Hybrid electric vehicle proportional control method and system based on neural network

Abstract

The invention relates to a hybrid electric vehicle proportional control method and system based on a neural network, the method is applied to a cloud end deployed with the neural network, and the method comprises the following steps: establishing connection with a vehicle end, and obtaining vehicle dynamics control data and road state data of a road where the vehicle is located; the vehicle dynamics control data and the road state data serve as input of a neural network, and influence coefficients of all types of sub-data on an engine are obtained through feature extraction and deep learning; and calculating the optimal control proportionality coefficient of the engine of the vehicle at the next moment based on the influence coefficient of each type of sub-data on the engine, and sending the optimal control proportionality coefficient to the vehicle end to realize the proportional control of the engine and the motor of the vehicle end. Compared with the prior art, the power output control proportionality coefficient of the engine can be dynamically adjusted, meanwhile, the power output control proportionality coefficient of the engine is prevented from being adjusted too frequently, and energy loss can be effectively reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of hybrid electric vehicles, and in particular, to a proportional control method and system for hybrid electric vehicles based on a neural network. Background Art

[0002] As a new type of energy-saving and environmental protection technology, hybrid electric vehicles can intelligently combine fuel and electricity, switch power sources according to different driving conditions, mainly rely on electricity when driving at low speeds, and switch to fuel drive when driving at high speeds, thereby reducing fuel consumption. However, this mode of switching driving power based on speed cannot flexibly adjust the output power of the engine and the motor according to the driver's driving conditions and different road conditions, resulting in poor overall energy-saving effects.

[0003] Chinese Patent Application Publication No. CN115071669A discloses a hybrid energy management system and method, which can adjust the power distribution of the electric drive and fuel drive of a hybrid electric vehicle in real time through a collaborative optimization method of vehicle navigation information and cloud computing, solve the energy management problem of hybrid electric vehicles under different driving conditions, and achieve a reduction in fuel consumption and carbon emissions. However, this application relies on the cooperation of driving conditions and PMP model calculation in the cloud, which poses high requirements on the cloud computing speed and computing power. How to achieve efficient deployment of computing resources and improve the response speed remains an urgent problem to be solved.

[0004] In summary, there is a need for a power output control system for the engine and the motor, which can better save energy loss under different road conditions and driving habits. Summary of the Invention

[0005] The purpose of the present invention is to overcome the above-mentioned defects existing in the prior art and provide a proportional control method and system for hybrid electric vehicles based on a neural network, so as to solve or partially solve the problem of inefficient energy loss saving under different road conditions and driving habits.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] In one aspect of the present invention, a proportional control method for hybrid electric vehicles based on a neural network is provided, which is applied to a cloud deployed with a neural network. The method includes the following steps:

[0008] Establish a connection with the vehicle terminal, and obtain vehicle dynamics control data and road state data of the road where the vehicle is located;

[0009] Take the vehicle dynamics control data and the road state data as the input of the neural network, and through feature extraction and deep learning, obtain the influence coefficients of various types of sub-data on the engine;

[0010] Based on the influence coefficients of each type of sub - data on the engine, calculate the optimal control ratio coefficient of the vehicle's engine at the next moment and send it to the vehicle end to achieve the proportional control of the vehicle's engine and motor.

[0011] As a preferred technical solution, the process of obtaining the influence coefficients of each type of sub - data on the engine includes the following steps:

[0012] Merge the vehicle dynamics control data and the road state data into a high - dimensional vector;

[0013] Based on the high - dimensional vector, calculate through the weight layer and the bias layer and input it into the LSTM module;

[0014] Based on the vector output by the LSTM module, obtain the output vector through another weight layer and bias layer;

[0015] Convert the output vector into the influence coefficients of each type of sub - data on the engine by splitting.

[0016] As a preferred technical solution, the calculation process of the output vector includes the following steps:

[0017] Concatenate the high - dimensional vector x t and the hidden state h at the previous moment t-1 to obtain the concatenated vector

[0018] Pass the vector to the forget gate, pass the information through the activation function Sigmoid, and calculate the forget vector where σ is the activation function Sigmoid, W f is the forget - gate weight matrix, and b f is the forget - gate bias matrix;

[0019] Pass the vector to the input gate, pass the information through the activation functions Sigmoid and tanh respectively to update the cell state, and calculate the vectors i t and

[0020]

[0021] where W i , W c are the input - gate weight matrices, and b i , b c are the input - gate bias matrices;

[0022] Multiply the output vectors i t and point - by - point to obtain the vector

[0023] Update unit status

[0024] Calculate the output gate vector where W o is the output gate weight matrix, b o is the output gate bias matrix;

[0025] Calculate the new hidden state h t = o t * tanh(c t );

[0026] Obtain the vector x output by the LSTM module t+1 = σ(W h · h t + b h ), where W h is the output weight matrix, b h is the output bias matrix.

