[0054] Therefore, the cabin air inlet temperature T at time k ain (k) can be represented as:

[0055] T ain (k)=γ6T evap (k)+γ7W bl (k)+τ3 (20)

[0056] Among them, γ6, γ7, and τ3 are identification parameters obtained by curve fitting using the least squares method from the cabin air inlet temperature output of the electric vehicle cooling model in AMESim. It can be seen that the control input in the air conditioning and cabin circuit is the blower air inlet mass flow rate W. bl ;

[0057] The fourth step is to establish a discrete temperature model for the liquid cooling circuit.

[0058] The battery pack cooling method in the liquid-cooled circuit is active air cooling and non-contact liquid cooling; the coolant inlet temperature and coolant outlet temperature of the battery pack in the cooling circuit can be defined as follows:

[0059]

[0060] Among them, T liq_in (k+1) represents the coolant inlet temperature of the battery pack in the cooling circuit at time k+1. Let be the mass flow rate of the coolant in the liquid cooling circuit at time k, and χ1, χ2, and The identification parameters are obtained by curve fitting using the least squares method based on the coolant inlet temperature output of the electric vehicle cooling model in AMESim.

[0061]

[0062] Among them, T liq_out (k+1) represents the coolant outlet temperature of the battery pack in the cooling circuit at time k+1, A liq h is the convective heat transfer area between the coolant and the battery pack. liq C is the convective heat transfer coefficient between the coolant and the battery pack. liq This refers to the specific heat capacity of the coolant.

[0063]

[0064] Among them, V pump η is the pump displacement.pump ρ is the volumetric efficiency of the water pump. liq ω is the liquid density of the coolant. pump (k) represents the control quantity, the pump speed, at time k.

[0065] Therefore, based on the thermal model of the lithium-ion battery pack given by equation (9), the heat Q of the active air-cooled battery pack at time k is obtained. bat_liq (k) and the heat Q of the liquid-cooled absorption battery pack bat_cab (k) are respectively represented as:

[0066]

[0067]

[0068] Where, ω fan (k) represents the rotational speed of the active air-cooled fan at time k. Here, the model parameter μ1 is 0.2671 and the model parameter μ2 is 2677.3.

[0069] Step 5: Cooling power calculation

[0070] The main energy consumer in a car's air conditioning circuit is the air conditioning compressor, specifically the air conditioning compressor's power consumption P. comp The main energy consumer in the cabin circuit is the blower, i.e., the blower power consumption P. bl The main energy consumption for cooling the battery pack comes from the water pump and the air-cooled fan; that is, the water pump power consumption is P. pump Power consumption of air-cooled fan P fan ;

[0071] First, calculate the pump power consumption P at time k. pump (k):

[0072]

[0073] Among them, P pump_m (k) represents the mechanical power of the pump at time k, η m The power conversion efficiency of the water pump is ΔP. pump For water pump pressure drop;

[0074] Secondly, calculate the compressor power consumption P at time k. comp Within a certain timeframe, the power required to lower both the battery pack temperature and the cabin temperature from their initial values ​​to their target temperatures comes from the air conditioning compressor. Therefore, assuming the air conditioning compressor's energy consumption consists of two parts: one for lowering the cabin air temperature and the other for lowering the coolant temperature in the cooling circuit, ignoring other low-power components, the power consumption P of the air conditioning compressor at time k can be obtained from the energy balance relationship. comp (k) is:

[0075]

[0076] Among them, C comp η is the specific heat capacity of air. comp W is the coefficient of performance for air conditioning. bl (k) represents the blower inlet mass flow rate at time k, T amb (k) represents the ambient temperature at time k, T ain (k) represents the cabin air inlet temperature at time k, C liq The specific heat capacity of the coolant. Let T be the mass flow rate of the coolant in the liquid cooling circuit at time k. liq_in (k) represents the coolant inlet temperature of the battery pack in the cooling circuit at time k, T liq_out (k) represents the coolant outlet temperature of the battery pack in the cooling circuit at time k;

[0077] Next, calculate the cabin air intake blower power consumption P at time k. bl (k) and fan power consumption P fan (k); Blower power P bl It controls the cabin air intake mass flow rate W bl The function, fan power P fan It controls the fan speed ω fan The functions, after data fitting, are estimated as follows:

[0078]

[0079]

[0080] Among them, K bl For the fan parameters, λ1, λ2, λ3, α1, α2, α3 and α4 are all identification parameters obtained by curve fitting using the least squares method based on the blower power and fan power output of the electric vehicle cooling model in AMESim.

[0081] Therefore, the cabin and battery thermal management system cooling power P at time k is defined. cooling (k) is:

[0082] P cooling (k)=P pump (k)+P comp (k)+P fan (k)+P bl (k) (30)

[0083] From equations (26), (27), (28), and (29), it can be seen that the cooling power P of the cabin and battery thermal management system is... cooling It is the water pump speed ωpump Fan speed ω fan blower inlet air mass flow rate W bl Battery pack temperature T bat and cabin temperature T cab The function, therefore the cabin and battery thermal management system cooling power P at time k. cooling (k) can be represented as:

[0084] P cooling (k)=f(T bat (k),T cab (k),ω pump (k),ω fan (k),W bl (k)) (31)

[0085] Similarly, we can conclude that the temperature change of the battery pack is related to the water pump speed ω. pump Fan speed ω fan The cabin temperature change is a function of the predicted vehicle speed V and the temperature T of the interior items. in Cabin outer shell temperature T shell And blower inlet air mass flow rate W bl Therefore, equations (13) and (18) are discretized according to the sampling time ΔT to obtain the discrete model T of the battery pack temperature. bat Discrete model T of (k+1) and cabin temperature cab (k+1) are respectively:

[0086] T bat (k+1)=T bat (k)+f(T bat (k),ω pump (k),ω fan (k),V(k))·ΔT (32)

[0087] T cab (k+1)=T cab (k)+f(T cab (k),T in (k),T shell (k),W bl (k))·ΔT (33)

[0088] Step 6: Construct a two-layer model predictive control framework

[0089] Using equations (31), (32), and (33), two state variables of the cabin and battery thermal management system are selected as battery pack temperature T. bat and cabin temperature T cab That is, it can be expressed as x = [T] bat ,T cab] T The three control variables are the liquid cooling water pump speed ω. pump Air-cooled fan speed ω fan And blower inlet air mass flow rate W bl That is, it can be expressed as u = [ω pump ,ω fan W bl ] T ;

[0090] To construct a two-layer model predictive control framework, the upper-layer objective function and constraints are first defined as follows:

[0091]

[0092] Where k+s|k represents the prediction of time k+s at time k;

[0093] P cooling (k+s|k) represents the prediction of the cabin and battery thermal management system cooling power at time k+s.

[0094] T bat (k+s|k) represents the prediction of the battery pack temperature at time k+s at time k;

[0095] T bat (k+s-1|k) represents the prediction of the battery pack temperature at time k+s-1 at time k;

[0096] f Tbat (k+s-1|k) is the battery pack temperature change function at time k+s-1;

[0097] T cab (k+s|k) represents the prediction of the cabin temperature at time k+s at time k;

[0098] T cab (k+s-1|k) represents the prediction of the cabin temperature at time k+s-1.

