An electric tractor's electronically controlled suspension system and its suspension height control method

Abstract

This invention belongs to the field of electric tractor technology, specifically relating to an electric tractor's electronically controlled suspension system and its suspension height control method. The invention provides an electric tractor's electronically controlled suspension system and its suspension height control method. The electronically controlled suspension system of this invention has a compact structure and high reliability, employing an electric push rod to drive the suspension mechanism. For this novel electronically controlled suspension system, a comprehensive intelligent suspension height control method based on digital twins, taking into account multiple factors, is designed. This overcomes the problems of high energy consumption, poor impact resistance, and high failure rate of existing tractor suspension systems during tillage, as well as the operational quality and agronomic level issues caused by poor tillage depth.

Description

Technical Field

[0001] This invention belongs to the field of electric tractor technology, specifically relating to an electric tractor electronically controlled suspension system and its suspension height control method. Background Technology

[0002] Electric tractors have opened up new avenues for the development of green agricultural machinery technology and will play a crucial role in future agricultural production. Traditional fuel-powered tractors typically use hydraulic control for their suspension systems. However, limitations in the response speed and precision of hydraulic components make precise control of the suspension system difficult. Furthermore, hydraulic suspension systems are structurally complex, requiring numerous hydraulic components and pipelines, which increases manufacturing costs, system failure rates, and maintenance costs. These drawbacks make it difficult to directly apply hydraulic suspension systems to electric tractors.

[0003] Electric tractors employing electric suspension systems offer unique advantages. Powered by the tractor's battery pack, they eliminate the need for an additional hydraulic system, reducing redundancy and complexity. The lifting force is provided by a separate power unit, independent of the drive system, ensuring uninterrupted traction during operation. Furthermore, electric suspension systems enable faster response times, allowing for quicker and smoother adjustments when attaching implements, and operate without noise.

[0004] Currently, the most commonly used methods for controlling tillage depth in tractor suspension systems include position adjustment, force adjustment, and a combination of force and position adjustment. Electric suspension systems are complex nonlinear systems influenced by various factors. In actual tillage operations, factors such as tillage position, tillage resistance, soil moisture, ambient temperature, soil type, equipment parameters, and performance all affect work quality and efficiency. Under these circumstances, traditional force and position adjustment methods cannot effectively guarantee work quality. Therefore, considering the excellent response speed of electric suspension systems, there is an urgent need to develop an intelligent suspension height control method that comprehensively considers multiple factors. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an electric tractor's electronically controlled suspension system and its suspension height control method. The electronically controlled suspension system of this invention features a compact structure and high reliability, employing an electric push rod to drive the suspension mechanism. For this novel electronically controlled suspension system, a digital twin-based intelligent suspension height control method considering multiple factors is designed. This overcomes the problems of high energy consumption, poor impact resistance, and high failure rate of existing tractor suspension systems during tillage, as well as the operational quality and agronomical issues caused by poor tillage depth.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] An electric control suspension system for an electric tractor, comprising a low-voltage servo motor 1, a lower baffle 2, a frame 3, an upper baffle 4, a first cylindrical pin 5, an electric push rod 6, an "L"-shaped sliding frame 7, a tie rod 8, a connecting pin 9, a suspension frame 10, a GPS locator 11, a main controller 12, a tillage depth controller 13, an electric push rod driver 14, a second cylindrical pin 15, a connecting pin 16, a pin shaft force sensor 17, and a reduction mechanism 18;

[0008] The frame 3 is a rectangular frame structure, including a left column and a right column arranged on the left and right sides, and a top plate arranged between the upper ends of the left column and the right column.

[0009] The “L”-shaped sliding frame 7 includes a bottom rectangular frame, a left upright beam, and a right upright beam. The bottom rectangular frame includes a front beam, a rear beam, a left side beam, and a right side beam. The left and right upright beams are perpendicular to the bottom rectangular frame and extend upwards. The bottom of the left and right upright beams are connected to the rear ends of the left and right side beams. The front ends of the left and right side beams are connected to the lower end of the suspension frame 10.

