Control device and method for controlling a drive system of a hybrid electric vehicle

Abstract

Method for controlling a drive system (202) of a hybrid electric vehicle (200), the method comprising: ◯ Determining auxiliary optimization data (105h) based on a first model (M1) by means of a first optimization method (OPT1), wherein the first model (M1) comprises: a first continuous state parameter representing a charge / discharge state of an energy storage device (216) of the drive system (202), a second continuous state parameter representing a travel time (t), a first control parameter set representing continuous control variables of the drive system (202), and wherein the auxiliary optimization data (105h) represent a profile of the first state parameter and the second state parameter associated with a predefined travel route,wherein the first optimization method (OPT1) comprises sequential quadratic programming (SQP) for determining the auxiliary optimization data (105h) based on route data (103) and actual control variable data (105i) taking into account the first and second continuous state parameters and the first control parameter set; , ◯ Determining predefined control variable data (105s) based on a second model (M2) by means of a second optimization method (OPT2), wherein the second model (M2) comprises: a first state parameter representing the charge / discharge state of the energy storage device (216), a second state parameter representing the travel time (t), a first control parameter set representing the continuous control variables of the drive system (202), and a second control parameter set representing discrete control variables of the drive system (202), wherein the predefined control variable data (105s) represent a predefined course of the continuous control variables and the discrete control variables of the drive system (202) for a portion of the predefined travel route, wherein the second optimization method (OPT2) comprises dynamic programming (DP) using Pontrjagin's maximum principle (PMP),and wherein, by means of the determined auxiliary optimization data (105h), target values ​​for the first state parameter and the second state parameter of the second model (M2) are determined for the part of the predefined route in order to determine the specified control variable data (105s) based on the route data (103), the actual control variable data (105i) and the determined auxiliary optimization data (105h), taking into account the first and second continuous state parameters and the first and second control parameter sets, ◯ wherein at least one control parameter of the second control parameter set is not optimized in the first optimization method (OPT1), ◯ wherein the auxiliary optimization data (105h) represent target specifications for the charge / discharge state of the energy storage device (216) and the travel time (t) in relation to a section of the predefined route taken into account in the second optimization method (OPT2), and ◯ wherein the second optimization method (OPT2) is arranged to solve a 2-point boundary value problem, wherein the auxiliary optimization data (105h) represent objectives for the 2-point boundary value problem; and o Outputting the default control variable data (105s) for operating the drive system (202) based on the output default control variable data (105s).

Description

[0001] Various embodiments relate to a control device for controlling a drive system of a hybrid electric vehicle and a method for controlling a drive system of a hybrid electric vehicle.

[0002] In general, cruise control systems can be used in automotive engineering to assist the driver in longitudinal control. Conventional cruise control systems, for example, regulate the speed to a static setpoint with or without consideration of other road users, but pay little or no attention to the drive energy used and thus the operating costs.

[0003] To calculate an optimal speed, for example, a so-called optimal control problem can be solved. However, particularly for drive systems of hybrid vehicles on comparatively long journeys (e.g., journeys of more than 50 km, e.g., more than 100 km, or more than 200 km), this involves too high a computational effort to perform these calculations in a reasonable time, e.g., in less than 10 minutes, less than 1 minute, or less than 10 seconds, especially using the computing power currently available in hybrid vehicles. For example, a computer system with a computing power in the range of approximately 1 GFLOPS to approximately 100 TFLOPS can currently be used in a hybrid vehicle.

[0004] DE 10 2014 209 687 A1 describes a method for the predictive operation of a motor vehicle based on a specification of several driving-relevant control variables.

[0005] DE 10 2008 038 753 A1 describes a method for controlling a hybrid drive in a rail vehicle, in which an electronic route timetable is specified as route-section-related speeds via a train control unit, and route-section-related drive types of the rail vehicle are predictively determined by means of the electronic route timetable before the start of the journey.

[0006] DE 10 2011 085 454 A1 describes a method for controlling a hybrid drive with an electrical energy storage device for a rail vehicle, in which an electronic route timetable is specified for each section of the route via a train control unit, and a section-related drive type of the rail vehicle is predictively determined by means of the electronic route timetable.

[0007] DE 102 22 699 A1 describes approximation models that are created for real parameter target range points that were generated, for example, by tests / simulations.

[0008] US 2003 / 0 018 400 A1 describes a method for determining an input command profile for essentially any dynamic system that can be modeled as a linear system, wherein the input command profile serves to transition an output of the dynamic system from one state to another state.

[0009] Sager, Sebastian. "Numerical methods for mixed-integer optimal control problems." Tönning: Der Andere Verlag, 2005, describes the development of numerical methods for solving mixed-integer optimal control problems.

[0010] Guo, Lulu, et al. “On-line optimal control of the gearshift command for multispeed electric vehicles.” IEEE / ASME Transactions on Mechatronics 22.4 (2017): 1519-1530, describes a method for shift strategy in propulsion systems to investigate the energy saving potential of an electric vehicle equipped with a multi-speed automated manual transmission.

[0011] DE 10 2013 016 569 A1 describes an operating method for a hybrid drive of a motor vehicle that can be operated with several operating modes.

[0012] The invention is based on the object of providing a control device and a method for controlling a drive system of a hybrid electric vehicle, which can provide a solution to the optimal control problem with low computational effort. This object is achieved by the independent claims. Advantageous further developments arise from the dependent claims.

[0013] According to various embodiments, a control device and a method for controlling a drive system of a hybrid electric vehicle are provided. The control device and the method are based on solving an optimal control problem and are configured, for example, such that the computational effort for solving the optimal control problem is kept so small that the optimal control problem can be calculated using an in-vehicle computing system and is thereby solved almost optimally. For this purpose, a model-based predictive control (abbreviated as MPC and referred to as model predictive control) with two optimization stages is implemented.A first (upper) MPC stage is used, which, for example, takes into account the entire planned route in order to calculate partial route target values ​​using a sequential quadratic program. The partial route target values ​​are used in a second (lower) MPC stage to determine the optimal control of the vehicle for partial routes (i.e., with a shorter horizon than the first MPC stage) using the partial route target values. The optimal control problem is solved in the lower MPC stage using a combination of dynamic programming and Pontrjagin's maximum principle. The low computational effort of the model predictive control with two optimization stages described here can be used in a hybrid electric vehicle (e.g.in a parallel hybrid electric vehicle) can achieve energy and cost savings of up to 38% compared to a conventional cruise control system with essentially constant speed settings.

[0014] According to various embodiments, a hybrid electric vehicle may include an internal combustion engine and an electric motor, which may be used alternately or simultaneously to power the hybrid electric vehicle. It is understood that the principles described herein may also be applied to similar vehicles with two different powertrains.

[0015] One aspect of various embodiments relates to model predictive control (MPC) of a drive train of a vehicle with two different drive types, e.g., a hybrid electric vehicle (HEV). Due to the two different drive types, the overall model used for the model predictive control or regulation is comparatively complex, i.e., the parameter space for describing the drive system is comparatively large (e.g., 5 to 20 parameters or more than 20 parameters are used) and, in particular, of a mixed integer type. Due to the complexity, the computational effort can conventionally be a limiting factor for the use of such control or regulation systems in a vehicle, provided that a computing system of the vehicle itself is to be used to perform the calculations of the model predictive control or regulation.

[0016] According to various embodiments, model predictive control takes place in two optimization stages (also referred to as MPC stages). The first optimization stage, for example, provides an approximation of lower quality than the second optimization stage and, in contrast, has a larger horizon than the second optimization stage. Thus, in the first optimization stage (albeit with a loss in quality), a large total distance can be taken into account, e.g., more than 200 km, and in the second optimization stage, a calculation with a higher quality can be carried out for various route sections (referred to as sub-sections) of the total distance (e.g., without a loss in quality). In particular, the second optimization stage is set up in such a way that the optimal control problem is completely solved, i.e., all model parameters used are also taken into account in the second optimization stage.

[0017] According to various embodiments, the first stage of the model predictive control is based on an approximation using SQP, i.e., a sequential quadratic program. The second stage is based on a calculation using a combination of dynamic programming (DP) and the Pontryagin Maximum Principle (PMP). SQP, for example, is sufficiently well suited to approximating the relevant continuous variables with nonlinear behavior, e.g., the efficiency maps of various drives, the state of charge of an energy storage device (e.g., a battery), the fill level of a fuel tank, etc. The combination of PMP and DP, for example, allows a sufficiently good calculation of a mixed (integer / continuous) state space. Integer states include, for example, the gear of the vehicle transmission, the two states of a start-stop function, etc.