[0099] f Tcab (k+s-1|k) is the cabin temperature change function at time k+s-1;

[0100] f Pcooling (k+s|k) is the cooling power function of the cabin and battery thermal management system at time k+s;

[0101] ω pump (k+s-1|k) is the liquid cooling water pump speed at time k+s-1, calculated based on the prediction of time k+s at time k;

[0102] ω fan(k+s-1|k) is the air-cooled fan speed at time k+s-1, calculated based on the prediction of time k+s at time k;

[0103] W bl (k+s-1|k) is the blower intake mass flow rate at time k+s-1, calculated based on the prediction of time k+s at time k;

[0104] ΔT u The sampling time of the upper layer, The target temperature for the battery pack. The target temperature for the cabin, N u For the upper-level prediction field of view, s = 1, 2, ..., N u K ub Weights are assigned to the degree to which the battery pack temperature approaches the target temperature of the battery. K uc The weighting of the degree to which the cabin temperature approaches the target cabin temperature. This is the lower limit of the battery pack's temperature. This represents the upper limit of the battery pack temperature. This is the lower limit of the cabin temperature. This represents the upper limit of cabin temperature; ω pump_min ω represents the minimum pump speed. pump_max ω represents the maximum speed of the water pump. fan_min ω represents the minimum fan speed. fan_max W represents the maximum fan speed. bl_min W represents the minimum inlet mass flow rate of the blower. bl_max To maximize the mass flow rate of the blower intake, the optimal state output x = [T] was obtained by optimizing the upper-level objective function. bat ,T cab ] T Two of the state variables T bat Let T be the desired temperature of the battery pack. bat_ref T cab The desired cabin temperature is denoted as T. cab_ref ;

[0105] Next, the lower-level objective function and constraints are defined as follows:

[0106]

[0107]

[0108]

[0109] ω pump_min ≤ω pump (k+r-1|k)≤ω pump_max

[0110] ω fan_min ≤ω fan (k+r-1|k)≤ω fan_max

[0111] W bl_min ≤W bl (k+r-1|k)≤W bl_max

[0112] Where k+r|k represents the prediction of time k+r at time k;

[0113] P cooling (k+r|k) is the prediction of the cabin and battery thermal management system cooling power at time k+r.

[0114] T bat (k+r|k) is the prediction of the battery pack temperature at time k+r at time k;

[0115] T bat_ref (k+r|k) is the prediction of the expected temperature of the battery pack at time k+r at time k;

[0116] T bat (k+r-1|k) is the prediction of the battery pack temperature at time k+r-1.

[0117] Let be the function of battery pack temperature change at time k+r-1;

[0118] T cab (k+r|k) represents the prediction of the cabin temperature at time k+r.

[0119] T cab_ref (k+r|k) is the prediction of the desired cabin temperature at time k+r.

[0120] T cab (k+r-1|k) represents the prediction of the cabin temperature at time k+r-1.

[0121] f Tcab (k+r-1|k) is the cabin temperature change function at time k+r-1;

[0122] ω pump (k+r-1|k) is the liquid cooling water pump speed at time k+r-1, which is obtained by predicting time k+r at time k;

[0123] ω fan (k+s-1|k) is the air-cooled fan speed at time k+r-1, calculated based on the prediction of time k+r at time k;

[0124] W bl (k+s-1|k) is the blower intake mass flow rate at time k+s-1, calculated based on the prediction of time k+r at time k;

[0125] ΔT l ΔT is the sampling time of the lower layer. u >ΔT l N l For the lower-level prediction field of view, r = 1, 2, ..., N l N here u >N l ;K lb K is the weight for how close the battery pack temperature is to the desired battery pack temperature. lb =1 / max[T bat (r)-T bat_ref (r)] 2 K lc K is the weight given to the degree to which the cabin temperature reaches the desired cabin temperature. lc =1 / max[T cab (r)-T cab_ref (r)] 2 ;

[0126] Step 7: Obtain the optimal control quantity and complete the optimization.

[0127] The optimal control variable sequence u is obtained through the two-level model prediction framework constructed in step six. * for:

[0128] u * =argminJ (36)

[0129] When s = 1, the optimal control quantity ω at time k is calculated. pump (k), ω fan (k) and W bl Substituting (k) into the system model, we can calculate the state output T at time k+1. bat (k+1) and T cab (k+1), substituting it back into the upper-level objective function, we obtain the optimal control quantity ω at s=2, i.e., time k+1. pump (k+1), ω fan (k+1) and W bl (k+1), and calculate the state output T at time k+2. bat (k+2) and T cab (k+2), in the upper prediction field N uThe internal rolling repetitive calculation yields the optimal state output at each time step. This optimal state output is then substituted into the lower-level objective function as the target values ​​for cabin temperature and battery pack temperature for temperature tracking. This is performed within the lower-level prediction field N. l Within this process, the lower-level objective function is then subjected to rolling optimization calculations to ultimately obtain the optimal control variable sequence u. * This is to achieve the goal of rolling optimization.

[0130] Compared with the prior art, the beneficial effects of the present invention are:

[0131] 1. This invention provides a cooling optimization method for the cabin and battery thermal management of a connected electric vehicle. It employs a two-layer model predictive optimization method. Considering the relatively slow dynamic characteristics of large battery packs, optimization over a long field of view is required to achieve better energy efficiency. While model predictive control in a short field of view can achieve high control accuracy, it consumes a large amount of energy. To address this need, a two-layer model predictive control framework is proposed. This framework is used to rapidly cool the battery pack while also considering passenger comfort requirements. The upper layer optimizes a low-consumption cooling trajectory within a long field of view, while the lower layer uses shorter sampling times and higher density to accurately track temperature, realizing the energy-saving potential of electric vehicles.

[0132] 2. The present invention provides a cooling optimization method for the cabin and battery thermal management of a connected electric vehicle. It adopts a vehicle speed prediction method based on extreme learning machine. In view of the uncertainty of future vehicle speed, it combines network information with electric vehicle thermal management. The prediction of vehicle speed distribution can more comprehensively estimate the thermal load demand of the battery, realize more precise adjustment of cooling strategy, and provide more sustainable and superior performance for electric vehicles.

[0133] 3. This invention provides a cooling optimization method for the thermal management of the cabin and battery in a connected electric vehicle. It employs a simplified, control-oriented thermal model of the battery pack and cabin. Considering that the electrical energy for cooling the battery pack comes from the battery itself, and that the cooling process originates partly from the liquid coolant and partly from the cabin's cool air, ultimately being absorbed by the air conditioning system, a simplified, control-oriented thermal model of the battery pack and cabin was established and validated to address the complex feedback between the three loops. This facilitates optimization research on electric vehicle battery packs that are simultaneously cooled by active air and liquid.

[0134] 4. The present invention provides a cooling optimization method for the cabin and battery thermal management of a connected electric vehicle. It adopts a multi-view, hierarchical optimization method to address the problems of long calculation time and high energy consumption of centralized model predictive control. It adopts a two-layer control framework with different prediction and control times. The upper layer can combine the traffic information of the connected vehicle to optimize the temperature reference trajectory with the minimum energy consumption in the long view, while the lower layer completes trajectory tracking in the short view. This not only reduces the computational complexity but also improves the temperature control accuracy, thereby improving the vehicle's energy efficiency. Attached Figure Description

[0135] The invention will now be further described with reference to the accompanying drawings:

[0136] Figure 1 This is a schematic diagram of the cabin and battery thermal management system, which consists of an air conditioning circuit, a battery pack liquid cooling circuit, and a cabin air circuit, and is based on the cooling optimization method for cabin and battery thermal management of a connected electric vehicle provided by the present invention.

[0137] Figure 2 The present invention provides a flowchart of a cooling optimization method for thermal management of the passenger compartment and battery of a connected electric vehicle.

[0138] Figure 3 The Extreme Learning Machine method is used to predict the vehicle speed change curve over 10 seconds in the optimization process of this method.

[0139] Figure 4 This is a graph showing the temperature change of the battery pack during the optimization process using this method.

[0140] Figure 5 This is a graph showing the cabin temperature changes during the optimization process using this method.

[0141] Figure 6 The graph shows the pump speed curve during the optimization process using this method.

[0142] Figure 7 This is a graph showing the fan speed during the optimization process using this method.

[0143] Figure 8 This is a graph showing the blower mass flow rate during the optimization process using this method.

[0144] Figure 9 The diagram shows the cooling power curve of the thermal management system during the optimization process using this method.

[0145] Figure 10 The physical parameters of the vehicle and battery during the optimization process using this method.

