Method for optimizing energy consumption of motor vehicle

Abstract

The invention relates to a method for optimizing the energy consumption of a motor vehicle (2) comprising a fuel tank or hydrogen tank, a battery and / or supercapacitor (30), a heat engine (M) or fuel cell (P), an electric machine (ME), a plurality of devices, each characterized by at least one state variable; and a computer (4) configured to control the traction chain of the motor vehicle (2) over a predetermined distance and capable of controlling the heat engine (M) or fuel cell (P), the electric machine (ME) and these devices by issuing a series of setpoints. The invention also relates to a computer (4) and a computer program product for implementing such a method, and to a motor vehicle (2) comprising such a computer (4).

Description

Technical field

[0001] The present invention relates to optimizing the energy consumption of a motor vehicle, and more particularly to a method for optimizing the energy consumption of a motor vehicle over a predefined distance. The object of the present invention is in particular to generate an optimized setpoint for controlling the vehicle over said distance to be traveled, while minimizing the energy consumption of the vehicle. "Energy consumption of the vehicle" according to the present invention means the consumption of fuel, hydrogen, electrical energy, the current supplied by the branch of the electronic charger of the vehicle battery, a certain number of restart operations of the generator set in a series hybrid vehicle (also called "range extender"), or also the combined consumption of several of these elements. Background art

[0002] In motor vehicles, it is known to optimize the energy consumption of the traction chain over a given or predicted distance. This type of optimization can be achieved for fuel, for electrical energy, for hydrogen consumption (in the case of vehicles equipped with a fuel cell), or simultaneously for two or three of these criteria.

[0003] In a known manner, the optimization can be achieved by using the principle known under the name of the Pontryagin maximum principle (PMP). This method consists of the following steps: minimizing the Hamiltonian function based on the criterion to be optimized (e.g., the amount of fuel or hydrogen) or also the electrical energy consumed, and the description of the dynamics of the system. The dynamics of the system are defined based on the state of the different variables of the vehicle (the speed of the vehicle, the charge state of the battery and / or supercapacitor, the temperature, etc.) and the different inputs or setpoints (the torque setpoint for the heat engine or for the (multiple) electric motors to be applied to the wheels of the vehicle, the torque setpoint for the motor of the generator set to be applied, and / or the power or current setpoint for the fuel cell, and / or also the heating setpoint for the catalytic converter, and / or the control setpoint for the cooling circuit, etc.). The Hamiltonian function is minimized in order to determine these setpoints, so as to make it possible to obtain a minimum consumption of fuel or hydrogen, and / or electrical energy.

[0004] Each input or setpoint depends on the state of certain variables. For example, the torque setpoint for the electric motor to be applied to the wheels of the vehicle depends on the speed of the vehicle and the charge state of the battery, the control setpoint for the cooling circuit depends on the real-time temperature in the cooling circuit, and the heating setpoint for the catalytic converter depends on the real-time temperature in the catalytic converter.

[0005] Then, the thus determined Hamiltonian function is minimized. In other words, the value of the setpoint with the lowest Hamiltonian value is selected and applied to the vehicle. Thus, the setpoint value is determined in real time according to the current state of the system.

[0006] Thus, the application of this type of PMP model uses an internal model that describes the dynamics of the system to be controlled (in this case, a motor vehicle and its traction chain) via differential equations (also called "state equations"), and these differential equations provide a representation of the system. Thus, this type of PMP model is mainly based on an optimal control strategy by means of open loop, which calculates the optimal control solution based on the prediction of the internal model. Equations called "optimality conditions" using "associated states" (also called "co-states", "associated parameters", "Lagrange parameters" or also "dual variables") are also used to solve the optimization problem. These associated states are associated with the state equations (the states representing the dynamic behavior of the physical system) and allow the optimization problem to be fully solved. In fact, the associated states must comply with the gradient equations in order to ensure the optimal nature of the global solution provided by the PMP model.

[0007] However, the initialization problem of these associated states is a key problem to be solved in order to improve the accuracy of the final result obtained for the optimization and to prevent the internal model from deviating too much from the real physical system. The re-initialization (or re-calibration) of the associated states is in fact necessary at certain moments or time periods, especially when the traffic conditions change (especially due to external disturbances, new obstacles detected on the road, change of itinerary, etc.). For this purpose, it is known to implement the initialization of the associated states via a method called the "shooting method". This type of shooting method is based on iterative simulations carried out on a common prediction scenario (in other words, a predictive horizon, which, in the case of a motor vehicle, particularly indicates the curves and gradients of the road, mandatory traffic light stops, traffic density, etc. when this type of predictive horizon is available for the system). More specifically, the target values of the final states of the state variables or associated states are initially defined in the on-board computer in the vehicle (this type of target value must be reached at the end of the predicted common scenario). Then, simulations of optimal control are carried out for different initialization values of the associated states based on this predicted common scenario. Each simulation (or "shot") consists of a numerical simulation of the internal model that starts with the initialization value of the associated state and is applied throughout the predicted common scenario. At the end of each "shooting", the calculated final state of the state variable or auxiliary state is then compared with the target value predefined in the computer. Then, the correction for the next initialization value of the associated state is calculated and applied to the equations of the model. The "shooting" is repeated until the target value is reached with an acceptable accuracy (determined by a predefined accuracy threshold) or until the maximum number of iterations has been carried out. Then, the initialization value that generates the "best" final state of the state variable or auxiliary state of the auxiliary state is provided as the input to the PMP model in order to continue the "true" optimal control of the energy consumption of the vehicle.