[0010] The upper ends of the two diagonal tie rods 8 are connected to the upper end of the suspension frame 10 by a hook pin 9, and the lower ends of the two diagonal tie rods 8 are respectively hinged to the rear of the left and right beams of the "L"-shaped sliding frame 7 by a connecting pin 16; the pin force sensor 17 is set at the connection between the left and right beams of the "L"-shaped sliding frame 7 and the suspension frame 10.

[0011] The upper end of the electric push rod 6 is hinged to the top plate of the frame 3 via the first cylindrical pin 5, and the lower end of the electric push rod 6 is hinged to the rear beam of the "L"-shaped sliding frame 7 via the second cylindrical pin 15.

[0012] The electric actuator 6 has a built-in magnetostrictive displacement sensor 23;

[0013] The lower baffle 2 is located in front of the frame 3 and is connected to the left and right pillars of the frame 3 respectively; the upper baffle 4 is located in front of the frame 3 and is connected to the left and right pillars of the frame 3 respectively; the upper baffle 4 is located above the lower baffle 2.

[0014] The left and right upright beams of the “L”-shaped sliding frame 7 are located between the left and right uprights of the frame 3 and the lower baffle 2 and the upper baffle 4.

[0015] The low-voltage servo motor 1 and the reduction mechanism 18 are connected to the frame 3 in sequence;

[0016] GPS locator 11 is installed on the vehicle body and connected to main controller 12 via cable;

[0017] The main controller 12 collects data from the GPS positioning device 11, the built-in magnetostrictive displacement sensor 23 of the electric push rod 6, and the pin force sensor 17 through a dedicated signal line to obtain information on the tillage position and the current status of the electric suspension system.

[0018] Data transceiver 24 is connected to main controller 12; data transceiver 24 remotely transmits and receives data collected by main controller 12;

[0019] The main controller 12, tillage depth controller 13, electric actuator driver 14 and electric actuator 6 are connected in sequence.

[0020] The electronically controlled suspension system also includes a long nylon pad 19 and a short nylon pad 22; the long nylon pad 19 is fixed to the front surfaces of the left and right pillars of the frame 3; the short nylon pad 22 is fixed to the rear surfaces of the lower baffle 2 and the upper baffle 4.

[0021] The electronically controlled suspension system also includes a bend plate 20 and a rolling bearing 21. The bend plate 20 is fixed on the left and right pillars of the frame 3, and the rolling bearing 21 is installed on the bend plate 20. The bearing 21 contacts the left and right beams of the "L"-shaped sliding frame 7, and plays a guiding and limiting role to ensure that the "L"-shaped sliding frame 7 rises and falls smoothly in the vertical direction.

[0022] Among them, the extension and retraction of the electric actuator 6 can drive the "L"-shaped sliding frame 7 to achieve lifting and lowering within a range of 37cm.

[0023] A method for controlling the suspension height of an electric tractor's electronically controlled suspension system, wherein the method includes the following steps:

[0024] Step S1: Establish a digital twin system, which includes an electronically controlled suspension system, farmland monitoring equipment, a digital twin, twin data, a digital twin service platform, and the connections between the various parts of the digital twin system;

[0025] The electric suspension system includes a magnetostrictive displacement sensor 23, a GPS locator 11, a pin force sensor 17, an electric actuator 6 as a suspension height actuator, and a data transceiver 24 built into the electric actuator 6;

[0026] The GPS locator 11 can measure the tractor's location and tillage position; the pin force sensor 17 can measure tillage resistance; the electric push rod 6 can perform lifting and lowering actions; and the data transceiver 24 can complete remote transmission and reception of information.

[0027] Farmland monitoring equipment collects weather and soil information in real time and transmits the weather and soil information to a digital twin service platform;

[0028] The weather conditions information includes rain / snow information and temperature;

[0029] The soil information includes soil moisture, soil temperature, soil type, and soil hardness;

[0030] A digital twin is a dynamic simulation model of the plowing process built on a server;

[0031] Based on the real-time status information of the electric tractor and its suspension system, the collected weather conditions and soil information, the digital twin runs synchronously in the virtual world. The digital twin simulates and generates dynamic and soil model data, and supports data query and status review.

[0032] Twin data is a collection of historical and real-time data generated by the electronically controlled suspension system, digital twin, and digital twin services. It includes tillage depth data measured by the magnetostrictive displacement sensor 23 built into the electric actuator 6, geographical location information provided by the GPS locator 11, force data acquired by the pin force sensor 17, and dynamic and soil model data simulated by the digital twin. This data is fused and stored in the server for subsequent analysis and decision-making.

