Range extender controller application software layer control strategy simulation test system

Abstract

The invention discloses a simulation test system for a control strategy of an application software layer of a range extender controller. The simulation test system comprises a virtual control system, a whole vehicle controlled object and a whole vehicle longitudinal dynamic model, the virtual control system comprises a virtual driver module, a virtual vehicle control unit, a virtual main drive motor controller, a range extender controller ASW, a virtual engine controller (ECU), a virtual generator controller (GCU), a virtual electric accessory controller and a virtual battery management system (BMS). The virtual whole vehicle controlled object comprises a virtual engine, a virtual generator, a virtual power battery, a virtual high-voltage electric accessory and a virtual chassis system. According to the invention, the test work which can only be carried out after hardware integration is completed in the past can be moved forward, the development period can be shortened, and the development cost is reduced.

Description

Technical Field

[0001] The invention relates to the technical field of control strategy simulation technology for extended-range electric vehicles, and in particular to a control strategy simulation test system for an application software layer of a range extender controller. Background Art

[0002] The software development of the range extender controller of an extended-range electric vehicle generally follows a V-shaped process, and corresponding testing and verification work is carried out at different stages of software development to ensure the quality of software development. Under the AUTOSAR standard, the control logic and implementation algorithms of each specific function in the range extender controller software are encapsulated in the application software layer (ASW), and the defined data interface is opened.

[0003] Correspondingly, the software integration test phase in the V-shaped process mainly verifies whether the control strategy integrated by the ASW meets the functional requirements of the software. The main test forms are Model in the Loop (MIL) and Hardware in the Loop (HIL). The test system or test environment used in the software integration test phase should have good consistency, and try to ensure that the test environment gap between the controller under test in the MIL and HIL stages is small to ensure the comparability of the test results; in addition, the controlled object should have good portability to be suitable for electric vehicle testing of different configurations.

[0004] Existing technical solutions:

[0005] 1. Invention CN105629960A provides an electric vehicle controller model-in-the-loop test system.

[0006] 2. Invention CN105446316A provides a pure electric vehicle whole vehicle controller model in-loop testing device and testing method thereof.

[0007] The above scheme is based on the MIL environment. No standardized data interface is established between the test environment and the controller under test, and there is a certain degree of coupling between the control system and the controlled object, which leads to significant changes in the subsequent HIL environment relative to the MIL environment and poor consistency.

[0008] 1. Invention CN114265393A provides a hybrid vehicle control strategy testing system and method based on hardware-in-the-loop.

[0009] 2. Invention CN113126593A provides a system and method for testing the energy management strategy of an extended-range electric vehicle.

[0010] The above solution is based on the HIL environment and relies on a real range extender controller. It must be tested after the hardware development is completed. It is impossible to discover and eliminate errors in advance, which prolongs the development cycle and increases development costs. Summary of the invention

[0011] In order to solve the above technical problems, the present invention provides a range extender controller application software layer control strategy simulation test system, which can move forward the testing work that could only be carried out after the hardware integration was completed, shorten the development cycle and reduce the development cost.

[0012] In order to achieve the above-mentioned invention object, the present invention proposes a range extender controller application software layer control strategy simulation test system, including: a virtual control system, a whole vehicle controlled object and a whole vehicle longitudinal dynamics model; the virtual control system is respectively connected to the whole vehicle controlled object and the whole vehicle longitudinal dynamics model;

[0013] The virtual control system includes a virtual driver module, a virtual vehicle controller, a virtual main drive motor controller, a range extender controller ASW, a virtual engine controller (ECU), a virtual generator controller (GCU), a virtual electric auxiliary component controller and a virtual battery management system (BMS);

[0014] The controlled objects of the virtual vehicle include: virtual engine, virtual generator, virtual power battery, virtual high-voltage electric auxiliary parts, and virtual chassis system;

[0015] The vehicle longitudinal dynamics model includes a model for calculating the vehicle's driving resistance during driving and a model for calculating the vehicle speed.