[0008] When direct mathematical resolution cannot be obtained due to the excessive complexity of the system involved, initializing the associated state via the "shooting method" thus remains the main known method for controlling the parameters of the PMP model, which has a very large impact (or sensitivity level) on the final state of the system. Therefore, even if the energy consumption of the vehicle is correctly minimized to the predicted common scenario, this initialization requires maximum precision. However, in solutions according to the prior art, it often happens that this initialization of the associated state is relatively imprecise and / or becomes incorrect in actual traffic situations with variable or unexpected disturbances. This results in imprecise and / or incorrect final values of the state variables or associated state, which may be very different from the target values. Therefore, optimality is not achieved for the expected traffic situation.

[0009] Therefore, there is a need to be able to have an optimal control method that allows optimizing the energy consumption of a motor vehicle by implementing the PMP model, thus providing a more accurate and reliable initialization of the associated state (despite potential differences between the internal model and the real physical system), and making it possible to maintain the optimality of the control, regardless of the traffic situation. Summary of the Invention

[0010] For this purpose and according to a first aspect, the invention relates to a method for optimizing the energy consumption of a vehicle implemented in an on-board computer of a motor vehicle, the vehicle comprising a tank for fuel or hydrogen, a battery and / or a supercapacitor capable of supplying electrical energy, a thermal engine supplied by the fuel tank or a fuel cell supplied by the hydrogen tank, at least one electric motor supplied with electrical energy provided by the battery and / or the supercapacitor, at least one device associated with the thermal engine or the fuel cell, at least one device associated with the electric motor, and at least one device associated with the battery or the supercapacitor, the journey of the vehicle over a predetermined distance and the traffic conditions being predefined or predicted in the computer, the computer being configured to control the traction chain of the motor vehicle over a predetermined distance and being able to control the thermal engine or the fuel cell, the electric motor and / or these devices by emitting a series of setpoints, the thermal engine or the fuel cell, the electric motor and these devices being each characterized by at least one state variable, each state variable making it possible to describe the operating state of the device it characterizes, the assembly constituted by the thermal engine or the fuel cell, the electric motor and these devices being represented by a system of state equations modeling the dynamics of the vehicle, the state equations depending at least on the instantaneous setpoint values and the state variables, a series of associated states being associated with the state equations and representing the conditions of the dynamic behavior of the vehicle, the energy consumption to be optimized being defined in a criterion equation as being firstly the cumulative instantaneous consumption of fuel or hydrogen of the thermal engine or the fuel cell and secondly the sum of a terminal correction term representing the electrical energy withdrawn from or stored in the battery or the supercapacitor, the terminal correction term depending on the difference in the state of charge between firstly the state of charge of the battery or the supercapacitor at the end of the predetermined distance and secondly the state of charge of the battery or the supercapacitor at the start of the predetermined distance, the correction term being a linear function defined by a scaling factor, the method being implemented during a period divided into constant sampling instants, the method comprising, at each sampling instant, the following steps:

[0011] - For each setpoint, calculating all the possible values of the Hamiltonian function of the setpoint using at least the criterion equation; and

[0012] - Determining the value of each setpoint for which the Hamiltonian function is the weakest,

[0013] The computer is configured to define target values for the final states of the state variables or the associated states according to the traffic conditions predefined or predicted in the computer, a common state variable being selected by the computer as the initial state, and the method further comprises a phase of updating the criterion equation, the phase comprising the following steps:

[0014] - Selecting a first initial value for the associated state;

[0015] - Based on the initial state, the first initial value selected for the associated state, and the internal model pre-implanted in the computer, calculate the first simulation value and the first pair of values of the internal model for the traffic condition predefined or predicted in the computer. The first pair of values is formed by the energy consumption of the vehicle and the difference in the charge state of the battery or supercapacitor obtained when the step of calculating the first simulation value is completed;

[0016] - If the target value of the state variable or the final state of the associated state is not obtained when the step of calculating the first simulation value is completed, then:

[0017] o Select a second initial value different from the first initial value for the associated state;

[0018] o Based on the initial state, the second initial value selected for the associated state, and the internal model, calculate the second simulation value and the second pair of values of the internal model for the traffic condition predefined or predicted in the computer. The second pair of values is formed by the energy consumption of the vehicle and the difference in the charge state of the battery or supercapacitor obtained when the step of calculating the second simulation value is completed;

[0019] - If the target value of the state variable or the final state of the associated state is not obtained when the step of calculating the second simulation value is completed, then:

[0020] o Calculate the first estimated value of the scaling factor using a method that estimates the gradient in a predefined manner based on the first pair of values and the second pair of values formed by the energy consumption of the vehicle and the difference in the charge state of the battery or supercapacitor. This first estimated value of the scaling factor is used to update the criterion equation of the energy consumption to be optimized; and

[0021] - Loop back to the previous steps until the target value of the state variable or the final state of the associated state is obtained.

[0022] In the above method, the steps other than the update phase are iteratively looped back at each new sampling moment. Additionally, the method is implemented for a predefined / given distance of the vehicle. This type of distance is predicted, for example, by a system of the "electronic horizon information" (or eHorizon) type, which conventionally is based on the ADASIS (Advanced Driver Assistance System Interface Specification) data format standard for predictive driver assistance systems or on any other type of device connected to the vehicle computer.

[0023] The method according to the invention makes it possible to define a torque setpoint in order to provide the power and acceleration required by the driver when the vehicle is running over a given or planned distance, while minimizing the energy consumption of the vehicle. Additionally, during the update phase, the method according to the invention makes it possible to adjust the estimate of the terminal correction term and to extract the estimate of the scaling factor in an integrated "observer" strategy. As a result, thanks to the update phase of the method, the initialization of the associated state is more accurate and reliable (even though there may be differences between the internal model and the real physical system), and it makes it possible to maintain the optimality of the control, regardless of the traffic conditions.

