Tractor powertrain system based on improved ECVT and its control method

Abstract

The present disclosure relates to the technical field of vehicle power units, and proposes a tractor powertrain system based on an improved ECVT and its control method. By optimizing the ECVT system structure and control strategy, the fuel economy, adaptability, power transmission efficiency and system life of the tractor are improved. The engine does not directly output power and can always operate within the optimal operating condition range, enhancing fuel economy. The control device based on deep learning can accurately predict the power demand and dynamically adjust the output power of the motor, improving the adaptability of the tractor in complex operating environments and avoiding system shocks caused by changes in slope or soil resistance. The motor replaces the hydraulic system, reducing the energy loss of hydraulic transmission, improving the overall transmission efficiency, and using the fast response characteristics of the motor to make power adjustment more precise and flexible, optimizing the operating performance of the tractor.

Description

Technical Field

[0001] The present disclosure relates to the technical field of vehicle power devices, and more specifically, to a tractor powertrain system based on an improved ECVT and a shifting control method. Background Art

[0002] The statements in this section merely provide background technical information related to the present disclosure and do not necessarily constitute prior art.

[0003] With the continuous development of agricultural mechanization, as an important agricultural machinery, the performance of the transmission system of tractors directly affects the operation efficiency and fuel economy. Traditional mechanical shifting technologies use sliding gears and engagement sleeves. Due to problems such as complex structure, uneven shifting, low reliability, and high noise, it is difficult to adapt to the complex environment of farmland operations. Continuously Variable Transmission (CVT) technology has gradually been applied in agricultural machinery. Among them, Hydrostatic Transmission (HST) has smooth shifting, but has problems such as low efficiency, large energy loss, and easy oil leakage. Electrically Controlled Continuously Variable Transmission (ECVT) uses an electric motor to replace the hydraulic system for power transmission, which not only improves efficiency but also can recover braking energy, and has better energy-saving performance. Therefore, it has become an important development direction for hybrid tractors.

[0004] During the operation of tractors, affected by factors such as soil humidity, operation depth, and the state of agricultural implements, the operation resistance shows complex non-linear dynamic changes, which puts higher requirements on the response speed and adaptability of the transmission system. Existing ECVT solutions still have many problems, mainly including: (1) The engine burden is too heavy, the fuel utilization rate is low, the fuel consumption is high, and thus the fuel economy is poor. Current ECVT solutions require the engine to supply energy to the wheels, the operation load, and the generator at the same time, resulting in a large operation load and the operating point deviating from the optimal working condition, affecting the fuel efficiency; (2) The operation environment of tractors varies, such as slope changes and soil resistance changes. In existing ECVT solutions, the motor cannot adjust the output power in a timely and accurate manner, which may cause impacts on the transmission system, reduce the service life, and damage the motor or gearbox when the load suddenly changes. Summary of the Invention

[0005] To solve the above problems, the present disclosure proposes a tractor powertrain system and a transmission control method based on an improved ECVT. By improving the ECVT scheme, optimizing the coordinated control of the engine and the motor, and introducing neural network predictive control, the precise control of the output power of the tractor transmission system is improved, and the operating efficiency, reliability, and stability of the power system are enhanced.

[0006] To achieve the above object, the present disclosure adopts the following technical solutions:

[0007] One or more embodiments provide a tractor powertrain system based on an improved ECVT, including a transmission system and a control device. The transmission system includes a generator, a mechanical coupling device, and at least two motors. The generator is used to supply electrical energy to the motors, and the power outputs of the motors are coupled through the mechanical coupling device.

[0008] The control device is configured to: predict the required power of the transmission system based on a deep learning model, dynamically allocate and adjust the output powers of the mutually coupled motors based on the predicted required power, and achieve adaptive power output adjustment.

[0009] One or more embodiments provide a control method for a tractor powertrain system based on an improved ECVT, including the following steps:

[0010] According to the obtained control instruction, control the tractor powertrain system to operate in the corresponding working mode;

[0011] When in the dual-motor parallel drive mode, obtain the tractor working state data and the images around the vehicle, and perform preprocessing;

[0012] Transmit the preprocessed data to the deep learning model for prediction to obtain the predicted required power of the transmission system;

[0013] Construct an objective function for power distribution with the goals of minimizing the deviation between the output power and the required power and minimizing the sudden change in the battery discharge rate;

[0014] Solve the objective function according to the obtained power battery parameters and the predicted required power to obtain a power distribution scheme.

[0015] Compared with the prior art, the beneficial effects of the present disclosure are as follows:

[0016] By optimizing the ECVT system structure and control strategy, the present disclosure improves the fuel utilization rate, adaptability, power transmission efficiency, and the lifespan of the entire power system of the tractor. The engine does not directly output power but is used for power generation, so that the engine can always operate within the optimal working condition range during operation, avoiding the decrease in fuel efficiency caused by load fluctuations, thereby improving fuel economy.

[0017] The control device based on deep learning can accurately predict power demand, dynamically adjust the output power of the motor, improve the adaptability of the tractor in complex operating environments, and avoid system shocks caused by changes in slope or soil resistance.

[0018] By replacing the hydraulic system with a motor, the energy loss of hydraulic transmission is reduced, the overall transmission efficiency is improved, and the fast response characteristics of the motor are utilized to make power adjustment more accurate and flexible, optimizing the operating performance of the tractor.

[0019] The advantages of the present disclosure and the advantages of additional aspects will be described in detail in the following specific embodiments. Brief Description of the Drawings

[0020] The accompanying drawings forming a part of this disclosure are used to provide a further understanding of the disclosure. The illustrative embodiments of the disclosure and their descriptions are used to explain the disclosure and do not constitute a limitation to the disclosure.

[0021] Figure 1 is a schematic structural diagram of the tractor powertrain system of Embodiment 1 of the present disclosure;

[0022] Figure 2 is a flowchart of the method of Embodiment 2 of the present disclosure;

[0023] Wherein: 1. Generator, 2. First motor, 3. Second motor, 4. First power battery, 5. Second power battery, 6. Planetary gear mechanism, 61. Sun gear, 62. Planet gear, 63. Planet carrier, 64. Ring gear, 7. Gear A, 8. Second gear, 9. First gear. Detailed Description of the Embodiments

[0024] The present disclosure will be further described below in conjunction with the drawings and embodiments.