[0146] Figure 11The physical parameters of the coolant, water pump, and compressor during the optimization process using this method are given. Detailed Implementation

[0147] The present invention will now be described in detail with reference to the accompanying drawings:

[0148] To address the need for passenger cabin temperature comfort in high-temperature environments, this invention seeks an optimized cooling strategy for rapidly reducing battery pack temperature, thereby minimizing the energy consumption of the thermal management system. The system comprises three main parts: an air conditioning system loop, a battery pack liquid cooling loop, and a cabin air loop. A schematic diagram of the system is shown below. Figure 1 As shown. In the air conditioning circuit, the air conditioning refrigeration cycle consists of a compressor, condenser, expansion valve, and evaporator. The battery pack liquid cooling circuit consists of a water pump, coolant pipes, and an evaporator shared with the air conditioning circuit. In the battery pack liquid cooling circuit, to increase heat dissipation efficiency, the battery pack temperature is simultaneously cooled by liquid cooling and active air cooling. During liquid cooling, the air conditioning refrigerant absorbs heat from the battery pack in the coolant in the liquid cooling pipes. During air cooling, a variable-speed fan blows cool air from the cabin into the battery pack to help lower the battery temperature. In the cabin air circuit, after the outside ambient temperature enters, it absorbs heat through the evaporator of the air conditioning circuit to lower the air temperature. The cool air is then delivered into the cabin by a blower, lowering the cabin's ambient temperature and simultaneously supplying it to the battery pack for air cooling.

[0149] Considering the potential impact of uncertainties in future vehicle speed changes on the thermal management of electric vehicle systems, this study utilizes intelligent connected vehicle-to-everything (V2X) communication technology to acquire and analyze traffic information in real time. Vehicle speed data is then input into a vehicle speed prediction model based on extreme learning machine to predict the future driving speed of electric vehicles. Introducing real-time optimization using connected information allows for better assessment of the heat generation and dissipation requirements of the battery pack within the prediction field of view, thereby maximizing the energy efficiency of the battery thermal management system.

[0150] Based on high-quality traffic data from the connected vehicle network, the upper layer optimizes and calculates the minimum power requirement of the vehicle within a longer prediction horizon, and plans the temperature trajectory of the battery pack and passenger cabin. Due to the sequential structure of the two-layer architecture, the integrated model of the lower layer's air conditioning circuit, battery pack liquid cooling circuit, and passenger cabin air circuit predicts and tracks the temperature reference input from the upper layer, achieving further energy savings. The flowchart of the cooling optimization method for passenger cabin and battery thermal management of a connected electric vehicle proposed in this invention is shown below. Figure 2 As shown in the accompanying drawings. The invention will now be described in detail with reference to the accompanying drawings.

[0151] The first step is to establish a vehicle speed prediction model based on the Extreme Learning Machine algorithm.

[0152] Historical vehicle speed data is obtained using connected vehicle communication technology, and a prediction model for future vehicle speed within the field of view is constructed based on the Extreme Learning Machine (ELM) algorithm. This method employs a training algorithm for a single-hidden-layer feedforward neural network, which consists of a three-layer structure: an input layer, a hidden layer, and an output layer. Historical vehicle speeds are used as the input to the ELM algorithm's input layer, and the predicted future vehicle speed is used as the output. The input layer has 10 neurons, corresponding to the historical vehicle speeds of the previous 10 seconds. The number of output layer neurons corresponds to the time domain length of the speed prediction, also set to 10, corresponding to the future vehicle speeds 10 seconds after the current moment. The number of hidden layer neurons (i.e., the number of hidden layer nodes), L, directly affects the computational complexity and fitting accuracy of the ELM network. The root mean square error (RMSE) between the predicted and actual speed data is used as the accuracy criterion. Under the premise of minimizing computational scale, an RMSE less than 0.5 satisfies the accuracy requirement, thus determining the optimal number of hidden layer neurons and setting L = 500.

[0153] The first step is the vehicle speed training process based on the Extreme Learning Machine algorithm, given a known vehicle speed training set S = {(X...} j ,t j )|X j ∈R n ,t j ∈R m Let X = {j = 1, 2, ..., N}, where X = ... j =[x j1 ,x j2 ,...,x jn ] T Given an n-dimensional historical vehicle speed sequence, t j =[t j1 ,t j2 ,...,t jm ] T Let R be an m-dimensional target vehicle speed sequence, N be the number of historical vehicle speed samples in the vehicle speed training set, j be a natural number from 1 to N, and R be a variable. n Let R be the set of n-dimensional real numbers. m Given an m-dimensional set of real numbers, the output o of a single-hidden-layer feedforward neural network j Represented as:

[0154]

[0155] Where L is the number of hidden layer nodes, 0 < L < N, W i =[w i1 ,w i2 ,...,w in ] T Let b be the input weight of the i-th hidden layer node. i β is the bias of the i-th hidden layer node.i Let g(·) be the output weight of the i-th hidden layer node, i = 1, 2, ..., L, and g(·) be the activation function of the hidden layer, a nonlinear function that provides the nonlinear mapping for the system. Let g(·) be the Sigmoid function, in the form of: During training, the input weights W of the hidden layer nodes i and the bias b of the hidden layer nodes i Since the output is randomly generated and requires no updating, the desired output o of a single-hidden-layer feedforward neural network is... j The output weights β of the hidden layer nodes need to be calculated. i ;

[0156] Furthermore, because when the number of hidden layer nodes L approaches the number of historical vehicle speed samples N in the vehicle speed training set infinitely, the output o of the feedforward neural network... j It can approach the target vehicle speed t in the training set infinitely. j That is, satisfying:

[0157]

[0158] From equations (1) and (2), we can obtain:

[0159]

[0160] The above formula can be expressed in matrix form as follows:

[0161] Hβ=T (4)

[0162] Where H is the output matrix containing L hidden layer nodes, β is the output weight matrix between the hidden layer and the output layer, and T is the desired output matrix, then matrices H, β, and T are expressed as follows:

[0163]

[0164]

[0165] Since the number of historical vehicle speed samples N in the training set is greater than the number of hidden layer nodes L, the goal of training the neural network is to minimize the approximation error. Finding the output weight matrix β is equivalent to finding the least-squares solution for the linear system Hβ = T, i.e.:

[0166] ||Hβ * -T||=min β ||Hβ-T||,β∈R L×m (6)

[0167] The least squares solution β is then obtained. * This is the optimal solution to equation (4):

[0168] β * =H+ T (7)

[0169] Among them, H + Let H be the Moore-Penrose generalized inverse matrix;

[0170] Secondly, the vehicle speed prediction process is based on the Extreme Learning Machine (ELM) algorithm. The previous step, the vehicle speed training process based on the ELM algorithm, determined the input weight matrix W = [W1, W2, ..., W...]. L ] T And the bias matrix b = [b1, b2, ..., b L ] T The output weight matrix β = β is calculated. * The test vehicle speed sequence X is now being presented. q =[x q1 ,x q2 ,...,x qm ] T Substituting q = 1, 2, ..., M into equation (1), we obtain the predicted vehicle speed output sequence o. q =βg(WX q +b), where o q To predict the vehicle speed output sequence, M is the number of test vehicle speed sequence samples, and q is a natural number from 1 to M. Therefore, the predicted vehicle speed V obtained by the Extreme Learning Machine algorithm is denoted as:

[0171] V=o q (8)

[0172] In this embodiment, the model training uses vehicle speed information from six driving cycles—New European Driving Standard (NEDC), High-Speed ​​Driving Cycle (HWFET), US Emissions Test Cycle (FTP75), China Light Vehicle Driving Cycle (CLTC), and Urban Driving Cycle (UDDS)—as the training dataset for the Extreme Learning Machine algorithm. Three driving cycles—aggressive driving cycle (US06), New York City Driving Cycle (NYCC), and Los Angeles Driving Cycle (LA92)—are selected as the test dataset. The resulting vehicle speed curves with a 10-second prediction field of view are shown below. Figure 3 As shown, the maximum deviation is less than 2 m / s.

[0173] The second step is to establish a thermal model of the battery pack.