[0024] According to a first variant embodiment, the update phase is carried out at a predefined regular time interval.

[0025] According to another variant embodiment, the method further comprises a step of detecting at least one predefined condition, said at least one predefined condition relating to the itinerary and / or traffic conditions of the vehicle predefined or predicted within the computer, and the update phase is carried out only when the computer has detected said at least one predefined condition. This type of predefined condition is for example related to external disturbances or new obstacles detected on the road, etc.

[0026] Advantageously, the traffic conditions are predicted within the computer according to a predefined time horizon via a static and / or dynamic data management system related to the road and / or the road traffic infrastructure connected to the computer, wherein the computer is configured to receive the traffic conditions according to a sliding time window throughout the itinerary of the vehicle over said predefined distance. This makes it possible to maintain the optimality of the control and to calibrate the co-state when predicting the traffic conditions according to a limited time horizon.

[0027] Preferably, the method for estimating the predefined gradient is a linear regression method, in particular the recursive least squares method. This type of recursive least squares method makes it possible to filter the non-linearity in the pair of values (i.e., the difference in the energy consumption of the vehicle and the charge state of the battery or supercapacitor obtained when completing the different shots carried out during the update phase).

[0028] Also preferably, when the method for estimating the predefined gradient is the recursive least squares method, the final estimate of the scaling factor calculated during the update phase is stored in the computer. This makes it possible to significantly reduce the amount of data stored in the memory of the computer at each new iteration of the update phase.

[0029] According to an embodiment of the present invention, the predetermined distance that the vehicle has to travel is divided by a computer into N consecutive distance segments, where N is a predefined integer, and during an update phase, the computer defines one or more target values for the associated states on N - 1 of the distance segments and one or more target values for the final state of the state variables on the remaining distance segment.

[0030] According to a specific example of this embodiment, the N - 1 distance segments correspond to the first N - 1 segments of the predetermined distance that the vehicle has to travel, and the remaining distance segment continues to correspond to the final segment of the predetermined distance.

[0031] Preferably, according to this embodiment, the computer enforces convergence of the associated states on the N - 1 distance segments towards the target values, the value of which corresponds to the scale factor.

[0032] According to an embodiment of the present invention, at each sampling instant, the method further comprises the following steps:

[0033] - determining an applicable setpoint field that includes a series of values for each setpoint;

[0034] - for each setpoint and within the determined applicable setpoint field, calculating all possible state gradients of the setpoint using at least the state equation;

[0035] - for each setpoint and within the determined applicable setpoint field, calculating all possible values of the energy consumption to be optimized using at least the criterion equation.

[0036] The invention also relates to a computer for controlling a traction chain of a motor vehicle over a predetermined distance, the vehicle comprising, among other things, and the computer: a fuel or hydrogen tank, a battery and / or a supercapacitor capable of supplying electrical energy, a thermal engine supplied by the fuel tank or a fuel cell supplied by the hydrogen tank, at least one electric motor supplied with electrical energy provided by the battery and / or the supercapacitor, at least one device associated with the thermal engine or the fuel cell, at least one device associated with the electric motor, and at least one device associated with the battery or the supercapacitor, the journey of the vehicle over the predetermined distance and the traffic conditions being predefined or predicted in the computer, the computer being capable of controlling the thermal engine or the fuel cell, the electric motor and / or these devices by issuing a series of setpoints, the thermal engine or the fuel cell, the electric motor and these devices each being characterized by at least one state variable, each state variable making it possible to describe the operating state of the device it characterizes, the assembly formed by the thermal engine or the fuel cell, the electric motor and these devices being represented by a system of state equations modeling the dynamics of the vehicle, said state equations depending at least on the instantaneous setpoint values and the state variables, a series of associated states being associated with the state equations and representing the conditions of the dynamic behavior of the vehicle, the computer being configured to implement the steps of the method as previously described.

[0037] The invention also relates to a motor vehicle comprising: a fuel or hydrogen tank, a battery and / or a supercapacitor capable of supplying electrical energy, a thermal engine supplied by the fuel tank or a fuel cell supplied by the hydrogen tank, at least one electric motor supplied with electrical energy provided by the battery and / or the supercapacitor, at least one device associated with the thermal engine or the fuel cell, at least one device associated with the electric motor, at least one device associated with the battery or the supercapacitor, and a computer for controlling the traction chain of the motor vehicle over a predetermined distance, the journey of the vehicle over the predetermined distance and the traffic conditions being predefined or predicted in the computer, the computer being capable of controlling the thermal engine or the fuel cell, the electric motor and / or these devices by issuing a series of setpoints, the thermal engine or the fuel cell, the electric motor and these devices each being characterized by at least one state variable, each state variable making it possible to describe the operating state of the device it characterizes, the assembly formed by the thermal engine or the fuel cell, the electric motor and these devices being represented by a system of state equations modeling the dynamics of the vehicle, said state equations depending at least on the instantaneous setpoint values and the state variables, wherein the computer is a computer as previously described.