[0033] The pin pressure measured by the pin force sensor 17 is converted into the tillage resistance experienced by the electric tractor in real time; the displacement signal of the magnetostrictive displacement sensor 23 built into the electric push rod 6 is converted into the tillage depth in real time.

[0034] The digital twin service platform runs on a server and can provide real-time monitoring and data acquisition functions, prediction and simulation functions, and support remote monitoring and management;

[0035] The connections between the various parts of a digital twin system are accomplished through TCP / IP and MQTT protocols;

[0036] Step S2: Suspension height control: Based on the historical weather conditions, soil information, and tillage depth data collected and recorded in Step S1, establish an online control problem model for the electric suspension system. Use the deep deterministic strategy gradient algorithm to train and obtain the optimal tillage depth. Set the suspension height of the electric tractor's electric suspension system to the tillage depth to achieve a control strategy with consistent tillage depth and low energy consumption.

[0037] Step S2.1: Establish the control strategy model for the electronically controlled suspension system.

[0038] The online control problem model for the electronically controlled suspension system is established as follows:

[0039] s t =[xy FrθT S lv] Formula 1

[0040] a t =[h] Formula 2

[0041] In Formula 1 and Formula 2, s t Let a be the state observation at time t; t t represents the motion at time t; x and y represent the horizontal and vertical coordinates of the current tillage position in the geodetic coordinate system, in meters; Fr represents the current tillage resistance, in N; θ represents soil moisture, in %mass; T represents soil temperature, in °C; S represents soil type; l represents the current tillage depth, in cm, which is an important constraint in suspension control; v represents the current vehicle speed, in m / s; h represents the current suspension lift, in cm.

[0042] Step S2.2: Train the control strategy model of the electronically controlled suspension system established in step S2.1 using the deep deterministic policy gradient algorithm.

[0043] The Deep Deterministic Policy Gradient Algorithm (DPRQA) employs a dual-network model, with both the Actor and Critic networks equipped with a target network of the same structure. Therefore, the DPRQA contains four networks. The presence of the target network reduces the fluctuation of network parameters, makes the learning process more stable, and makes the model easier to converge.

[0044] Step S2.3: Set the reward function and evaluate the performance of the electronically controlled suspension system control strategy model trained in Step S2.2.

[0045] The reward variable R is used to evaluate the characteristics of the electronically controlled suspension system and a reward function is established:

[0046] R = [lv] Formula 3

[0047] In Formula 3, R is the reward variable; l is the tillage depth in cm, which is an important constraint in suspension control; v is the vehicle speed in m / s.

[0048] The instant reward function is expressed as follows:

[0049] r = -(k1|l-l0| 2 +k2|v-v0| 2 ) Formula 4

[0050] In Formula 4, r is the immediate reward; k1 and k2 are the weighting coefficients of tillage depth and vehicle speed, respectively; l is the expected tillage depth in cm; l0 is the target tillage depth in cm; v is the expected vehicle speed in m / s; v0 is the target vehicle speed in m / s.

[0051] The model is trained using the historical weather conditions, soil information, and tillage depth data collected and recorded in step S1. During the training process, the model learns the optimal suspension height control strategy by interacting with the digital twin system in order to maximize the cumulative reward of the reward function.

[0052] Step S2.4: Set the suspension height of the electric tractor's electronically controlled suspension system to the tillage depth of the electronically controlled suspension system control strategy model obtained after training in step S2.3, so as to achieve a control strategy with consistent tillage depth and low energy consumption.

[0053] Step S3: Apply the digital twin system in actual farming to determine the optimal tillage depth based on real-time environmental parameters and historical data; continuously collect actual farming data to iteratively improve the digital twin system and reinforcement learning model in order to improve the accuracy of suspension height prediction and operational efficiency.

[0054] Step S4: In actual farming, by monitoring environmental parameters and farming resistance in real time, and combining the digital twin system improved in step S3 with the reinforcement learning model trained by the deep deterministic policy gradient algorithm, the optimal suspension height is predicted and determined.