[0016] Furthermore, the driving resistance model includes

[0017] (1) Calculation of vehicle air resistance model:

[0018]

[0019] Where: F w (t) is the air resistance of the vehicle at time t, in N; C D is the vehicle drag coefficient; A is the vehicle frontal area, unit: m2; ρ a is the air density, unit is kg / m3; v rel (t) is the relative wind speed at time t, in m / s. The relative wind speed is the vehicle speed v(t) relative to the stationary road surface and the wind speed v relative to the stationary road surface at time t. w The algebraic sum of (t);

[0020] (2) Calculation of vehicle ramp resistance model:

[0021] F ε (t) = mg·sin(α(t))

[0022] Where: F ε (t) is the vehicle slope resistance at time t, in N; m is the vehicle mass, in kg; g is the acceleration of gravity, generally g = 9.8N / kg; α(t) is the road slope angle at time t, in degrees;

[0023] (3) Calculation of vehicle braking force model:

[0024] F brk (t) = n wh ·f(r brk (t))·A

[0025] Where: F brk (t) is the vehicle braking force at time t, in N; n wh is the number of wheels involved in braking; r brk (t) is the brake pedal opening at time t, ranging from 0 to 1; f(r brk (t)) represents the MAP of the brake pressure applied by the brake according to the brake pedal opening, in Pa; A is the brake working area, in m2.

[0026] Beneficial effects of the present invention

[0027] The virtual test environment architecture proposed in the present invention has the following technical features:

[0028] (1) The present invention uses a fully simulated method to build a simulation verification system, which can directly perform functional testing and verification on the control strategy integrated in the application software layer during the range extender controller software development phase, and discover problems in advance. Compared with traditional HIL bench and vehicle testing, the present invention can perform simulation testing on a computer without any additional hardware, and can move forward the testing work that could only be carried out after hardware integration was completed, shortening the development cycle and reducing development costs.

[0029] (2) The present invention defines a standardized signal interface between the range extender controller ASW and the test environment, which is divided into a hard-wire signal interface and a CAN message signal interface according to the signal type of the real range extender controller. Therefore, the test environment can be transplanted into the HIL bench without much modification, and the real range extender controller can be connected to carry out the test to ensure the consistency of the test environment.

[0030] The virtual controlled object of the extended-range electric vehicle proposed by the present invention has the following technical features:

[0031] (1) The whole vehicle controlled object is built based on Cruise M commercial software, which can accurately restore the physical state of key components such as the engine, generator and power battery during vehicle operation, ensuring the simulation accuracy of the controlled object.

[0032] (2) The controlled object model built based on Cruise M supports multiple environments from MIL to SIL to HIL, and can quickly change vehicle configuration and component parameters according to test requirements to ensure the portability of the controlled object.

[0033] The vehicle longitudinal dynamics model provided by the present invention has the following technical features:

[0034] (1) The neural network method is used to predict rolling resistance. It does not involve empirical parameters such as friction coefficient, rolling resistance coefficient, slip coefficient, etc. in the conventional rolling resistance calculation formula, which are difficult to accurately obtain. Instead, it is based on motor output torque, ramp angle, vehicle speed and braking torque, which can be accurately obtained. These physical quantities are used for training and prediction, thereby deriving a more accurate rolling resistance.

[0035] (2) Use a DAE-BiLSTM cascade network. Input the motor torque, ramp angle, vehicle speed and brake torque parameters that are damaged by noise, train a denoising autoencoder to restore the original samples that have not been damaged, and improve the robustness and anti-interference ability of the network; perform feature dimensionality reduction on the input, reduce the input four-dimensional features into one-dimensional abstract features, and reduce the difficulty of training; use BiLSTM to train the input time series signal to capture the dependency between the input samples before and after, and improve the accuracy of prediction;

[0036] (3) The vehicle's driving resistance and speed are calculated by combining physical models with neural networks, which ensures the accuracy of the calculation results while reducing the complexity of the model. Air resistance, ramp resistance, and braking force are calculated using physical formulas, and rolling resistance is predicted using a neural network. The vehicle's current external traction and acceleration are calculated and integrated to obtain the vehicle speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 Provides a simulation test system for the range extender application layer;

[0038] Figure 2 It is the data interface between the test environment and the range extender controller ASW;

[0039] Figure 3 Calculation process of vehicle speed of the controlled object;

[0040] Figure 4 It is the denoising autoencoder network structure;

[0041] Figure 5 It is a bidirectional long short-term memory network structure;

[0042] Figure 6 Predicting network structures for rolling resistance. DETAILED DESCRIPTION

[0043] In order to explain the technical content, achieved objectives and effects of the present invention in detail, the following is an explanation in combination with the implementation modes and the accompanying drawings.