[0027] As a preferred technical solution, the influence coefficients of the output vector on the engine in various types of sub - data are realized by splitting using the following formula:

[0028] Split the output vector v into [v1, v2, v3], which respectively represent the influence of various types of data on the engine proportionality coefficient;

[0029] Calculate the optimal control proportionality coefficient k of the engine required at time t + 1 = v1 + v2 + v3.

[0030] As a preferred technical solution, the road state data includes road congestion index sub - data and slope sub - data.

[0031] As a preferred technical solution, the road congestion index sub - data is calculated using the following formula:

[0032]

[0033] where c is the road congestion index, L1 is the actual traffic flow of the road, L2 is the capacity of the road, is the average speed of the road, and v1 is the speed when the road is unobstructed.

[0034] As a preferred technical solution, the vehicle dynamics control data includes accelerator pedal sub - data and brake pedal sub - data.

[0035] As a preferred technical solution, the process of realizing the proportional control of the vehicle - end engine and motor based on the optimal control proportionality coefficient at the vehicle end includes the following steps:

[0036] Calculate the difference ratio between the proportional coefficient of the engine required at time t+1 and the proportional coefficient of the engine required at time t;

[0037] Based on the difference ratio, calculating a final engine proportional coefficient;

[0038] Based on the final engine scaling factor, the motor scaling factor is calculated.

[0039] As a preferred technical solution, the final engine proportional coefficient and motor proportional coefficient are calculated using the following formula:

[0040]

[0041] k' t+1 =1-k t+1

[0042] Among them, k ε is the engine proportional coefficient k required at time t+1 and the engine proportional coefficient k required at time t t The difference ratio, k t+1 , k' t+1 are the final engine proportional coefficient and motor proportional coefficient respectively, and ε is the threshold for whether to adjust the engine proportional coefficient.

[0043] Another aspect of the present invention provides a hybrid vehicle proportional control system based on a neural network, which is used to implement the aforementioned hybrid vehicle proportional control method based on a neural network. The system includes a cloud data processing center (6) on which a neural network is deployed and a vehicle end, wherein the vehicle end includes:

[0044] at least one camera (8);

[0045] A GPS module (9) is used to obtain the congestion index of the road where the vehicle is located, the location of the road and the slope data;

[0046] Millimeter wave radar (10);

[0047] An accelerator pedal (11), used for collecting accelerator pedal sub-data;

[0048] A brake pedal (12), used for collecting brake pedal sub-data;

[0049] The vehicle-mounted computer (5) is used to obtain data from the GPS module (9), the camera (8), the millimeter wave radar (10), the accelerator pedal (11) and the brake pedal (12), and transmit the data to the cloud data processing center (6) via a wireless network;

[0050] A proportional control unit (7) is configured to control the output power and target gear of the engine (3) and the motor (4) respectively through the engine controller (1) and the motor controller (2) based on the optimal control proportional coefficient obtained from the cloud data processing center (6).

[0051] Compared with the prior art, the present invention has at least one of the following beneficial effects:

[0052] (1) Reducing energy loss while ensuring sufficient power output of the hybrid vehicle: The neural network unit based on LSTM provided by the present invention processes the data from existing sensors at the input end of the neural network with simple processing, reducing the transformation cost caused by the secondary installation and debugging of vehicle sensors. The neural network unit based on LSTM can make full use of data information such as the accelerator pedal, brake pedal, road congestion index, and slope to calculate the power output control ratio of the hybrid vehicle during driving. While considering the driving habits of the driver, it also takes into account the current driving conditions of the vehicle, thereby obtaining the optimal engine power output control ratio, and further achieving the goal of reducing energy loss while ensuring sufficient power output of the hybrid vehicle.

[0053] (2) Maximizing the utilization of vehicle data: The present invention can collect sensor data such as the accelerator pedal, brake pedal, congestion coefficient, and slope during the vehicle driving process in real time, and classify and upload the data to the cloud to realize the real-time online update of the data based on the LSTM neural network, efficiently utilize the data during the vehicle driving process, and can also provide auxiliary driving information to the driver.

