WO2022013416A1

WO2022013416A1 - Method for operating a vehicle

Info

Publication number: WO2022013416A1
Application number: PCT/EP2021/069940
Authority: WO
Inventors: Gala JOVANOVIC; Katharina Hauer; James Girard
Original assignee: Robert Bosch Gmbh
Priority date: 2020-07-16
Filing date: 2021-07-16
Publication date: 2022-01-20
Also published as: DE102020208886A1; EP4182198A1

Abstract

The invention relates to a method for operating a vehicle comprising at least one electric machine designed as a drive unit, an internal combustion engine, and an electric energy storage system for the at least one electric machine designed as a drive unit. Using a method based on artificial intelligence, in particular an artificial neural network (NN1), in which the relationship between the efficiency of the internal combustion engine and the efficiency of the electric system, comprising the at least one electric machine designed as a drive unit and the electric energy storage system, is taken into consideration, an operating strategy (B) is determined for the vehicle, thereby optimizing the usage efficiency of the internal combustion engine and the electric system, and the vehicle is operated according to the operating strategy (B). The invention also relates to another method in addition thereto and to a method for teaching a neural network.

Description

description

title

Method of operating a vehicle

The present invention relates to a method for operating a vehicle with a drive train that has a plurality of drive units, at least one of which is designed as an electric machine and one as an internal combustion engine, a method for teaching in an artificial neural network, and a computing unit and a computer program for carrying it out.

State of the art

In addition to motor vehicles with only one internal combustion engine, there are also more and more motor vehicles with one or more electric drive units in addition to the internal combustion engine. Such vehicles are then so-called hybrid vehicles. There are also vehicles with only electric drives.

In vehicles with multiple drive units, it is desirable to find an operating strategy that is as optimal as possible for dividing a required torque or a required power between the drive units. This should typically be done with the aim of optimizing energy efficiency, which usually also includes an energy storage system. The latter can also be relevant for purely electrically operated vehicles with only one or more electric drive units (but eg no internal combustion engine that transmits torque directly to the drive wheels) and, for example, several energy storage units (eg also including fuel cells). Disclosure of Invention

According to the invention, a method for operating a vehicle, a method for teaching in an artificial neural network and a computing unit and a computer program for carrying them out with the features of the independent patent claims are proposed. Advantageous events are the subject of the dependent claims and the description below.

The present invention relates to a method for operating a vehicle with at least one electrical machine designed as a drive unit, an internal combustion engine and an electrical energy storage system (e.g. with one or more batteries) for the at least one electrical machine designed as a drive unit. A drive unit is to be understood here in particular as meaning that a torque for driving at least one wheel of the vehicle can be transmitted to this wheel. In this respect, in the case of electrical machines, such electrical machines come into consideration that can be operated as a motor. However, it is expedient if such electrical machines can then also be operated as generators.

A so-called hybrid vehicle, in which the internal combustion engine is (likewise) designed as a drive unit, comes into consideration here as a vehicle. In particular, a transmission with, for example, several gears can then also be provided here. However, the specific topology of the hybrid vehicle (eg a so-called P2 topology) is not relevant here, ie one or more electric machines can be installed before and/or after a transmission and/or on one axis and/or on one or more be arranged wheels. However, a vehicle is also conceivable in which the internal combustion engine is only provided (e.g. via a coupling with an electrical machine or a generator) to generate electrical energy for charging the electrical energy storage system or for operating the electrical machine used as a drive unit. to provide. In the hybrid vehicles mentioned in particular, in which the state of charge of the electrical energy storage system, i.e. of a battery in particular, can also vary, there are several degrees of freedom that can be taken into account when a required torque is to be distributed to the drive units. In particular due to the degree of freedom of the state of charge or its change, emissions, in particular of carbon dioxide, can be reduced. For example, the internal combustion engine can be operated at a load point that is as optimal as possible with regard to low emissions. To compensate for the torque supplied by the internal combustion engine and the torque requested by the vehicle, the electric machine can be operated as a generator or as a motor, depending on the situation, ie the battery is charged or discharged.