[0044] A range extender controller application software layer control strategy simulation test system. The test system provided consists of three parts: a virtual control system, a vehicle controlled object, and a vehicle longitudinal dynamics model. The system structure is as follows: Figure 1 shown.

[0045] 1. Virtual control system

[0046] The virtual control system proposed in the present invention adopts a forward simulation architecture, such as Figure 1 As shown, it includes a driver module, a virtual vehicle controller, a virtual main drive motor controller, a range extender controller ASW, a virtual engine controller (ECU), a virtual generator controller (GCU), a virtual electric auxiliary component controller and a virtual battery management system (BMS).

[0047] (1) Driver module: simulates real driving behavior, including key switch ignition, electric auxiliary switch, and accelerator / brake pedal control functions. The driver module adjusts the position of the accelerator / brake pedal according to the current vehicle speed and target vehicle speed feedback from the system to control the driving speed.

[0048] (2) Virtual vehicle controller: simulates the control logic related to the range extender, main drive motor, BMS and high-voltage electrical accessories in the real vehicle controller, such as enabling high-voltage relays, enabling various high-voltage equipment, power generation requests, driving requests, etc. At the same time, it receives feedback signals from various control units during the driving process of the vehicle.

[0049] (3) Virtual engine controller: This controller simulates the control strategy of the real ECU based on the speed or torque request, outputs the target throttle opening to the engine controlled object, and provides feedback on the engine's current operating status, actual speed, combustion torque, friction torque, effective output torque, and instantaneous fuel consumption.

[0050] (4) Virtual generator controller: It simulates the control strategy of outputting the target torque to the generator controlled object based on the speed or torque request in the real GCU, and feeds back the current working status, actual speed, effective torque, mechanical power, electrical power, bus voltage and bus current of the generator.

[0051] (5) Virtual main drive motor controller: This controller simulates the control strategy of outputting the target torque to the main drive motor controlled object based on the accelerator pedal opening signal sent by the driver in the real motor controller, and provides feedback on the motor's current working status, actual speed, effective torque, mechanical power, electrical power, bus voltage, and bus current.

[0052] (6) Virtual battery management system: used to simulate the real battery management system's limitation on the maximum charge and discharge power of the power battery at different temperatures and states of charge (SOC), and to provide feedback on the current SOC, battery voltage, charge and discharge status and power, battery pack temperature, internal resistance and other signals.

[0053] (7) Virtual electrical auxiliary component controller: used to simulate the relay control logic of each high-voltage electrical auxiliary component in a real vehicle, receive the switch enable signal from the vehicle controller, and feedback the actual status, power and current signals of each electrical auxiliary component.

[0054] (8) Range extender controller ASW: The data interface between the encapsulated range extender controller ASW model and the test system is as follows: Figure 2 As shown in the figure, according to the signal transmission form of the real controller, the signal interface can be divided into hard-wired signals and CAN message signals. The following signal interactions exist between the test system and the range extender controller ASW:

[0055] 1) The range extender controller ASW receives control signals from the test system, including: start-up instructions, working status instructions, required power generation instructions, etc. from the vehicle controller; it also receives feedback signals from other virtual controllers, including: engine speed and torque signals from the virtual ECU, generator speed, torque and power signals from the virtual GCU, voltage, current, SOC status, etc. from the virtual BMS.

[0056] 2) The test system receives control signals from the range extender controller ASW, including: operating status instructions for controlling the virtual ECU and virtual GCU, required speed, required torque, etc.; it also receives feedback signals from the range extender, such as: operating status of the range extender, actual speed, actual voltage, actual power, etc.

[0057] 2. The object of the vehicle charge

[0058] The whole vehicle controlled object is built based on Cruise M software, including the core components of the extended-range electric vehicle, mainly including the engine, generator, power battery, high-voltage electric auxiliary components (such as electric air compressor, electric steering pump and other high-voltage power-consuming equipment), and transmission system. The whole vehicle controlled object receives the control signal from the virtual control system, calculates the driving status of the whole vehicle, the working status of the range extender, the operating status of the main drive motor, and the charging and discharging status of the battery, and forms feedback signals to input into the virtual control system to achieve closed-loop control.

[0059] (1) The engine model is based on a map (MAP) model. The engine's external characteristics, friction torque, and fuel consumption MAP are input, and the effective output torque and energy consumption of the engine and motor at each speed are derived in the form of interpolation.