[0174] First, a thermal model of lithium-ion batteries is a fundamental condition for studying the thermal management of electric vehicle battery packs. Since the cooling process is a single discharge process, battery aging and capacity degradation can be neglected. Furthermore, the difference between the surface temperature and the core temperature of the battery pack is ignored, and the battery pack is treated as a point mass for derivation, rather than focusing on individual cells. Moreover, the model must reflect both the electrical and thermal characteristics of the battery. Therefore, this invention, based on the analysis of the hydrodynamic laws of liquid and air-cooled battery packs, and according to the law of conservation of energy, establishes a thermal model of lithium-ion battery packs:

[0175]

[0176] in, I represents the temperature change of the battery pack per unit time. bat R represents the current of the battery pack during vehicle operation. bat m is the internal resistance of the battery pack. bat For the mass of the battery pack, c bat Q represents the specific heat capacity of the battery pack. bat_liq Q is used to cool and absorb heat from the battery pack in the liquid cooling circuit. bat_cab Active air cooling absorbs heat from the battery pack; battery pack voltage U bat Able to be determined by open-circuit voltage U oc and battery pack current I bat With the internal resistance R of the battery pack bat The product is represented as:

[0177] U bat =U oc -I bat R bat (10)

[0178] Battery pack current I bat The function to meet the success rate requirement is as follows:

[0179] I bat =(P tr +P comp +P bl +P pump +P fan ) / U bat (11)

[0180] Among them, P tr For the power consumption of vehicle traction, P comp For the power consumption of the air conditioner compressor, P bl For the power consumption of the cabin air intake blower, P pump For the power consumption of the liquid cooling circuit water pump, P fan The power consumption of the air-cooled fan is used; therefore, the battery pack current I is obtained from equations (10) and (11). bat for:

[0181]

[0182] Substituting equation (12) into equation (9), the battery thermal model is updated as follows:

[0183]

[0184] In equation (13), the traction power consumption P tr Represented as:

[0185]

[0186] Where V is the vehicle's speed. Let F be the acceleration of the car. Due to the uncertainty of the future speed, the car's speed is the predicted speed V obtained in the first step based on the extreme learning machine algorithm. m is the mass of the car. a It is the aerodynamic drag during driving, F roll It is the rolling resistance of the car, F gxt The slope resistance is expressed as follows:

[0187]

[0188] F roll =mgC r (16)

[0189] F gxt =mgsinφ (17)

[0190] Among them, C d ρ is the aerodynamic drag coefficient. a It is air density, A f It is the frontal area of ​​the vehicle when it is in motion, C r Here, φ is the rolling resistance coefficient, and φ is the slope angle. We assume φ = 0 here, so F... gxt =0;

[0191] The third step is to establish temperature prediction models for the air conditioning loop and cabin loop, oriented towards control.

[0192] The cooling optimization strategy of this method aims to rapidly reduce the battery pack temperature at high temperatures and maintain it within an ideal range, preventing excessively high battery temperatures that could shorten battery life or cause damage. Simultaneously, it seeks to ensure driver comfort by quickly lowering the cabin air temperature to within the perceived temperature range. Therefore, the battery pack temperature T... bat Cabin air temperature T cab The air conditioning loop is configured as two state variables. Due to the complex thermodynamics and fluid dynamics involved, it is unsuitable for controller design. Therefore, a predictive model for the air conditioning loop and cabin temperature, oriented towards control, is needed.

[0193] As can be seen from 1, the air temperature T inside the cabin cab The air conditioning system is cooled by a compressor in the air conditioning circuit. Liquid cooling of the battery pack involves heat transfer between the refrigerant in the liquid cooling circuit and the refrigerant in the air conditioning circuit. Air cooling of the battery pack uses a variable-speed fan to blow cabin air into the battery pack to lower its temperature. Because the simulation model of the air conditioning system involves complex thermodynamics and fluid dynamics, it is unsuitable for controller design. Therefore, a discrete model of the air conditioning circuit and cabin temperature needs to be established for control purposes. The discrete model of the cabin temperature can be expressed as:

[0194]

[0195] Where k = 0, 1, 2..., when k = 0 is the initial time, T cab (k) represents the cabin temperature at time k, T cab (k+1) represents the cabin temperature at time k+1, T in (k) represents the temperature of the contents inside the cabin at time k, T shell (k) represents the cabin outer shell temperature at time k, W. bl (k) represents the blower inlet mass flow rate at time k, T ain (k) represents the cabin air inlet temperature at time k, where it is assumed that the temperature of the interior contents is T. in and cabin shell temperature T shell Since the changes in operating conditions are not significant, both are set as constants. γ1, γ2, γ3 and τ1 are identification parameters obtained by curve fitting using the least squares method based on the cabin temperature output of the electric vehicle cooling model in AMESim.

[0196] Since the discrete model of evaporator temperature can be expressed as:

[0197]

[0198] Among them, T evap (k) represents the evaporator temperature at time k, T evap (k+1) represents the evaporator temperature at time k+1. γ4, γ5, and τ2 are the target temperature of the evaporator, and the identification parameters are obtained by curve fitting using the least squares method through the evaporator output of the electric vehicle cooling model in AMESim.

[0199] Therefore, the cabin air inlet temperature T at time k ain (k) can be represented as:

[0200] T ain (k)=γ6T evap (k)+γ7Wbl (k)+τ3 (20)

[0201] Among them, γ6, γ7, and τ3 are identification parameters obtained by curve fitting using the least squares method from the cabin air inlet temperature output of the electric vehicle cooling model in AMESim. It can be seen that the control input in the air conditioning and cabin circuit is the blower air inlet mass flow rate W. bl ;

[0202] The fourth step is to establish a discrete temperature model for the liquid cooling circuit.

[0203] The battery pack cooling method in the liquid-cooled circuit is active air cooling and non-contact liquid cooling; the coolant inlet temperature and coolant outlet temperature of the battery pack in the cooling circuit can be defined as follows:

[0204]

[0205] Among them, T liq_in (k+1) represents the coolant inlet temperature of the battery pack in the cooling circuit at time k+1. Let be the mass flow rate of the coolant in the liquid cooling circuit at time k, and χ1, χ2, and The identification parameters are obtained by curve fitting using the least squares method based on the coolant inlet temperature output of the electric vehicle cooling model in AMESim.

[0206]

[0207] Among them, T liq_out (k+1) represents the coolant outlet temperature of the battery pack in the cooling circuit at time k+1, A liq h is the convective heat transfer area between the coolant and the battery pack. liq C is the convective heat transfer coefficient between the coolant and the battery pack. liq This refers to the specific heat capacity of the coolant.

[0208]

[0209] Among them, V pump η is the pump displacement. pump ρ is the volumetric efficiency of the water pump. liq ω is the liquid density of the coolant. pump (k) represents the control quantity, the pump speed, at time k.

[0210] Therefore, based on the thermal model of the lithium-ion battery pack given by equation (9), the heat Q of the active air-cooled battery pack at time k is obtained. bat_liq (k) and the heat Q of the liquid-cooled absorption battery pack bat_cab (k) are respectively represented as:

[0211]

[0212]

[0213] Where, ω fan (k) represents the rotational speed of the active air-cooled fan at time k. Here, the model parameter μ1 is 0.2671 and the model parameter μ2 is 2677.3.

[0214] Step 5: Calculate cooling power.

[0215] The entire system comprises three main circuits: the air conditioning circuit (compressor circuit), the cabin circuit (cabin air supply circuit), and the liquid cooling circuit (battery pack liquid cooling circuit). This invention does not approximate a proportional relationship between thermal management cooling power and the maximum heat dissipation of the battery pack. An overly simplified underlying thermal circuit cannot simultaneously reflect temperature changes in the cabin circuit, as well as the relationship between the air conditioning circuit and the liquid cooling circuit. Because heat transfer passes through actuators that generate the main energy consumption of the system, the primary energy consumer in the automotive air conditioning circuit is the air conditioning compressor, i.e., the air conditioning compressor power consumption P. comp The main energy consumer in the cabin circuit is the blower, i.e., the blower power consumption P. bl The main energy consumption for cooling the battery pack comes from the water pump and the air-cooled fan; that is, the water pump power consumption is P. pump Power consumption of air-cooled fan P fan ;

[0216] First, calculate the pump power consumption P at time k. pump (k):

[0217]

[0218] Among them, P pump_m (k) represents the mechanical power of the pump at time k, η m The power conversion efficiency of the water pump is ΔP. pump For water pump pressure drop;

[0219] Secondly, calculate the compressor power consumption P at time k. comp Within a certain timeframe, the power required to lower both the battery pack temperature and the cabin temperature from their initial values ​​to their target temperatures comes from the air conditioning compressor. Therefore, assuming the air conditioning compressor's energy consumption consists of two parts: one for lowering the cabin air temperature and the other for lowering the coolant temperature in the cooling circuit, ignoring other low-power components, the power consumption P of the air conditioning compressor at time k can be obtained from the energy balance relationship. comp (k) is:

[0220]

[0221] Among them, C comp η is the specific heat capacity of air.comp W is the coefficient of performance for air conditioning. bl (k) represents the blower inlet mass flow rate at time k, T amb (k) represents the ambient temperature at time k, T ain (k) represents the cabin air inlet temperature at time k, C liq The specific heat capacity of the coolant. Let T be the mass flow rate of the coolant in the liquid cooling circuit at time k. liq_in (k) represents the coolant inlet temperature of the battery pack in the cooling circuit at time k, T liq_out (k) represents the coolant outlet temperature of the battery pack in the cooling circuit at time k;

[0222] Next, calculate the cabin air intake blower power consumption P at time k. bl (k) and fan power consumption P fan (k); Blower power P bl It controls the cabin air intake mass flow rate W bl The function, fan power P fan It controls the fan speed ω fan The functions, after data fitting, are estimated as follows:

[0223]

[0224]

[0225] Among them, K bl For the fan parameters, λ1, λ2, λ3, α1, α2, α3 and α4 are all identification parameters obtained by curve fitting using the least squares method based on the blower power and fan power output of the electric vehicle cooling model in AMESim.

[0226] Therefore, the cabin and battery thermal management system cooling power P at time k is defined. cooling (k) is:

[0227] P cooling (k)=P pump (k)+P comp (k)+P fan (k)+P bl (k) (30)

[0228] From equations (26), (27), (28), and (29), it can be seen that the cooling power P of the cabin and battery thermal management system is... cooling It is the water pump speed ω pump Fan speed ω fan blower inlet air mass flow rate W bl Battery pack temperature T batand cabin temperature T cab The function, therefore the cabin and battery thermal management system cooling power P at time k. cooling (k) can be represented as:

[0229] P cooling (k)=f(T bat (k),T cab (k),ω pump (k),ω fan (k),W bl (k)) (31)

[0230] Similarly, we can conclude that the temperature change of the battery pack is related to the water pump speed ω. pump Fan speed ω fan The cabin temperature change is a function of the predicted vehicle speed V and the temperature T of the interior items. in Cabin outer shell temperature T shell And blower inlet air mass flow rate W bl Therefore, equations (13) and (18) are discretized according to the sampling time ΔT to obtain the discrete model T of the battery pack temperature. bat Discrete model T of (k+1) and cabin temperature cab (k+1) are respectively:

[0231] T bat (k+1)=T bat (k)+f(T bat (k),ω pump (k),ω fan (k),V(k))·ΔT (32)

[0232] T cab (k+1)=T cab (k)+f(T cab (k),T in (k),T shell (k),W bl (k))·ΔT (33)

[0233] Step 6: Construct a two-layer model predictive control framework

[0234] For cooling in high-temperature environments, i.e. when the initial temperature of the battery pack and cabin is higher than the reasonable operating range, the control objective of the system is to reduce the battery temperature and keep it operating within the ideal operating temperature range, reduce the cabin temperature, quickly reach the set comfortable temperature, and minimize the cumulative power consumption of the entire system.

[0235] Because temperature is a slow-heating characteristic, its optimization requires a relatively long field of view. To address this requirement, the proposed two-layer architecture model predictive control optimization strategy employs different prediction fields, sampling times, and constraints. This allows the upper layer to optimize energy consumption within its long field of view, while the remaining layers dynamically track the temperature. This problem is a multi-state, multi-constraint nonlinear dynamic optimization problem.

[0236] Using equations (31), (32), and (33), two state variables of the cabin and battery thermal management system are selected as battery pack temperature T. bat and cabin temperature T cab That is, it can be expressed as x = [T] bat ,T cab ] T The three control variables are the liquid cooling water pump speed ω. pump Air-cooled fan speed ω fan And blower inlet air mass flow rate W bl That is, it can be expressed as u = [ω pump ,ω fan W bl ] T ;

[0237] To construct a two-layer model predictive control framework, the upper-layer objective function and constraints are first defined as follows:

[0238]

[0239] Where k+s|k represents the prediction of time k+s at time k;

[0240] P cooling (k+s|k) represents the prediction of the cabin and battery thermal management system cooling power at time k+s.

[0241] T bat (k+s|k) represents the prediction of the battery pack temperature at time k+s at time k;

[0242] T bat (k+s-1|k) represents the prediction of the battery pack temperature at time k+s-1 at time k;

[0243] f Tbat (k+s-1|k) is the battery pack temperature change function at time k+s-1;

[0244] T cab (k+s|k) represents the prediction of the cabin temperature at time k+s at time k;

[0245] T cab (k+s-1|k) represents the prediction of the cabin temperature at time k+s-1.

[0246] f Tcab (k+s-1|k) is the cabin temperature change function at time k+s-1;

[0247] f Pcooling (k+s|k) is the cooling power function of the cabin and battery thermal management system at time k+s;

[0248] ω pump (k+s-1|k) is the liquid cooling water pump speed at time k+s-1, calculated based on the prediction of time k+s at time k;

[0249] ω fan (k+s-1|k) is the air-cooled fan speed at time k+s-1, calculated based on the prediction of time k+s at time k;

[0250] W bl (k+s-1|k) is the blower intake mass flow rate at time k+s-1, calculated based on the prediction of time k+s at time k;

[0251] ΔT u The sampling time of the upper layer, The target temperature for the battery pack. The target temperature for the cabin, N u For the upper-level prediction field of view, s = 1, 2, ..., N u K ub Weights are assigned to the degree to which the battery pack temperature approaches the target temperature of the battery. K uc The weighting of the degree to which the cabin temperature approaches the target cabin temperature. This is the lower limit of the battery pack's temperature. This represents the upper limit of the battery pack temperature. T is the lower limit of cabin temperature. ca U b L is the upper limit of cabin temperature; ω pump_min ω represents the minimum pump speed. pump_max ω represents the maximum speed of the water pump. fan_min ω represents the minimum fan speed. fan_max W represents the maximum fan speed. bl_min W represents the minimum inlet mass flow rate of the blower. bl_max To maximize the mass flow rate of the blower intake, the optimal state output x = [T] was obtained by optimizing the upper-level objective function. bat ,T cab ] T Two of the state variables T bat Let T be the desired temperature of the battery pack. bat_ref T cabThe desired cabin temperature is denoted as T. cab_ref ;

[0252] Next, the lower-level objective function and constraints are defined as follows:

[0253]

[0254]

[0255]

[0256] ω pump_min ≤ω pump (k+r-1|k)≤ω pump_max

[0257] ω fan_min ≤ω fan (k+r-1|k)≤ω fan_max

[0258] W bl_min ≤W bl (k+r-1|k)≤W bl_max

[0259] Where k+r|k represents the prediction of time k+r at time k;

[0260] P cooling (k+r|k) is the prediction of the cabin and battery thermal management system cooling power at time k+r.

[0261] T bat (k+r|k) is the prediction of the battery pack temperature at time k+r at time k;

[0262] T bat_ref (k+r|k) is the prediction of the expected temperature of the battery pack at time k+r at time k;

[0263] T bat (k+r-1|k) is the prediction of the battery pack temperature at time k+r-1.

[0264] Let be the function of battery pack temperature change at time k+r-1;

[0265] T cab (k+r|k) represents the prediction of the cabin temperature at time k+r.

[0266] T cab_ref (k+r|k) is the prediction of the desired cabin temperature at time k+r.

[0267] T cab(k+r-1|k) represents the prediction of the cabin temperature at time k+r-1.

[0268] f Tcab (k+r-1|k) is the cabin temperature change function at time k+r-1;

[0269] ω pump (k+r-1|k) is the liquid cooling water pump speed at time k+r-1, which is obtained by predicting time k+r at time k;

[0270] ω fan (k+s-1|k) is the air-cooled fan speed at time k+r-1, calculated based on the prediction of time k+r at time k;

[0271] W bl (k+s-1|k) is the blower intake mass flow rate at time k+s-1, calculated based on the prediction of time k+r at time k;

[0272] ΔT l ΔT is the sampling time of the lower layer. u >ΔT l N l For the lower-level prediction field of view, r = 1, 2, ..., N l N here u >N l ;K lb K is the weight for how close the battery pack temperature is to the desired battery pack temperature. lb =1 / max[T bat (r)-T bat_ref (r)] 2 K lc K is the weight given to the degree to which the cabin temperature reaches the desired cabin temperature. lc =1 / max[T cab (r)-T cab_ref (r)] 2 ;

[0273] Step 7: Obtain the optimal control quantity and complete the optimization.