[0025] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present disclosure belongs.

[0026] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof. It should be noted that, without conflict, the various embodiments and features in the present disclosure can be combined with each other. The embodiments will be described in detail below with reference to the drawings.

[0027] Embodiment 1

[0028] In the technical solutions disclosed in one or more embodiments, as Figure 1 shown, the tractor powertrain system based on the improved ECVT includes a transmission system and a control device. The transmission system includes a generator, a mechanical coupling device, and at least two electric motors. The generator is used to provide electrical energy for the electric motors, and the electric motors achieve the coupling of power output through the mechanical coupling device;

[0029] The control device is configured to: predict the required power of the transmission system based on a deep learning model, and dynamically allocate and adjust the output power of the mutually coupled electric motors based on the predicted required power to achieve adaptive power output regulation.

[0030] In this embodiment, the tractor powertrain system uses electric motors to replace the hydraulic pump and hydraulic motor in the traditional continuously variable transmission system to achieve a more efficient power transmission method. The power of the system comes from the engine. The engine does not directly drive the wheels, but drives the generator. The generator converts mechanical energy into electrical energy for the electric motors to use. At the same time, a power battery is provided to store excess electrical energy to further improve the energy utilization rate. During the operation of the tractor, the electric motors act together through the mechanical coupling device to achieve optimal distribution of power output.

[0031] The deep learning model in the control device predicts the future power demand based on the real-time operating state of the tractor, and dynamically allocates and adjusts the output power of the cooperating electric motors. In this way, when the operation resistance suddenly changes, the system can quickly adjust the power output to prevent impact damage to the transmission system, and at the same time improve the response speed and adaptability, enabling the tractor to more smoothly adapt to different operating environments.

[0032] In the above solution of this embodiment, the power is not directly output by the engine. The generator provided converts the power into electrical energy and then realizes power transmission through multiple electric motors, improving the flexibility of output energy regulation.

[0033] The number of electric motors can be set as needed. Specifically, asFigure 1 As shown, in this embodiment, three motors are introduced into the continuously variable transmission system of the tractor, including a generator 1, a first motor 2, and a second motor 3; the power output ends of the first motor 2 and the second motor 3 are coupled through a mechanical coupling device to provide driving force for the vehicle wheels;

[0034] In some embodiments, the power output end of at least one motor is connected to the power output shaft (PTO Out) of the vehicle operation end device. In this embodiment, two motors are provided, and the output end of the first motor 2 is connected to the power output shaft (PTO Out) for operation through a clutch CPTO;

[0035] Optionally, a power battery is arranged between the generator and the motor. The generator stores electrical energy in the power battery, and the power battery provides electrical energy for the motor;

[0036] In this embodiment, specifically, the power battery includes a first power battery 4 and a second power battery 5. The generator 1, the first power battery 4, and the first motor 2 are connected in sequence, and the generator 1, the second power battery 5, and the second motor 3 are connected in sequence;

[0037] A feasible technical solution is that the mechanical coupling device can adopt a planetary gear mechanism 6 to couple the power outputs of the first motor 2 and the second motor 3 through the planetary gear mechanism 6;

[0038] Specifically, for the planetary gear structure with two motors set as Figure 1 shown, it includes a sun gear 61, a planetary gear 62, a planetary carrier 63, and a ring gear 64; the power output end of the second motor 3 is connected to the ring gear 64, and a lock LOCK is arranged on the ring gear 64; the first motor 2 is connected to the sun gear 61 of the planetary gear structure through a gear structure, and the sun gear 61 and the ring gear 64 respectively transmit power to the planetary carrier 63 through the planetary gear 62, and the power is output through the planetary carrier 63;

[0039] In this embodiment, the power coupling of the two motors is achieved through the planetary gear mechanism 6. The planetary gear mechanism 6 is used to couple the power of the two motors. The motor that was originally used as a generator still has to undertake the power generation task and is still connected to the engine. It is necessary to keep the engine operating in the optimal operating range, which requires this generator to be in a relatively stable operating state and not change its output power constantly like the drive motor. In this way, the generator does not have the ability to adjust its own speed to adapt to the changes in the current working conditions. Moreover, the operating conditions of the tractor are sometimes complex. If there is only one drive motor, its continuously variable transmission ability may not be able to meet the requirements of the complex working conditions of the tractor. In this embodiment, the planetary gear mechanism 6 is used as a speed converging device to couple the power of the two motors, which can adjust the output speed more freely and flexibly and better realize the function of continuously variable transmission. The planetary gear mechanism 6 has the function of synthesizing and decomposing the torques provided by the two motors according to a specific ratio; this is a function that cannot be achieved by simply connecting the two motors in series or each acting on the wheels separately.

[0040] When the power demand is large, multiple motors provide power simultaneously. In order to maintain the relatively stable operating state of the motors, the planetary gear mechanism 6 adopted in this embodiment well converges the speeds and torques of the two motors to cope with complex working conditions.

[0041] Specifically, the first motor 2 is connected to the sun gear 61 of the planetary gear structure through a gear structure. The gear structure includes a first gear 9 and a second gear 8 that are meshed and connected.

[0042] In a specific implementation scheme, the output end of the planet carrier 63 of the planetary gear structure is connected to the drive shaft (CVT Out) of the vehicle wheel through a gear A7, as Figure 1 shown. In the figure, DR (Drive Reverse) and DF (Drive Forward) respectively represent the reverse gear (reverse drive) and forward gear (forward drive) of the transmission system.

[0043] The above-mentioned tractor powertrain system based on the improved ECVT can achieve one or more of the following working modes by controlling the working devices of the motor and the generator:

[0044] Working mode 1: Single-motor drive mode. The generator 1 and the first motor 2 work, and the second motor 3 does not work. The lock LOCK on the planetary gear mechanism 6 is in the locked state; the first motor 2 is used to drive the tractor wheels or / and the power output shaft of the working end equipment.

[0045] In the first working mode, when the stored power of the first power battery 4 used to supply power to the first motor 2 is lower than the set threshold, the generator starts to work to charge the first power battery 4; when there is also a load at the working end, it is provided by the first motor 2.