[0038] The invention also relates to a computer program product, characterized in that it comprises a series of program code instructions which, when executed by one or more processors, cause the (s) processor(s) to be configured to implement the method as previously described. Description of the Drawings

[0039] In the following, embodiments of the present invention will be described by way of non - limiting examples with reference to the accompanying drawings, in which:

[0040] - Figure 1 schematically illustrates a vehicle according to the present invention, the vehicle being provided with an on - vehicle computer; and

[0041] - Figure 2 represents a flowchart of a method for optimizing the energy consumption of a vehicle implemented by a Figure 1 computer according to the present invention. Detailed Description

[0042] Referring Figure 2 , the present invention relates to a method for optimizing the energy consumption of a motor vehicle 2 (shown in Figure 1 ) implemented in an on - vehicle computer 4 of the motor vehicle 2. The motor vehicle 2 is, for example (but not limited to), a hybrid vehicle. In addition to the computer 4, this type of hybrid vehicle 2 also conventionally includes a thermal engine M (in the case of a "thermoelectric" type hybrid vehicle 2) (often referred to as an ICE (internal combustion engine)) or a fuel cell P operating on hydrogen (in the case of a "hydro - electric" type hybrid vehicle 2). This type of hybrid vehicle 2 also includes at least one electric machine M E (often referred to as an EMA), a fuel tank or a hydrogen tank ( Figure 1 not shown in the figures) and a power battery 30 (or a supercapacitor in a non - shown variant).

[0043] The motor vehicle 2 also includes at least one device associated with the thermal engine M and in particular a device 10 for cooling the thermal engine M and the catalytic converter 20, or at least one device associated with the fuel cell P (this type of device is not shown in Figure 1 ). The cooling device 10 makes it possible to reduce the temperature of the thermal engine M during its use. In particular, the cooling device 10 includes a coolant. The catalytic converter 20 connected to the thermal engine M via an exhaust system can reduce the amount of pollutant products in the exhaust gases emitted by the thermal engine M before these exhaust gases are discharged to the outside of the vehicle. The catalytic converter 20 also includes a heating device that can increase the temperature in the catalytic converter 20 in order to carry out the de - pollution of the exhaust gases. The heating device of the catalytic converter 20 must be supplied with electrical energy in order to operate. The motor vehicle 2 also includes at least one device associated with the electric machine M E and in particular a series of voltage converters ( Figure 1 not shown in the figures), these voltage converters making it possible to convert the voltage between the battery 30 and the electric machine M E . The motor vehicle 2 also includes at least one device associated with the battery 30 (or supercapacitor), for the sake of clarity, this type of device is not shown inFigure 1 is not shown in

[0044] The heat engine M is particularly suitable for being supplied with fuel from a fuel tank, and the fuel cell P can be supplied with hydrogen from a hydrogen tank. The heat engine M also includes an exhaust system for the exhaust gases emitted during the combustion of the air-fuel mixture in the heat engine M. The electric machine M E can be supplied with electrical energy from the storage battery 30.

[0045] The vehicle 2 may include other devices associated with the heat engine M or the fuel cell P, and other devices associated with the electric machine M E associated therewith.

[0046] "System" means a series of elements assembled in the vehicle 2, which elements consume or produce electrical energy, fuel or hydrogen. For example, the system includes the series of devices previously described, namely the heat engine M or the fuel cell P, the electric machine M E , the cooling device 10, the catalytic converter 20 and the storage battery 30.

[0047] Each device is characterized by at least one state variable, so that it is possible to describe the operating state of the device. For example, the cooling device 10 is characterized by the coolant temperature. Also, for example, the catalytic converter 20 is characterized by an internal temperature value. Again, for example, the storage battery 30 is characterized by the state variable of the load, the heat engine M is characterized by the rotational speed, the fuel cell P is characterized by the temperature or pressure present in the circuit for supplying hydrogen and oxygen, etc.

[0048] The energy consumption to be optimized (to be optimized via the method according to the invention) of the vehicle 2 is represented by the criterion equation g(u, q). The criterion equation g(u, q) depends at least on the instantaneous setpoint value u and the state variable q. Also preferably, the criterion equation also depends on the disturbances and / or setpoints w applied to the vehicle 2 and is then written as g(u, q, w). The energy consumption J to be optimized over a given time period T is associated with the criterion equation via the following equation:

[0049]

[0050] In addition, the energy consumption J to be optimized is defined as the sum of first the cumulative instantaneous consumption Ji of fuel or hydrogen of the heat engine M or the fuel cell P and second the terminal correction term Jc representing the electrical energy extracted from or stored in the storage battery 30 (or supercapacitor). This terminal correction term Jc can be based on first the state of charge SoC of the storage battery 30 (or supercapacitor) at the end of the distance traveled by the vehicle 2 final and second the state of charge SoC of the storage battery 30 (or supercapacitor) at the start of the distance traveled by the vehicle 2 initialcalculated based on the difference in the state of charge. Therefore, the terminal correction term Jc is then expressed as:

[0051] J c = k.(SoC final -SoC initial )

[0052] where k is a proportionality factor, and if the difference (SoC final -SoC initial ) is positive, the value of Jc is positive (in this case, electrical energy is stored in the battery 30 or the supercapacitor, and the correction term Jc is subtracted from the cumulative instantaneous consumption Ji in the expression for J), and if the difference (SoC final -SoC initial ) is negative, the value of Jc is negative (in this case, electrical energy is extracted from the battery 30 or the supercapacitor, and the correction term Jc is added to the cumulative instantaneous consumption Ji in the expression for J).

[0053] Then, the energy consumption J to be optimized is expressed in the criterion equation as:

[0054] J = J i - k.(SoC final -SoC initial )

[0055] The system is represented by a system of state equations f(u,q) that models the dynamics of the vehicle 2. The state equation f(u,q) depends at least on the instantaneous setpoint value u and the state variable q. Also preferably, the state equation additionally depends on the disturbances and / or setpoints w applied to the vehicle 2 and is then written as f(u,q,w).