In order to always make the operation of the vehicle as efficient as possible in this respect, a physical model can be used in which the drive train (with the drive units) and, if necessary, the energy storage system (and possibly also an internal combustion engine not used for the drive) are mapped will. For example, using a cost function, as in DE 102005 044268 A1, the most efficient operating strategy for the vehicle with the electric machines and the internal combustion engine, and thus in particular also the drive train, can be determined. With such a cost function, the "cost" (for example the total energy consumption taking into account efficiency) of the power that can be made available by the battery can be related to the "cost" of the power that can be made available by the internal combustion engine. The respective costs can be appropriately weighted. This is also referred to as the "Equivalent Consumption Minimization Strategy" (ECMS).

Irrespective of whether a physical model is used or not, when using a cost function in order to optimize consumption efficiency (this relates in particular to energy efficiency, i.e. the ratio of usable to supplied energy, but possibly also taking undesirable emissions into account ) of the internal combustion engine and the electrical system to determine an operating strategy for the vehicle is a crucial one Point is a reference variable that relates an efficiency (the efficiency in particular with regard to energy consumption) of the internal combustion engine and an efficiency of the electrical system, comprising the at least one drive unit designed as an electrical machine and the energy storage system. This reference value is required (in connection with the cost function) in order to compare the efficiency of the internal combustion engine (here fuel consumption and possibly exhaust gas emissions are relevant) with the efficiency of the electrical system (here, in particular, the state of charge of the electrical energy storage system is not too full and should not be too empty, relevant, possibly also a loss in the inverter and the like) to be able to compare with each other in order to be able to weigh up against each other. This reference value can also be referred to as an equivalence factor, since it can be used to determine values for consumption that are equivalent to one another for the electrical system and the internal combustion engine.

A previous option was to estimate such a reference variable as precisely as possible (particularly taking into account the average efficiency ratios of the electrical system and the internal combustion engine) and then set it to a specific value. However, as has now been shown, these efficiencies or efficiency ratios depend very heavily on the operating point, in particular on the internal combustion engine. If a current driving situation deviates greatly from the driving situation assumed for determining the reference variable, it will therefore not be possible to determine an optimal operating strategy with the fixed reference variable.

Against this background, it is proposed that the reference variable (possibly only implicitly) be determined using a method based on artificial intelligence, in particular an artificial neural network, specifically repeatedly, for example at certain time intervals or according to other suitable methods Criteria such as distance traveled. An operating strategy for the drive train can then be determined by using the reference variable while optimizing the consumption efficiency of the internal combustion engine and the electrical system. The drive train is then operated according to the (thus determined) operating strategy (e.g. with a corresponding distribution of torque ments on the drive units or electric machine (s) and internal combustion engine).

By means of artificial intelligence or an (artificial) neural network, the reference variable can be determined particularly precisely and dynamically. The neural network can, for example, be trained or taught based on a consumption efficiency for different scenarios of the use of electric machine and internal combustion engine and for different driving situations or driving cycles. With a neural network that has been taught in once, the computing effort during operation of the vehicle can also be kept low. This is because the learning can take place initially before the vehicle is put into operation for the first time and, if necessary, also later at certain time intervals (even if the vehicle is otherwise not being used).

While the reference variable is determined using a method based on artificial intelligence, in particular an artificial neural network, the operating strategy can be determined (with optimization of consumption efficiency) from the reference variable, for example using (or based on) a Cost function (e.g. the ECMS mentioned at the beginning) can be determined. This means, for example, that a cost function that has already been implemented can be retained, while the reference value used in it is always determined up-to-date.

However, it is also preferred if the determination of the operating strategy (while optimizing the consumption efficiency) is also determined using a (further) method based on artificial intelligence, in particular an artificial neural network. Here, for example, two different (artificial) neural networks can be used, with the reference variable from the first neural network being used as the input for the second neural network. This enables the operating strategy to be determined particularly precisely and with little computational effort.

However, it is also particularly useful if the operating strategy directly represents the output variable of the method based on artificial intelligence. This combines the determination of the reference value and the determination of the operating strategy in a method based on artificial intelligence.

The reference variable no longer has to be available as an output variable.

