THERMAL SIMULATION DEVICE FOR ELECTRIC VEHICLE BATTERIES BY AUTOMATIC LEARNING, METHOD, VEHICLE COMPRISING SUCH A DEVICE

Abstract

The invention relates to a thermal simulation device (1) for a traction battery (2) of an automobile, comprising: - a battery (2) cooling system (3) which includes a cooler (3a), a pump (6) for circulating a refrigerant (4), and a heat exchanger (7); - a conditioning means (8, 8a) for applying a given temperature to the battery (2) for given times; - at least one simulation means (10) for varying operating parameters of the battery (2) and the pump (6); - a means for determining thermal equilibrium and / or power reduction; - a machine learning module (12) configured to integrate the data corresponding to said operating parameters and determine a temperature variation law during battery (2) use. Fig. 1

Description

Title of the invention: THERMAL SIMULATION DEVICE FOR AN ELECTRIC VEHICLE BATTERY BY AUTOMATIC LEARNING, METHOD, VEHICLE COMPRISING SUCH A DEVICE

[0001] The invention relates to the field of devices and systems for estimating the temperature of the battery during use.

[0002] Accurate temperature estimation makes it possible to anticipate potential premature battery deterioration and to react, for example, by cooling the battery. Furthermore, battery overheating leads to excessive electrical energy consumption and results in premature discharge.

[0003] Simulation solutions have been proposed for predicting the temperature of the traction battery. However, these solutions are not satisfactory. They are based on a simplification and modeling of the physical characteristics of the battery, including the geometry, the thermal characteristics (in particular the thermal capacity, thermal conductivity and density), as well as the source of the battery's heat, which depends on several factors (in particular the temperature, the electrical power, the state of charge (SOC) and the state of health (SOH) of the battery).

[0004] These models oversimplify the physical characteristics, resulting in significant inaccuracies and in reliability. In reality, it is impossible to obtain sufficiently accurate models for reliable simulation in the prior art. For example, for the heat source, the battery supplier does not have reliable measurements, and the data provided is insufficient to develop a model that covers the entire range of the most impactful operating parameters (in particular temperature, state of charge, and health status).

[0005] In addition, these models require very long experiments involving excessively long periods of time.

[0006] Thus, a first objective of the present invention is to propose a modeling solution that is substantially accurate and reliable under different battery usage conditions.

[0007] A second objective is to propose a simulation solution that allows modeling of battery temperature evolutions while significantly limiting the number and complexity of experiments.

[0008] To achieve these objectives, the invention proposes a thermal simulation device for a motor vehicle traction battery, comprising - a battery cooling system that includes an attached cooler against at least one side of the battery, a refrigerant circulation pump, and a heat exchanger; - a conditioning method for applying a given temperature to the battery for given times; - at least one simulation method to vary battery and pump usage parameters; - a means to determine if a thermal equilibrium of the battery is reached and / or a means to determine if the battery reduces its power; - a machine learning module configured to integrate the data corresponding to said usage parameters and determine a temperature variation law during battery use.

[0009] The machine learning module is, in particular, a deep machine learning module. This module is based on a deep neural network. It is an artificial intelligence module. This module is implemented by computer means and is known in itself.

[0010] Advantageously, the device according to the invention involves the realization of a precise and reliable model taking into account the physical characteristics of the battery through machine learning.

[0011] Furthermore, the simulation device according to the invention does not require lengthy and complex experiments, but only tests where certain parameters are modified. This implies a significantly reduced time for optimal efficiency.

[0012] According to other aspects taken in isolation, or combined according to all technically feasible combinations: - the machine learning module determines a variation law including a temperature gradient as a function of time; and / or - The law of variation includes at least one of the following formulas:

[0013] Qgen = qx Tbr x P,

[0014] Tht-Thh^c^Qgen,

[0015] Q = c^'f^(Thl-T , A; LJ ^trans ^3 J \ bb cornant J [0 ° 16] ïq q \ x A? > \ ^gen ^trans / 4 \ 2 2 /

[0017] T _ i grad ----&---

[0018] where ô^est the heat generated by the battery; ci is a constant coefficient; Tbt is the temperature measured outside the battery-cooler interface; P is the electrical power of the battery; Tbb is the temperature at the battery-cooler interface; 62 is a constant coefficient representing the internal thermal resistance in the battery; Q is the heat transferred by the battery; ^trans f is the flow rate of the cooling fluid; Tcooiant is the inlet temperature of the battery cooler; c3 is the heat transfer coefficient;