[0054] (3) Capturing time series features fully: The neural network unit based on LSTM provided by the present invention uses not only the current driving data of the vehicle but also the historical driving data of the vehicle in the LSTM module adopted in the neural network, which can well capture the long-term dependence relationship in the time series data and better predict the optimal engine control ratio at the next moment.

[0055] (4) Optimizing the problem of frequent output ratio adjustment: The proportional control unit provided by the present invention compares the optimal engine power output control ratio calculated by the neural network with the engine power output control ratio at the previous moment to obtain the engine power output control ratio at the next moment, and then calculates the motor power output ratio, effectively preventing the problem of frequent adjustment of the engine power output ratio, and at the same time achieving the purpose of saving energy loss. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a flowchart of the proportional control method for a hybrid vehicle based on a neural network in the embodiment;

[0057] Figure 2Schematic diagram of the proportional control system of a hybrid vehicle based on a neural network in the embodiment;

[0058] Figure 3 Structural diagram of the neural network based on LSTM in the embodiment;

[0059] Figure 4 Structural diagram of the LSTM module of the neural network unit in the embodiment;

[0060] Figure 5 Schematic diagram of the electronic device in the embodiment,

[0061] wherein, 1. Engine controller, 2. Motor controller, 3. Engine, 4. Motor, 5. On-vehicle computer, 6. Cloud data processing center, 7. Proportional control unit, 8. Camera, 9. GPS module, 10. Millimeter-wave radar, 11. Accelerator pedal, 12. Brake pedal. Specific implementation manner

[0062] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0063] Embodiment 1

[0064] In view of the problems existing in the foregoing prior art, this embodiment provides a proportional control method for a hybrid vehicle based on a neural network, which is applied to the cloud data processing center 6. The center is provided with a Figure 3 recurrent neural network based on LSTM as shown. The recurrent neural network based on LSTM inputs data such as the accelerator pedal, brake pedal, congestion index, and slope of the passing road transmitted by the on-vehicle computer 5 into the corresponding neural network for data feature extraction. Then, after being processed by a weight matrix and a bias matrix, it is calculated by the LSTM module unit and multiplied by the corresponding weight. Finally, the influence coefficients of each type of data on the engine are summed to obtain the optimal engine power output proportional control coefficient, and then transmitted to the on-vehicle computer. Specifically, see Figure 1 , the method includes the following steps:

[0065] Step S001, combine the data such as the accelerator pedal, brake pedal, congestion index, and slope obtained by the vehicle to form high-dimensional vector information x, and use the combined data information x as the input end of the recurrent neural network unit based on LSTM.

[0066] Among them, the congestion index calculation formula:

[0067]

[0068] Wherein, c is the congestion index of the passing road, L1 is the actual traffic flow of the passing road, and L2 is the capacity of the passing road. is the average speed of the passing road, and v1 is the speed when the passing road is unobstructed, which can be obtained by searching historical data.

[0069] Step S002: After calculating the high-dimensional vector information x through the weight layer w1 and the bias layer b1, it is used as the input vector u of the LSTM module:

[0070] u = w1x + b1.

[0071] Step S003: Input the vector information u into the LSTM module to calculate the output vector v. See Figure 4 for the schematic diagram of the LSTM module. Specifically, this step includes the following steps S301 to S309:

[0072] Step S301: Connect the hidden state h t-1 at the previous moment and the current input x t together, and output the vector as:

[0073]

[0074] In S302: Pass the vector to the forget gate, pass the information through the activation function Sigmoid, and calculate the forget vector f t , which is used to determine the information to be discarded or retained, and the output value is (0, 1). Close to 0 means forgetting, and close to 1 means retaining:

[0075]

[0076] Wherein, σ is the activation function Sigmoid, and W f is the forget gate weight matrix, and b f is the forget gate bias matrix.

[0077] Step S303: Pass the vector to the input gate, and pass the information through the activation functions Sigmoid and tanh respectively to update the cell state, and calculate the vector i t and Sigmoid is used to determine which values to update, and tanh adjusts the values to (-1, 1):

[0078]

[0079] Wherein, W i, W c is the input gate weight matrix, b i , b c is the input gate bias matrix.