[0046] The first working mode is used to be turned on when the power requirement for the tractor's work is relatively low. The generator 1 and the first motor 2 are connected in series through the first power battery 4, and the second motor 3 does not work in this mode. The energy flow direction of this working mode is: from the engine to the generator 1 and then to the first power battery 4, the first power battery supplies energy to the first motor 2, and the first motor 2 drives the wheels and the PTO to work.

[0047] In the state of the first working mode, the engine does not directly bear the working load, so it can continuously operate in the optimal working range, which can significantly improve fuel economy. The power source of the drive motor that directly bears the working load is the power battery, and the power battery can achieve relatively stable power supply regardless of the working conditions of the tractor.

[0048] Working mode two: Dual-motor parallel drive mode. The generator 1, the first motor 2, and the second motor 3 all work, and the lock LOCK on the planetary gear mechanism 6 is in the open state; the first motor 2 and the second motor 3 jointly drive the wheels, or / and the first motor 2 drives the power output shaft connected to the working end equipment.

[0049] In the second working mode, when the power of the power battery is lower than the set threshold, the power battery is charged by the generator 1.

[0050] The second working mode is started when the power demand for the tractor's work is high. By releasing the lock LOCK, the second motor 3 can also be added to the driving work, that is, the parallel mode of the two motors. The generator 1 receives the power provided by the engine and charges the first power battery 4 and the second power battery 5, so that the driving work can be shared by the two motors to achieve the energy supply for the tractor's work requirements.

[0051] The energy flow direction in the state of the second working mode is: starting from the generator 1, it is divided into two parts. One part flows to the first power battery 4 (in the charging mode), and the other part flows to the second power battery 5 (in the charging mode). The first power battery 4 supplies power to the first motor 2, and the second power battery 5 supplies power to the second motor 3. The power of the first motor 2 and the second motor 3 is coupled through the planetary gear mechanism 6 and jointly drives the wheels. If there is a load on the PTO, it is still borne by the first motor 2.

[0052] Adopting this dual-motor parallel drive mode, the load is distributed to two drive motors, significantly reducing the burden on a single motor and ensuring the stability and reliability of the system under high-load conditions. The planetary gear mechanism 6 can synthesize and decompose the output torques of the two drive motors according to a specific ratio, ensuring a smoother and more efficient power output of the system under high-load conditions. The control device equipped with a deep learning model in this mode can predict the output power in real time according to the working environment of the tractor and dynamically adjust the output powers of the two motors, ensuring the adaptability of the system under different working conditions, reducing the impact on the motors and the gearbox, and extending the system life.

[0053] Working mode three, braking power generation mode:

[0054] The rotation of the wheel drives the rotation of the planet carrier 63, and the planet carrier 63 drives the planet gear 62 and the sun gear 61 to rotate in sequence, driving the first motor 2 to work and reverse through the gear structure, and charging the first power battery 4;

[0055] When the tractor brakes, if there is no hybrid power system, the energy consumed by braking will be completely wasted. However, the hybrid power system in this embodiment can recover a part of the braking energy. When the tractor brakes, the deceleration of the wheel will drive the reverse rotation of the planetary gear mechanism 6, which in turn drives the reverse rotation of the motor. At this time, the original drive motor will become a generator, and a part of the energy consumed by braking will be converted into electrical energy and stored in the power battery connected to the motor. In this way, a part of the energy lost during braking is recovered, the energy utilization efficiency is improved, the battery endurance mileage can be extended, and the optimization of the tractor energy management strategy can be realized.

[0056] During the uniform driving process of the tractor, the resistance overcome is usually non-linearly changing. In the braking power generation mode, the deep learning model of the control device plays a very important role. When the resistance suddenly becomes smaller, that is, the current driving force of the tractor is too large and the resistance is small, the deep learning model will predict the next change in the output power and adjust the actual operating power of the motor. However, if the tractor still does not quickly drop to the target vehicle speed, braking is required, and at this time, the braking power generation mode is activated. It can not only quickly reach the target vehicle speed but also improve the operating efficiency of the system, making the tractor driving safer and more energy-efficient.

[0057] Table 1 Operating conditions of components in each working mode;

[0058]

[0059] In the above embodiments, the three working modes set for the different working environments faced by the tractor can well address the problem of large variations in power demand when the tractor is working in the field. When in series, when the power battery is in the charging mode, the generator 1 supplies power to the power battery to provide a stable and lasting power source for the motor. In the parallel mode, the planetary gear mechanism 6 is used to enable the generator 1 to still cooperate well with the motor to complete stepless speed change under a relatively stable working state, and the total load of the tractor is shared by the two motors together. The switching between these two modes gives full play to the role of the three motors, and there is also a braking energy recovery mode that well improves the energy utilization rate of the tractor.

[0060] In some embodiments, the deep learning model can adopt a neural network model;

[0061] When the hybrid powertrain system is in the dual-motor parallel drive mode, the control device configured as a neural network model needs to control the dual motors, which involves the problem of energy distribution between the two motors. This mode is also the mode with the highest operating frequency of the control device. Taking this mode as an example, this design elaborates in detail how the control device works.

[0062] In some embodiments, the control device is configured to: predict the required power of the speed change system based on the deep learning model, dynamically distribute and adjust the output power of the mutually coupled motors based on the predicted required power, and achieve adaptive power output adjustment. This method can be executed in the working mode two state and includes the following steps:

[0063] Step 1: Obtain the tractor working state data and the images around the vehicle, and perform preprocessing;

[0064] Step 2: Transmit the preprocessed data to the deep learning model for prediction to obtain the predicted required power of the speed change system That is, the total power to be output;

[0065] Step 3: Construct an objective function for power distribution with the goal of minimizing the deviation between the output power and the required power and minimizing the sudden change in the battery discharge rate;

[0066] Step 4: Solve the objective function according to the obtained power battery parameters and the predicted required power to obtain a power distribution plan;

[0067] Further, when the power of the power battery is lower than the set threshold, start the engine and the generator 1 for charging;

[0068] In step 1, obtain the tractor working state data and the images around the vehicle;

[0069] Optionally, the tractor working state data may include vehicle speed, load, etc.;

[0070] Specifically, the tractor working state data is numerical data, and the preprocessing includes:

[0071] Step 11, data cleaning: handle missing values through filling, deletion or interpolation, and remove extreme outliers;

[0072] Step 12, normalize the data so that the data is converted to a value within 0 to 1;

[0073] A further technical solution is that the deep learning model includes two branches, including a parallel convolutional neural network (CNN) and a multi-layer perceptron (MLP). The output ends of the convolutional neural network (CNN) and the multi-layer perceptron (MLP) are connected to a splicing module and a fully connected layer; feature extraction is performed on the image data through the convolutional neural network, and feature extraction is performed on the preprocessed tractor working state data through the multi-layer perceptron; the extracted features are spliced, and then a fully connected operation is performed through the fully connected layer to perform data dimensionality reduction to obtain a prediction result.