[0060] Tq eng=r th ·f c (n eng )-f r (n eng )

[0061] b eng =f(n eng ,Tq eng )

[0062] Where: Tq eng is the effective output torque of the engine, in Nm; r th is the engine throttle opening, ranging from 0 to 1; n eng is the engine speed, in rpm; f c (n eng ) The maximum combustion torque at the corresponding speed is derived from the engine external characteristic curve table, in Nm; f r (n eng ) is the engine friction torque at the corresponding speed derived from the table, in Nm; b eng It is the engine specific fuel consumption obtained by checking MAP at a specific engine speed and output torque, in g / kWh.

[0063] (2) The power battery is used to simulate the charge and discharge characteristics of the power battery in a real vehicle. It is an equivalent circuit model established by connecting multiple first-order RC circuits in series. The number of battery strings, capacity, charge and discharge internal resistance, and open-circuit voltage are input, and the current and voltage response of the power battery at a specific state of charge (SoC) is derived in the form of interpolation.

[0064] (3) The generator and main drive motor are modeled based on the map. The external characteristics and efficiency MAP of the motor at different bus voltages are input, and the output torque of the engine and motor at different voltages and speeds are derived in the form of interpolation.

[0065] (4) The high-voltage electrical accessories are designed as ideal current sources, which change the direction of electrical energy flow by controlling the current size and positive and negative current: when the current is positive, the current source is the electrical appliance, simulating the process of the range extender / power battery supplying power to the electrical accessories; when the current is negative, the current source is the power supply, simulating the process of the vehicle plugged in an external charger feeding power to the range extender / power battery.

[0066] (5) The chassis system is used to receive the driving torque output by the main drive motor and to provide feedback on the driving status of the vehicle based on the values ​​of air resistance, slope resistance, rolling resistance, etc. calculated and derived from the longitudinal dynamics model of the vehicle.

[0067] 3. Vehicle longitudinal dynamics model

[0068] like Figure 3The vehicle longitudinal dynamics model is used to calculate the vehicle's driving resistance and speed during driving. Air resistance, ramp resistance and braking resistance are calculated based on the physical model, and rolling resistance is trained and predicted using a neural network model.

[0069] 3.1 Calculation method of driving resistance parameters based on physical model:

[0070] (1) Calculate the vehicle's air resistance:

[0071]

[0072] Where: F w (t) is the air resistance of the vehicle at time t, in N; C D is the vehicle drag coefficient; A is the vehicle frontal area, unit: m2; ρ a is the air density, unit is kg / m3; v rel (t) is the relative wind speed at time t, in m / s. The relative wind speed is the vehicle speed v(t) relative to the stationary road surface and the wind speed v relative to the stationary road surface at time t. w The algebraic sum of (t).

[0073] (2) Calculate the vehicle slope resistance:

[0074] F ε (t) = mg·sin(α(t))

[0075] Where: F ε (t) is the vehicle slope resistance at time t, in N; m is the vehicle mass, in kg; g is the acceleration due to gravity, generally g = 9.8 N / kg; α(t) is the road slope angle at time t, in degrees.

[0076] (3) Calculate vehicle braking force:

[0077] F brk (t) = n wh ·f(r brk (t))·A

[0078] Where: F brk (t) is the vehicle braking force at time t, in N; n wh is the number of wheels involved in braking; r brk (t) is the brake pedal opening at time t, ranging from 0 to 1; f(r brk (t)) represents the MAP of the brake pressure applied by the brake according to the brake pedal opening, in Pa; A is the brake working area, in m2.

[0079] 3.2 Rolling resistance prediction method based on neural network:

[0080] The vehicle rolling resistance prediction network structure proposed by the present invention is as follows: Figure 6 As shown in the figure, the network consists of two parts in series: a denoising autoencoder (DAE) feature dimension reduction network and a bidirectional long short-term memory (BiLSTM) regression prediction network. The DAE encoder is used for input, normalization and feature dimension reduction of each signal, and the BiLSTM network is used to predict and output the encoded one-dimensional abstract features.

[0081] (1) DAE feature dimension reduction network structure:

[0082] The DAE input feature dimensionality reduction network provided by the present invention consists of two parts: an encoder and a decoder, and the structure is as follows: Figure 4 The encoder part includes the input layer, encoding layer 1, and encoding layer 2. The 4 neuron nodes in the input layer are connected to the 3 neuron nodes S in the encoding layer 1 in a fully connected form. E11 ~S E13 , and then connected to the two neuron nodes S of the encoding layer 2 in a fully connected form E21 ~S E22 , and finally connected to the neuron node C in the middle of the network where the encoding is completed. The functional expression of the encoding process is as follows:

[0083] C=f(W E x+b E )

[0084] Where: f is the encoding function, x is the input matrix of the input layer, W E and b E are the weight parameters and bias in the coding layer respectively.