[0056] The computer 4 forms, for example, part of a data processing unit that stores application programs or computer programs that can cooperate with the computer 4 (for the sake of clarity, the data processing unit and the application program or computer program are not shown in Figure 1 ). As a variant, the application program or computer program is directly stored in the computer 4. The computer 4 is connected to the heat engine or fuel cell P, the electric motor M E , the battery 30, and the series of devices described previously (specifically including the cooling device 10 and the catalytic converter 20). The computer 4 is also connected to a static and / or dynamic data management system related to the road and / or road traffic infrastructure (this type of system is in Figure 1(not shown). The static and / or dynamic data management system is constructed, for example, according to a cloud architecture and makes it possible to provide (or predict) a time range of the operating cycle (or state change period) of road infrastructure elements detected in front of it during a vehicle journey, in particular depending on the speed of the vehicle and the traffic density on the road. The management system also makes it possible to predict the distance, curves and gradients of the road, forced traffic light stops, traffic density, etc. The management system is, for example, a system of the "electronic horizon information" (or eHorizon) type, which is conventionally based on the ADASIS (Advanced Driver Assistance System Interface Specification) data format standard for predictive driver assistance systems or based on any other type of device. In a known manner, this type of "eHorizon" type system makes it possible to control both static data related to road infrastructure (such as, for example, the nature of the road, intersections, applied legal speed limits, etc.) and dynamic data (average speed of vehicles on the road, traffic density, dynamic data related to road infrastructure elements, etc.), where the static data and the dynamic data represent the traffic situation as a whole. This type of "eHorizon" type system can receive these data, decode these data (via a decoder), reconstruct these data (via a data reconstructor), and transmit these data to the computer 4, and can implement a prediction algorithm for the vehicle distance using, for example, the concept of "most likely distance or path" (most likely path). The computer 4 is configured to receive the traffic situation predicted by the management system according to a sliding time window of the entire journey of the vehicle 2 along the distance traveled by the vehicle 2.

[0057] The computer 4 can receive measurements of each value of the state variable associated with each device. Additionally, the computer 4 can control the device by issuing setpoints according to the (multiple) values of the variables of each device to which it is connected.

[0058] Thus, for example, the setpoint issued to the thermal engine M specifies the value of the torque to be applied to the thermal engine M and depends in particular on the speed of the vehicle and the power demand from the driver of the vehicle. Also, for example, the setpoint issued to the fuel cell P specifies the power that the fuel cell P needs to supply and depends on the charge state of the battery 30 and / or the supercapacitor and the power demand from the driver. Also, for example, the setpoint issued to the electric motor M E specifies the torque to be applied to the electric motor M E and depends on the charge state of the battery 30 and the power demand from the driver.

[0059] The setpoint sent to the catalytic converter 20 relates to the temperature in the catalytic converter 20 and depends on the temperature measured in the catalytic converter 20. The setpoint sent to the cooling device 10 relates to the temperature of the coolant and depends on the measured temperature of the coolant.

[0060] The computer 4 is also configured to determine an applicable setpoint field that includes a series of values for each setpoint. The computer 4 is also configured to implement the principle of the PMP method (in other words, the Pontryagin maximum principle) by determining the Hamiltonian function H(x, u * , λ) based on different setpoint values of the applicable setpoint field. At this time, the notation u * is introduced, which represents the optimal control. The associated state λ (also referred to as the "associated parameter", "Lagrange parameter", "associated vector" or also as the "co-state vector") is also introduced. These associated states are associated with the state equations representing the conditions of the dynamic behavior of the physical system and will make it possible to fully solve the optimization problem. The computer 4 is also configured to define the target value of the final state of the state variable q or the associated state λ according to the traffic conditions predefined or predicted within the computer 4.

[0061] The computer 4 includes a processor that can execute a series of instructions, making it possible to implement these functions.

[0062] Reference Figure 2 , now an embodiment of the method according to the invention for optimizing the energy consumption of the vehicle 2 implemented by the computer 4 as previously described will be described.

[0063] For the sake of simplicity of description, the variable considered is the state of charge of the battery 30 or the supercapacitor. The different control setpoints considered are, for example: the torque of the thermal engine M and the torque of the electric motor M E . Other parameters can be considered, which relate to at least one device related to the electric motor M E or to parameters related to the fuel cell P, or to the at least one device related to the fuel cell P, or also to parameters related to the at least one device related to the battery 30 or the supercapacitor.

[0064] The method is implemented during a period divided into constant sampling instants and includes steps that are iteratively looped back at each new sampling instant. The sampling period between two consecutive sampling instants depends on the dynamics of the system and can be chosen, for example, between 10 ms and 500 ms, typically equal to 100 ms.

[0065] According to Figure 2In the specific embodiment illustrated, the method includes an initial step 22 during which the computer 4 determines an applicable setpoint field u, including a series of values u for each setpoint to be applied to the variable.

[0066] Then, the method includes a subsequent step 24 during which, for each setpoint and within the determined applicable setpoint field u, the computer 4 calculates all possible state gradients of the setpoint using at least the state equation f(u, q, w) describing the system.

[0067] The method includes a parallel or subsequent step 26 during which, for each setpoint and within the determined applicable setpoint field u, the computer 4 calculates all possible values of the criterion or combination of criteria to be optimized using at least the criterion equation g(u, q, w). As a variant not shown, the calculation steps 24, 26 can be reversed.

[0068] Then, the method includes a subsequent step 28 during which, for each setpoint and within the determined applicable setpoint field u, the computer 4 calculates all possible values H of the Hamiltonian function H(u, q, λ, w) of the setpoint using the value of the state gradient and the possible values of the energy consumption to be optimized determined for the setpoint. For this purpose, the computer 4 uses the principle of the PMP method (in other words, the method of the Pontryagin maximum principle).