[0019] Tgrad is the temperature gradient; A t is the time step in the simulation; and The indices "new" and "old" indicate the temperature of the current time step and the temperature of the previous time step; and / or

[0020] - said simulation means is configured to apply several powers of battery, multiple pump operating flow rates, multiple battery charge states, multiple battery health states, or a combination of at least two of these parameters; and / or

[0021] - the simulation device further includes an insulation means for thermally isolating mically at least one side of the battery where the cooler is not attached; and / or

[0022] - the conditioning means for applying a given temperature is configured to achieve a uniform internal battery temperature; and / or

[0023] - said applied temperature is 30°C.

[0024] The invention further relates to a method for thermal simulation of a motor vehicle traction battery, comprising - a step to apply a given temperature to the battery for given times; - a step to vary several parameters of the battery and the pump; - a step to cool the battery; - a step to determine if a thermal equilibrium is reached or if the battery reduces its power; - a step to integrate the data from the different parameters and determine a temperature variation law during battery use by means of a machine learning module.

[0025] Another object of the invention relates to a computer program comprising program code instructions for the execution of at least some of the steps of the simulation method according to the invention, when said program is running on at least one computer.

[0026] The invention also relates to a motor vehicle platform comprising a simulation device according to the invention.

[0027] Another object of the invention relates to a motor vehicle comprising a cooling control unit connected to a simulation device according to the invention so as to be able to exchange data between said control unit and said simulation device, and control the pump according to the law of temperature variation.

[0028] The invention will be further detailed by describing non-limiting embodiments, and based on the accompanying figures illustrating variants of the invention, in which: - [Fig.l] schematically illustrates a thermal simulation device according to a preferred variant of the invention; - [Fig. 2] schematically illustrates a comparison between machine learning simulation data and actual measurements under initial conditions; and - [Fig.3] schematically illustrates a comparison similar to that of the previous figure under second conditions.

[0029] The invention relates to a thermal simulation device 1 for a traction battery 2 of a motor vehicle. It is a system for evaluating the thermal variations of said battery 2 under specified conditions. The battery 2 is an electric traction motor battery of a motor vehicle.

[0030] The simulation device 1 includes a cooling system 3. The cooling system 3 is configured to cool the battery 2. The cooling system 3 includes a cooler 3a affixed to at least one side of the battery 2. The cooler 3a can be fixed to, for example glued to, said side of the battery 2.

[0031] In particular, the cooler 3a is configured to perform heat exchange with the battery 2 via a contact interface on said side. The interface comprises a surface of the battery 2 having a form-matching relationship with a corresponding surface of the cooler 3a.

[0032] In the illustrated variant, the cooler 3a has a plate shape.

[0033] Preferably, the cooler 3a is configured to have the same heat transfer coefficient (or thermal transfer coefficient) as the battery 2.

[0034] The cooler 3a is traversed by a refrigerant 4 which is preferably a liquid or cooling fluid.

[0035] The cooling system 3 further includes a cooling circuit 5 in which the refrigerant 4 circulates.

[0036] The cooling system 3 further includes a circulation pump 6 of the Refrigerant 4. Pump 6, once activated, allows the refrigerant 4 to circulate in the cooling circuit 5. In the illustrated variant, pump 6 is located downstream of cooler 3a in the cooling circuit 5. This is in particular the hot part 5a of the cooling circuit 5.

[0037] The cooling system 3 further includes a heat exchanger 7. The heat exchanger 7 is configured to absorb heat from the cooler 3a following heat exchange with the coil 2. The heated refrigerant 4 is cooled at the heat exchanger 7 so that the remainder of the cooling circuit 5 is the cold part 5b. The cooled refrigerant 4 can then recirculate through the cooler 3a for further heat exchange between the cooler 3a and the coil 2. The cycle can then begin again. In the illustrated embodiment, the heat exchanger 7 is located downstream of the pump 6 in the cooling circuit 5.

[0038] The cooling system 3 further includes a conditioning means 8, 8a for applying a given temperature to the battery 2 for given times. This includes, in particular, elements for simulating ambient temperatures and / or heating of the battery 2 under operating conditions. This could be a chamber equipped with a heating device 8a or an air conditioning system. The control 8 can be computer-controlled. The battery 2 is thus subjected to specific temperature conditions that can be varied. In particular, a given temperature is applied for a given time during one test, then a different temperature is applied for a given time during another test, and so on. For example, depending on the test, the temperature could be 10, 15, 20, 25, 30°C, or other temperature values.The conditioning means 8, 8a is implemented so as to have a homogeneous internal temperature.