[0018] According to various embodiments, a combination of SQP and PMP-DP is used in a two-stage optimization principle for energy-optimized control of the longitudinal guidance and powertrain of a hybrid electric vehicle. This results in the advantage, for example, that a complete optimization of all relevant parameters is possible in the second stage (PMP-DP) and can be performed in such a computationally efficient manner that the calculation can be carried out in a corresponding vehicle. The result of the two-stage optimization is, for example, of high quality compared to an optimal calculation based on DP.

[0019] According to various embodiments, a control / regulating device can be configured to control / regulate a drive system of a hybrid electric vehicle and can comprise at least one processor. The at least one processor can, for example, be configured to: (I) receive route data, wherein the route data, for example, associated with a predefined route, represent a predefined gradient profile and a predefined speed limit profile; (II) receive actual control variable data, wherein the actual control variable data represent an actual state of at least one continuous control variable and at least one discrete control variable of the drive system; (III) determine auxiliary optimization data based on a first model by means of a first optimization method, wherein the first model comprises: a first state parameter,which represents a charge / discharge state of an energy storage device of the drive system, a second state parameter representing a travel time, and at least one continuous control parameter representing the at least one continuous control variable of the drive system, and wherein the auxiliary optimization data represent a profile of the first state parameter (i.e., the charge / discharge state of the energy storage device) and the second state parameter (i.e., the travel time), wherein the first optimization method comprises sequential quadratic programming (optimization, SQP); (IV) to determine, based on a second model, by means of a second optimization method, default control variable data, wherein the second model comprises: a first state parameter representing the charge / discharge state of the energy storage device, a second state parameter representing the travel time,at least one continuous control parameter representing the at least one continuous control variable of the drive system, and at least one integer control parameter representing the at least one discrete control variable of the drive system, wherein the specified control variable data represent a specified curve of the at least one continuous control variable and the at least one discrete control variable of the drive system for a respective section of the predefined route, wherein the second optimization method comprises dynamic programming (optimization, DP) using the Pontrjagin Maximum Principle (PMP),and wherein, using the determined auxiliary optimization data, target values ​​for the first state parameter and the second state parameter of the second model are determined for the respective section; and (V) outputting the specified control variable data for operating the drive system based on the outputted specified control variable data.

[0020] Furthermore, another aspect of various embodiments can be clearly seen in providing a control / regulating device which is configured to: (I) receive route data, wherein the route data represent at least one predefined gradient profile and a predefined speed limit profile associated with a predefined route; (II) receive actual control variable data, wherein the actual control variable data represent a respective actual state of a control parameter set, wherein the control parameter set comprises continuous control variables and discrete control variables of the drive system; (III) determine auxiliary optimization data based on a first model by means of a first optimization method, wherein the first model comprises: a first continuous state parameter representing a charge / discharge state of an energy storage device of the drive system,a second continuous state parameter representing a travel time, a first control parameter set representing the continuous control variables of the drive system, and wherein the auxiliary optimization data represent a profile of the first state parameter and the second state parameter associated with the predefined travel route, wherein the first optimization method comprises sequential quadratic programming (optimization, SQP) for determining the auxiliary optimization data based on the received travel route data and actual control variable data, taking into account the first and second continuous state parameters and the first control parameter set; (IV) to determine predefined control variable data based on a second model by means of a second optimization method, wherein the second model comprises: a first state parameter representing the charge / discharge state of the energy storage device,a second state parameter representing the travel time, a first control parameter set representing the continuous control variables of the drive system, and a second control parameter set representing discrete control variables of the drive system, wherein the specified control variable data represent a specified curve of the at least one continuous control variable and the at least one discrete control variable of the drive system for a portion of the predefined route, wherein the second optimization method comprises dynamic programming (optimization, DP) using Pontrjagin's Maximum Principle (PMP), and wherein target values ​​for the first state parameter and the second state parameter of the second model for the partial route are determined using the determined auxiliary optimization data in order to determine the specified control variable data based on the received route data,the actual control variable data and the determined auxiliary optimization data, taking into account the first and second continuous state parameters and the first and second control parameter sets; and (V) output the specified control variable data for operating the drive system based on the outputted specified control variable data.

[0021] Furthermore, another aspect of various embodiments can be clearly seen in providing a method for controlling a drive system of a hybrid electric vehicle, wherein the method can comprise, for example: (I) determining auxiliary optimization data based on a first model by means of a first optimization method, wherein the first model comprises: a first continuous state parameter representing a charge / discharge state of an energy storage device of the drive system, a second continuous state parameter representing a travel time, a first control parameter set representing continuous control variables of the drive system, and wherein the auxiliary optimization data represent a profile of the first state parameter and the second state parameter associated with a predefined travel route,wherein the first optimization method comprises sequential quadratic programming for determining the auxiliary optimization data based on route data and actual control variable data, taking into account the first and second continuous state parameters and the first control parameter set; (II) determining default control variable data based on a second model by means of a second optimization method, wherein the second model comprises: a first state parameter representing the charge / discharge state of the energy storage device, a second state parameter representing the travel time, a first control parameter set representing the continuous control variables of the drive system, and a second control parameter set representing discrete control variables of the drive system,wherein the specified control variable data represents a specified curve of the continuous control variables and the discrete control variables of the drive system for a portion of the predefined route, wherein the second optimization method comprises dynamic programming using Pontrjagin's maximum principle, and wherein, using the determined auxiliary optimization data, target values ​​for the first state parameter and the second state parameter of the second model are determined for the portion of the predefined route to determine the specified control variable data based on the route data, the actual control variable data, and the determined auxiliary optimization data, taking into account the first and second continuous state parameters and the first and second control parameter sets; and (III) outputting the specified control variable data for operating the drive system based on the outputted specified control variable data.

[0022] Furthermore, another aspect of various embodiments can be clearly seen in providing a method for controlling a drive system of a hybrid electric vehicle, wherein the method can comprise, for example: (I) determining auxiliary optimization data based on a first model by means of a first optimization method, wherein the first optimization method comprises sequential quadratic programming for determining the auxiliary optimization data based on route data and actual control variable data, taking into account parameters which represent the drive system of the hybrid electric vehicle, and wherein the auxiliary optimization data represent a profile of a subset of the parameters assigned to a predefined route;(II) Determining specified control variable data based on a second model by means of a second optimization method, wherein the specified control variable data represent a specified curve of control variables of the drive system for a partial section of the predefined route, wherein the second optimization method comprises dynamic programming using Pontrjagin's maximum principle, and wherein, based on the determined auxiliary optimization data, target values ​​for one or more parameters of the second model for the partial section are determined in order to determine the specified control variable data based on the route data, the actual control variable data, and taking into account parameters that represent the drive system of the hybrid electric vehicle; and (III) Outputting the specified control variable data for operating the drive system based on the outputted specified control variable data.

[0023] Examples of embodiments are shown in the figures and are explained in more detail below.

[0024] It shows Fig. 1 a control device in a schematic representation, according to various embodiments; Fig. 2 shows a hybrid electric vehicle with a control device in a schematic representation, according to various embodiments; Fig. 3A shows an example of an efficiency map of an internal combustion engine, which can be taken into account in the control by means of the control device, according to various embodiments; Fig. 3B shows an example of an efficiency map of an electric motor, which can be taken into account in the control by means of the control device, according to various embodiments; Fig. 4 shows the principle of an optimal control problem that can be solved by two optimization stages based on SQP and PMP-DP, according to various embodiments; Fig. 5 and Fig. 6 each shows a solution principle of an optimal control problem using PMP-DP, according to various embodiments; Fig. 7 shows exemplary route data which can be supplied to the control device according to various embodiments; Fig. 8 shows a relationship between the solution quality of a PMP-DP calculation of an optimal control problem and the computational time, according to various embodiments; Fig. 9 shows, by way of example, a fuel consumption map for an internal combustion engine-transmission unit under a fuel-optimal gear selection, according to various embodiments; Fig. 10 shows a two-stage model-based predictive control device in a schematic representation, according to various embodiments; Fig. 11 is a schematic representation of a second stage of a two-stage model-based predictive control device, according to various embodiments; and Fig. 12 illustrates an exemplary tangent function that may be used in a controller to avoid violations of energy storage limits, according to various embodiments.