[0074] Furthermore, the training process of the deep learning model is as follows:

[0075] Step S1, obtain the historical working state data of the tractor, the images around the vehicle, and the corresponding output power, construct a data set and divide it;

[0076] Specifically, divide the training set and the test set according to a set ratio;

[0077] Step S2, initialize the parameters of the neural network;

[0078] Specifically, set parameters such as the number of nodes in the input layer, hidden layer and output layer of the neural network, and the learning rate of the neural network. Optionally, the learning rate can be set to 0.1;

[0079] Step S3, use the historical working state data of the tractor and the images around the vehicle as inputs, and the output power as the output, and use the training set data to iteratively train the deep learning model, and adjust the network parameters through the backpropagation algorithm until the set iteration cut-off condition is met to obtain a trained deep learning model;

[0080] During the training process, the change of the loss function and the prediction performance on the validation set can be monitored to prevent the situation of overfitting of the data from causing too poor training effect.

[0081] Furthermore, use the test set to verify and evaluate the trained deep learning model.

[0082] Evaluation metrics can include prediction error, accuracy, recall, etc.; if the gap between the results calculated by the model and the actual data is still too large, methods such as adjusting the network structure, changing the activation function, and optimizing the algorithm can be used to further optimize the network model. The trained deep learning model is applied to tractor control, and data is collected again after each operation to correct the deep learning model.

[0083] In the dual-motor parallel drive mode, the control objective of the control device is to dynamically allocate the output power according to the real-time load demand, so that the output power P2 of the first motor 2 and the output power P3 of the second motor 3 can quickly match to meet the total demand for output power, avoiding power supply delay or insufficient power.

[0084] The required power to be output predicted based on the deep model includes wheel drive power and operation drive power. The wheel drive power is expressed as , and the operation drive power of the operation end device is expressed as , and the required power output is the sum of the two:

[0085] ;

[0086] The state of charge of the first power battery 4 is expressed as SOC4, and the state of charge of the second power battery 5 is expressed as SOC5. The state of charge is used to reflect the remaining power of the two batteries and is used to constrain the discharge capacity; the instantaneous maximum discharge power of the first power battery 4 is expressed as , and the instantaneous maximum discharge power of the second power battery 5 is expressed as ;

[0087] In step 3, the objective function for power distribution construction is:

[0088] In step 3, the objective function for power distribution construction is:

[0089] ;

[0090] ;

[0091] ;

[0092] ;

[0093] ;

[0094] ;

[0095] Among them, is the discharge power ratio of the first motor 2, is the discharge power ratio of the second motor 3; Output power allocated to the first motor 2; Output power allocated to the second motor 3; Predicted total demand power; And Motor power correction factors for the first motor 2 and the second motor 3 respectively: Is the power tracking weight, which gives priority to ensuring the motor power demand and its value is dynamically adjusted according to the load change rate; Is the discharge smoothness weight, which suppresses the sudden change of the battery discharge power and prevents the motor efficiency from decreasing due to voltage fluctuation.

[0096] After the neural network predicts that the next output power demand changes, it will adjust And Parameters. For example, if the PTO demand power increases, Increases, that is, the first motor 2 responsible for driving the PTO undertakes more power and the wheel driving resistance is greater. By adjusting That is, the second motor 3 undertakes more power;

[0097] Solving the above objective function and dynamically adjusting And When the magnitudes are such that the sum of the two terms in the formula brackets is minimized as much as possible to achieve dynamic optimization of the output power. The constructed objective function for power distribution includes:

[0098] The first term is to minimize the deviation between the actual output power of the motor and the demand power;

[0099] The second term is to suppress the sudden change of the battery discharge rate and avoid the instability of motor control caused by a sudden drop in voltage.

[0100] Furthermore, the constraint conditions include:

[0101] 1) Meeting the constraints of the power battery discharge capacity:

[0102] ;

[0103] ;

[0104] Among them, , Are the maximum instantaneous discharge powers of the first power battery 4 and the second power battery 5, And Are the discharge efficiency coefficients of the first power battery 4 and the second power battery 5 respectively, and C4 and C5 are the rated capacities of the first power battery 4 and the second power battery 5 respectively.

[0105] 2) Safety constraints for power battery discharge, i.e., when the state of charge of the power battery is lower than the set safety threshold, stop discharging:

[0106] SOC4≥20%, SOC5≥20%;

[0107] In the objective function constructed in the above solution of this embodiment, the motor power correction coefficient is added and , which can compensate for the dynamic delay of battery discharge. The sudden increase in battery discharge will cause a slight voltage drop, resulting in a deviation in the actual output power, and can further adjust the accuracy of output power distribution.

[0108] Furthermore, it also includes a method for solving the objective function, including the following steps:

[0109] Step 31: Parameter initialization: According to the power battery parameters obtained in real time and the predicted total power demand , initialize the proportionality coefficients and , set the initial weights and , and determine the reference values of the motor power correction coefficients and ; among them, the power battery parameters include the state of charge SOC4 of the first power battery 4, the state of charge SOC5 of the second power battery 5, and the maximum instantaneous discharge power of the battery and , specifically:

[0110] (1) Based on the current SOC state of the power battery, proportionally allocate the initial values of and ;

[0111] Specifically, if the remaining power of the first power battery 4 is more, the initial value is set higher;

[0112] (2) Set the initial weights and according to the load change rate;

[0113] Optionally, when the load change rate is greater than the set value, is set to 0.9, is set to 0.1;

[0114] (3) Calculate the motor power correction coefficients and according to the battery discharge efficiency and , specifically:

[0115] ;

[0116] ;

[0117] Among them, and are efficiency compensation coefficients.