[0055] In step S1, the digital twin is constructed as follows:

[0056] Step S1.1: Use MATLAB to construct a dynamic model of the electric suspension actuator, including the motor model, transmission system model and load model of the electric actuator 6;

[0057] Step S1.2: Using the soil information data obtained from farmland monitoring equipment, establish a soil mechanical property model in MATLAB, including the soil shear stress-strain relationship;

[0058] Step S1.3: Use Unity3D to build a visualization model to present the real-time status and operation of the electric tractor and its suspension system. The visualization model is synchronized with the plowing process and updated in real time to ensure that the model's predictions are consistent with the actual operation.

[0059] In step S2.2, the Actor network uses a six-layer neural network; the input is the system state s, and the input layer has 8 nodes; the activation function of the first hidden layer uses the ReLU function, and the number of nodes is 240; the activation function of the second hidden layer also uses the ReLU function, and the number of nodes is 200; the output layer is action a, and it has 1 node; the learning rate of the network is set to 0.001. The Critic network also uses a four-layer neural network; the input is the set of system state s and action a, so the input layer contains 9 nodes; the number of nodes in the two hidden layers are 200 and 150 respectively; the activation function of the hidden layers is the ReLU function; the output layer is the value function Q(s,a), and it has 1 node; the learning rate of the network is set to 0.002.

[0060] In step S3, the iterative improvement steps include: data collection and analysis, model evaluation, model update, and iterative loop;

[0061] Data collection and analysis: Collect real-time tillage data, including tillage depth, tillage resistance, soil moisture, soil temperature, soil type, and vehicle speed; clean and preprocess the collected data to remove outliers and noise, and perform feature extraction and dimensionality reduction for subsequent analysis.

[0062] Model evaluation: Tillage depth consistency and energy consumption are used as indicators to evaluate model performance, assessing the prediction accuracy and control effect of digital twin systems and reinforcement learning models;

[0063] Model Update: Based on model evaluation results and problem diagnosis, adjust the parameters of the digital twin system model and the reinforcement learning model trained by the deep deterministic policy gradient algorithm;

[0064] Iterative loop: The updated model is trained using a deep deterministic policy gradient algorithm, and new data is continuously collected for iterative training and optimization; through data feedback loop, the control policy of the reinforcement learning model is continuously optimized to improve the model's prediction accuracy and control effect.

[0065] Through the above iterative improvement steps, the digital twin system and the reinforcement learning model trained by the deep deterministic policy gradient algorithm are continuously optimized to achieve more efficient suspension height control and operational efficiency.

[0066] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0067] This invention provides an electric control suspension system for an electric tractor, which uses an electric push rod to drive an "L"-shaped sliding frame, enabling more efficient use of electrical energy and reducing energy waste. Precise suspension height adjustment can be achieved through digital control, allowing for accurate adjustments based on different operating conditions and needs. Compared to traditional hydraulic suspension systems, this invention's electric control suspension system has higher energy efficiency, faster response speed, and higher precision. The electric suspension system eliminates the need for hydraulic oil, reducing the environmental pollution risk associated with hydraulic oil, and the electric components are relatively simple, resulting in lower maintenance costs.

[0068] Meanwhile, this invention proposes a suspension height control method that considers multiple factors, based on a digital twin control method. This method comprehensively considers factors such as tillage location, tillage resistance, soil moisture, ambient temperature, and soil type, and predicts the optimal suspension height through a digital twin model. This enables the suspension system to adaptively adjust according to different operating environments and requirements. By combining historical tillage data and environmental parameters, intelligent suspension height control is achieved. Compared with traditional force and position adjustment methods, it can more accurately respond to the needs under different tillage conditions, improving tillage efficiency and operation quality.

[0069] Before using an electric tractor for tillage, a digital twin system was first established. This system includes an electric suspension system, a digital twin, twin data, and a digital twin service platform. The digital twin is a dynamic simulation model of the tillage process built on a server, including a dynamic model of the electric suspension actuator, a soil mechanics model, and a visualization model. Next, a reinforcement learning model was trained. Based on data recorded over the years, such as tillage location, tillage resistance, soil moisture, ambient temperature, soil type, tillage depth, and vehicle speed, the optimal suspension height control strategy was obtained using a reinforcement learning algorithm. During tillage, real-time data was monitored and analyzed through the digital twin service platform. Based on historical data and the suspension height control strategy trained by the reinforcement learning algorithm, the "L"-shaped sliding frame 7 was adjusted in real-time via the electric push rod 6 to adjust the suspension height. The digital twin system uses real-time data to drive the digital twin to run synchronously in the virtual world, while also supporting data querying and status review.