[0274] The optimal control variable sequence u is obtained through the two-level model prediction framework constructed in step six. * for:

[0275] u * =arg min J (36)

[0276] When s = 1, the optimal control quantity ω at time k is calculated. pump (k), ω fan (k) and W blSubstituting (k) into the system model, we can calculate the state output T at time k+1. bat (k+1) and T cab (k+1), substituting it back into the upper-level objective function, we obtain the optimal control quantity ω at s=2, i.e., time k+1. pump (k+1), ω fan (k+1) and W bl (k+1), and calculate the state output T at time k+2. bat (k+2) and T cab (k+2), in the upper prediction field N u The internal rolling repetitive calculation yields the optimal state output at each time step. This optimal state output is then substituted into the lower-level objective function as the target values ​​for cabin temperature and battery pack temperature for temperature tracking. This is performed within the lower-level prediction field N. l Within this process, the lower-level objective function is then subjected to rolling optimization calculations to ultimately obtain the optimal control variable sequence u. * This is to achieve the goal of rolling optimization.

[0277] The following further elaborates on the verification results of the cooling optimization method for cabin and battery thermal management of a connected electric vehicle provided by this invention:

[0278] To further verify the effectiveness of the present invention, the proposed dual-frame model predictive control optimization strategy for the cabin and battery thermal management of connected electric vehicles was verified under the SC03 operating condition (a specific driving cycle under the condition of full-load operation of air conditioning in summer high temperature).

[0279] The battery pack consists of 13 battery cells connected in parallel and series configurations (30*10). The batteries are 26650 lithium iron phosphate batteries, each with a voltage of 3.3V and a capacity of 2300mAh. The liquid cooling circuit uses a 50% volume concentration ethylene glycol aqueous solution as the coolant. The initial battery pack temperature T is set to [value missing] when k=0. bat (0) Cabin temperature T cab (0) Ambient temperature T amb (0) Temperature of interior items in the cabin T in (0) Cabin outer shell temperature T shell (0) and the coolant inlet temperature T of the battery pack in the cooling circuit. liq_in (0) is 308K, the target temperature of the battery pack. The target temperature in the cabin is 303K. It is 298K;

[0280] State constraints include: battery pack temperature upper limit 313K, lower limit 288K; cabin temperature upper limit 308K, lower limit 293K. Control constraints include: water pump speed upper limit 8000r / min, lower limit 0; fan speed upper limit 200r / min, lower limit 0; blower mass flow rate upper limit 0.15kg / s, lower limit 0.01kg / s.

[0281] The proposed speed prediction (first step) and two-layer model predictive control optimization method (sixth and seventh steps) are compared with the commonly used centralized model predictive control optimization method in the prior art. The battery pack temperature change curve is shown in the figure. Figure 4 As shown, the cabin temperature change curve is as follows: Figure 5 As shown, the control flow rate pump speed curve is as follows: Figure 6 As shown, the fan speed curve is as follows: Figure 7 As shown in the figure, the curve of the blower's air mass flow rate is as follows: Figure 8 As shown, the cooling power curve of the thermal management system is as follows: Figure 9 As shown; from Figure 4 Battery pack temperature T bat The comparison curves show that, during the driving cycle, compared to centralized model predictive control, the mode using a two-layer frame for optimization and tracking, although initially experiencing a slightly higher temperature, ultimately reached the target temperature, with the error remaining within 1K; from Figure 5 It can be seen that the cabin temperature was rapidly reduced, with a maximum temperature error of 0.7K; this is because in a hot environment, the external ambient temperature T... amb It must be higher than the cabin air intake temperature T. ain Cars typically use an internal air circulation system to quickly reach the comfort requirements of the passenger cabin; however, this also increases the system's energy consumption. Figure 6 It can also be seen that the blower intake mass flow rate W under the method of the present invention bl More stable; Figure 7 The rotational speed ω of the medium-cooled water pump pump Ultimately, it operates at 2500 r / min, while centralized model predictive control methods mostly operate at 3800 r / min; Figure 8 The fan speed ω can be seen from this. fan The operating frequency is also significantly lower than the speed under centralized model predictive control, resulting in higher operating efficiency; therefore, it can be seen that the cooling power of the thermal management system is significantly lower than that of single-layer model predictive control, such as... Figure 9 As shown; the total energy consumption of centralized model predictive control is 357.73 kJ, while the energy consumption of the model predictive control optimization method based on a two-layer architecture is 340.77 kJ, reducing cooling energy consumption by 9.3%; other vehicle and battery physical parameters are as follows: Figure 10 As shown; parameters for coolant, water pump, and compressor are as follows: Figure 11 As shown; In summary, in hot and high-temperature environments with the air conditioning running at full load, the cooling optimization method for cabin and battery thermal management of connected electric vehicles proposed in this invention can quickly reduce the operating temperature of the battery pack and meet the requirements for cabin air comfort, while reducing the cooling energy consumption of the thermal management system and improving the energy efficiency of the electric vehicle.