[0085] The neuron structure of the decoder is mirror-symmetric to that of the encoder. Neuron node C is fully connected to the two neuron nodes S in the decoding layer 1. D11 ~S D12 , and then connected to the three neuron nodes S of the decoding layer 2 in a fully connected form D21 ~S D23 , and finally connected to the 4 neuron nodes of the output layer in a fully connected form. The functional expression of the decoding process is as follows:

[0086] y=g(W D C+b D )

[0087] Where: g is the decoding function, y is the output matrix of the output layer, W D and b D are the weight parameters and bias in the decoding layer respectively.

[0088] The activation function between each fully connected layer uses the Tanh function:

[0089]

[0090] (2) DAE network training method:

[0091] The DAE network provided by the present invention adds noise to the input samples and then trains the network to restore the original input samples before the noise is added, thereby improving the robustness and anti-interference ability of the network. The training process is as follows:

[0092] 1) Construct the input matrix:

[0093] x={T DM ,α,v,T brk}

[0094] Where: T DM is the output torque of the main drive motor, in Nm; α is the road slope, in degrees; v is the vehicle speed, in m / s; T brk is the braking torque, in Nm.

[0095] 2) Use the zero distribution normalization method to normalize the input samples:

[0096]

[0097] Where: I j is any element in the input sample, is the normalized sample element, I max and I min are the maximum and minimum values ​​in the sample respectively. After normalizing the input matrix, the following results are obtained

[0098]

[0099] 3) Add destructive noise to the sample:

[0100]

[0101] Where: is the normalized input matrix after adding destructive noise, {e 1 ,e 2 ,e 3 ,e 4} is a destructive noise that causes some values ​​of the input matrix to be randomly lost.

[0102] 4) Define the training target loss function:

[0103]

[0104] Where: L D is the loss value of the DAE network; is the i-th output matrix in the output sequence; is the i-th input matrix in the input sequence; n is the sequence length.

[0105] 5) Divide the training set, test set and validation set, and train the DAE network using the gradient descent method, with the loss function minimized as the training goal, and iterate the training until L D Decrease and converge to a satisfactory range.

[0106] 6) L D After convergence, take the encoder part of the DAE network and use the neuron node C as the encoding output result.

[0107] (3) BiLSTM regression prediction network structure:

[0108] The structure of the BiLSTM regression prediction network proposed in the present invention is as follows Figure 5 As shown in the figure, a forward layer and a reverse layer are set in the hidden layer of the network. The transmission direction of the forward layer is the same as time, and it is responsible for processing the input layer data of the previous moment. The transmission direction of the reverse layer is opposite to time, and it is responsible for processing the input layer data of the future moment. The forward layer and the reverse layer are connected to the output layer together, and the prediction results corresponding to each moment are output. The prediction results are denormalized as the final output of the network.

[0109] The input parameters of the input layer are the one-dimensional abstract features after partial compression of the DAE encoder. The output results of the output layer are the rolling resistance prediction results at each moment. The Tanh function is used as the activation function between the input and output layers and the hidden layer.

[0110] (4) DAE-BiLSTM rolling resistance prediction network training process:

[0111] 1) Constructing original data: The time signal generated during the actual vehicle driving process is used as the input sample, which needs to include the following prior data: the output torque T of the main drive motor DM (t), unit Nm; road slope α(t), unit °; vehicle speed v(t), unit m / s; braking torque T brk (t), unit Nm; actual vehicle α v (t) Acceleration, unit: m / s2.

[0112] 2) Calculate rolling resistance based on data samples and physical formulas:

[0113]

[0114] Where: F r (t) is the rolling resistance at time t, in N; R wh is the effective rolling radius of the wheel, in m; m is the vehicle mass, in kg; η FD is the total efficiency of the transmission system; iFD is the total transmission ratio of the transmission system.

[0115] 3) The calculated sample rolling resistance is normalized using the zero distribution normalization method:

[0116]

[0117] Where: is the normalized rolling resistance at time t, and the normalized sample values ​​are distributed between -1 and 1; F rmax and F rmin They are the maximum and minimum values ​​of the rolling resistance of the sample, respectively, in N.