[0118] Step 32: Construct a dynamic optimization model: Transform the objective function and constraint conditions into an optimization problem with constraints, and use the Lagrange multiplier method to construct an augmented objective function:

[0119] ;

[0120] Among them, is the Lagrange multiplier for power balance constraint, is the multiplier for inequality constraints (such as battery discharge capacity, SOC safety threshold), is the corresponding constraint function.

[0121] In this embodiment, the augmented objective function set considers the constraints in advance in the solution of the objective function. Specifically, in the last two terms of the function, the augmented objective function is used as the object of positive optimization, and the constraint conditions are considered in advance, which can prevent the final solution value of the objective function from not satisfying the constraint conditions. In addition, a correction coefficient is added, taking into account the slight change in voltage caused by the change in battery discharge amount, further improving the accuracy of the solution.

[0122] Step 33: Use the gradient descent method to iteratively solve and adaptively adjust the power tracking weight, discharge smoothness weight, and motor power correction coefficient, and perform constraint verification and feasibility correction to obtain the solution result; specifically:

[0123] 1) Gradient descent iteration: Calculate the partial derivatives of the objective function with respect to and , and update the parameters along the negative gradient direction:

[0124] ;

[0125] ;

[0126] Among them, η is the learning rate, which is dynamically adjusted by the Armijo criterion to accelerate convergence.

[0127] The Armijo criterion is a strategy for dynamically adjusting the step size, commonly used in the gradient descent method. The purpose is to select a suitable step size in each step to ensure sufficient descent of the objective function, thereby accelerating the convergence speed and avoiding inefficiencies caused by oscillations or too small step sizes.

[0128] 2) Weight dynamic adjustment: Online adjust the power tracking weight and the discharge smoothness weight according to the real-time load change rate and the battery discharge stability index. If the power deviation :

[0129] exceeds the set threshold, increase and decrease . ;

[0130] If the mutation term of the discharge rate exceeds the set threshold, increase and decrease .

[0131] 3) Correction coefficient compensation: Update the motor power correction coefficient according to the real-time feedback of the battery voltage fluctuation and :

[0132] ;

[0133] where , is the deviation between the battery voltage and the rated value, and is the proportional coefficient.

[0134] 4) Conduct constraint verification and feasibility correction. After each iteration, verify whether the solution meets the constraint conditions:

[0135] If P4 or P5 exceeds the instantaneous discharge capacity of the battery, reduce and proportionally to ensure that:

[0136] ;

[0137] ;

[0138] If SOC4 or SOC5 is lower than 20%, force to stop discharging and trigger the generator charging mode;

[0139] If α + β ≠ 1, normalize and to ensure power balance.

[0140] Furthermore, based on the actual tractor power system, conduct real-time feedback and update:

[0141] Feed the solution result back to the control system to drive the motor to perform power distribution. At the same time, store the current working condition data (images, loads, battery status, etc.) and the distribution result in the historical database, and update the deep learning model parameters using the incremental learning algorithm every set period (such as 24 hours) to improve the prediction accuracy.

[0142] In this embodiment, based on the wheel drive power demand predicted in real time and the PTO load power , combined with the state of charge (SOC) and maximum discharge capacity of the first power battery 4 and the second power battery 5, a power distribution ratio coefficient is generated through a dynamic optimization algorithm α and β , to ensure that the total output powers P2 and P3 accurately track the required power under the combined drive mode . Using the weight coefficients and<> to dynamically balance the power tracking accuracy and the smoothness of the discharge rate: when the load suddenly changes, the priority is increased to reduce the power deviation, while the discharge mutation is suppressed to ensure the system stability. At the same time, the motor power correction coefficients and are introduced to correct the motor power output in real time, offsetting the power loss caused by the battery terminal voltage fluctuation. By restricting the battery SOC safety limit (≥20%) and the power balance condition (α + β = 1), the adaptive energy distribution under the dual-motor parallel drive mode can be achieved. On the premise of ensuring the drive response speed, the battery discharge process is maintained smoothly, avoiding the dynamic performance deterioration caused by the power distribution mismatch.

[0143] During use, when the tractor is about to climb a steep slope, the tractor suddenly needs to increase the power from 30 kW to 60 kW. After capturing the image of the current scene, the neural network algorithm is used to predict the upcoming change in power demand of the tractor, that is, a higher wheel drive power is required , then the proportional distribution is controlled. For example, through optimization calculation, is increased from 0.5 to 0.8, making the output power of the first motor 2 higher; is increased from 0.5 to 0.9 to give priority to ensuring sufficient power; is decreased from 0.5 to 0.1 to allow the battery discharge to mutate. In addition, the sudden increase in the discharge amount of the first power battery 4 will cause a slight voltage drop. Then, is increased from 1.0 to 1.05, which can supplement 5% of the power to the first motor 2. Through the cooperation of these three groups of parameters: and , and , and , the controller not only ensures the power response but also protects the stability and safety of the battery and the motor during operation.

[0144] Further technical solution: when the reduction of the driving resistance of the tractor is greater than the set value and the target vehicle speed is not reached within the set time after the resistance reduction, the braking power generation mode is started;

[0145] For the other two working modes, the single-motor drive mode and the dual-motor parallel drive mode, the switching can be realized according to the manual switching of the driver.

[0146] In this embodiment, a control device based on a neural network algorithm is set up to enable the output power of the motor to better adapt to the changes in road conditions, predict the changes in the working conditions of the tractor, enable the driving motor to make preparations for the next working environment, and reduce the impact on the motor caused by the sudden changes in the tractor environment.