Claims

1. A cooling optimization method for the thermal management of the passenger compartment and battery of a connected electric vehicle. In this method, the passenger compartment and battery thermal management system includes an air conditioning circuit, a battery pack liquid cooling circuit, and a passenger compartment air circuit. Battery pack cooling is achieved simultaneously through non-contact liquid cooling and active air cooling. Non-contact liquid cooling uses a water pump to drive the circulation of coolant in the liquid cooling pipes, exchanging heat with the air conditioning refrigerant to reduce the battery pack temperature. Active air cooling uses a variable-speed fan to blow cool air from the passenger compartment into the battery pack to assist in reducing the battery temperature. Passenger compartment cooling is achieved by a blower delivering air cooled by the air conditioning into the passenger compartment. Therefore, this method achieves rapid reduction of the battery pack and passenger compartment temperature and reduces energy consumption by controlling the water pump speed, variable-speed fan speed, and blower air mass flow rate. The specific steps are as follows: The first step is to establish a vehicle speed prediction model based on the Extreme Learning Machine algorithm. Based on historical vehicle speed data, an extreme learning machine algorithm is constructed to establish a prediction model for future vehicle speed within the field of view. This method is a training algorithm for a single hidden layer feedforward neural network, which consists of three layers: an input layer, a hidden layer, and an output layer. The first step is the vehicle speed training process based on the Extreme Learning Machine algorithm, given a known vehicle speed training set S = {(X...} j ,t j )|X j ∈R n ,t j ∈R m Let X = {j = 1, 2, ..., N}, where X = ... j =[x j1 ,x j2 ,...,x jn ] T Given an n-dimensional historical vehicle speed sequence, t j =[t j1 ,t j2 ,...,t jm ] T Let R be an m-dimensional target vehicle speed sequence, N be the number of historical vehicle speed samples in the vehicle speed training set, j be a natural number from 1 to N, and R be a variable. n Let R be the set of n-dimensional real numbers. m Given an m-dimensional set of real numbers, the output o of a single-hidden-layer feedforward neural network j Represented as: Where L is the number of hidden layer nodes, 0 < L < N, W i =[w i1 ,w i2 ,...,w in ] T Let b be the input weight of the i-th hidden layer node. i β is the bias of the i-th hidden layer node. i Let g(·) be the output weight of the i-th hidden layer node, i = 1, 2, ..., L, and g(·) be the activation function of the hidden layer, a nonlinear function that provides the nonlinear mapping for the system. Let g(·) be the Sigmoid function, in the form of: During training, the input weights W of the hidden layer nodes i and the bias b of the hidden layer nodes i Since the output is randomly generated and requires no updating, the desired output o of a single-hidden-layer feedforward neural network is... j The output weights β of the hidden layer nodes need to be calculated. i ; Furthermore, because when the number of hidden layer nodes L approaches the number of historical vehicle speed samples N in the vehicle speed training set infinitely, the output o of the feedforward neural network... j It can approach the target vehicle speed t in the training set infinitely. j That is, satisfying: From equations (1) and (2), we get: The above formula can be expressed in matrix form as follows: Hβ=T (4) Where H is the output matrix containing L hidden layer nodes, β is the output weight matrix between the hidden layer and the output layer, and T is the desired output matrix, then matrices H, β, and T are expressed as follows: Since the number of historical vehicle speed samples N in the training set is greater than the number of hidden layer nodes L, the goal of training the neural network is to minimize the approximation error. Finding the output weight matrix β is equivalent to finding the least-squares solution for the linear system Hβ = T, i.e.: ||Hβ * -T||=min β ||Hβ-T||,β∈R L×m (6) The least squares solution β is then obtained. * This is the optimal solution to equation (4): b * =H + T (7) Among them, H + Let H be the Moore-Penrose generalized inverse matrix; Secondly, the vehicle speed prediction process is based on the Extreme Learning Machine (ELM) algorithm. The previous step, the vehicle speed training process based on the ELM algorithm, determined the input weight matrix W = [W1, W2, ..., W...]. L ] T And the bias matrix b = [b1, b2, ..., b L ] T The output weight matrix β = β is calculated. * The test vehicle speed sequence X is now being presented. q =[x q1 ,x q2 ,...,x qm ] T Substituting q = 1, 2, ..., M into equation (1), we obtain the predicted vehicle speed output sequence o. q =βg(WX q +b), where o q To predict the vehicle speed output sequence, M is the number of test vehicle speed sequence samples, and q is a natural number from 1 to M. Therefore, the predicted vehicle speed V obtained by the Extreme Learning Machine algorithm is denoted as: V=o q (8) The second step is to establish a thermal model of the battery pack. The thermal model of a battery pack must reflect both the electrical and thermal characteristics of the battery. Therefore, based on the law of conservation of energy, a thermal model for a lithium-ion battery pack is established: in, I represents the temperature change of the battery pack per unit time. bat R represents the current of the battery pack during vehicle operation. bat m is the internal resistance of the battery pack. bat For the mass of the battery pack, c bat Q represents the specific heat capacity of the battery pack. bat_liq Q is used to cool and absorb heat from the battery pack in the liquid cooling circuit. bat_cab Active air cooling absorbs heat from the battery pack; battery pack voltage U bat Able to be determined by open-circuit voltage U oc and battery pack current I bat With the internal resistance R of the battery pack bat The product is represented as: U bat =U oc -I bat R bat (10) Battery pack current I bat The function to meet the success rate requirement is as follows: I bat =(P tr +P comp +P bl +P pump +P fan ) / U bat (11) Among them, P tr For the power consumption of vehicle traction, P comp For the power consumption of the air conditioner compressor, P bl For the power consumption of the cabin air intake blower, P pump For the power consumption of the liquid cooling circuit water pump, P fan The power consumption of the air-cooled fan is used; therefore, the battery pack current I is obtained from equations (10) and (11). bat for: Substituting equation (12) into equation (9), the battery thermal model is updated as follows: In equation (13), the traction power consumption P tr Represented as: Where V is the vehicle's speed. Let F be the acceleration of the car. Due to the uncertainty of the future speed, the car's speed is the predicted speed V obtained in the first step based on the extreme learning machine algorithm. m is the mass of the car. a It is the aerodynamic drag during driving, F roll It is the rolling resistance of the car, F gxt The slope resistance is expressed as follows: F roll =mgC r (16) F gxt =mgsinφ (17) Among them, C d ρ is the aerodynamic drag coefficient. a It is air density, A f It is the frontal area of ​​the vehicle when it is in motion, C r Here, φ is the rolling resistance coefficient, and φ is the slope angle. We assume φ = 0 here, so F... gxt =0; The third step is to establish discrete temperature models for the air conditioning loop and cabin loop, oriented towards control. Because air conditioning simulation models involve complex thermodynamics and fluid dynamics, they are unsuitable for controller design. Therefore, it is necessary to establish a discrete model of the air conditioning loop and cabin temperature for control purposes. The discrete model of cabin temperature can be represented as... Where k = 0, 1, 2..., when k = 0 is the initial time, T cab (k) represents the cabin temperature at time k, T cab (k+1) represents the cabin temperature at time k+1, T in (k) represents the temperature of the contents inside the cabin at time k, T shell (k) represents the cabin outer shell temperature at time k, W. bl (k) represents the blower inlet mass flow rate at time k, T ain (k) represents the cabin air inlet temperature at time k, where it is assumed that the temperature of the interior contents is T. in and cabin shell temperature T shell Since the changes in operating conditions are not significant, both are set as constants. γ1, γ2, γ3 and τ1 are identification parameters obtained by curve fitting using the least squares method based on the cabin temperature output of the electric vehicle cooling model in AMESim. Since the discrete model of evaporator temperature can be expressed as: Among them, T evap (k) represents the evaporator temperature at time k, T evap (k+1) represents the evaporator temperature at time k+1. γ4, γ5, and τ2 are the target temperature of the evaporator, and the identification parameters are obtained by curve fitting using the least squares method through the evaporator output of the electric vehicle cooling model in AMESim. Therefore, the cabin air inlet temperature T at time k ain (k) can be represented as: T ain (k)=γ6T evap (k)+γ7W bl (k)+τ3 (20) Among them, γ6, γ7, and τ3 are identification parameters obtained by curve fitting using the least squares method from the cabin air inlet temperature output of the electric vehicle cooling model in AMESim. It can be seen that the control input in the air conditioning and cabin circuit is the blower air inlet mass flow rate W. bl ; The fourth step is to establish a discrete temperature model for the liquid cooling circuit. The battery pack cooling method in the liquid-cooled circuit is active air cooling and non-contact liquid cooling; the coolant inlet temperature and coolant outlet temperature of the battery pack in the cooling circuit can be defined as follows: Among them, T liq_in (k+1) represents the coolant inlet temperature of the battery pack in the cooling circuit at time k+1. Let be the mass flow rate of the coolant in the liquid cooling circuit at time k, and χ1, χ2, and The identification parameters are obtained by curve fitting using the least squares method based on the coolant inlet temperature output of the electric vehicle cooling model in AMESim. Among them, T liq_out (k+1) represents the coolant outlet temperature of the battery pack in the cooling circuit at time k+1, A liq h is the convective heat transfer area between the coolant and the battery pack. liq C is the convective heat transfer coefficient between the coolant and the battery pack. liq This refers to the specific heat capacity of the coolant. Among them, V pump η is the pump displacement. pump ρ is the volumetric efficiency of the water pump. liq ω is the liquid density of the coolant. pump (k) represents the control quantity, the pump speed, at time k. Therefore, based on the thermal model of the lithium-ion battery pack given by equation (9), the heat Q of the active air-cooled battery pack at time k is obtained. bat_liq (k) and the heat Q of the liquid-cooled absorption battery pack bat_cab (k) are respectively represented as: Where, ω fan (k) represents the rotational speed of the active air-cooled fan at time k. Here, the model parameter μ1 is 0.2671 and the model parameter μ2 is 2677.