[0025] In the following detailed description, reference is made to the accompanying drawings, which form a part hereof, and in which is shown by way of illustration specific embodiments in which the invention may be practiced. It is understood that other embodiments may be utilized and structural or logical changes may be made without departing from the scope of the present invention. It is to be understood that the features of the various exemplary embodiments described herein may be combined with one another unless specifically stated otherwise. The following description, therefore, is not to be taken in a limiting sense, and the scope of the present invention is defined by the appended claims.

[0026] In at least partially autonomous vehicles, for example, the driver can no longer freely choose the speed. This allows, for example, the speed to be automatically adjusted so that the energy costs required for a journey are minimized as much as possible or another destination can be optimized. The method or control device described here, for example, determines this optimized speed (or the control variables required to set the optimized speed) for every possible powertrain configuration (electric vehicle, conventional vehicle, hybrid electric vehicle).

[0027] According to various embodiments, a method for generating an optimized energy management and driving strategy is described. The method for generating an energy management and driving strategy relates to route data. The route data can, for example, represent a predetermined gradient and a predetermined permissible speed range along the upcoming route. The predetermined permissible speed range can, for example, have respective maximum and minimum speeds along the upcoming route. Furthermore, the travel time can be predetermined as a parameter. According to various embodiments, the cost-optimized values ​​of the control variables (e.g., gear selection, engine start / stop control, braking, combustion engine, and / or electric motor torque) are determined depending on a vehicle model.In a control loop, for example, the actual control variables are determined (represented by corresponding actual control variable data) and the control variables are adjusted accordingly according to the specified control variables (represented by corresponding specified control variable data). The specified control variables are determined based on a two-stage optimization process. This allows a clear reduction in the deviation of the actual control variables from the specified control variables.

[0028] Based on a corresponding model of a hybrid electric vehicle's drive system, a complex mixed-integer optimal control problem may arise that cannot be solved with sufficient accuracy and low computational effort using conventional solution methods. For example, conventional solution methods make simplifications, such as not considering the travel time as a state, in order to make the simplified problem computable using dynamic programming. Overall, conventional solution methods are not suitable for solving the complete optimization problem with such low computational effort that it can be calculated while the vehicle is moving.

[0029] The control device and the method described herein relate to optimal control in a two-stage, model-based, predictive control (MPC). In the first stage of the MPC, the problem approximated as a sequential quadratic program (SQP) is solved for the entire upcoming driving path. The state profiles determined in this way for the travel time and the energy storage energy are used in the second stage to solve the complete optimal control problem for short upcoming sections of the route. This second stage uses a combination of dynamic programming (DP) and the Pontryagin maximum principle (PMP), called PMP-DP. The combination of the SQP in the first stage of the MPC and the PMP-DP in the second stage of the MPC solves the complete optimal control problem in the second stage (e.g., in contrast to conventional bilevel MPC).Using the described combination of the algorithms SQP and PMP-DP, the computational effort for the complete optimal solution can be reduced for the first time to such an extent that the method can be calculated on an ECU for vehicle use, e.g. on an ECU with a computing power of 1 GLOPS to 100 TFLOPS.

[0030] Fig. 1 illustrates a control device 100 in a schematic view, according to various embodiments. The control device 100 can, for example, be used to control a drive system of a hybrid electric vehicle, e.g., as described herein, with an internal combustion engine and an electric drive configured in parallel thereto. It is understood that the control device 100 can also be used in the same or similar manner to control another drive system, e.g., a drive system of a hybrid fuel cell vehicle, or generally a vehicle with more than two different drive types. Furthermore, it is understood that the control device 100 described herein can also be configured as a control device, i.e., the actual control variables are not readjusted while the vehicle is traveling, but are, for example, determined only once in advance.Such control can be helpful in that, for example, routes that have been driven several times can be optimized in advance.

[0031] According to various embodiments, the control device 100 may have at least one processor 102. The at least one processor 102 or the control device 100 may, for example, be configured to receive route data 103. The route data 103 may, for example, represent a predefined gradient profile and / or a predefined speed limit profile associated with a predefined route (see, for example, Fig. 7). It is understood that the route data 103 can also represent other conditions related to the route. The predefined speed limit profile can be at least a profile of a maximum speed to be driven relative to the predefined route, e.g., relative to various sections of the predefined route. The predefined speed limit profile can also be a profile of a minimum speed to be driven relative to the predefined route, e.g., relative to various sections of the predefined route.

[0032] According to various embodiments, the at least one processor 102 of the control device 100 can be configured such that it receives (instead of the predefined speed limit profile or in addition thereto) route characteristics (e.g. curve radii and / or traffic sign information) and determines or adapts the predefined speed limit profile based on the route characteristics.

[0033] Furthermore, the at least one processor 102 or the control device 100 can be configured, for example, to output preset control variable data 105s (also referred to as target control variable data) and to receive actual control variable data 105i. According to various embodiments, the drive system of a hybrid electric vehicle can be operated based on control variables, wherein the actual control variable data 105i represents the respective actual values ​​(which are the currently available values) of the control variables and the target control variable data 105s represents the respective target values ​​(clearly adapted to the route) of the control variables. The control variables can have continuous variables (referred to as continuous control variables), such as the drive torque of the respective drive, the braking torque of the braking system, etc. Furthermore, the control variables can be discrete control variables, such as the state of a transmission (e.g.the gear of a gearshift), the state of a start / stop system for starting and stopping an internal combustion engine, etc.

[0034] According to various embodiments, the received actual control variable data 105i can represent an actual state of at least one continuous control variable and at least one discrete control variable of the drive system. The at least one processor 102 or the control device 100 can be configured to determine auxiliary optimization data 105h based on a first model (M1) using a first optimization method (OPT1).

[0035] According to various embodiments, the first model can comprise the following: a first state parameter, which represents, for example, a charge / discharge state of an energy storage device of the drive system, a second state parameter, which represents a travel time, and at least one continuous control parameter, which represents the at least one continuous control variable of the drive system. According to various embodiments, the first state parameter can assume a comparatively large number (e.g., more than 100, more than 500, or more than 1000) of mutually different values. According to various embodiments, the continuous control parameter can assume a comparatively large number (e.g., more than 100, more than 500, more than 1000) of mutually different values.The fact that, according to various embodiments, the travel time is included as a state parameter in the first model (M1) may, for example, make it possible to take into account corresponding boundary conditions for the travel time during the optimization.

[0036] According to various embodiments, the auxiliary optimization data 105h can represent a profile of the first state parameter (i.e., illustratively, a profile of the charge / discharge state of an energy storage device) and the second state parameter (i.e., illustratively, the travel time). The profile can be determined, for example, for the entire predefined travel distance, e.g., for a travel distance of more than 50 km, more than 100 km, or more than 200 km. It is understood that the profile can also be determined for a travel distance of less than 50 km. Within the scope of a control system, the profile can be determined, for example, for the entire predefined travel distance at predefined times.

[0037] According to various embodiments, the first optimization method may comprise or be sequential quadratic programming (optimization, SQP). By means of sequential quadratic programming, in particular models with continuous (e.g., non-integer, e.g., fractional-rational) parameters can be efficiently optimized. According to various embodiments, the first model (M1) may not optimize integer control parameters and / or integer state parameters. According to various embodiments, an integer parameter (e.g., state parameter, control parameter, etc.) may assume a comparatively small number (e.g., less than 50, less than 25, or less than 10) of mutually different values. The term program, as used herein with reference to a sequential quadratic program or a dynamic program, may also be referred to as programming, e.g.,sequential quadratic programming or dynamic programming.

[0038] Furthermore, the at least one processor 102 or the control device 100 can be configured to determine preset control variable data based on a second model (M2) by means of a second optimization method (OPT2).

[0039] According to various embodiments, the second model (M2) may, for example, comprise: a first state parameter representing the charge / discharge state of the energy storage device, a second state parameter representing the travel time, at least one continuous control parameter representing the at least one continuous control variable of the drive system, and at least one integer control parameter representing the at least one discrete control variable of the drive system.