[0118] 4) Divide the training samples: divide the output torque T of the main drive motor DM (t), road slope α(t), vehicle speed v(t) and braking torque T brk (t) as input, normalized result As output, the training, test, and validation sets are divided proportionally.

[0119] 5) With the goal of minimizing the loss function and maximizing the prediction accuracy, the training is iterated until the loss function decreases and converges, and the prediction accuracy of the training set, test set, and validation set reaches a satisfactory range. Finally, the following prediction model is obtained:

[0120]

[0121] Where: F r,p is the predicted value corresponding to the input matrix; W is the weight coefficient matrix of the trained DAE-BiLSTM network, and b is the bias matrix of the network.

[0122] 6) Denormalization of prediction results:

[0123]

[0124] Where: F r,p is the actual rolling resistance value predicted by the model, in N; F rmax and F rmin They are the maximum and minimum rolling resistance values ​​calculated from the original input samples, in N.

[0125] 3.3 Calculation of acceleration and vehicle speed results:

[0126] (1) Calculate and derive F based on the physical formula w (t), F ε (t), F brk (t).

[0127] (2) Derivation of rolling resistance F based on the neural network prediction modelr,p (t)

[0128] (3) Calculate the vehicle acceleration and speed:

[0129]

[0130] Where: v(t) is the vehicle speed at time t, in m / s; v 0 is the initial vehicle speed, in m / s.

[0131] 4. Signal interaction between the virtual control system, the vehicle controlled object and the vehicle longitudinal dynamics model (1) The following are the signals input from the virtual control system to the vehicle controlled object model:

[0132]

[0133]

[0134]

[0135]

[0136]

[0137] (2) The following are the signals fed back by the vehicle controlled object model to the virtual control system

[0138] Variable Name Variable Description unit Feedback Source iISGAct_A Actual motor current A ISG Motor pwrISGElec_kW Motor power kW ISG Motor pwrISGMech_kW Motor mechanical power kW ISG Motor nISGAct_rpm Actual motor speed rpm ISG Motor trqISGAct_Nm Actual motor torque Nm ISG Motor uISGAct_V Actual motor voltage V ISG Motor tISGAct_℃ Actual motor temperature ℃ ISG Motor

[0139] Variable Name Variable Description unit Feedback Source rBatSOC_perc Battery Pack SOC ％ Power battery iBatAct_A Battery pack current A Power battery uBatAct_V Battery pack voltage V Power battery pwrBatAct_kW Battery pack power kW Power battery pwrLossBat_kW Battery pack power loss kW Power battery rBatAct_Ω Battery pack internal resistance Ω Power battery tBatAct_℃ Battery pack temperature ℃ Power battery uBatLoss_V Battery pack voltage loss V Power battery uBatCell_V Battery pack single string voltage V Power battery rBatCell_Ω Internal resistance of a single battery string Ω Power battery uBatCellRes_V Battery pack single string internal resistance voltage V Power battery

[0140] Variable Name Variable Description unit Feedback Source iMotAct_A Actual motor current A Main drive motor pwrMotElec_kW Motor power kW Main drive motor pwrMotMech_kW Motor mechanical power kW Main drive motor nMotAct_rpm Actual motor speed rpm Main drive motor trqMotAct_Nm Actual motor torque Nm Main drive motor uMotAct_V Actual motor voltage V Main drive motor tMotAct_degC Actual motor temperature ℃ Main drive motor

[0141] Variable Name Variable Description unit Feedback Source iAuxAct_A Actual current A Electrical accessories uAuxAct_V Actual voltage V Electrical accessories pwrAuxAct_kW Actual electrical power kW Electrical accessories

[0142] (3) The following are the signals input from the vehicle controlled object model to the vehicle longitudinal dynamics model:

[0143] Variable Name Variable Description unit Control Object lWhlBase_mm Wheelbase mm Kinetic model mConVeh_kg Vehicle weight kg Kinetic model arConVeh_m2 Windward area <![CDATA[m 2 ]]> Kinetic model mConAddLoad_kg Additional weight kg Kinetic model rConTire_m Tire radius m Kinetic model rhoConAir_kgm3 Air density <![CDATA[Kg / m 3 ]]> Kinetic model valTireFricCoeff Tire friction coefficient none Kinetic model cwConVeh Drag coefficient / Kinetic model iConGear Main reducer speed ratio / Kinetic model jConInGear_kgm2 Main reducer input shaft inertia <![CDATA[Kg·m 2 ]]> Kinetic model jConOutGear_kgm2 Main reducer output shaft inertia <![CDATA[Kg·m 2 ]]> Kinetic model effGear_perc Main reducer efficiency ％ Kinetic model rRoadSlope_perc slope ％ Kinetic model trqVehDrvAct_Nm Actual driving torque of the vehicle Nm Kinetic model

[0144] (2) The following are the signals fed back by the vehicle longitudinal dynamics model:

[0145] Variable Name Variable Description unit Feedback Source vVehAct_kmh Actual vehicle speed km / h Kinetic model accVehAct_ms2 Actual vehicle acceleration <![CDATA[m / s 2 ]]> Kinetic model fVehResAct_N Actual driving resistance of the vehicle N Kinetic model mVehAct_kg Actual vehicle mass kg Kinetic model lDrvAct_km Actual vehicle mileage km Kinetic model nWhlRLAct_rpm Actual left rear wheel speed rpm Kinetic model nWhlFLAct_rpm Actual left front wheel speed rpm Kinetic model nWhlFRAct_rpm Actual right front wheel speed rpm Kinetic model nWhlRRAct_rpm Actual right rear wheel speed rpm Kinetic model

[0146] 4. Specific process of simulation test

[0147] (1) Vehicle start:

[0148] The driver module sends a key ON signal, and the virtual vehicle controller sends the following control signals to the vehicle controlled object model:

[0149] The power battery main positive relay is ON;

[0150] ISG motor enable command ON;

[0151] Main drive motor enable command ON;

[0152] The electrical auxiliary relay control command is ON;

[0153] At this point, the vehicle's high-voltage system is ready and simulation testing can begin.

[0154] (2) Range extender start-up and power generation:

[0155] The driver module issues the range extender power generation command and the required power generation, and the range extender controller ASW issues the following control signals

[0156] Send an engine start command to the virtual ECU, calculate and give the engine target torque;

[0157] Send an enable command to the virtual GCU to calculate and give the generator target speed;

[0158] After receiving the above instructions, the virtual ECU and virtual GCU send the following control signals to the vehicle controlled object model:

[0159] The virtual ECU sends a start command and the calculated throttle opening to the engine;

[0160] The virtual GCU sends an enable command and the calculated required torque to the ISG motor.

[0161] At this time, the engine and generator respond to the target speed and torque and enter the power generation condition.

[0162] (3) Vehicle driving:

[0163] The driver module shifts into the D gear, releases the handbrake, and steps on the accelerator or brake pedal. The virtual vehicle controller sends the following control signals to the vehicle controlled object model:

[0164] Send an enable command to the virtual main drive motor controller, calculate and give the main drive motor target load percentage;

[0165] Sends the calculated brake pressure percentage to the virtual electric auxiliary controller.

[0166] Virtual main drive motor controller and virtual electric auxiliary component controller:

[0167] The virtual main drive motor controller sends an enable command and calculated required torque to the main drive motor;

[0168] The virtual electric auxiliary controller sends calculated brake pressure to the brakes.

[0169] (4) The vehicle longitudinal dynamics model is calculated as follows:

[0170] The air resistance, slope resistance and brake resistance are calculated according to the air density, slope and brake pressure at the previous moment;

[0171] According to the wheel speed at the previous moment, the driving torque of the main drive motor and the resistance calculated above, the trained neural network calculates the rolling resistance, and then the total external force of the vehicle at the current moment is obtained;

[0172] The vehicle's current speed and acceleration are calculated based on the resultant external force.

[0173] The virtual test environment architecture proposed in the present invention has the following technical features:

[0174] (1) The present invention uses a fully simulated method to build a simulation verification system, which can directly perform functional testing and verification on the control strategy integrated in the application software layer during the range extender controller software development phase, and discover problems in advance. Compared with traditional HIL bench and vehicle testing, the present invention can perform simulation testing on a computer without any additional hardware, and can move forward the testing work that could only be carried out after hardware integration was completed, shortening the development cycle and reducing development costs.

[0175] (2) The present invention defines a standardized signal interface between the range extender controller ASW and the test environment, which is divided into a hard-wire signal interface and a CAN message signal interface according to the signal type of the real range extender controller. Therefore, the test environment can be transplanted into the HIL bench without much modification, and the real range extender controller can be connected to carry out the test to ensure the consistency of the test environment.