[0070] To further optimize the suspension height control strategy, the predictive function of the digital twin system can be used. By viewing the prediction results of the virtual model, the operating effect of the tractor under different suspension heights can be understood. In this way, the suspension height can be adjusted according to the prediction results to achieve the best operating effect. Attached Figure Description

[0071] Figure 1 This is a schematic diagram of the overall structure of the electric tractor electronically controlled suspension system of the present invention;

[0072] Figure 2 This is a structural schematic diagram of frame 3;

[0073] Figure 3 A schematic diagram of an electrically controllable lifting "L"-shaped sliding frame;

[0074] Figure 4 This is a schematic diagram of the suspension height control method.

[0075] The reference numerals in the attached figures are:

[0076] 1. Low-voltage servo motor; 2. Lower baffle plate

[0077] 3. Chassis 4. Upper fender

[0078] 5. First cylindrical pin; 6. Electric actuator.

[0079] 7. "L"-shaped sliding frame 8. Diagonal tie rod

[0080] 9. Hanging pin 10. Hanging bracket

[0081] 11. GPS locator 12. Main controller

[0082] 13. Tillage depth controller 14. Electric actuator

[0083] 15. Second cylindrical pin 16. Connecting pin

[0084] 17. Pin force sensor 18. Reduction mechanism

[0085] 19. Long nylon pad 20. Curved plate

[0086] 21. Rolling bearing; 22. Short nylon pad.

[0087] 23. Magnetostrictive displacement sensor 24. Data transceiver equipment Detailed Implementation

[0088] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0089] like Figures 1-3 As shown, an electric control suspension system for an electric tractor includes a low-voltage servo motor 1, a lower baffle plate 2, a frame 3, an upper baffle plate 4, a first cylindrical pin 5, an electric push rod 6, an "L"-shaped sliding frame 7, a tie rod 8, a connecting pin 9, a suspension frame 10, a GPS locator 11, a main controller 12, a tillage depth controller 13, an electric push rod driver 14, a second cylindrical pin 15, a connecting pin 16, a pin shaft force sensor 17, a reduction mechanism 18, a long nylon pad 19, a curved plate 20, a rolling bearing 21, and a short nylon pad 22.

[0090] like Figure 2 As shown, the frame 3 is a rectangular frame structure, including a left column and a right column arranged on the left and right sides, and a top plate arranged between the upper ends of the left column and the right column.

[0091] like Figure 3 As shown, the "L"-shaped sliding frame 7 includes a bottom rectangular frame, a left upright beam, and a right upright beam. The bottom rectangular frame includes a front beam, a rear beam, a left side beam, and a right side beam. The left and right upright beams are perpendicular to the bottom rectangular frame and extend upwards, with their bottom ends connected to the rear ends of the left and right side beams. The front ends of the left and right side beams are connected to the lower end of the suspension frame 10.

[0092] The upper ends of the two diagonal tie rods 8 are connected to the upper end of the suspension frame 10 via hook pins 9, and the lower ends of the two diagonal tie rods 8 are respectively hinged to the rear parts of the left and right beams of the "L"-shaped sliding frame 7 via connecting pins 16. The pin force sensor 17 is installed at the connection between the left and right beams of the "L"-shaped sliding frame 7 and the suspension frame 10.

[0093] like Figure 1 and Figure 3As shown, the upper end of the electric push rod 6 is hinged to the top plate of the frame 3 via a first cylindrical pin 5, and the lower end of the electric push rod 6 is hinged to the rear beam of the "L"-shaped sliding frame 7 via a second cylindrical pin 15. The extension and retraction of the electric push rod 6 can drive the "L"-shaped sliding frame 7 to achieve lifting and lowering within a range of 37cm.

[0094] The electric push rod 6 has a built-in magnetostrictive displacement sensor 23, which obtains the corresponding tillage depth value by detecting the change in push rod displacement.