[0040] According to various embodiments, the specified control variable data can represent a specified curve of the at least one continuous control variable and the at least one discrete control variable of the drive system for a portion of the predefined route. The second optimization method (OPT2) can comprise or be dynamic programming (optimization, DP) using the Pontryagin Maximum Principle (PMP).

[0041] Using the determined auxiliary optimization data 105h, for example, target values ​​for the first state parameter and the second state parameter of the second model (M2) can be determined for the portion of the predefined route. Thus, the second optimization method (OPT2) can be used efficiently to completely solve the problem, whereby, for example, integer and non-integer parameters (e.g., control parameters and / or state parameters) can be optimized simultaneously. The integer parameters can be used, for example, to model discrete variables, and the non-integer parameters can be used, for example, to model continuous variables.

[0042] According to various embodiments, the default control variable data 105s can be output by the at least one processor 102 or the control device 100 (e.g., to a control system) for operating the drive system of a hybrid electric vehicle based on the output default control variable data 105s.

[0043] According to various embodiments, actuating data 107 can be calculated based on the actual control variable data 105i and the specified control variable data 105s, for example by means of the at least one processor 102 of the control device 100 or by means of another computer system 112, wherein corresponding actuating devices 114 implement the actuating data 107 in such a way that a deviation of the actual control variable data 105i from the specified control variable data 105s is reduced.Actuating devices 114 may, for example, comprise at least one actuator for braking the hybrid electric vehicle, at least one actuator for changing the speed of the internal combustion engine of the hybrid electric vehicle, at least one actuator for changing the speed of the electric motor of the hybrid electric vehicle, at least one controller for changing a state of a start / stop system of the hybrid electric vehicle, and at least one actuator for changing the state of the transmission of the hybrid electric vehicle.

[0044] Fig. 2 illustrates components of a drive system 202 of a hybrid electric vehicle 200 that may be considered in the first model (M1) and / or second model (M2), according to various embodiments.

[0045] The drive system 202 of the hybrid electric vehicle 200 comprises, for example, an internal combustion engine 204 and an associated fuel tank 214. By means of the internal combustion engine 204, a drive torque (a torque) can be generated, wherein the efficiency is dependent on the speed, as is shown, for example, in Fig. 3A is illustrated using an example characteristic field 300v.

[0046] The internal combustion engine 204 can be started and stopped using a start / stop system 224. In particular, when the hybrid electric vehicle 200 does not require torque provided by the internal combustion engine 204, the start / stop system 224 and the internal combustion engine 204 can be configured to deactivate the internal combustion engine 204. A control of a clutch 208 of the drive system 202 can be coupled to the start / stop system 224, so that, for example, the internal combustion engine 204 is decoupled when it is stopped.

[0047] The drive system 202 of the hybrid electric vehicle 200 further comprises, for example, at least one electric motor 206 and at least one associated energy storage device 216. By means of the electric motor 206, a drive torque (a torque) can be generated, wherein the efficiency is also dependent on the speed, as is the case, for example, in Fig. 3B is illustrated using an exemplary characteristic curve field 300e.

[0048] Optionally, the hybrid electric vehicle 200 may include a converter 226, e.g., an alternating current (AC) / direct current (DC) converter for operating the electric motor 206 using the corresponding energy from the energy storage device 216.

[0049] The motor shafts of the electric motor 206 and the combustion engine 204 can be selectively coupled or uncoupled from each other by means of the clutch 208.

[0050] The drive system 202 of the hybrid electric vehicle 200 further comprises, for example, a braking system 210 by means of which a braking torque can be applied to the front and / or rear wheels (for the sake of clarity, the braking system 210 is shown only schematically for the front wheels).

[0051] According to various embodiments, the drive system 202 of the hybrid electric vehicle 200 may include a transmission 230 (e.g., an automatic transmission). The transmission 230 may, for example, be selectively operated in multiple gear ratios (e.g., 5, 6, 7, 9, or more than 9). On the input side, torque may be provided to the transmission from the electric motor 206 and / or the internal combustion engine 204.

[0052] The hybrid electric vehicle 200 may, for example, include the control device 100 described herein for controlling / regulating the operation of the components of the drive system 202. Thus, the hybrid electric vehicle 200 may be driven at least partially autonomously or fully autonomously.

[0053] According to various embodiments, the gear selection of the transmission 230, the start / stop control, the torque of the internal combustion engine 204, the torque of the electric motor 206, and the torque of the braking system 210 can be control variables of the control device 100. In the second model (M2), for example, the gear selection of the transmission 230 and the start / stop control can each be described using integer parameters, and the torques, for example, can be described using continuous (e.g., fractionally rational) parameters. In the first model, for example, the gear selection of the transmission 230 and the start / stop control can be disregarded.

[0054] System states can be, for example, the energy stored, the kinetic energy (or velocity), and time. In both models (M1, M2), the energy stored, the kinetic energy (or velocity), and time can be described using continuous (e.g., fractionally rational) state parameters.

[0055] Fig. Figure 4 illustrates a calculation of a profile of the charge / discharge state of the energy storage device 216 (e.g., the first state parameter of the first model M1) based on the first model (M1) and SQP, according to various embodiments. The profile of the charge / discharge state of the energy storage device 216 determined using the first model (M1) and SQP can be used to determine a target energy for respective sections and to optimize these sections using the second model (M2) and PMP-DP.

[0056] The target quantity can be a target state (e.g. the target energy) related to a section of the route for travel time (the second state parameter) and charge / discharge state of the energy storage (the first state parameter).

[0057] According to various embodiments, the optimal control problem can be mathematically formulated as follows: minimize ∫sose(Ffuel costs(u→,x→,s)+FBattery energy costs(u→,x→,s))ds, where s represents the route with S0 as the starting point and S e as endpoint, where u→ represents the control variables, and where x→ represents the system states. Instead of fuel costs and energy storage energy costs, other quantities can also be minimized, e.g., fuel consumption and energy consumption. In the mathematical formulas described here, the vectors u→ and x→ can also be written as u and x.

[0058] In the following, details of the modeling of the drive system 202 of the hybrid electric vehicle 200 and the control / regulation aspects of the control device 100 or corresponding control methods are described by way of example.

[0059] According to various embodiments, the state of the combustion engine or the start / stop system, σ, can be a discrete system state, e.g., either on or off. The corresponding control variable u σcontrols σ and simultaneously the state of clutch 208. The state of clutch 208 is thus also discrete; accordingly, clutch 208 is considered either closed or open. When the transmission is in neutral, the drive wheels of hybrid-electric vehicle 200 can move independently of transmission 230 and drive units 204, 206. This property is utilized in so-called coasting. If clutch 208 is closed, combustion engine 204 and electric motor 206 have, for example, the same rotational speed or angular velocity, or at least a fixed ratio of the respective rotational speeds or angular velocities ω(v, g).

[0060] The angular velocity ω(v, g) depends on the longitudinal vehicle speed v, the dynamic wheel radius r, and the transmission ratio (g). The transmission ratio (g) is defined, for example, by the respective gear, which represents another discrete system state. The corresponding control variable u g controls g.

[0061] Both discrete control variables u g and u σ can be used in the vector and→ and the corresponding states g and σ in the vector xd→.

[0062] For the torques, M E , the combustion engine 204 and, M M , of the electric motor 206, for example, corresponding forces F E = M E / r and F M = M M / r which together with the force of the conventional brake F B represent continuous control variables and in the vector uc→ The resulting control variable vector is u→=(uc→, ud→).

[0063] Furthermore, according to various embodiments, the efficiency of the transmission 230 (η g ) must be taken into account, as well as the driving resistances (e.g. caused by inertia, air resistance, rolling resistance, road gradient, and braking force).

[0064] Overall, the balance of forces on the wheel, F W (v), as follows: FW(v)=(FM+FEσ−FT,d(u))γ(g)∀s∈[s0, sf], FW(v)=mvdvds+cav2+cα+FB, where F T,d (u) describes the dissipative force of the transmission. Furthermore, m is the vehicle mass, c a a constant coefficient of air resistance and c α a factor that depends on the gradient of the track.

[0065] By introducing the kinetic energy E VNonlinearities can be eliminated and the state differential equation of the kinetic energy, f V , is as follows: EV'(u,EV,g)=fv(u,EV,σ,g)=(FM+FEσ−FT,d(u))γ(g)−FB−2caEV / m=−cα∈m[amin, amax],

[0066] The change in kinetic energy can be caused by a minimal acceleration a min and a maximum acceleration a max be limited.