[0069] Then, the Hamiltonian function H(u, q, λ, w) is expressed as:

[0070] H ( u , q, λ, w) = [g( u , q, w)] + λ T ·[f( u , q, w)]

[0071] where λ T is the transpose of the associated vector λ.

[0072] Then, during a subsequent step 31, the computer 4 determines a series of so-called "optimal" setpoints u by minimizing the Hamiltonian function calculated during the previous step 28 for each of the setpoints. * . In the embodiment considered, the computer 4 thus determines a first optimal "thermal" setpoint intended to control the heat engine M, and a second optimal "electrical" setpoint intended to control the electric motor M E .

[0073] According to the invention, the phase 32 of updating the criterion equation implemented by the computer 4 can also be implemented during the method. According to a first variant embodiment ( Figure 2(not shown in the figure), the update phase 32 is implemented at a predetermined regular time interval spaced apart from each other, for example, by a few minutes. According to Figure 2 Another variant embodiment shown in the figure, the method includes step 34, during which the computer 4 detects at least one predetermined condition. The (multiple) predetermined conditions generally relate to the itinerary of the vehicle 2 predefined or predicted within the computer 4 and / or traffic conditions. This type of predetermined condition is related, for example, to external interference, itinerary changes, or newly detected obstacles on the road, etc. Therefore, the update phase 32 is only implemented when the computer 4 has detected the (multiple) predetermined conditions.

[0074] Initially, during the update phase 32, the computer 4 selects the common state variable q as the initial state. In addition, the computer 4 defines the target value of the final state of the state variable q or the associated state λ according to the traffic conditions predefined or predicted within the computer 4. The update phase 32 corresponds to the "shooting method" and includes a first step 36, in which the computer 4 selects a first initial value for the associated state λ.

[0075] The update phase 32 includes a subsequent step 38, during which the computer 4 calculates the first simulated value of the internal model for the traffic conditions predefined or predicted within the computer 4 according to the initial state, the first initial value selected for the associated state λ, and the internal model pre-implanted in the computer 4. Therefore, at the completion of this calculation step 38, a first pair of values is obtained, that is, the energy consumption J1 of the vehicle 2 and the charging state difference ΔSoC1 of the battery 30 (or supercapacitor); where ΔSoC1 = (SOC final1 -SoC initial ).

[0076] If the target value of the final state of the state variable q or the associated state λ has been obtained at the completion of the step 38 of calculating the first simulated value, the update phase 32 ends. In other aspects, the update phase 32 includes a subsequent step 40, during which the computer 4 selects a second initial value different from the first initial value for the associated state λ. According to a first advantageous embodiment, the second initial value is selected by changing the step size depending on the difference between the first initial value and the target value for the associated state λ and by carefully adjusting the value of the step size. According to another advantageous embodiment, the second initial value is selected by reversing the model of the neural network (in this case, the model presupposes a phase before training and learning the neural network).

[0077] The update phase 32 includes the subsequent step 42, during which the computer 4 calculates a second simulated value of the internal model based on the initial state, the second initial value selected for the associated state λ, and the internal model for the traffic conditions predefined or predicted within the computer 4. Thus, upon completion of this calculation step 42, a second pair of values is obtained, namely, the energy consumption J2 of the vehicle 2 and the charge state difference ΔSoC2 of the battery 30 (or supercapacitor); where ΔSoC2 = (SoC final2 -SoC initial ).

[0078] If the target value of the final state of the state variable q or the associated state λ is obtained upon completion of the step 42 of calculating the second simulated value, the update phase 32 ends. Otherwise, the update phase 32 includes the subsequent step 44, during which the computer 4 calculates a first estimated value of the scaling factor k based on the first pair of values and the second pair of values (i.e., the energy consumptions J1, J2 of the vehicle 2 and the charge state differences ΔSoC1, ΔSoC2 of the battery 30 (or supercapacitor)). In fact, each simulation of the internal model implemented provides a pair of values, namely, the energy consumption of the vehicle 2 and the charge state difference of the battery 30 (or supercapacitor); and the relationship between these two variables is a refinement function. The slope of this refinement function is the scaling factor k, where the original y-axis is the value of the energy consumption of the vehicle 2 when the final charge state of the battery 30 is equal to its initial charge state. The first estimated value of the scaling factor k (provided by using an integrated "observer" strategy) is then used to update the criterion equation (previously described) for the energy consumption J to be optimized, where the thus updated criterion equation is reused in the calculation step 26. The steps 36, 38, 40, 42, 44 of the update phase 32 are then looped back using the updated criterion equation until the target value of the final state of the state variable q or the associated state λ is obtained. Then, the update phase 32 ends.

[0079] The computer 4 uses a method for estimating a predefined gradient to implement the step 44 of calculating the estimated value of the scaling factor k. Preferably, the method for estimating the gradient can be a linear regression method formulated in the form of recursive least squares. In this case, the final estimated value of the scaling factor k calculated during the calculation step 44 is saved in the computer 4. Additionally, this type of recursive least squares method requires the initialization of parameters after a second emission implemented during the step 42 of calculating the second simulated value. As a variant, the method for estimating the gradient in a predefined manner can be a method of averages, a method of generalized least squares, or any other known method for estimating the gradient.

[0080] This method can be iterated repeatedly during the use of the vehicle.