[0095] The lower bulkhead 2 is located at the front of the frame 3 and is connected to the left and right pillars of the frame 3. The upper bulkhead 4 is located at the front of the frame 3 and is connected to the left and right pillars of the frame 3. The upper bulkhead 4 is located above the lower bulkhead 2.

[0096] The left and right upright beams of the “L”-shaped sliding frame 7 are located between the left and right uprights of the frame 3 and the lower and upper baffles 2 and 4.

[0097] Long nylon pads 19 are fixed to the front surfaces of the left and right uprights of the frame 3. Short nylon pads 22 are fixed to the rear surfaces of the lower baffle 2 and the upper baffle 4. During operation, the long nylon pads 19 and short nylon pads 22 are used for shock absorption and cushioning between the frame 3 and the "L"-shaped sliding frame 7.

[0098] The bending plate 20 is fixed to the left and right uprights of the frame 3, and the rolling bearing 21 is installed on the bending plate 20. The bearing 21 contacts the left and right beams of the "L"-shaped sliding frame 7, and plays a guiding and limiting role to ensure that the "L"-shaped sliding frame 7 can be raised and lowered smoothly in the vertical direction.

[0099] The low-voltage servo motor 1 and the reduction mechanism 18 are connected to the frame 3 in sequence. The low-voltage servo motor 1 and the reduction mechanism 18 serve as auxiliary power output devices for the electric tractor.

[0100] The GPS locator 11 is installed on the vehicle body and connected to the main controller 12 via a cable.

[0101] The main controller 12 collects data from the GPS positioning device 11, the built-in magnetostrictive displacement sensor 23 of the electric push rod 6, and the pin force sensor 17 through a dedicated signal line to obtain the tillage position and the current status information of the electric control suspension system.

[0102] Data transceiver 24 is connected to the main controller 12. Data transceiver 24 remotely transmits and receives data collected by the main controller 12. For example, data transceiver 24 can communicate remotely with a remote server.

[0103] The main controller 12, tillage depth controller 13, electric push rod driver 14, and electric push rod 6 are connected in sequence. The electric push rod 6 controls the displacement of the push rod by receiving signals from the electric push rod driver 14. When the push rod of the electric push rod 6 moves, it drives the "L"-shaped sliding frame 7 to rise and fall, thereby driving the agricultural implement to rise and fall.

[0104] The working process of this invention is as follows:

[0105] Based on the optimal tillage depth information provided by the digital twin system, the main controller 12 sends a control signal to the electric push rod 6 via the electric push rod driver 14. After receiving the control signal, the electric push rod 6 drives the push rod to extend and retract. The extension and retraction of the push rod drives the "L"-shaped sliding frame 7 to rise and fall, thereby driving the agricultural implement to rise and fall, and realizing the adjustment of the suspension height.

[0106] The magnetostrictive displacement sensor 23 built into the electric push rod 6 monitors the displacement change of the push rod in real time; the magnetostrictive displacement sensor 23 converts the push rod displacement into an electrical signal and transmits it to the main controller 12; the main controller 12 calculates the current tillage depth based on the displacement signal and uses it as tillage depth feedback information for closed-loop control.

[0107] The pin force sensor 17 is installed at the connection between the "L"-shaped sliding frame 7 and the suspension frame 10. When the implement is subjected to soil resistance, the pin force sensor 17 detects the corresponding force change and converts it into an electrical signal, which is then transmitted to the main controller 12. The main controller 12 calculates the current tillage resistance based on the force signal and uses it as tillage resistance feedback information for closed-loop control.

[0108] like Figure 4 As shown, a suspension height control method according to the electric tractor's electronically controlled suspension system includes the following steps:

[0109] Step S1: Establish a digital twin system, which includes an electronically controlled suspension system, farmland monitoring equipment, a digital twin, twin data, a digital twin service platform, and the connections between the various parts of the digital twin system;

[0110] The electric suspension system includes a magnetostrictive displacement sensor 23, a GPS locator 11, a pin force sensor 17, an electric actuator 6, and a data transceiver 24.

[0111] The GPS locator 11 can measure the tractor's location information and tillage position; the pin force sensor 17 can measure data such as tillage resistance; the electric push rod 6 can perform lifting and lowering actions; and the data transceiver 24 can complete the remote transmission and reception of information.