[0080] Step S304: Multiply the output vector i t from the previous step pointwise with to obtain the vector for calculating the important information retained in the vector ;

[0081]

[0082] Step S305: Multiply the cell state c t-1 pointwise with the forget vector f t , and then perform pointwise addition with the vector to update the cell state to a new value relevant to what the neural network discovers. The calculation formula for c t is:

[0083]

[0084] Step S306: Pass the hidden state h t-1 from the previous moment and the current input into the activation function Sigmoid to calculate the output gate vector o t :

[0085]

[0086] where, W o is the output gate weight matrix, b o is the output gate bias matrix

[0087] Step S307: Pass the newly modified cell state c t into the tanh function, and then multiply the tanh output vector pointwise with the output vector o t to determine the information that the hidden state should carry, and output the new hidden state h t :

[0088] h t = o t *tanh(c t )

[0089] Step S308: Pass the new hidden state h t into the activation function Sigmoid to calculate the current output vector x t+1 , and use it as the input vector for the next moment. Its calculation method is:

[0090] x t+1 = σ(W h·h t +b h )

[0091] Where, W h is the output weight matrix, and b h is the output bias matrix.

[0092] Step S309: Repeat steps S301 to S308 for a total of t times, where the number of times t can be 4; when t = 1, c t-1 and h t-1 are zero matrices.

[0093] Step S004: Pass the vector x t+1 obtained by calculating the vector u through the LSTM module to the weight layer w2 and the bias layer b2 to obtain the output vector v:

[0094] v = w2·x t+1 +b2

[0095] Where, w2 is the weight matrix of each type of data, and b2 is the bias matrix of each type of data

[0096] Step S005: Split the calculated v data into [v1, v2, v3], which respectively represent the influence of each type of data on the engine ratio coefficient;

[0097] Step S006: Calculate the optimal control ratio coefficient k of the engine required at time t + 1:

[0098] k = v1 + v2 + v3

[0099] After obtaining the optimal control ratio coefficient, the cloud data processing center 6 sends it to the vehicle end, and performs ratio control through the following process:

[0100] Step S101, the ratio control unit at the vehicle end predicts the target power and target gear required at the next moment of the current road condition according to the collected actual torque, actual vehicle speed, actual throttle opening, and actual brake pedal data.

[0101] The ratio control unit 7 compares and calculates the optimal control ratio calculated by the recurrent neural network based on LSTM and the control ratio at the previous moment, calculates the ratio of the power outputs of the engine 3 and the motor 4 required at the next moment, and then calculates the power required to be provided by the engine and the motor respectively according to the target power and target gear, so as to calculate the fuel injection amount of the engine 3 and the power consumption of the motor 4, and outputs them to the engine controller 1 and the motor controller 2 respectively, and then controls the operation of the engine 3 and the motor 4.

[0102] The calculation formula for the required engine 3 ratio coefficient is:

[0103]

[0104] In the formula, k t+1 is the engine proportionality coefficient required at time t + 1, and k t is the engine proportionality coefficient required at time t. k is the optimal engine control ratio calculated by the recurrent neural network based on LSTM. k ε is the difference ratio between the engine proportionality coefficient required at time t + 1 and the engine proportionality coefficient required at time t. ε is the threshold for whether to adjust the engine proportionality coefficient. The value of ε can be taken from [0.05, 0.2], and it can be adjusted according to the differences between the engine 3 and the motor 4 of the hybrid vehicle, which can effectively avoid frequently adjusting the proportionality coefficients of the engine 3 and the motor 4, resulting in increased energy consumption.

[0105] k ε The calculation formula of

[0106]

[0107] Compare the optimal engine control proportionality coefficient at each moment with the engine control proportionality coefficient at the previous moment. If k ε < ε, it means that the difference between the current optimal engine control proportionality coefficient and the engine control proportionality coefficient at the previous moment is not large. To prevent frequent adjustment of power output, increase the failure rate and fail to achieve the purpose of energy conservation, the engine control proportionality coefficient at this moment remains unchanged; when k ε ≥ ε, it means that the difference between the current optimal engine control coefficient and the engine control proportionality coefficient at the previous moment is relatively large, indicating that there are relatively large changes such as acceleration, deceleration or changes in driving conditions at present. At this moment, the engine control proportionality coefficient is adjusted to the optimal engine control proportionality coefficient.

[0108] The calculation formula for the motor proportionality coefficient required at the next moment is:

[0109] k' t+1 = 1 - k t+1

[0110] In the formula, k' t+1 is the motor proportionality coefficient required at time t + 1.