The operating strategy or the reference variable is preferably determined for a given driving situation and based on previous driving data. In addition or as an alternative, it is also preferred if the operating strategy or the reference variable is determined for a given driving situation and based on future driving data, then, for example, using navigation data and/or environment sensors on the vehicle, for example for a specific route and /or Duration of the expected driving situation (e.g. with regard to expected speed, expected necessary torque and the like, possibly also based on the route with possible curves and speed restrictions) reliably and as accurately as possible in advance.

Information, e.g. about the distance traveled and/or distance still to be travelled, one or more of which are preferably taken into account as an input variable:

- a temporal proportion of high torque (e.g. above a specified threshold value, in particular over a certain route),

- a temporal proportion of low torque (e.g. below a specified threshold value, in particular over a certain distance),

- a temporal proportion of high engine speed (e.g. above a specified threshold value, in particular over a certain route),

- a period of low engine speed (e.g. below a specified threshold value, in particular over a certain route),

- a standard deviation of the speed over a certain driving distance and

- a standard deviation of the total requested torque over a certain distance ne.

The operating strategy or the reference variable can thus be determined particularly precisely. Information about the current driving situation, a required total torque, a battery state of charge (in the case of the electrical energy storage system), a speed of the internal combustion engine and the electrical machine as well as other operating limits of the vehicle, in particular from Electrical machine, internal combustion engine and electrical energy storage system into consideration. The information mentioned above for determining the reference variable can also be used for this purpose. The operating strategy, in particular with an optimal distribution of the total torque to the electric machine(s) and the internal combustion engine, can then be obtained directly as an output—and thus particularly quickly.

The invention also relates to a method for learning (also referred to as training) an artificial neural network, which can be used to determine the reference variable or directly the operating strategy in a method explained above. In this case, driving data is simulated for different periods of time, which reference value results in the optimum operating strategy for the vehicle for each period of time in terms of consumption efficiency of the internal combustion engine and electrical system. If the operating strategy is determined directly, driving data can also be used to simulate which operating strategy is optimal for the vehicle in terms of consumption efficiency of the internal combustion engine and electrical system for each time segment.

This makes it possible to train the neural network in order to determine the reference variable or the operating strategy, which as a result provides the best possible distribution of a total torque to the at least one electric machine and the internal combustion engine - i.e. the operating strategy - in particular based on a current driving situation.

A computing unit according to the invention, for example a control unit of a motor vehicle, is set up, in particular in terms of programming, to carry out a method according to the invention. Although the artificial neural network can also be taught in on a control unit of a motor vehicle, but also on an external (particularly very powerful) processing unit such as a computer.

The implementation of a method according to the invention in the form of a computer program or computer program product with program code for carrying out all method steps is advantageous because this causes particularly low costs, especially if an executing control device is also used for other tasks and is therefore available anyway. Suitable data carriers for providing the computer program are, in particular, magnetic, optical and electrical storage devices such as hard drives, flash memories, EEPROMs, DVDs, etc. It is also possible to download a program via computer networks (Internet, intranet, etc.).

Further advantages and refinements of the invention result from the description and the attached drawing.

The invention is illustrated schematically in the drawing using exemplary embodiments and is described below with reference to the drawing.

Brief description of the drawings

FIG. 1 schematically shows a vehicle in which a method according to the invention can be carried out.

FIG. 2 shows the determination of an operating strategy as part of a method according to the invention in a preferred embodiment.

FIGS. 3 to 5 show schematic sequences of methods according to the invention in various preferred embodiments.

FIG. 6 schematically shows a sequence of a method according to the invention in a further preferred embodiment. embodiment(s) of the invention

FIG. 1 shows a vehicle 100 in which a method according to the invention can be carried out. The vehicle 100 has two axles 110 and 120, the axle 120 as a drivable axle—with correspondingly drivable wheels—being connected to a drive train 101. The vehicle 100 or the drive train 101 has an internal combustion engine 130 designed as a drive unit and an electric machine 140 designed as a drive unit, which can be connected by means of a clutch 131 in a torque-transmitting manner.

Furthermore, an electrical energy storage system 150 designed as a battery or having a battery is provided, which is electrically connected to the electrical machine 140 . In the drive train 101, a transmission 160 is also provided, by means of which different gears can be set or can be selected.