[0067] By formulating it in the path domain, the travel time t results as a further system state, whose state differential equation can be formulated as follows: ft(EV)=t'(EV)=1 / v=m2EV.

[0068] The energy storage energy E S can also be a system state and with the other continuous system states a vector xc→ be summarized to xc→=(ES, EV, t). This vector can be combined with the vector of discrete system states xd→ in a state vector x→=(xc→, xd→) be summarized.

[0069] The state vector x→ can be used in the state differential equation of energy storage energy as follows: fS(FM, x)=ES'(FM, x)=−PS(FM, x) / 2EV / m, where the chemical (ie internal) energy storage power P S is defined as negative when the energy storage device is being charged. For accumulators, the energy storage power can also be referred to as battery power.

[0070] The energy storage model can comprise a series circuit of an ideal voltage source and an ohmic resistance. This means, for example, that the chemical energy storage capacity, P S , which are determined, for example, by the constant power limits P Smin and P Smaxmay be limited, the sum of the dissipative energy storage capacity P S,d and the electrical power.

[0071] The energy storage performance can be described as follows: PS(FM, x)=PS,d(FM, x)+PM(FM, x)+PM,d(FM, x)∈[PSmin, PSmax].

[0072] The clamping power results, for example, from the mechanical power, M M , of the electric motor as follows: PM(FM, x)=MM(FM)ω(x)=FM2EV / mγ(g), and its dissipative performance PM,d(FM, x→), given, for example, as a static characteristic map (see, for example, Fig. 3B).

[0073] The losses of the energy storage device can be determined by the electrical power, the ohmic resistance R(FM, x→) and the open circuit voltage U(E S ), for example as follows: PS,d(FM, x)=(U(ES)−U2(ES)−4R(FM, x)(PM(FM, x)+PM,d(FM, x)))24R(FM, x).

[0074] The electrical resistance of the energy storage device (e.g. the battery resistance) can, for example, take two discrete values, depending on whether the energy storage device is being charged or discharged, e.g. as follows: R(FM, x)={R+,for PS(FM, x)≤0R−,for PS(FM, x)>0.

[0075] According to various embodiments, for the system states (x→), Energy storage energy (E S ), kinetic energy (E V ), travel time (t), start / stop system (σ), and gear (g), taking into account the control variables (u) of the drive train of the hybrid electric vehicle under consideration, namely the power of the combustion engine (F E ), power of the electric motor (F M ), brake force (F E ), control of the start / stop system (u σ ) and gear control (u g ).

[0076] The quality measure can be formulated, for example, as follows: J(u, x)=κS(ES0−ES(Sf))+∫S0Sf(κE(FE+FE,d(FE, x))σγ(g)+|uσ|βσ+|ug|βg+a2 / 2EV / mβa)ds.

[0077] This is an example of a so-called Bolza quality measure, which consists of a cost component for the target state and an integral cost component. The terms β σ and β g in the quality measure represent the energy loss due to the starting process of the combustion engine and friction losses in the clutch when shifting gears. The term β a adapts the driving behavior and prevents or allows excessive acceleration of the hybrid electric vehicle. It is understood that other formulations of the model and the performance measure are also possible, leading to the same or similar results.

[0078] Using the quality measure, the optimal control problem can be summarized as follows: minimize J(u, x) under the conditions xc'=fc(u, x) xd+=xd+ud x(s0)=x0, t(sf)<=tf EV'∈m[amin, amax] ES'∈−[PSmin, PSmax] / 2EV / m x∈[xmin, xmax] uc∈[uc,min(x), uc,max(x)] uσ, ug∈{−1, 0, 1}.

[0079] The function f combines c (u,x) = (f S , f V , f t ) the state differential equations of all continuous system states.

[0080] To solve the formulated optimal control problem with the discrete solution method, for example, the path coordinate can be discretized and the continuous system states x C be quantized.

[0081] This results in a discrete optimal control problem as follows: Minimize J(u(k), x(k), k)=κS(ES0−ES(Nk))+∑k=1Nk(κE(FE(k)+FE,d(FE(k), x(k))σ(k)γ(g(k))+a2(k) / 2EV(k) / mβa)Δs(k)+∑k=1Nk(|uσ(k)|βσ+|ug(k)|βg) under the conditions xc(k+1)=xc(k)+fc(u, x)Δs xd(k+1)=xd(k)+ud x(0)=x0, t(Nk)<=tf fV(k)∈m[min, max] fS(k)∈−[PSmin, PSmax] / 2EV(k) / m x∈[xmin, xx] uc∈ [uc,min(x), ucmax(x)] ug, uσ∈{−1, 0, 1}, each is a discrete Point of Coordinate.

[0082] The discrete optimal control problem can, for example, be calculated using a numerical algorithm (e.g., DP). This DP algorithm starts with k = 1 and calculates, for each state x(k + 1), the sum of the state transition costs based on the possible predecessor states at k and the cumulative costs from the start to the respective predecessor state at k. For example, only the predecessor states that satisfy the constraints can be considered in order to keep the computation time small. The state transition costs depend, for example, on the control variables, which are determined by inverting the state differential equations. This means that u is inferred from the difference x(k+1)-x(k). For example, it is assumed that the control variables are in the interval [s k , s k+1] does not change, which corresponds to a sampling of the instantaneous value. Once all possible total costs for a state x(k) have been calculated, the lowest value is stored in the cost tensor and the index of the corresponding predecessor. This algorithm can be implemented, for example, with nested loops that examine all possible state transitions to calculate the state change costs. For this purpose, for example, all possible E V (k), E S (k), σ(k) and g(k) are iterated.

[0083] According to various embodiments, in the second stage of the MPC, the optimization is based on a combined solution of the DP with Pontrjagin's maximum principle.

[0084] The following provides an example of the Pontryagin Maximum Principle (PMP). Since the optimum is usually defined as a minimum in optimal control, the maximum principle can be transformed into a minimum principle by changing its sign. It can be intuitively understood as the extremum principle.

[0085] According to various embodiments, a corresponding Hamiltonian function may be defined as follows: H(x,ψ,u,s)=F(x,u,s)+ψTf(x,u,s) where F(x, u, t) is the quality measure, which is merged with the state differential equation f(x, u, t) via Lagrange multipliers (ψ). The vector of Lagrange multipliers ψ(s) = (ψ S (s), ψ V (s), ψ t (s)) contains the costates ψ S (s), ψ V (s) and ψ t (s) for the respective system states E S , E V and t.

[0086] To apply the PMP to the optimal control problem, the discrete states g and σ can be neglected and only their control variables retained in the problem description. An adapted quality measure can be written as follows: L(x,u,s)=κSFS(s)+κE(FE(s)+FE,d(FE,x,s))σ(s)γ(g(s))+|uσ|βσ+|ug|βg+a2(s) / 2EV(s) / mβa.

[0087] The optimal control problem can therefore be formulated as follows: minimize H(x,ψ,u,s) under the conditions ψ'(s)=−∂H(x,ψ,u,s)∂xc(s) xd+=xd+ud x(s0)=x0, t(sf)<=tf EV'(s)∈m[amin, amax] ES'(s)∈−[PSmin, PSmax] / 2EV(s) / m x(s)∈[xmin, xmax] uc(s)∈[uc,min(x), ucmax(x)] ug, uσ∈{−1, 0, 1}.

[0088] In order to determine the piecewise constant course of the respective costate, the problem can be solved with a recursive algorithm that divides the horizon into segments at the points where the violation of the energy storage boundaries with a constant ψ S is largest in the segment. This results, for example, in a piecewise constant costate, which can represent the globally optimal solution.

[0089] In the following, an exemplary algorithm is described in which the energy storage energy and the travel time can be removed from the optimal control problem, for example by changing values ​​for the corresponding costates ψ S and ψ t This specification of the costates ψ S and ψ t can be clearly realized by the first stage of MPC described here based on an SQP-based algorithm.

[0090] According to various embodiments, a higher-level algorithm may solve the costates by solving a 2-point boundary value problem with horizon segmentation. This algorithm is described in Fig. 5 and Fig. 6 is shown as an example, where the optimal control problem 500p is solved.