[0039] According to one embodiment, the simulation device 1 further comprises an insulation means 9 for thermally insulating at least one side of the battery 2 where the cooler 3a is not attached. This is, for example, a case made of insulating material or combined with an insulating material. Advantageously, this makes it possible to focus the heat dissipation from the battery 2 towards the cooler 3a, and to improve the efficiency of the cooler 3a.

[0040] The cooling system 3 further includes at least one simulation means 10. The simulation means are used during the tests. The simulation means allow for varying the operating parameters of the battery 2 and the operating parameters of the pump 6.

[0041] The operating parameters of battery 2 may include the power in use. Thus, several battery power ratings can be applied in different tests via a simulation of battery 2's use to power the motor of traction. For example, the degree to which the accelerator pedal is pressed helps determine the battery power required during testing.

[0042] The battery 2 usage parameters may include the battery state of charge (or State of Charge - SOC). Thus, tests may be carried out with several different battery states of charge, for example at 100%, 75%, 50%, 25%, 10% or other state of charge values.

[0043] The same applies to the health status of battery 2, reflecting the aging of battery 2. The health status can be expressed in discrete values, for example from 1 to 10. The tests can thus take into account the health status of battery 2 in simulation.

[0044] The operating parameters of the pump 6 may include the activations and deactivations of the pump 6 and the refrigerant flow rates 4.

[0045] In other words, more generally, the simulation means can be configured to apply several battery 2 powers, several battery 2 charge states, several battery 2 health states, several pump 6 operating rates, or a combination of at least two of these parameters.

[0046] The cooling system 3 further includes a means 11, configured to determine whether thermal equilibrium of the battery 2 has been reached. In particular, this means is implemented under the various test conditions. It allows for determining whether a test can be stopped. In other words, the test is stopped when the temperature of the battery 2 is stable. This refers specifically to the internal temperature of the battery 2.

[0047] This means may include a temperature sensor or probe.

[0048] Alternatively or in combination, the cooling system 3 further comprises a, or preferably said means 11, configured to determine whether the battery 2 is reducing its power, i.e., whether it is beginning to derate. The means 11 may include a computer-type module analyzing parameters of the battery 2 for this purpose. Similarly, this means 11 can be implemented under the various test conditions. It allows determining whether a test can be stopped. In other words, the test is stopped when the power of the battery 2 decreases.

[0049] This means 11 may include a power measurement device.

[0050] The cooling system 3 further comprises a machine learning module 12. Preferably, this is a deep machine learning module. It is an artificial intelligence module. This module 12 is based on a deep neural network in the illustrated variant. This module 12 is implemented by computer means and is known per se.

[0051] The machine learning module 12 is configured to integrate the data corresponding to the aforementioned usage parameters resulting from the various tests.

[0052] The machine learning module 12 is further configured to determine a temperature variation law during battery 2 use based on this data.

[0053] Advantageously, the simulation device 1 involves the realization of a precise and reliable model via this variation law from the machine learning module 12. This model takes into account the physical characteristics of the battery 2 under reproducible operating conditions thanks to the data from the different tests by taking advantage of artificial intelligence.

[0054] Furthermore, the simulation device 1 does not require lengthy and complex experiments, but only tests where certain parameters, particularly the use of the battery 2 and the pump 6, are modified. This implies simple experiments with significantly reduced experimentation times and results in optimal efficiency.

[0055] Preferably, the machine learning module 12 determines a variation law including a temperature gradient as a function of time. Advantageously, this allows for better prediction of the temperature variations of battery 2.

[0056] The invention further relates to a thermal simulation method for a traction battery 2 of an automobile. The simulation method makes it possible to evaluate the thermal variations of said battery 2 under specified conditions. The simulation method can be implemented using the thermal simulation device 1 as described above. In this case, the steps consist of implementing some or all of the actions of the various elements of the simulation device 1.

[0057] The simulation method includes a step for applying a given temperature to the battery 2 for given times. The step is implemented via a means for applying a temperature such as the conditioning means 8, 8a described previously.

[0058] The simulation method further includes a step to vary several parameters of the battery 2 and the pump 6. These include in particular the usage parameters of the battery 2 and / or the usage parameters of the pump 6 described above.