Furthermore, a computing unit 180 designed as a control unit is provided, by means of which the drive units, the clutch and, if applicable, the transmission can be controlled. It goes without saying that several computing units, which then communicate with one another, can also be provided for this purpose.

The vehicle 100 is therefore a hybrid vehicle. The invention is to be explained here by way of example. It goes without saying that, as mentioned at the outset, other types of vehicles or topologies can also be used.

In Figure 2, the determination of an operating strategy is shown using a diagram in which a consumption V (also representative of costs) over a torque M in Nm of the electric machine (designated M (E)) and the internal combustion engine (designated M (V) designated) are applied. The associated curves for the consumption are shown with V _E for the electrical machine (or the entire electrical system) and with Vv for the internal combustion engine. By way of example, a total torque to be provided by 100 Nm, for example, at a rotational speed (the engine) of 1000 min ^-1.

The curves show that when the share of the internal combustion engine increases, the consumption (of fuel) increases, while the consumption (of electrical energy) of the electric machine decreases or even becomes negative (since its share of the total torque decreases accordingly). ). The same applies vice versa. The course K now represents a cost function (for example within the framework of the ECMS mentioned at the outset) in which the consumption of the internal combustion engine and the consumption of the electrical machine are added in a weighted manner. This means that K=Vv+S*V _E , where S is the reference value used within the scope of the invention (and as previously with ECMS), which relates the two loads to one another.

In order to now - with a given reference variable S - find an operating strategy, i.e. a division of the total torque of 100 Nm between the internal combustion engine and the electric machine, while consumption or consumption efficiency is optimized, a minimum of the cost function K can be sought. This minimum is labeled P here and results in a torque of approximately 160 Nm for the internal combustion engine and -60 Nm for the electrical machine, i.e. the electrical energy store is charged.

As already mentioned, the use of a fixed value for the reference variable can lead to undesirable (or non-optimal) results in the operating strategy, so within the scope of the present invention, this reference variable using a method based on artificial intelligence, in particular an artificial neural network, for example, is regularly redetermined. Using this reference variable or its value, however, the operating strategy can then be determined, at least according to one embodiment, as explained with reference to FIG. 2, as also explained below.

FIGS. 3 to 5 show schematic sequences of methods according to the invention in various preferred embodiments. In Figure 3 is next shown a variant in which the reference variable S is determined using an artificial neural network NNi.

For this purpose, the artificial neural network NNi receives as inputs, for example, the variables for driving behavior DSI of the driver ("Driver Style Indication"), which indicates, for example, typical behavior with regard to acceleration, etc., an average speed Vmean, a maximum speed v _max , a minimum speed V _min , a state of charge SoC of the battery or of the electrical energy storage system ("state of charge"), an average total torque requirement of the driver Mc.mean, a maximum total torque requirement of the driver Mc.max, a minimum total torque requirement of the driver MG, min and a height difference DH to be overcome on the route, if available (e.g. with the integration of navigation data). These input variables in the sense of driving data can be determined and used for a specific past period of time and/or—then predictively, for example based on navigation data—for a specific future period of time.

The reference variable S determined in this way is then used in a cost function K in order—as already explained with reference to FIG. 2—to determine the operating strategy B and the optimal costs K′. For this purpose, the cost function K receives a required total _{torque M G} and a current speed n of the internal combustion engine as an input variable. The operating strategy B as an output variable specifies in particular a torque to be set by the electric machine and a torque by the internal combustion engine, which are then set by suitable activation.

These costs K′ then physically determine the optimum torque of the internal combustion engine and the electric machine in order to optimally meet a total torque that is immediately required by the driver in terms of energy. This specific division of the torques between the internal combustion engine and the electric machine is then coordinated in, for example, a computing unit designed as an engine control unit and ultimately supplied to the respective machines as torque requirements. By constantly redetermining the reference variable S using the artificial neural network NNi, the operating strategy can always be operated as optimally as possible in terms of consumption efficiency (ie high efficiency in terms of energy or fuel consumption and possibly emissions). A particular advantage of using machine learning methods lies in the simple and fast calculation (or, in the case of repeated calculation, the adaptation) of the reference value if the method, e.g. the artificial neural network, has been trained or taught (in advance). is.