[0091] For example, the optimal control problem 500p can be formulated as follows: Minimize ∑k=k0kf(L(u,x,k)+ψS fS(u,x)+ψt ft(u,x,k))Δs under the conditions EV(k+1)=EV(k)+fV(k)Δs xd(k+1)=xd(k)+ud(k) x(k0)=x0, t(Nk)≈tf ES(Nk)≈ES,f fV(k)∈m[amin, amax] fS(k)∈−[PSmin, PSmax] / 2EV(k) / m x⌣∈[x⌣min, x⌣max] uc(k)∈[uc,min(x), ucmax(x)] ug, uσ∈{−1, 0, 1}, where the solution is iterative with: x⌣=(EV,xd).

[0092] The target conditions for travel time and energy storage energy are specified, since each value for the co-states causes exactly one target value for the respective state.

[0093] Fig. 7 illustrates, by way of example, route data 103 with a predefined gradient and a predefined speed limit.

[0094] Fig. 8 illustrates an example diagram 800 which represents the solution quality relative to the computational effort of the PMP-DP for a predefined route and a corresponding discretization Δs of the path.

[0095] With the goal of further reducing the computational effort required to solve the optimal control problem, the optimal control problem is formulated below as a quadratic problem (QP). By solving the QP in an SQP scheme, the linearization error caused by approximations can be minimized.

[0096] For the SQP, the optimal control problem can be formulated as follows: Minimize f0(x˜) under the conditions fi(x˜)≤0 hj(x˜)=0 x˜∈C. where the states x̃ must lie in the convex set C.

[0097] In a QP, for example, integer states cannot be considered, which is why the start / stop system and the transmission gear selection are removed from the mixed-integer problem. For this purpose, the combustion engine is assumed to be permanently switched on (σ = 1) and a gear change is assumed to be instantaneous without a penalty term g.

[0098] These assumptions allow, for example, the gear to be selected in advance so that fuel consumption is minimized during pure combustion engine operation (i.e., neglecting the influence of the electric motor). For this purpose, instead of the consumption map of the combustion engine, a corresponding map for the combustion engine-transmission unit, which determines the driving force F̃ E,W generated on the wheels. Fig. Figure 9 illustrates, by way of example, a characteristic map 900 (e.g., a fuel consumption map) for the combustion engine-transmission unit for fuel-optimized gear selection. The characteristic map 900 was approximated, for example, by an analytical function.

[0099] The quality measure can then be formulated as follows: J(t˜,E˜V,F˜E,W,E˜S)=κS(E˜S0−E˜S(sf))+κE∫s0sfP˜Evds=κS(E˜ S0−E˜S(sf))+κEζ0∫s0sf1vds+κEζ1∫s0sfds+κEm2∫s0sf(ζ2E˜V+ζ3F ˜E,W+ζ4F˜E,W2+ζ5F˜V2)ds=κS(E˜S0−E˜S(sf))+κEζ0(t˜(sf)−t˜0 )+κEζ1(sf−s0)+κEm2∫s0sf(ζ2E˜V+ζ3F˜E,W+ζ4F˜E,W2+ζ5F˜V2)ds, where the costs can be expressed as a quadratic convex function.

[0100] The maximum power delivered by the engine-transmission unit to the wheels can be a highly nonlinear and piecewise discrete function. This can be accounted for, for example, by performing a piecewise nonlinear inner approximation, e.g., in the form: F˜E,Wmax=min(ζW,1+ζW,2E˜VζW,3,ζW,4+ζW,5 / E˜V)

[0101] The electric motor can also be combined with the transmission to form a single unit. The upper limit of the power of this EM transmission unit F˜M,Wmax=min(ζM,1,ζM,2+ζM,3 / E˜V,) and the lower limit F˜M,Wmin=max(ζM,4,ζM,5+ζM,6 / E˜V) can, for example, be approximated with two sections.

[0102] According to the forces for the combustion engine-gearbox unit and the electric motor-gearbox unit, an energy storage force F̃ S = P S / v, which can be understood as the longitudinal force acting on the vehicle when drive power is taken from the energy storage device.

[0103] Consequently, the electrical power balance can be calculated by modelling the energy storage with a constant efficiency for charging η S,chr and unloading η S,dis can be described as a balance of forces, e.g. as follows: F˜S≥max(F˜M,W / (ηM,motηgηS,dis),F˜M,WηM,genηgηS,chr).

[0104] In addition to the energy storage force, for example, another control signal, z, can be used to form the state differential equation as follows: t˜'=zz≥1 / 2E˜V / m.

[0105] The nonlinear control signal, z, can be approximated, for example, using two affine parts: z˜≥max(ζt,1+ζt,2E˜V,ζt,3+ζt,4E˜V).

[0106] The resulting optimal control problem can thus have linear state differential equations, e.g. as follows: E˜S'=−F˜S E˜V'=F˜M,W+F˜E,W−F˜B−2caE˜V / m−cα t˜'=z˜.

[0107] By linearization around a reference trajectory Ẽ V a convex inner approximation can be formulated as follows: 1 / E˜V≈flin(E˜V, E˜¯V).

[0108] This means that the model is completely approximated by quadratic functions, for example, which makes it possible to develop an SQP.

[0109] For this purpose, the state vector x̃ = (Ẽ S , Ẽ V , t̃) and the control vector ũ = (F̃ E,W , FM,W , F B , F S , z̃) is defined as described above, and s is discretized using instantaneous sampling, analogously to the DP and PMP-DP. Consequently, the QP solved in each SQP iteration can be described as follows: minimize J(x˜(k˜),u˜(k˜))+Q(x˜(k˜),u˜(k˜)) under the conditions x˜(k˜+1)=A(k˜)x˜(k˜)+B(k˜)u˜(k˜)+w(k˜) C(k˜)x˜(k˜)+D(k˜)u˜(k˜)≤b(k˜) x˜(0)=x˜0, t˜(Nk)<=t˜f x˜(k˜)∈[x˜min(k˜), x˜max(k˜)], u˜(k˜)∈[u˜min(k˜), u˜max(k˜)].

[0110] The matrices A, B, C, D and the vectors w, b, x̃ min , x max as described above. These matrices and vectors depend, for example, on k̃, since the slope and the limits of Ẽ Vas well as the approximation of ž depend on the path. The interval between two samples does not have to be constant and can be adjusted depending on the reference speed; for example, a coarser sampling can be used for higher speeds.

[0111] According to various embodiments, the SQP and the PMP-DP can be implemented as a pure control method. To incorporate model deviations, a simulation model based on the previously described control model and supplemented by additional dynamics can be used. These additional dynamics, for example, cause a deviation between the control and simulation models, which is why a pure control that calculates the states using the control model causes an error in the prediction of the states resulting in the simulation model.

[0112] Due to the discrepancy between reality and the control model, it may be helpful to update the specifications for the control variables at a comparatively high frequency (e.g., more than 0.5 Hz). For this purpose, the two-stage MPC described herein can be used, which can be used as an energy management strategy with efficient longitudinal control online in the hybrid electric vehicle 200. The division of tasks between the upper (high-level MPC) and the lower (low-level MPC) stage is Fig. 10 is illustrated schematically.

[0113] The route data 103 (e.g., trajectories for gradient, minimum and maximum kinetic energy) as well as the current system states can be used as input data. The first (upper) MPC stage can, for example, be configured to determine the general trend of the energy storage energy and the travel time trajectory until the end of the route. The second (lower) MPC stage can, for example, be configured to determine targets for a shorter forecast horizon based on the trajectories determined by the first stage, thus allowing the control variables, including the discrete decisions, to be determined with a higher level of detail.

[0114] In the second (low-level) MPC stage, for example, the horizon can be divided into sections that are short enough to be calculated, for example, using PMP-DP in a predefined time (e.g., in less than 1 s). The problem can, for example, be solved in sections, i.e., by continuously optimizing short look-ahead horizons, so it is helpful, for example, to specify values ​​for the desired target time and target energy storage energy for the respective section. The specification of target conditions is carried out, for example, by the first (high-level) MPC stage, which calculates the energy management for a horizon that extends to the end of the route. The optimized results of the first MPC stage, for example, are used to determine the target values ​​for the energy storage energy Ẽ S,f and the target time t̃ ffor the limited look-ahead horizon of the second MPC stage. According to various embodiments, both the first and second MPC stages can take into account the current system states, which are fed back as starting states. The update rate of the current system states can be adapted to the duration of the iterations. Additional input signals can be, for example, the trajectories of the gradient (α) and the speed limits (E Vmin and E Vmax ) for the current route.