[0059] The simulation method further includes a step to cool the battery 2. For example, the battery 2 is cooled if it reaches a given threshold depending on the tests. This step is carried out using a battery cooler 3a such as the one described above.

[0060] The simulation method further includes a step to determine whether thermal equilibrium is reached and / or whether battery 2 reduces its power. In particular, The process includes carrying out several tests. The tests are stopped when the internal temperature of battery 2 is stable or if battery 2 loses power due to derating phenomena.

[0061] The tests result in data, for example in matrix form.

[0062] The processing of this data is carried out using a machine learning module 12. Thus, the method further includes a step to integrate this data from the different parameters, and on the basis of this data, determine a law of temperature variation during the use of the battery 2.

[0063] Regarding now an example of embodiment, the simulation process includes a step to install a cell or battery tray module 2 on a cooler 3a having the same heat transfer coefficient as battery 2.

[0064] Another step consists of thermally insulating all surfaces of the battery 2 except for the interface with the cooler 3a. This step can implement the insulation means 9 described previously.

[0065] Another step consists of conditioning the battery 2 to a temperature (30°C for example depending on the requirement) for a sufficient time to have a homogeneous internal temperature.

[0066] Another step involves starting battery 2 at a constant power and starting pump 6 at a constant speed. The test matrix is ​​a combination of the power of battery 2 and the fluid flow rate. The number of test points must be sufficient to represent a characteristic curve, which is often non-linear over the entire test range.

[0067] The state of charge (SOC) and state of health (SOH) of battery 2 also influence heat production. Consequently, the test matrix is ​​repeated for different SOC and SOH values.

[0068] The tests stop as soon as thermal equilibrium is reached, or as soon as battery 2 begins to reduce its power due to the beginning of derating.

[0069] A machine learning model with a deep neural network architecture is to be installed. The number of layers and the number of nodes in each layer are adapted to achieve sufficient accuracy.

[0070] To demonstrate the performance of this model, a transient simulation is carried out on the basis of the simulation device setup as described above, which can be illustrated by [Fig.1].

[0071] The electrical efficiency of battery 2 is simplified as a dependence on the temperature of battery 2, excluding the influence of the state of charge. The heat production is therefore described by the equation below:

[0072] Q^c^T^P, Or ■ Ô?«;est 'a heat generated by battery 2 ; - c'i is simplified as a constant coefficient; - Thl is the temperature measured at the top of battery 2; and - P is the electrical power of battery 2.

[0073] The temperature of battery 2 at the interface with the cooler 3a is much lower than Tbf due to the internal thermal resistance in the cell of battery 2, hence:

[0074] Tbt-Tbh = c2xQgen, Or - Tbb is the temperature at the heat transfer interface; and - c2 is a constant coefficient representing the internal thermal resistance in battery 2.

[0075] The heat transfer from battery 2 to cooler 3a is modeled by the law below. A simplification is made for the correlation between the heat transfer coefficient and the fluid flow rate, hence: 100761 d™, = '-3 * / X ( Tbh - Ta„,„„ ), OR - Qf is the heat transferred; ^trans ' - f is the flow rate of the cooling fluid; - Tflow is the inlet temperature of cooler 3a of battery 2; and - c3 is the heat transfer coefficient, simplified as a constant coefficient.

[0077] The transient evolution of the temperatures of battery 2 is calculated using the following equation:

[0078] n A v A t - rv (Tbt new+Thbjww Tfa bh_plü Y, V Ugen - ) XA? -q oh 2 / Or - AZ is the time step in the simulation; - The indices "new" and "old" indicate the temperature of the current time step and the temperature of the previous time step; and - is a coefficient relating to the heat capacity of battery 2, constant by simplification.

[0079] The temperature gradient is therefore 100801 = A /

[0081] The equations are solved implicitly with a time step such as 1s for example. Preferably, the coefficients are chosen to have temperature values ​​in a real range.

[0082] The input data for training the machine learning model are the flow rate and temperature of the fluid, as well as the power and temperature of battery 2. The label of this model is the temperature gradient. After obtaining the model with sufficient accuracy, a simulation with this model and an integrator is performed to compare measurements and simulation results.

[0083] The comparison is shown in Figures 2 and 3. These figures illustrate, on the y-axis, the battery temperature resulting from the thermal system simulation (S) and from the machine learning model combined with an integrator (AI). The x-axis represents the analysis time in seconds (s).