[0176] The virtual controlled object of the extended-range electric vehicle proposed by the present invention has the following technical features:

[0177] (1) The whole vehicle controlled object is built based on Cruise M commercial software, which can accurately restore the physical state of key components such as the engine, generator and power battery during vehicle operation, ensuring the simulation accuracy of the controlled object.

[0178] (2) The controlled object model built based on Cruise M supports multiple environments from MIL to SIL to HIL, and can quickly change vehicle configuration and component parameters according to test requirements to ensure the portability of the controlled object.

[0179] The vehicle longitudinal dynamics model provided by the present invention has the following technical features:

[0180] (1) The neural network method is used to predict rolling resistance. It does not involve empirical parameters such as friction coefficient, rolling resistance coefficient, slip coefficient, etc. in the conventional rolling resistance calculation formula, which are difficult to accurately obtain. Instead, it is based on motor output torque, ramp angle, vehicle speed and braking torque, which can be accurately obtained. These physical quantities are used for training and prediction, thereby deriving a more accurate rolling resistance.

[0181] (2) Use a DAE-BiLSTM cascade network. Input the motor torque, ramp angle, vehicle speed and brake torque parameters that are damaged by noise, train a denoising autoencoder to restore the original samples that have not been damaged, and improve the robustness and anti-interference ability of the network; perform feature dimensionality reduction on the input, reduce the input four-dimensional features into one-dimensional abstract features, and reduce the difficulty of training; use BiLSTM to train the input time series signal to capture the dependency between the input samples before and after, and improve the accuracy of prediction;

[0182] (3) The vehicle's driving resistance and speed are calculated by combining physical models with neural networks, which ensures the accuracy of the calculation results while reducing the complexity of the model. Air resistance, ramp resistance, and braking force are calculated using physical formulas, and rolling resistance is predicted using a neural network. The vehicle's current external traction and acceleration are calculated and integrated to obtain the vehicle speed.

[0183] Although the present invention has been described in detail above with specific embodiments, it is obvious to those skilled in the art that some modifications or improvements can be made to the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all belong to the scope of protection claimed by the present invention.

Claims

1. A range extender controller application software layer control strategy simulation test system, characterized by: It includes a virtual control system, a whole vehicle controlled object and a whole vehicle longitudinal dynamics model; the virtual control system is connected to the whole vehicle controlled object and the whole vehicle longitudinal dynamics model respectively; The virtual control system includes a virtual driver module, a virtual vehicle controller, a virtual main drive motor controller, a range extender controller ASW, a virtual engine controller (ECU), a virtual generator controller (GCU), a virtual electric auxiliary component controller and a virtual battery management system (BMS); The controlled objects of the virtual vehicle include: virtual engine, virtual generator, virtual power battery, virtual high-voltage electric auxiliary parts, and virtual chassis system; The vehicle longitudinal dynamics model includes a model for calculating the vehicle's driving resistance during driving and a model for calculating the vehicle speed.

2. A range extender controller application software layer control strategy simulation test system according to claim 1, characterized in that: The driving resistance model includes (1) Calculation of vehicle air resistance model: Where: F w (t) is the air resistance of the vehicle at time t, in N; C D is the vehicle drag coefficient; A is the vehicle frontal area, unit: m2; ρ a is the air density, unit is kg / m3; v rel (t) is the relative wind speed at time t, in m / s. The relative wind speed is the vehicle speed v(t) relative to the stationary road surface and the wind speed v relative to the stationary road surface at time t. w The algebraic sum of (t); (2) Calculation of vehicle ramp resistance model: F ε (t)=mg·sin(α(t)) Where: F ε (t) is the vehicle slope resistance at time t, in N; m is the vehicle mass, in kg; g is the acceleration of gravity, generally g = 9.8N / kg; α(t) is the road slope angle at time t, in degrees; (3) Calculation of vehicle braking force model: F brk (t)=n wh ·f(r brk (t))·A Where: F brk (t) is the vehicle braking force at time t, in N; n wh is the number of wheels involved in braking; r brk (t) is the brake pedal opening at time t, ranging from 0 to 1; f(r brk (t)) represents the MAP of the brake pressure applied by the brake according to the brake pedal opening, in Pa; A is the brake working area, in m2.

Images