[0081] According to an embodiment of the invention, the distance that the vehicle 2 has to travel is divided by the computer 4 into N consecutive distance segments, where N is a predefined integer. During the update phase 32, thus, the computer 4 defines one or more target values for the associated state λ on N - 1 distance segments and one or more target values for the final state of the state variable q on the remaining distance segment. This remaining distance segment corresponds, for example, to the final distance segment that the vehicle 2 has to cover, where the N - 1 distance segments thus correspond to the first N - 1 distance segments of this distance. In this case, for example, the final state of the charge state of the accumulator 30 is imposed on the remaining distance segment. As a variant, the remaining distance segment can correspond to any other segment of the distance that the vehicle 2 has to travel. In this case, usually when the vehicle 2 reaches the entrance of the urban area or a recharging terminal during this distance, for example, the final state of the charge state of the accumulator 30 is imposed on this distance segment. As a variant, after a given minimum time period of temperature increase during the heating phase, the final temperature in the passenger compartment of the vehicle 2 or in the catalytic converter can be imposed.

[0082] Preferably, according to this embodiment, the computer 4 enforces the convergence of the associated state λ on the N - 1 distance segments where the final state of the state variable q remains free, towards a target value whose value corresponds to the scaling factor k. In fact, in order to obtain the minimum value of the energy consumption J to be optimized at the end of the prediction scenario, without any constraint imposed on the final state of the state variable q, the PMP method implies that the associated state λ converges towards a target value at the end of this scenario, whose value corresponds to the scaling factor k (the proof of this theorem is known in the prior art).

[0083] Thus, the method makes it possible to obtain a setpoint value for the minimum energy consumption of the vehicle 2.

[0084] In addition, during the update phase, the method according to the invention makes it possible to adjust the estimated value of the terminal correction term and to extract the estimated value of the scaling factor in an integrated "observer" strategy. As a result, thanks to the update phase 32 of the method, the initialization of the associated state λ is more accurate and reliable (although there may be differences between the internal model and the real physical system), and this makes it possible to maintain the optimality of the control, regardless of the traffic conditions.

Claims

1. A method for optimizing the energy consumption of a vehicle (2) implemented in an on-board computer (4) of the motor vehicle (2), the vehicle (2) comprising a fuel tank or a hydrogen tank, a battery and / or a supercapacitor (30) capable of supplying electrical energy, a thermal engine (M) supplied by the fuel tank or a fuel cell (P) supplied by the hydrogen tank, at least one electric motor (M E ) supplied with electrical energy provided by the battery and / or the supercapacitor (30), at least one device (10, 20) associated with the thermal engine (M) or the fuel cell (P), at least one device (30) associated with the electric motor (M E ), and at least one device associated with the battery or the supercapacitor (30), the travel of the vehicle (2) over a predetermined distance and the traffic conditions being predefined or predicted within the computer (4), the computer (4) being configured to control the traction chain of the motor vehicle (2) over a predetermined distance, the predetermined distance that the vehicle (2) has to travel being divided by the computer (4) into N consecutive distance segments, N being a predefined integer, the N - 1 distance segments corresponding to the first N - 1 segments of the predetermined distance that the vehicle (2) has to travel, the remaining distance segment corresponding to the final segment of the predetermined distance, the computer (4) being able to control the thermal engine (M) or the fuel cell (P), the electric motor (M E ) and / or the devices (10, 20, 30) by issuing a series of setpoints, the thermal engine (M) or the fuel cell (P), the electric motor (M E ) and the devices (10, 20, 30) each being characterized by at least one state variable, each state variable making it possible to describe the operating state of the device it characterizes, the thermal engine (M) or the fuel cell (P), the electric motor (M E ) and the components constituted by the said devices (10, 20, 30) are represented by a system of state equations for the dynamics modelling of the said vehicle (2), the said state equations depending at least on instantaneous setpoint values and state variables, a series of associated states being associated with the said state equations and representing the conditions of the dynamic behaviour of the said vehicle (2), the energy consumption to be optimized being defined in a criterion equation as being firstly the cumulative instantaneous consumption of fuel or hydrogen of the said heat engine (M) or the said fuel cell (P) and secondly the sum of corrective terminal terms representing the electrical energy withdrawn from or stored in the said accumulator battery or the said supercapacitor (30), the said corrective terminal terms depending on the difference in the state of charge between firstly the state of charge of the said accumulator battery or the said supercapacitor (30) at the end of the said predetermined distance and secondly the state of charge of the said accumulator battery or the said supercapacitor (30) at the start of the said predetermined distance, the said corrective terms being linear functions defined by a scaling factor, the said method being implemented during a period divided into constant sampling instants, the said method comprising the following steps at each sampling instant: - For each setpoint, calculate (28) all possible values of the Hamiltonian function for said setpoint using at least said criterion equation; and - Determine (31) the value of each setpoint for which said Hamiltonian function is the weakest, characterized in that the computer (4) is configured to define one or more target values of the associated state on N-1 distance segments and one or more target values of the final state of the state variables on said remaining distance segments according to traffic conditions predefined or predicted within the computer (4), a common state variable being selected by the computer (4) as the initial state, and in that the method further comprises a phase (32) of updating said criterion equation, said phase (32) comprising the following steps: - Select (36) a first initial value for said associated state; - Calculate (38) a first simulated value, a first pair of values, of said internal model for said traffic conditions predefined or predicted within the computer (4) according to said initial state and the first initial value selected for said associated state and an internal model pre-implanted in the computer (4), said first pair of values being formed by the energy consumption of the vehicle (2) obtained when completing the step (38) of calculating said first simulated value and the difference in the state of charge of the accumulator or the supercapacitor (30); - If the target value of the final state of said state variable or associated state is not obtained when completing the step (38) of calculating said first simulated value, then: o Select (40) a second initial value different from said first initial value for said associated state; o Calculate (42) a second simulated value, a second pair of values, of said internal model for said traffic conditions predefined or predicted within the computer (4) according to said initial state and the second initial value selected for said associated state and said internal model, said second pair of values being formed by the energy consumption of the vehicle (2) obtained when completing the step (42) of calculating said second simulated value and the difference in the state of charge of the accumulator or the supercapacitor (30); - If the target value of the final state of said state variable or associated state is not obtained when completing the step (42) of calculating said second simulated value, then: o Calculate (44) a first estimated value of said scaling factor using a method of estimating the gradient in a predefined manner according to the first pair of values and the second pair of values formed by the energy consumption of the vehicle (2) and the difference in the state of charge of the accumulator or the supercapacitor (30), said first estimated value of said scaling factor being used to update the criterion equation of the energy consumption to be optimized; and - Loop back to the foregoing steps (36, 38, 40, 42, 44) until the target value of the final state of said state variable or associated state is obtained, the computer (4) being configured to enforce convergence of the associated state on the N-1 distance segments towards the target value, the value of said target value corresponding to said scaling factor.