[0084] In [Fig. 2], the electrical power is fixed at 50 kW and the flow rate of fluid 4 varies according to a sinuous curve from 0 to 10 l / min. In [Fig. 3], the flow rate of fluid 4 is fixed at 10 l / min and the electrical power varies according to a sinuous curve from 0 to 100 kW.

[0085] The period of the sinuous curves is 25 s. The temperature of the cooling fluid 4 remains at 20 °C.

[0086] It is clear from Figures 2 and 3 that this machine learning model reproduces the evolution of the temperature of the simulation.

[0087] The invention further relates to a computer program comprising program code instructions for executing computer-implemented steps of a simulation method as described above, when said program is running on at least one computer. For example, this could be a simulation program to be installed in a corresponding control unit of a motor vehicle platform 13, such as a test platform.

[0088] Another object of the invention relates to a vehicle platform 13 comprising a simulation device 1 as described above. Such a platform is used for the production of a category of motor vehicles. The invention naturally covers several platforms for several categories of motor vehicles.

[0089] The invention also relates to a motor vehicle comprising a cooling control unit connected to a simulation device 1 as described above so as to be able to exchange data between said control unit and said simulation device 1. Such a control unit can be configured to control the pump 6 taking into account the machine learning model.

Claims

Demands

1.

2.

3. Thermal simulation device (1) for a motor vehicle traction battery (2), comprising: - the battery (2), - a battery cooling system (3) which includes a re-cooler (3a) affixed against at least one side of the battery (2), a pump (6) for circulating a refrigerant (4), and a heat exchanger (7); - a conditioning means (8, 8a) for applying a given temperature to the battery (2) for given times; - at least one simulation means (10) to vary the usage parameters of the battery (2) and the pump (6); - a means to determine if a thermal equilibrium of the battery (2) is reached and / or a means to determine if the battery (2) reduces its power; - a machine learning module (12) configured to integrate the data corresponding to said usage parameters and determine a temperature variation law during battery (2) use. Simulation device (1) according to claim 1, characterized in that the machine learning module (12) determines a variation law including a temperature variation gradient as a function of time. Simulation device (1) according to any one of claims 1 to 2, characterized in that the variation law comprises at least one of the following formulas: Qgen “ C1 XP bt XP ; P ht " Pbb - C2 X Qgen' ; QIrons X f (zi za \ . A . . , / bt tuiw^bb neyv bt bb old \ .* ~Qt x A / = x —=—5—=---' ^gen ^trans / \ 2 Z y OR 1 grad — / y G^est is the heat generated by the battery (2); is a constant coefficient; Tht is the temperature measured outside the battery (2)-cooler (3a) interface; P is the electrical power of the battery (2); Tbb is the temperature at the interface at the battery (2)-cooler (3a) interface; c2 is a constant coefficient representing the internal thermal resistance in the battery (2); Of is the heat transferred by the battery (2); f is the flow rate of the cooling fluid; TcOOiant is the inlet temperature of the cooler (3a) of the battery (2); c3 is the heat transfer coefficient; Tgrad is the temperature gradient; At is the time step in the simulation; and the subscripts "new" and "old" indicate the temperature of the current time step and the temperature of the previous time step.

4. Simulation device (1) according to any one of claims 1 to 3, characterized in that said simulation means (10) is configured to apply several battery powers (2), several pump operating rates (6), several battery charge states (2), several battery health states (2), or a combination of at least two of these parameters.

5. Simulation device (1) according to any one of claims 1 to 4, further comprising an insulation means (9) for thermally insulating at least one side of the battery (2) where the cooler (3a) is not affixed.

6. Simulation device (1) according to any one of claims 1 to 5, characterized in that the conditioning means (8, 8a) for applying a given temperature is configured to implement a homogeneous internal temperature of the battery (2).

7. Method for thermal simulation of a motor vehicle traction battery (2), comprising: - a step for applying a given temperature to the battery (2) for given times; - a step for varying several parameters of the battery (2) and of a pump (6) of a battery cooling system (3); - a step for cooling the battery (2); - a step for determining whether a thermal equilibrium is reached and / or whether the battery (2) reduces its power; - a step to integrate the data of the different parameters and determine a law of temperature variation during battery use (2) by means of a machine learning module.

8. Computer program comprising program code instructions for performing the steps of the simulation method according to claim 7, when said program is running on at least one computer.

9. Motor vehicle platform comprising a simulation device (1) according to any one of claims 1 to 6.

Images