[0147] Embodiment 2

[0148] Based on Embodiment 1, a control method for the powertrain system of a tractor based on an improved ECVT is provided in this embodiment. As Figure 2 shown, it can be implemented in the control device and includes the following steps:

[0149] Step 1: According to the obtained control instruction, the control system works in the corresponding working mode;

[0150] Step 2: In the dual-motor parallel drive mode, obtain the working state data of the tractor and the images of its surrounding environment, and perform preprocessing;

[0151] Step 3: Transmit the preprocessed data to the deep learning model for prediction to obtain the required power of the predicted transmission system;

[0152] Step 4: With the goal of minimizing the deviation between the output power and the required power and minimizing the sudden change in the battery discharge rate, construct an objective function for power distribution;

[0153] Step 5: According to the obtained power battery parameters and the predicted required power, solve the objective function to obtain a power distribution plan;

[0154] In Step 1, the control instruction can be an operation instruction issued by the driver, which is directly used to select the working mode. When the system receives the instruction, it directly controls the corresponding components to act and execute the corresponding working mode, including Working Mode 1, Working Mode 2, and Working Mode 3;

[0155] In Step 2, obtain the working state data of the tractor and the images around the vehicle;

[0156] Optionally, the working state data of the tractor may include vehicle speed, load, etc.;

[0157] Specifically, the working state data of the tractor is numerical data, and the preprocessing includes:

[0158] Step 21, Data cleaning: Handle missing values by filling, deleting, or interpolation, and remove extreme outliers;

[0159] Step 22, Normalize the data so that the data is converted to a value within 0 to 1;

[0160] A further technical solution is that the deep learning model includes two branches, including a parallel convolutional neural network (CNN) and a multi-layer perceptron (MLP). The output ends of the convolutional neural network (CNN) and the multi-layer perceptron (MLP) are connected to a splicing module and a fully connected layer; feature extraction is performed on the image data through the convolutional neural network, and feature extraction is performed on the preprocessed tractor working state data through the multi-layer perceptron; the extracted features are spliced, and then a fully connected operation is performed through the fully connected layer to perform data dimensionality reduction to obtain a prediction result.

[0161] Furthermore, the training process of the deep learning model is as follows:

[0162] Step S1, Obtain the historical working state data of the tractor, the images around the vehicle, and the corresponding output power, construct a data set and divide it;

[0163] Specifically, divide the training set and the test set according to a set ratio;

[0164] Step S2, Initialize the parameters of the neural network;

[0165] Specifically, set the number of nodes in the input layer, hidden layer, and output layer of the neural network, as well as parameters such as the learning rate of the neural network. Optionally, the learning rate can be set to 0.1;

[0166] Step S3, Use the historical working state data of the tractor and the images around the vehicle as inputs, and the output power as the output. Iteratively train the deep learning model with the training set data, and adjust the network parameters through the backpropagation algorithm until the set iteration cut-off condition is met to obtain a trained deep learning model;

[0167] During the training process, the change of the loss function and the prediction performance on the validation set can be monitored to prevent the situation of overfitting of the data from causing too poor training effect.

[0168] Furthermore, use the test set to verify and evaluate the trained deep learning model.

[0169] Evaluation metrics can include prediction error, accuracy, recall, etc.; if the gap between the result calculated by the model and the actual data is still too large, methods such as adjusting the network structure, changing the activation function, and optimizing the algorithm can be used to further optimize the network model. The trained deep learning model is applied to tractor control, and data is collected again after each operation to correct the deep learning model.

[0170] In the dual-motor parallel drive mode, the control objective of the control device is: according to the real-time load demand, dynamically allocate the output power so that the output power P2 of the first motor 2 and the output power P3 of the second motor 3 can quickly match to meet the total demand of the output power, avoiding power supply delay or insufficient power.

[0171] The required power to be output predicted based on the deep model includes the wheel drive power and the operation drive power. The wheel drive power is expressed as , and the operation drive power of the operation end device is expressed as , and the required power output is the sum of the two:

[0172] ;

[0173] The state of charge of the first power battery 4 is expressed as SOC4, and the state of charge of the second power battery 5 is expressed as SOC5. The state of charge is used to reflect the remaining power of the two batteries and is used to constrain the discharge capacity; the instantaneous maximum discharge power of the first power battery 4 is expressed as , and the instantaneous maximum discharge power of the second power battery 5 is expressed as ;

[0174] In step 4, the objective function for power distribution construction is:

[0175] In step 3, the objective function for power distribution construction is:

[0176] ;

[0177] ;

[0178] ;

[0179] ;

[0180] ;

[0181] ;

[0182] Among them, is the discharge power ratio of the first motor 2, is the discharge power ratio of the second motor 3; Output power allocated to the first motor 2; Output power allocated to the second motor 3; Predicted total required power; And Motor power correction factors for the first motor 2 and the second motor 3 respectively: Is the power tracking weight, which gives priority to ensuring the motor power demand, and its value is dynamically adjusted according to the load change rate; Is the discharge smoothness weight, which suppresses the sudden change of the battery discharge power and prevents the motor efficiency from decreasing due to voltage fluctuation.

[0183] After the neural network predicts that the next output power demand changes, it will adjust And Parameters. For example, if the PTO required power increases, Increases, that is, the first motor 2 responsible for driving the PTO undertakes more power, and the wheel driving resistance is greater. By adjusting That is, the second motor 3 undertakes more power;

[0184] Solving the above objective function and dynamically adjusting And When the magnitudes are adjusted, the sum of the two terms in the formula brackets should be minimized as much as possible to achieve dynamic optimization of the output power. The constructed objective function of power distribution includes:

[0185] The first term is to minimize the deviation between the actual output power of the motor and the required power;

[0186] The second term is to suppress the sudden change of the battery discharge rate and avoid the instability of motor control caused by voltage drop.

[0187] Furthermore, the constraint conditions include:

[0188] 1) Meet the constraints of the power battery discharge capacity:

[0189] ;

[0190] ;

[0191] Among them, , Are the maximum instantaneous discharge powers of the first power battery 4 and the second power battery 5, And Are the discharge efficiency coefficients of the first power battery 4 and the second power battery 5 respectively, and C4 and C5 are the rated capacities of the first power battery 4 and the second power battery 5 respectively.

[0192] 2) Safety constraints for power battery discharge, that is, when the state of charge of the power battery is lower than the set safety threshold, stop discharging:

[0193] SOC4≥20%, SOC5≥20%;

[0194] In the objective function constructed in the above solution of this embodiment, a motor power correction coefficient is added and , which can compensate for the dynamic delay of battery discharge. The sudden increase in battery discharge will cause a slight drop in voltage, resulting in a deviation in the actual output power, and can further adjust the accuracy of output power distribution.