2. The method according to claim 1, wherein Said update phase (32) is implemented at predetermined regular time intervals.

3. The method according to claim 1, wherein The method additionally includes a step (34) of detecting at least one predefined condition, the at least one predefined condition relating to a journey and / or traffic situation of the vehicle predefined or predicted within the computer (4), and wherein the update phase (32) is only implemented when the computer (4) has detected the at least one predefined condition.

4. The method according to any one of claims 1 to 3, wherein The traffic situation is predicted within the computer (4) according to a predefined time range via a static and / or dynamic data management system associated with the road and / or the road traffic infrastructure connected to the computer (4), wherein the computer (4) is configured to receive the traffic situation according to a sliding time window over the journey of the vehicle (2) over the predefined distance.

5. The method according to any one of claims 1 to 4, wherein, The method for estimating the predefined gradient is a recursive least squares method.

6. The method according to claim 5, wherein, The final estimated value of the scale factor calculated during the update phase (32) is stored in the computer (4).

7. The method according to any one of claims 1 to 6, wherein At each sampling moment, the method further includes the following steps: - determining (22) the applicable setpoint field, the field including a series of values for each setpoint; - for each setpoint and within the determined applicable setpoint field, calculating (24) all possible state gradients of the setpoint using at least the state equation; - for each setpoint and within the determined applicable setpoint field, calculating (26) all possible values of the energy consumption to be optimized using at least the criterion equation.

8. A computer (4) for controlling a traction chain of a motor vehicle (2) over a predetermined distance, the vehicle (2) comprising, among other things, and the computer (4): a fuel tank or a hydrogen tank, a battery and / or a supercapacitor (30) capable of supplying electrical energy, a thermal engine (M) supplied by the fuel tank or a fuel cell (P) supplied by the hydrogen tank, at least one electric motor (M E ) supplied with electrical energy provided by the battery and / or the supercapacitor (30), at least one device (10, 20) associated with the thermal engine (M) or the fuel cell (P), at least one device (30) associated with the electric motor (M E ), and at least one device associated with the battery or the supercapacitor (30), the travel of the vehicle (2) over a predetermined distance and the traffic conditions being predefined or predicted within the computer (4), the computer (4) being capable of controlling the thermal engine (M) or the fuel cell (P), the electric motor (M E ) and / or the devices (10, 20, 30) by emitting a series of setpoints, the thermal engine (M) or the fuel cell (P), the electric motor (M E ) and the devices (10, 20, 30) each being characterized by at least one state variable, each state variable making it possible to describe the operating state of the device it characterizes, the assembly formed by the thermal engine (M) or the fuel cell (P), the electric motor (M E ) and the devices (10, 20, 30) being represented by a system of state equations modeling the dynamics of the vehicle (2), the state equations depending at least on instantaneous setpoint values and state variables, a series of correlations between states being associated with the state equations and representing the conditions of the dynamic behavior of the vehicle (2), the computer (4) being characterized in that it is configured to implement the steps of the method according to any one of claims 1 to 8.

9. A motor vehicle (2), comprising: A fuel tank or a hydrogen tank, a battery and / or a supercapacitor (30) capable of supplying electrical energy, a thermal engine (M) supplied by the fuel tank or a fuel cell (P) supplied by the hydrogen tank, at least one electric motor (M E ) supplied with electrical energy provided by the battery and / or the supercapacitor (30), at least one device (10, 20) associated with the thermal engine (M) or the fuel cell (P), at least one device (30) associated with the electric motor (M E ), at least one device associated with the battery or the supercapacitor (30), and a computer (4) for controlling the traction chain of the motor vehicle (2) over a predetermined distance, the journey of the vehicle (2) over the predetermined distance and the traffic conditions being predefined or predicted in the computer (4), the computer (4) being able to control the thermal engine (M) or the fuel cell (P), the electric motor (M E ) and / or the devices (10, 20, 30) by sending a series of setpoints, the thermal engine (M) or the fuel cell (P), the electric motor (M E ) and the devices (10, 20, 30) each being characterized by at least one state variable, each state variable making it possible to describe the operating state of the device it characterizes, the assembly formed by the thermal engine (M) or the fuel cell (P), the electric motor (M E ) and the devices (10, 20, 30) being represented by a system of state equations modeling the dynamics of the vehicle (2), the state equations depending at least on instantaneous setpoint values and state variables, the motor vehicle (2) being characterized in that the computer (4) for controlling the traction chain complies with claim 8.

10. A computer program product, characterized in that, It includes a series of program code instructions which, when executed by one or more processors, configure the (s) processor(s) to implement the method according to any one of claims 1 to 7.

Images