[0115] According to various embodiments, comparatively long look-ahead horizons can be calculated in the first MPC stage, which can be particularly desirable for the sensible use of high-capacity energy storage devices such as in plug-in hybrid vehicles, so that the horizon extends to the end of the route. Such an MPC can incur a high computational effort. Therefore, according to various embodiments, an SQP scheme can be used for this purpose. The SQP scheme can be implemented as a real-time iteration. The real-time iteration can, for example, utilize the iterative nature of the MPC by distributing the SQP iterations across the MPC updates. According to various embodiments, only one QP can be solved with each MPC update, which reduces the computational effort and enables the use of a conventional QP solver.

[0116] The underlying second MPC stage can be used to calculate shorter look-ahead horizons than those achieved with the first MPC stage. This allows the use of PMP-DP to calculate the actual control, which promises a significantly higher quality of the solution, since the discrete states and control variables are also optimized and can be calculated using a non-convex model.

[0117] To follow the objectives of the first MPC stage (MPC-1), the algorithm of the second MPC stage (MPC-2) can, for example, be set up to solve a 2-point boundary value problem to find ψ S and ψ t To do this, the bisection can be used as a single-shooting procedure. The single-shooting procedure is carried out, for example, while the hybrid electric vehicle 200 is moving. For this purpose, at each iteration of the MPC, a fixed value for ψ S and ψ tshot, allowing a prediction of the states at the end of the short horizon of the second MPC stage (MPC-2) to be calculated. For the next shot of the single-shot procedure, for example, the deviation of this prediction from the state specifications of the first MPC stage (MPC-1) can be stored and used the next time the second MPC stage (MPC-2) is called to either calculate ψ S or ψ t using bisection. For example, if ψ t was adjusted in the last iteration, a new value for ψ S be determined, and vice versa.

[0118] Fig. 11 shows the second MPC stage (MPC-2) of the control device 100 in a schematic representation, according to various embodiments.

[0119] For example, the second MPC stage (MPC-2) can include a nonlinear proportional controller (NPR), e.g., to prevent potential violations of the energy storage limits. Using the nonlinear proportional controller (NPR), a new value for the co-state W can be determined. S,lim which can guarantee a feasible control in the PMP-DP, for example. For this purpose, the non-linear proportional controller (NPR) receives the value for ψ S from the single-shot block (EVB), as well as the current charge / discharge state and the energy storage limits. For example, if the charge / discharge state of the energy storage is close to its lower limit, W S,lim and thus forces a charging of the energy storage, and vice versa. This controller behavior can be achieved by implementing a tangent function 1200, as in Fig. 12 shown, can be realized.

[0120] In the following, various examples are described which relate to what has been described above and what is shown as examples in the figures.

[0121] Example 1 is a method for controlling a drive system of a hybrid electric vehicle, the method comprising: determining auxiliary optimization data based on a first model by means of a first optimization method, wherein the first model comprises: a first continuous state parameter representing a charge / discharge state of an energy storage device of the drive system, a second continuous state parameter representing a travel time, a first control parameter set representing continuous control variables of the drive system, and wherein the auxiliary optimization data represent a profile of the first state parameter and the second state parameter associated with a predefined travel route,wherein the first optimization method comprises sequential quadratic programming for determining the auxiliary optimization data based on route data and actual control variable data, taking into account the first and second continuous state parameters and the first control parameter set; determining default control variable data based on a second model by means of a second optimization method, wherein the second model comprises: a first state parameter representing the charge / discharge state of the energy storage device, a second state parameter representing the travel time, a first control parameter set representing the continuous control variables of the drive system, and a second control parameter set representing discrete control variables of the drive system,wherein the specified control variable data represents a specified curve of the continuous control variables and the discrete control variables of the drive system for a portion of the predefined route, wherein the second optimization method comprises dynamic programming using Pontrjagin's maximum principle, and wherein, using the determined auxiliary optimization data, target values ​​for the first state parameter and the second state parameter of the second model are determined for the portion of the predefined route, for determining the specified control variable data based on the route data, the actual control variable data, and the determined auxiliary optimization data, taking into account the first and second continuous state parameters and the first and second control parameter sets; and outputting the specified control variable data for operating the drive system based on the outputted specified control variable data.

[0122] In Example 2, the method according to Example 1 may further comprise that the route data represent at least one predefined gradient profile and one predefined speed limit profile associated with the predefined route.

[0123] In example 3, the method according to example 1 or 2 may further comprise that the actual control variable data represent a respective actual state of the control variables of the first and second control parameter sets.

[0124] In Example 4, the method according to any one of Examples 1 to 3 may further comprise that at least one control parameter of the second control parameter set is not optimized in the first optimization method.

[0125] In example 5, the method according to any one of examples 1 to 4 may further comprise that in the first optimization method a horizon up to the end of the journey is taken into account, preferably up to a horizon of more than 50 km.

[0126] In Example 6, the method according to any one of Examples 1 to 5 may further comprise that a horizon of less than 10 km is taken into account in the second optimization method, preferably of less than 5 km.

[0127] In Example 7, the method according to any one of Examples 1 to 6 may further comprise that the first optimization method has an update interval in a range of 10 s to 300 s.

[0128] In Example 8, the method according to any one of Examples 1 to 7 may further comprise that the first optimization method comprises a discretization of the horizon in a range of 1 m to 100 m, preferably less than 50 m.

[0129] In Example 9, the method according to any one of Examples 1 to 8 may further comprise that the second optimization method has an update interval in a range of 0.01 s to 5 s, preferably less than 1 s.

[0130] In Example 10, the method according to any one of Examples 1 to 9 may further comprise that the second optimization method comprises a discretization of the horizon in a range of 10 m to 100 m, preferably less than 100 m.

[0131] In Example 11, the method according to any one of Examples 1 to 10 may further comprise that the second optimization method comprises a quantization of the acceleration in a range of 0.1 m / s 2 up to 1 m / s 2 , preferably less than 1 m / s 2 .

[0132] In Example 12, the method according to any one of Examples 1 to 11 may further comprise that the first control parameter set comprises one or more control parameters that represent one or more of the following continuous control variables: torque or force of an internal combustion engine of the drive system; torque or force of an electric motor of the drive system; torque or force of a braking system of the drive system.

[0133] In example 13, the method according to any one of examples 1 to 12 may further comprise that the second control parameter set comprises one or more control parameters that represent one or more of the following discrete control variables: state of a start / stop system of an internal combustion engine of the drive system; state of a transmission of the drive system.

[0134] In Example 14, the method according to any one of Examples 1 to 13 may further comprise that the control parameters of the second control parameter set are integers and wherein the control parameters of the first control parameter set are non-integers, preferably fractionally rational.

[0135] In Example 15, the method according to any one of Examples 1 to 14 may further comprise that the first and second optimization methods are performed by means of a computing system, wherein the computing system has a computing power of less than 100 TFLOPS.

[0136] In Example 16, the method according to any one of Examples 1 to 15 may further comprise that the second optimization method comprises a non-linear proportional controller for taking into account energy storage limits of the energy storage device.

[0137] In Example 17, the method according to any one of Examples 1 to 16 may further comprise that the second optimization method is configured to solve a 2-point boundary value problem, wherein the auxiliary optimization data represents objectives for the 2-point boundary value problem.

[0138] In example 18, the method according to any one of examples 1 to 17 may further comprise that the auxiliary optimization data represent target specifications for the charge / discharge state of the energy storage device and the travel time relative to a section of the predefined travel route taken into account in the second optimization method.