[0111] Embodiment 2

[0112] Based on Embodiment 1, this embodiment provides a proportional control system for a hybrid vehicle based on a neural network, which is used to implement the proportional control method for a hybrid vehicle based on a neural network as in Embodiment 1. The system is applied to existing hybrid vehicle products and adopts the principle of coordinated operation of the engine and the motor. Refer to Figure 2 , the system includes a cloud data processing center 6 deployed with a neural network and a vehicle end. The vehicle end includes:

[0113] 1. At least one camera 8;

[0114] 2. A GPS module 9, configured to obtain the congestion index, the location, and the slope data of the road where the vehicle is located;

[0115] 3. A millimeter-wave radar 10;

[0116] 4. An accelerator pedal 11, configured to collect sub-data of the accelerator pedal;

[0117] 5. A brake pedal 12, configured to collect sub-data of the brake pedal;

[0118] 6. An in-vehicle computer 5, configured to obtain the data of the GPS module 9, the camera 8, the millimeter-wave radar 10, the accelerator pedal 11, and the brake pedal 12, and send the data to the cloud data processing center 6 through a wireless network;

[0119] 7. A proportional control unit 7, configured to control the output power and the target gear of the engine 3 and the motor 4 respectively through the engine controller 1 and the motor controller 2 based on the optimal control proportional coefficient obtained from the cloud data processing center 6.

[0120] Wherein, the engine controller 1 and the motor controller 2 are mainly used to control the output power and the target gear of the engine 3 and the motor 4 respectively with the proportional control parameters obtained by the proportional control unit 7.

[0121] The in-vehicle computer 5 obtains the data of the GPS module 9, multiple cameras 8, the millimeter-wave radar 10, the accelerator pedal 11, and the brake pedal 12, and sends the data to the cloud data processing center 6 through a wireless network. Preferably, the in-vehicle computer 5 uses 4G / 5G communication to transmit data with the cloud data processing center 6, which can reduce the time consumed by data transmission and improve the timeliness of the proportional control system of the engine and the motor.

[0122] The GPS module 9 is mainly used to obtain data such as the congestion index, the location, and the slope of the passing road;

[0123] In the present invention, the hybrid vehicle adopts the working modes of the engine and the motor. Under different passing roads, according to the conditions of the accelerator pedal and the brake pedal driven by the driver, the optimal control ratio of the engine power output is calculated and compared with the engine power output control ratio at the previous moment, which can dynamically adjust the engine power output control ratio coefficient and prevent the engine power output control ratio coefficient from being adjusted too frequently, effectively saving energy loss.

[0124] Embodiment 3

[0125] Based on Embodiment 1, refer to Figure 5, this embodiment provides an electronic device, including: one or more processors and a memory. One or more programs are stored in the memory, and the one or more programs include instructions for executing the neural network-based proportional control method for hybrid electric vehicles as described in Embodiment 1.

[0126] As Figure 5 described, at the hardware level, the electronic device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory. Of course, it may also include other hardware required for other services. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to implement the above Figure 1 described method. Of course, in addition to the software implementation, the present invention does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc. That is to say, the execution subject of the following processing flow is not limited to each logic unit, and can also be hardware or a logic device.

[0127] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of, for example, read-only memory (ROM) or flash memory (flash RAM). The memory is an example of a computer-readable medium.

[0128] Embodiment 4

[0129] This embodiment provides a computer-readable storage medium, including one or more programs for execution by one or more processors of an electronic device. The one or more programs include instructions for executing the neural network-based proportional control method for hybrid electric vehicles as described in Embodiment 1.

[0130] Computer-readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology for information storage. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tapes, disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information accessible by a computing device. As defined herein, computer-readable media do not include transitory computer-readable media, such as modulated data signals and carrier waves.

[0131] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A proportional control method for a hybrid vehicle based on a neural network, characterized in that, Applied to the cloud where a neural network is deployed, the method includes the following steps: Establish a connection with the vehicle end, and obtain vehicle dynamics control data and road condition data of the road where the vehicle is located; Use the vehicle dynamics control data and the road condition data as the input of the neural network, and through feature extraction and deep learning, obtain the influence coefficients of various types of sub-data on the engine; Based on the influence coefficients of various types of sub-data on the engine, calculate the optimal control proportional coefficient of the vehicle's engine at the next moment, and send it to the vehicle end to achieve proportional control of the vehicle end's engine and motor.