[0195] Furthermore, the method for solving the objective function in step 5 includes the following steps:

[0196] Step 51: Parameter initialization: According to the power battery parameters obtained in real time and the predicted total power demand , initialize the proportionality coefficients and , set the initial weights and , and determine the reference values of the motor power correction coefficients and ; where the power battery parameters include the state of charge SOC4 of the first power battery 4, the state of charge SOC5 of the second power battery 5, and the maximum instantaneous discharge power of the battery and , specifically:

[0197] (1) Based on the current SOC state of the power battery, proportionally allocate the initial values of and ;

[0198] Specifically, if the remaining power of the first power battery 4 is more, the initial value is set higher;

[0199] (2) Set the initial weights and according to the load change rate;

[0200] Optionally, when the load change rate is greater than the set value, is set to 0.9, is set to 0.1;

[0201] (3) Calculate the motor power correction coefficients and according to the battery discharge efficiency and , specifically:

[0202] ;

[0203] ;

[0204] Among them, and are efficiency compensation coefficients.

[0205] Step 52: Construct a dynamic optimization model: Convert the objective function and constraints into an optimization problem with constraints, and use the Lagrange multiplier method to construct an augmented objective function:

[0206] ;

[0207] Among them, is the Lagrange multiplier of the power balance constraint, is the multiplier of the inequality constraints (such as battery discharge capacity, SOC safety threshold), is the corresponding constraint function.

[0208] In this embodiment, the augmented objective function set considers the constraints in advance in the solution of the objective function. Specifically, in the last two terms of the function, the augmented objective function is used as the object of positive optimization, and the constraint conditions are considered in advance, which can prevent the final solution value of the objective function from not satisfying the constraint conditions. In addition, a correction coefficient is added, taking into account the slight change in voltage caused by the change in battery discharge amount, further improving the accuracy of the solution.

[0209] Step 53: Use the gradient descent method to iteratively solve and adaptively adjust the power tracking weight, discharge smoothness weight, and motor power correction coefficient, and perform constraint verification and feasibility correction to obtain the solution result; specifically:

[0210] 1) Gradient descent iteration: Calculate the partial derivatives of the objective function with respect to and , and update the parameters along the negative gradient direction:

[0211] ;

[0212] ;

[0213] Among them, η is the learning rate, which is dynamically adjusted by the Armijo criterion to accelerate convergence.

[0214] [[ID=5|4]]2) Weight dynamic adjustment: According to the real-time load change rate and battery discharge stability index, online adjust the power tracking weight and the discharge smoothness weight :

[0215] If the power deviation exceeds the set threshold, then increase , decrease ;

[0216] If the discharge rate mutation term exceeds the set threshold, then increase , and decrease .

[0217] 3) Correction coefficient compensation: According to the real-time feedback of the battery voltage fluctuation, update the motor power correction coefficients and :

[0218] ;

[0219] Among them, , is the deviation between the battery voltage and the rated value, and is the proportional coefficient.

[0220] 4) Conduct constraint verification and feasibility correction. After each iteration, verify whether the solution satisfies the constraint conditions:

[0221] If P4 or P5 exceeds the instantaneous discharge capacity of the battery, then proportionally reduce and , ensuring satisfaction of:

[0222] ;

[0223] ;

[0224] If SOC4 or SOC5 is lower than 20%, then force the discharge to stop and trigger the generator charging mode;

[0225] If α + β ≠ 1, then perform normalization processing on and to ensure power balance.

[0226] Furthermore, based on the actual tractor power system, conduct real-time feedback and update:

[0227] Feed the solution result back to the control system to drive the motor to perform power distribution. At the same time, store the current working condition data (images, loads, battery status, etc.) and the distribution result in the historical database, and update the deep learning model parameters using the incremental learning algorithm every set period (such as 24 hours) to improve the prediction accuracy.

[0228] In this embodiment, based on the real-time predicted wheel drive power demand and the PTO load power , combined with the state of charge (SOC) and the maximum discharge capacity of the first power battery 4 and the second power battery 5, generate the power distribution proportional coefficients α and β, ensure that the total output powers P2 and P3 accurately track the required power under the combined drive mode. . Utilize the weight coefficients and to dynamically balance the power tracking accuracy and the discharge rate smoothness: When the load suddenly changes, the priority is increased to narrow the power deviation, then the discharge mutation is suppressed to ensure the system stability. Meanwhile, introduce the motor power correction coefficients and to correct the motor power output in real time and offset the power loss caused by the battery terminal voltage fluctuation. By restricting the battery SOC safety limit (≥20%) and the power balance condition (α + β = 1), the adaptive energy distribution under the dual-motor parallel drive mode can be achieved. On the premise of ensuring the drive response speed, the battery discharge process is maintained smoothly, and the dynamic performance deterioration caused by the power distribution mismatch is avoided.

[0229] During use, when the tractor is about to climb a steep slope, the tractor suddenly needs to increase the power from 30 kW to 60 kW. After capturing the image of the current scene, use the neural network algorithm to predict the upcoming power demand change of the tractor, that is, a higher wheel drive power is required , then control the proportional distribution, such as by optimizing the calculation to increase from 0.5 to 0.8, so that the output power of the first motor 2 is higher; increase from 0.5 to 0.9 to give priority to ensuring sufficient power; decrease from 0.5 to 0.1 to allow the battery discharge to mutate. In addition, the sudden increase in the discharge amount of the first power battery 4 will cause a slight voltage drop. Then, increase from 1.0 to 1.05 to supplement 5% of the power to the first motor 2. Through these three groups of parameters: and , and , and 's cooperation, the controller not only ensures the power response but also protects the stability and safety of the battery and motor operation.

[0230] Further technical solution, when the reduction of the tractor driving resistance is greater than the set value and the target vehicle speed is not reached within the set time after the resistance reduction, start the braking power generation mode;

[0231] For the other two working modes, the single-motor drive mode and the dual-motor parallel drive mode, they can be switched according to the driver's manual switching.

[0232] In this embodiment, a control device based on a neural network algorithm is set up to enable the output power of the motor to better adapt to changes in road conditions, predict changes in the working conditions of the tractor, enable the driving motor to make preparations for the next working environment, and reduce the impact on the motor caused by sudden changes in the tractor environment.