3. Step 5: Cooling power calculation The main energy consumer in a car's air conditioning circuit is the air conditioning compressor, specifically the air conditioning compressor's power consumption P. comp The main energy consumer in the cabin circuit is the blower, i.e., the blower power consumption P. bl The main energy consumption for cooling the battery pack comes from the water pump and the air-cooled fan; that is, the water pump power consumption is P. pump Power consumption of air-cooled fan P fan ; First, calculate the pump power consumption P at time k. pump (k): Among them, P pump_m (k) represents the mechanical power of the pump at time k, η m The power conversion efficiency of the water pump is ΔP. pump For water pump pressure drop; Secondly, calculate the compressor power consumption P at time k. comp Within a certain timeframe, the power required to lower both the battery pack temperature and the cabin temperature from their initial values ​​to their target temperatures comes from the air conditioning compressor. Therefore, assuming the air conditioning compressor's energy consumption consists of two parts: one for lowering the cabin air temperature and the other for lowering the coolant temperature in the cooling circuit, ignoring other low-power components, the power consumption P of the air conditioning compressor at time k can be obtained from the energy balance relationship. comp (k) is: Among them, C comp η is the specific heat capacity of air. comp W is the coefficient of performance for air conditioning. bl (k) represents the blower inlet mass flow rate at time k, T amb (k) represents the ambient temperature at time k, T ain (k) represents the cabin air inlet temperature at time k, C liq The specific heat capacity of the coolant. Let T be the mass flow rate of the coolant in the liquid cooling circuit at time k. liq_in (k) represents the coolant inlet temperature of the battery pack in the cooling circuit at time k, T liq_out (k) represents the coolant outlet temperature of the battery pack in the cooling circuit at time k; Next, calculate the cabin air intake blower power consumption P at time k. bl (k) and fan power consumption P fan (k); Blower power P bl It controls the cabin air intake mass flow rate W bl The function, fan power P fan It controls the fan speed ω fan The functions, after data fitting, are estimated as follows: Among them, K bl For the fan parameters, λ1, λ2, λ3, α1, α2, α3 and α4 are all identification parameters obtained by curve fitting using the least squares method based on the blower power and fan power output of the electric vehicle cooling model in AMESim. Therefore, the cabin and battery thermal management system cooling power P at time k is defined. cooling (k) is: P cooling (k)=P pump (k)+P comp (k)+P fan (k)+P bl (k) (30) From equations (26), (27), (28), and (29), it can be seen that the cooling power P of the cabin and battery thermal management system is... cooling It is the water pump speed ω pump Fan speed ω fan blower inlet air mass flow rate W bl Battery pack temperature T bat and cabin temperature T cab The function, therefore the cabin and battery thermal management system cooling power P at time k. cooling (k) can be represented as: P cooling (k)=f(T bat (k),T cab (k),ω pump (k),ω fan (k),W bl (k)) (31) Similarly, we can conclude that the temperature change of the battery pack is related to the water pump speed ω. pump Fan speed ω fan The cabin temperature change is a function of the predicted vehicle speed V and the temperature T of the interior items. in Cabin outer shell temperature T shell And blower inlet air mass flow rate W bl Therefore, equations (13) and (18) are discretized according to the sampling time ΔT to obtain the discrete model T of the battery pack temperature. bat Discrete model T of (k+1) and cabin temperature cab (k+1) are respectively: T bat (k+1)=T bat (k)+f(T bat (k),ω pump (k),ω fan (k),V(k))·ΔT (32) T cab (k+1)=T cab (k)+f(T cab (k),T in (k),T shell (k),W bl (k))·ΔT (33) Step 6: Construct a two-layer model predictive control framework Using equations (31), (32), and (33), two state variables of the cabin and battery thermal management system are selected as battery pack temperature T. bat and cabin temperature T cab That is, it can be expressed as x = [T] bat ,T cab ] T The three control variables are the liquid cooling water pump speed ω. pump Air-cooled fan speed ω fan And blower inlet air mass flow rate W bl That is, it can be expressed as u = [ω pump ,ω fan W bl ] T ; To construct a two-layer model predictive control framework, the upper-layer objective function and constraints are first defined as follows: Where k+s|k represents the prediction of time k+s at time k; P cooling (k+s|k) represents the prediction of the cabin and battery thermal management system cooling power at time k+s. T bat (k+s|k) represents the prediction of the battery pack temperature at time k+s at time k; T bat (k+s-1|k) represents the prediction of the battery pack temperature at time k+s-1 at time k; Let be the function of battery pack temperature change at time k+s-1; T cab (k+s|k) represents the prediction of the cabin temperature at time k+s at time k; T cab (k+s-1|k) represents the prediction of the cabin temperature at time k+s-1. Let be the cabin temperature change function at time k+s-1; Let be the cooling power function of the cabin and battery thermal management system at time k+s; ω pump (k+s-1|k) is the liquid cooling water pump speed at time k+s-1, which is obtained by predicting time k+s at time k. ω fan (k+s-1|k) is the air-cooled fan speed at time k+s-1, calculated based on the prediction of time k+s at time k; W bl (k+s-1|k) is the blower intake mass flow rate at time k+s-1, calculated based on the prediction of time k+s at time k; ΔT u The sampling time of the upper layer, The target temperature for the battery pack. The target temperature for the cabin, N u For the upper-level prediction field of view, s = 1, 2, ..., N u K ub Weights are assigned to the degree to which the battery pack temperature approaches the target temperature of the battery. K uc The weighting of the degree to which the cabin temperature approaches the target cabin temperature. This is the lower limit of the battery pack's temperature. This represents the upper limit of the battery pack temperature. This is the lower limit of the cabin temperature. This represents the upper limit of cabin temperature; ω pump_min ω represents the minimum pump speed. pump_max ω represents the maximum speed of the water pump. fan_min ω represents the minimum fan speed. fan_max W represents the maximum fan speed. bl_min W represents the minimum inlet mass flow rate of the blower. bl_max To maximize the mass flow rate of the blower intake, the optimal state output x = [T] was obtained by optimizing the upper-level objective function. bat ,T cab ] T Two of the state variables T bat Let T be the desired temperature of the battery pack. bat_ref T cab The desired cabin temperature is denoted as T. cab_ref ; Next, the lower-level objective function and constraints are defined as follows: oh pump_min ≤ω pump (k+r-1|k)≤ω pump_max oh fan_min ≤ω fan (k+r-1|k)≤ω fan_max W bl_min ≤W bl (k+r-1|k)≤W bl_max Where k+r|k represents the prediction of time k+r at time k; P cooling (k+r|k) is the prediction of the cabin and battery thermal management system cooling power at time k+r. T bat (k+r|k) is the prediction of the battery pack temperature at time k+r at time k; T bat_ref (k+r|k) is the prediction of the expected temperature of the battery pack at time k+r at time k; T bat (k+r-1|k) is the prediction of the battery pack temperature at time k+r-1. Let be the function of battery pack temperature change at time k+r-1; T cab (k+r|k) represents the prediction of the cabin temperature at time k+r. T cab_ref (k+r|k) is the prediction of the desired cabin temperature at time k+r. T cab (k+r-1|k) represents the prediction of the cabin temperature at time k+r-1. Let be the cabin temperature change function at time k+r-1; ω pump (k+r-1|k) is the liquid cooling water pump speed at time k+r-1, calculated based on the prediction of time k+r at time k; ω fan (k+s-1|k) is the air-cooled fan speed at time k+r-1, calculated based on the prediction of time k+r at time k; W bl (k+s-1|k) is the blower intake mass flow rate at time k+s-1, calculated based on the prediction of time k+r at time k; ΔT l ΔT is the sampling time of the lower layer. u >ΔT l N l For the lower-level prediction field of view, r = 1, 2, ..., N l N here u >N l ;K lb K is the weight for how close the battery pack temperature is to the desired battery pack temperature. lb =1 / max[T bat (r)-T bat_ref (r)] 2 K lc K is the weight given to the degree to which the cabin temperature reaches the desired cabin temperature. lc =1 / max[T cab (r)-T cab_ref (r)] 2 ; Step 7: Obtain the optimal control quantity and complete the optimization. The optimal control variable sequence u is obtained through the two-level model prediction framework constructed in step six. * for: he * =argminJ (36) When s = 1, the optimal control quantity ω at time k is calculated. pump (k), ω fan (k) and W bl Substituting (k) into the system model, we can calculate the state output T at time k+1. bat (k+1) and T cab (k+1), substituting it back into the upper-level objective function, we obtain the optimal control quantity ω at s=2, i.e., time k+1. pump (k+1), ω fan (k+1) and W bl (k+1), and calculate the state output T at time k+2. bat (k+2) and T cab (k+2), in the upper prediction field N u The internal rolling repetitive calculation yields the optimal state output at each time step. This optimal state output is then substituted into the lower-level objective function as the target values ​​for cabin temperature and battery pack temperature for temperature tracking. This is performed within the lower-level prediction field N. l Within this process, the lower-level objective function is then subjected to rolling optimization calculations to ultimately obtain the optimal control variable sequence u. * This is to achieve the goal of rolling optimization.

Images