[0139] Example 19 is a control device (100) for controlling a drive system (202) of a hybrid electric vehicle (200), wherein the control device (100) is configured to: receive route data (103), wherein the route data (103) represent a predefined gradient profile and a predefined speed limit profile associated with a predefined route; receive actual control variable data (105i), wherein the actual control variable data (105i) represent an actual state of at least one continuous control variable of the drive system (202) and at least one discrete control variable of the drive system (202); determine auxiliary optimization data (105h) based on a first model (M1) by means of a first optimization method (OPT1), wherein the first model (M1) comprises: a first state parameter representing a charge / discharge state of an energy storage device (216) of the drive system (202),a second state parameter representing a travel time (t), and at least one first control parameter representing the at least one continuous control variable of the drive system (202), and wherein the auxiliary optimization data (105h) represent a profile of the first state parameter and the second state parameter, wherein the first optimization method (OPT1) comprises sequential quadratic programming (SQP); based on a second model (M2), by means of a second optimization method (OPT2), to determine predefined control variable data (105s), wherein the second model (M2) comprises: a first state parameter representing the charge / discharge state of the energy storage device (216), a second state parameter representing the travel time (t), at least one first control parameter representing the at least one continuous control variable of the drive system (202),and at least one second control parameter representing the at least one discrete control variable of the drive system (202), and wherein the specified control variable data (105s) represent a specified curve of the at least one first control variable and the at least one second control variable of the drive system (202) for a respective section of the predefined route, wherein the second optimization method (OPT2) comprises dynamic programming (DP) using Pontrjagin's maximum principle (PMP), and wherein target values ​​for the first state parameter and the second state parameter of the second model (M2) for the respective section are determined using the determined auxiliary optimization data (105h); and outputting the specified control variable data (105s) for operating the drive system (202) based on the outputted specified control variable data (105s).

[0140] Example 20 is a method for controlling a drive system of a hybrid electric vehicle, the method comprising: determining auxiliary optimization data based on a first model by means of a first optimization method, wherein the first optimization method comprises sequential quadratic programming for determining the auxiliary optimization data based on route data and actual control variable data taking into account parameters which represent the drive system of the hybrid electric vehicle, and wherein the auxiliary optimization data represent a profile of a subset of the parameters assigned to a predefined route;Determining specified control variable data based on a second model by means of a second optimization method, wherein the specified control variable data represent a specified curve of control variables of the drive system for a section of the predefined route, wherein the second optimization method comprises dynamic programming using Pontrjagin's maximum principle, and wherein, based on the determined auxiliary optimization data, target values ​​for one or more parameters of the second model for the section are determined in order to determine the specified control variable data based on the route data, the actual control variable data, taking into account parameters that represent the drive system of the hybrid electric vehicle; and outputting the specified control variable data for operating the drive system based on the outputted specified control variable data.

[0141] Example 21 is a non-transitory storage medium comprising instructions that, executed by at least one processor, carry out the method for controlling a drive system of a hybrid electric vehicle according to any one of Examples 1 to 18 or 20.

[0142] It is understood that functions, algorithms, etc. described herein with reference to a method can also be implemented in the same way in a control device and vice versa.

Claims

[1] Method for controlling a drive system (202) of a hybrid electric vehicle (200), the method comprising: ◯ Determining auxiliary optimization data (105h) based on a first model (M1) by means of a first optimization method (OPT1), wherein the first model (M1) comprises: a first continuous state parameter representing a charge / discharge state of an energy storage device (216) of the drive system (202), a second continuous state parameter representing a travel time (t), a first control parameter set representing continuous control variables of the drive system (202), and wherein the auxiliary optimization data (105h) represent a profile of the first state parameter and the second state parameter associated with a predefined travel route,wherein the first optimization method (OPT1) comprises sequential quadratic programming (SQP) for determining the auxiliary optimization data (105h) based on route data (103) and actual control variable data (105i) taking into account the first and second continuous state parameters and the first control parameter set;, ◯ Determining predefined control variable data (105s) based on a second model (M2) by means of a second optimization method (OPT2), wherein the second model (M2) comprises: a first state parameter representing the charge / discharge state of the energy storage device (216), a second state parameter representing the travel time (t), a first control parameter set representing the continuous control variables of the drive system (202), and a second control parameter set representing discrete control variables of the drive system (202), wherein the predefined control variable data (105s) represent a predefined course of the continuous control variables and the discrete control variables of the drive system (202) for a portion of the predefined travel route, wherein the second optimization method (OPT2) comprises dynamic programming (DP) using Pontrjagin's maximum principle (PMP),and wherein, by means of the determined auxiliary optimization data (105h), target values ​​for the first state parameter and the second state parameter of the second model (M2) are determined for the part of the predefined route in order to determine the specified control variable data (105s) based on the route data (103), the actual control variable data (105i) and the determined auxiliary optimization data (105h), taking into account the first and second continuous state parameters and the first and second control parameter sets, ◯ wherein at least one control parameter of the second control parameter set is not optimized in the first optimization method (OPT1), ◯ wherein the auxiliary optimization data (105h) represent target specifications for the charge / discharge state of the energy storage device (216) and the travel time (t) in relation to a section of the predefined route taken into account in the second optimization method (OPT2), and ◯ wherein the second optimization method (OPT2) is arranged to solve a 2-point boundary value problem, wherein the auxiliary optimization data (105h) represent objectives for the 2-point boundary value problem; and o Outputting the default control variable data (105s) for operating the drive system (202) based on the output default control variable data (105s). [2] Method according to claim 1, wherein the route data (103) represent at least one predefined gradient profile and one predefined speed limit profile associated with the predefined route. [3] Method according to claim 1 or 2, wherein the actual control variable data (105i) represent a respective actual state of the control variables of the first and second control parameter sets. [4] Method according to one of claims 1 to 3, wherein the first control parameter set comprises one or more control parameters representing one or more of the following continuous control variables: • torque or power of an internal combustion engine (204) of the drive system (202); • torque or power of an electric motor (206) of the drive system (202); • torque or force of a braking system (210) of the drive system (202); and wherein the second control parameter set comprises one or more control parameters representing one or more of the following discrete control variables: • State of a start / stop system of an internal combustion engine (204) of the drive system (202); • Condition of a gearbox (230) of the drive system (202). [5] Method according to one of claims 1 to 4, wherein the control parameters of the second control parameter set are integers and wherein the control parameters of the first control parameter set are non-integers. [6] The method of claim 5, wherein the control parameters of the first control parameter set are fractionally rational. [7] Control device (100) for controlling a drive system (202) of a hybrid electric vehicle (200), wherein the control device (100) is configured: ◯ to receive route data (103), wherein the route data (103) represent a predefined gradient and a predefined speed limit associated with a predefined route; ◯ to receive actual control variable data (105i), wherein the actual control variable data (105i) represents an actual state of at least one continuous control variable of the drive system (202) and at least one discrete control variable of the drive system (202); ◯ to determine auxiliary optimization data (105h) based on a first model (M1) by means of a first optimization method (OPT1), wherein the first model (M1) comprises: • a first state parameter representing a charge / discharge state of an energy storage device (216) of the drive system (202), • a second state parameter representing a travel time (t), and • at least one first control parameter representing the at least one continuous control variable of the drive system (202), and wherein the auxiliary optimization data (105h) represent a profile of the first state parameter and the second state parameter, wherein the first optimization method (OPT1) comprises sequential quadratic programming (SQP); ◯ to determine target control variable data (105s) based on a second model (M2) by means of a second optimization method (OPT2), wherein the second model (M2) comprises: • a first state parameter representing the charge / discharge state of the energy storage device (216), • a second state parameter, which represents the travel time (t), • at least one first control parameter, which represents the at least one continuous control variable of the drive system (202), and • at least one second control parameter, which represents the at least one discrete control variable of the drive system (202), and wherein the specified control variable data (105s) represent a specified curve of the at least one first control variable and the at least one second control variable of the drive system (202) for a respective section of the predefined route, wherein the second optimization method (OPT2) comprises dynamic programming (DP) using the Pontrjagin maximum principle (PMP), and wherein target values ​​for the first state parameter and the second state parameter of the second model (M2) for the respective section are determined by means of the determined auxiliary optimization data (105h), wherein the at least one second control parameter of the second model (M2) is not optimized in the first optimization method (OPT1), wherein the auxiliary optimization data (105h) represent target specifications for the charge / discharge state of the energy storage device (216) and the travel time (t) in relation to a section of the predefined route taken into account in the second optimization method (OPT2), and wherein the second optimization method (OPT2) is arranged to solve a 2-point boundary value problem, wherein the auxiliary optimization data (105h) represent objectives for the 2-point boundary value problem; and ◯ to output the default control variable data (105s) for operating the drive system (202) based on the output default control variable data (105s). [8] Non-volatile storage medium comprising instructions which, executed by at least one processor, carry out the method for controlling a drive system (202) of a hybrid electric vehicle (200) according to one of claims 1 to 6.

Images