A further variant is shown in FIG. 4, in which the reference variable S is determined using the artificial neural network NNi. This can be done as explained with reference to FIG. 3, so that reference can be made to the description there.

However, the reference variable S is then not supplied to the cost function; instead, a (second or further) artificial neural network NN2 is used instead of the cost function. This receives as input in addition to the reference value S and the total moment M _G (as in the cost function), and n is an actual speed. As starting the artificial neural network NN2 then supplies the Be sales strategy B (as in the cost function), and the optimum costs K ' for the drive train or electric machine and internal combustion engine.

As mentioned, the current speed of the internal combustion engine can be directly taken into account in the artificial neural network NN2, in particular if the artificial neural network NN2 has been trained accordingly for different speeds. In this way, an optimal operating strategy can be found even more easily and quickly for different speeds.

In a simple form of an operating strategy, the torque distribution can be optimized for a given actual speed of the internal combustion engine and possibly the electric machine(s). The actual speed then results from the vehicle speed and the drive train transmission ratio for the current gear. A further execution of the operating strategy allows a joint optimization of torque split and gear. The speed is also optimized with the gear for a specific speed. To this end, for example, the costs are calculated for the optimal torque distribution for each gear, and the gear is then selected for which the costs are lowest.

Figure 5 shows a further variant in which _{the operating strategy is determined directly using an artificial neural network NN 3} , but the reference variable is not explicitly determined, but with a relationship between the efficiency of the internal combustion engine and the efficiency of the electrical system is implicitly taken into account in the artificial neural network, for example after suitable training.

In contrast to Figure 4, the artificial neural network NN ₃ itself receives the variables for the required total _{torque M G} and the speed n of the internal combustion engine as inputs, as well as other optional inputs for a time component f _M.Hi of the high torque (e.g above a specified threshold value, in particular over a specific route), a time component f _M,io of low torque (e.g. below a specified threshold value, in particular over a specific route), a time component f _n ,Hi of high engine speed (e.g. above a predetermined threshold value, in particular over a specific route) and a time proportion f _n , i_o at low speed (eg below a predetermined threshold value, in particular special over a specific route). These input variables in the sense of driving data can be determined and used for a specific past period of time and/or, then predictively, for a specific future period of time. The optional inputs mentioned can also be present in whole or in part in the embodiments of NN1 according to FIG. 3 or 4.

With comparable input variables, as are used in the variant according to FIG be, as is also the case with the variant according to Figure 4, but without ne explicit determination of the reference value. Even if such an artificial neural network may be somewhat more difficult to learn, an optimal operating strategy can be obtained particularly easily and quickly.

FIG. 6 schematically shows a sequence of a method according to the invention in a further preferred embodiment, namely for teaching in a neural artificial neural network, as is described, for example, with reference to FIG. 4 or 5.

First, in a step 600, a progression of a speed (over time) and, if desired, other measured variables for the vehicle in question is determined for a specific time segment. From these, in a step 610, using a (e.g. physical) model of the vehicle, in which e.g. driving dynamics properties and various gear ratios and the like are taken into account, various variables, if they were not measured, are determined.

In step 620, for example, a speed, a (total required) torque and operating limits of the electric machine and internal combustion engine, each over time, are provided as such variables that are then used in step 650 to optimize a cost function (e.g. ECMS ) be used. In addition, in a step 625, different values for the reference variable S are provided (by variation), which are also used to optimize the cost function. Likewise, in a step 630, a state of charge is provided and used to optimize the cost function.

In step 650, an optimal value for the reference variable is then sought for, for example, the past minute (i.e., for example the time period mentioned) for each value of the reference variable used, i.e. an operating strategy is determined from the various values from step 625 and checked with which Value for the reference value that consumption efficiency is optimal. Such an optimal value for the datum for that particular period of time is then provided in step 670 . In addition, in step 610 further variables are determined, which are provided in step 640, namely, for example, an average speed, an average, a maximum, and a minimum required total torque, each over the specific period of time, eg, over the last minute. These variables and the state of charge from step 630 are then used to find the optimal value of the reference variable in a step 660 for the artificial neural network, which is provided in step 680 . The results from step 670 are compared to the results from step 680 to train the neural network.