[0233] The above are only the preferred embodiments of the present disclosure and are not used to limit the present disclosure. For those skilled in the art, the present disclosure can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present disclosure shall be included within the protection scope of the present disclosure.

[0234] Although the specific implementation manners of the present disclosure have been described above in conjunction with the accompanying drawings, they do not limit the protection scope of the present disclosure. Those skilled in the art should understand that, based on the technical solutions of the present disclosure, various modifications or deformations that can be made by those skilled in the art without creative efforts are still within the protection scope of the present disclosure.

Claims

1. The tractor powertrain system based on the improved ECVT is characterized in that: It includes a speed change system and a control device. The speed change system includes a generator, a mechanical coupling device, and at least two motors. The generator is used to supply electrical energy to the motors, and the power outputs of the motors are coupled through the mechanical coupling device. It is set to two motors, including a first motor and a second motor. The control device is configured to: predict the required power of the speed change system based on a deep learning model, dynamically allocate and adjust the output powers of the mutually coupled motors based on the predicted required power, and achieve adaptive power output regulation. The control device predicts the output power based on a deep learning model, dynamically allocates and adjusts the output powers of the mutually coupled motors based on the predicted power, and achieves adaptive power output regulation, including the following steps: Obtain the tractor working state data and the images around the vehicle, and perform preprocessing. Transmit the preprocessed data to the deep learning model for prediction to obtain the predicted required power of the speed change system. Construct an objective function for power distribution with the goal of minimizing the deviation between the output power and the required power and minimizing the sudden change in the battery discharge rate. Among them, the constructed objective function for power distribution is: ； ; ; ; ; ; where, is the discharge power ratio of the first motor, is the discharge power ratio of the second motor; is the output power allocated to the first motor; is the output power allocated to the second motor; is the predicted total demand power; and are the motor power correction factors of the first motor and the second motor respectively: is the power tracking weight, which gives priority to ensuring the motor power demand, and its value is dynamically adjusted according to the load change rate; is the discharge smoothness weight, which suppresses the sudden change of the battery discharge power and prevents the motor efficiency from decreasing due to voltage fluctuation; Solve the objective function according to the obtained power battery parameters and the predicted required power to obtain a power distribution scheme. During the solution process of the objective function, correction coefficient compensation: Update the motor power correction coefficient according to the real-time feedback of battery voltage fluctuations and : ； Among them, , is the deviation of the battery voltage from the rated value, is the proportionality coefficient.

2. The tractor powertrain system based on the improved ECVT as claimed in claim 1, wherein: The mechanical coupling device adopts a planetary gear mechanism, and the power outputs of the first motor and the second motor are coupled through the planetary gear mechanism.

3. The tractor powertrain system based on the improved ECVT as described in claim 2, characterized in that: The planetary gear structure includes a sun gear, planetary gears, a planetary carrier, and a ring gear; the power output end of the second motor is connected to the ring gear, and a lock LOCK is provided on the ring gear; the first motor is connected to the sun gear of the planetary gear structure, and the sun gear and the ring gear respectively transmit power to the planetary carrier through the planetary gears and output power through the planetary carrier.

4. The tractor powertrain system based on the improved ECVT as claimed in claim 3, wherein: The speed change system includes the following working modes: Single-motor drive mode: The generator and the first motor work, the second motor does not work, and the lock LOCK on the planetary gear mechanism is in a locked state; the first motor is used to drive the tractor wheels or / and the power output shaft of the working end equipment. Dual-motor parallel drive mode: The generator, the first motor, and the second motor all work, and the lock LOCK on the planetary gear mechanism is in an open state; the first motor and the second motor jointly drive the wheels or / and the first motor drives the power output shaft connected to the working end equipment. Regenerative braking mode: The rotation of the wheels drives the rotation of the planetary carrier, the planetary carrier drives the planetary gears and the sun gear to rotate in sequence, and drives the first motor to work and reverse through the gear structure to charge the first power battery.

5. The tractor powertrain system based on the improved ECVT as described in claim 1, characterized in that: The constructed objective function for power distribution includes: The first term is to minimize the deviation between the actual output power of the motor and the required power. The second term is to suppress the sudden change in the battery discharge rate and avoid the instability of motor control caused by a sudden drop in voltage. Constraint conditions, including: satisfying the constraint of the power battery discharge capacity and the safety constraint of the power battery discharge.

6. The control method of the tractor powertrain system based on the improved ECVT according to any one of claims 1-5, characterized in that, It includes the following steps: According to the obtained control instructions, the control system operates in the corresponding working mode; When in the dual-motor parallel drive mode, the working state data of the tractor and the images around the vehicle are acquired and preprocessed; The preprocessed data is transmitted to the deep learning model for prediction to obtain the required power of the predicted transmission system; With the goal of minimizing the deviation between the output power and the required power and minimizing the mutation of the battery discharge rate, an objective function for power distribution is constructed; According to the obtained power battery parameters and the predicted required power, the objective function is solved to obtain a power distribution scheme.

7. The control method according to claim 6, wherein: The constructed objective function for power distribution includes: The first term is to minimize the deviation between the actual output power of the motor and the required power; The second term is to suppress the mutation of the battery discharge rate and avoid the instability of motor control caused by a sudden voltage drop; The constraint conditions include: the constraint of meeting the discharge capacity of the power battery and the safety constraint of the power battery discharge.

8. The control method according to claim 6, characterized in that, It includes the following steps: The deep learning model includes two branches, including a parallel convolutional neural network and a multi-layer perceptron. The output ends of the convolutional neural network and the multi-layer perceptron are connected to a splicing module and a fully connected layer; feature extraction is performed on the image data through the convolutional neural network, and feature extraction is performed on the preprocessed tractor working state data through the multi-layer perceptron; the extracted features are spliced, and then a fully connected operation is performed through the fully connected layer for data dimensionality reduction to obtain a prediction result.

9. The control method according to claim 6, characterized in that It includes the following steps: when the reduction in the driving resistance of the tractor is greater than the set value and the target vehicle speed is not reached within the set time after the resistance reduction, the braking power generation mode is started.

Images