2. The proportional control method for a hybrid vehicle based on a neural network according to claim 1, wherein The process of obtaining the influence coefficients of various types of sub-data on the engine includes the following steps: Merge the vehicle dynamics control data and the road condition data into a high-dimensional vector; Based on the high-dimensional vector, calculate through the weight layer and the bias layer, and input it into the LSTM module; Based on the vector output by the LSTM module, through another weight layer and bias layer, obtain the output vector; Convert the output vector into the influence coefficients of various types of sub-data on the engine through splitting.

3. The proportional control method for a hybrid vehicle based on a neural network according to claim 2, characterized in that The calculation process of the output vector includes the following steps: Concatenate the high-dimensional vector x t and the hidden state h at the previous moment t-1 to obtain the concatenated vector Pass the vector to the forget gate, pass the information through the activation function Sigmoid, and calculate the obtained forget vector where σ is the activation function Sigmoid, W f is the forget gate weight matrix, and b f is the forget gate bias matrix; Pass the vector to the input gate, pass information through the activation functions Sigmoid and tanh respectively, update the cell state, and calculate the vector i t and Among them, W i , W c are the input gate weight matrices, b i , b c are the input gate bias matrices; Multiply the output vector i t and point by point to obtain the vector Update unit status Calculate the output gate vector where W o is the output gate weight matrix, and b o is the output gate bias matrix; Calculate the new hidden state h t = o t * tanh(c t ); Obtain the vector x output by the LSTM module t+1 = σ(W h ·h t + b h ), where W h is the output weight matrix, and b h is the output bias matrix.

4. A proportional control method for a hybrid vehicle based on a neural network according to claim 2, characterized in that The conversion of the output vector into the influence coefficients of various types of sub-data on the engine through splitting is implemented by the following formula: Split the output vector v into [v1, v2, v3], which respectively represent the influence of various types of data on the engine proportional coefficient; Calculate the optimal control proportional coefficient k of the engine required at time t + 1 = v1 + v2 + v3.

5. A proportional control method for a hybrid vehicle based on a neural network according to claim 1, characterized in that The road condition data includes road congestion index sub-data and slope sub-data.

6. A proportional control method for a hybrid vehicle based on a neural network according to claim 5, characterized in that, The road congestion index sub-data is calculated by the following formula: Among them, c is the road congestion index, L1 is the actual traffic flow of the road, and L2 is the capacity of the road. is the average speed of the road, and v1 is the speed when the road is unobstructed.

7. A proportional control method for a hybrid vehicle based on a neural network according to claim 1, characterized in that, The vehicle dynamics control data includes throttle pedal sub-data and brake pedal sub-data.

8. A proportional control method for a hybrid vehicle based on a neural network according to claim 1, characterized in that, At the vehicle end, the process of implementing proportional control of the vehicle end's engine and motor based on the optimal control proportional coefficient includes the following steps: Calculate the difference ratio between the proportional coefficient of the engine required at time t + 1 and the proportional coefficient of the engine required at time t; Based on the difference ratio, calculate the final engine proportional coefficient; Based on the final engine proportional coefficient, calculate the motor proportional coefficient.

9. A proportional control method for a hybrid vehicle based on a neural network according to claim 8, characterized in that, The final engine proportional coefficient and the motor proportional coefficient are calculated by the following formula: k′ t+1 = 1 - k t+1 where k ε is the difference ratio between the engine proportionality coefficient k required at time t + 1 and the engine proportionality coefficient k required at time t, k t , k t+1 , k' t+1 are the final engine proportionality coefficient and the motor proportionality coefficient respectively, and ε is the threshold for whether to adjust the engine proportionality coefficient.

10. A proportional control system for a hybrid vehicle based on a neural network, characterized in that, For implementing the proportional control method of a hybrid electric vehicle based on a neural network as described in any one of claims 1-9, the system includes a cloud data processing center (6) where a neural network is deployed and a vehicle end, and the vehicle end includes: At least one camera (8); A GPS module (9) for obtaining the congestion index, location and slope data of the road where the vehicle is located; A millimeter-wave radar (10); A throttle pedal (11) for collecting throttle pedal sub-data; A brake pedal (12) for collecting brake pedal sub-data; An in-vehicle computer (5) for obtaining the data of the GPS module (9), camera (8), millimeter-wave radar (10), throttle pedal (11) and brake pedal (12), and sending it to the cloud data processing center (6) through a wireless network; A proportional control unit (7) is configured to control the output power and target gear of the engine (3) and the motor (4) respectively through the engine controller (1) and the motor controller (2) based on the optimal control proportionality coefficient obtained from the cloud data processing center (6).

Images