These steps, which are explained for a period of time, can now be carried out, for example on a suitable computer or possibly also on a control unit of a vehicle, for a large number of different periods of time, for example from a history of data from the vehicle (or a specific vehicle type). the one with which the artificial neural network is trained. As explained above, this can then be used during operation of the vehicle in order to always determine the current reference variable.

Claims

Expectations

1. A method for operating a vehicle (100) with at least one electric machine (140) designed as a drive unit, an internal combustion engine (130) and an electrical energy storage system (150) for the at least one electric machine designed as a drive unit, wherein using a Method based on artificial intelligence, in particular an artificial neural network (NNi, NN ₃ ), in which a relationship between an efficiency of the internal combustion engine (140) and an efficiency of the electrical system, comprising the at least one electrical machine (130) designed as a drive unit and the electrical energy storage system (150), is taken into account in relation to one another, an operating strategy (B) for the vehicle (100) is determined by optimizing a consumption efficiency (V) of the internal combustion engine and electrical system, and the vehicle (100) is determined according to the operating strategy (B ) is operated.

2. The method as claimed in claim 1, wherein at least one of the following _information is taken into account as an input variable for the method (NN1, NN ₃ ) based on artificial intelligence: a time component (f M.Hi ) of high torque, a time component ( fM,i_o) at low torque, a time component (f _n ,Hi) at high speed, a time component (f _n ,i_o) at low speed, a standard deviation of the speed over a specific driving distance, and a standard deviation of the total requested torque over a certain distance.

3. The method according to claim 1 or 2, wherein at least one of the following information is taken into account _{as an input variable in the method (NN1, NN 3 ) based on artificial intelligence: a driving behavior (DSI) of the} driver, an average speed (v _me an), a maximum speed (Vmax), a minimum speed (v _m m), a state of charge (SoC) of the electrical energy storage system, an average total torque request from the driver (Mc.mean), a maximum total driver torque requirement (Mc.max), a minimum total driver torque requirement (MG,min) and a height difference (DH) to be negotiated on the route.

4. The method according to any one of the preceding claims, wherein a required total torque (MG) and / or a speed (n) of the internal combustion engine when determining the operating strategy (B), in particular as an input variable based on artificial intelligence method (NIM2, NIM3), be taken into account.

5. The method according to any one of the preceding claims, wherein using the artificial intelligence-based method (NN1), in particular as the output variable, a reference variable (S) is determined, which determines the efficiency of the internal combustion engine (130) and the efficiency of the electric System, comprising the at least one electrical machine (140) designed as a drive unit and the electrical energy storage system (150), relating to one another, wherein the operating strategy (B) for the vehicle (100) is determined using the reference variable (S).

6. The method according to claim 5, wherein the operating strategy (B) is determined using a method based on artificial intelligence, in particular an artificial neural network (NN2).

7. The method according to any one of claim 5, wherein the operating strategy (B) using a cost function (K) is determined.

8. The method according to any one of claims 1 to 4, wherein the operating strategy (B) is determined as the output variable of the method (NN3) based on artificial intelligence.

9. The method as claimed in one of the preceding claims, in which the operating strategy (B) is determined for a predefined driving situation and based on past driving data and/or based on future driving data.

10. The method according to any one of the preceding claims, wherein the internal combustion engine (140) is designed as a drive unit or in which the internal combustion engine (140) only for generating electrical energy.

11. A method for teaching an artificial neural network (NNi, NN ₃ ) for a method according to any one of the preceding claims, wherein driving data is simulated for different periods of time, which reference variable for each period of time is the optimal operating strategy for the vehicle with regard to consumption efficiency of the internal combustion engine and electrical system results, or with driving data being simulated for different periods of time, which operating strategy is optimal for the vehicle in terms of consumption efficiency of the internal combustion engine and electrical system for each period of time.

12. Arithmetic unit (180) which is set up to carry out all method steps of a method according to one of the preceding claims.

13. Computer program that causes a processing unit (180) to carry out all method steps of a method according to one of claims 1 to 11 when it is executed on the processing unit (180).

14. Machine-readable storage medium with a computer program stored thereon according to claim 13.