[0112] Farmland monitoring equipment collects weather and soil information in real time and transmits the information to a digital twin service platform.

[0113] The weather conditions information includes rain / snow information and temperature.

[0114] The soil information includes soil moisture, soil temperature, soil type, and soil hardness.

[0115] A digital twin is a dynamic simulation model of the plowing process built on a server. The construction method of a digital twin is as follows:

[0116] Step S1.1: Use MATLAB to construct a dynamic model of the electric suspension actuator, including the motor model, transmission system model and load model of the electric actuator 6;

[0117] Step S1.2: Using the soil information data obtained from farmland monitoring equipment, establish a soil mechanical property model in MATLAB, including the soil shear stress-strain relationship;

[0118] Step S1.3: Use Unity3D to build a visualization model to present the real-time status and operation of the electric tractor and its suspension system. The visualization model is synchronized with the plowing process and updated in real time to ensure that the model's predictions are consistent with the actual operation.

[0119] Based on the real-time status information of the electric tractor and its suspension system, the collected weather conditions and soil information, the digital twin runs synchronously in the virtual world. The digital twin simulates and generates dynamic and soil model data, and supports operations such as data query and status review.

[0120] Twin data is a collection of historical and real-time data generated by the electronically controlled suspension system, digital twin, and digital twin services. It includes tillage depth data measured by the magnetostrictive displacement sensor 23 built into the electric actuator 6, geographical location information provided by the GPS locator 11, force data acquired by the pin force sensor 17, and dynamic and soil model data simulated by the digital twin. This data is fused and stored in the server for subsequent analysis and decision-making.

[0121] The pin pressure measured by the pin force sensor 17 is converted into the tillage resistance experienced by the electric tractor in real time; the displacement signal of the magnetostrictive displacement sensor 23 built into the electric push rod 6 is converted into the tillage depth in real time.

[0122] The digital twin service platform runs on a server and can provide real-time monitoring and data acquisition functions, prediction and simulation functions, and support remote monitoring and management.

[0123] The connections between the various parts of a digital twin system are accomplished through protocols such as TCP / IP and MQTT.

[0124] Step S2: Suspension Height Control: Based on the historical weather conditions, soil information, and tillage depth data collected and recorded in Step S1, an online control problem model for the electric suspension system is established. The optimal tillage depth is obtained by training the model using a deep deterministic strategy gradient algorithm. The suspension height of the electric tractor's electric suspension system is then set to the tillage depth to achieve a control strategy that ensures consistent tillage depth and low energy consumption.

[0125] Step S2.1: Establish the control strategy model for the electronically controlled suspension system.

[0126] The core of this method is to utilize a digital twin as a dynamic learning environment to verify the feasibility and performance of suspension height control. Strategies learned from the digital twin environment reduce the gap between simulation and reality, thereby enabling the learning of more realistic suspension height control strategies.

[0127] To improve the convergence speed of reinforcement learning training in digital twin systems with higher degrees of freedom (larger state-action space), a suspension control height and vehicle speed control strategy based on a deep deterministic policy gradient algorithm is proposed. Considering the observability of the digital twin system and the control characteristics of the electronically controlled suspension system in a real environment, the online control problem model of the electronically controlled suspension system is established as follows:

[0128] s t =[xy FrθT S lv] Formula 1

[0129] a t =[h] Formula 2

[0130] In Formula 1 and Formula 2, s t Let a be the state observation at time t; t t represents the motion at time t; x and y represent the horizontal and vertical coordinates of the current tillage position in the geodetic coordinate system, in meters; Fr represents the current tillage resistance, in N; θ represents soil moisture, in %mass; T represents soil temperature, in °C; S represents soil type; l represents the current tillage depth, in cm, which is an important constraint in suspension control; v represents the current vehicle speed, in m / s; and h represents the current suspension lift, in cm.

[0131] Step S2.2: Train the control strategy model of the electronically controlled suspension system established in step S2.1 using the Deep Deterministic Policy Gradient Algorithm (DDPG).

[0132] DDPG employs a dual-network model, with both the Actor and Critic networks equipped with a target network of the same structure, resulting in four networks in DDPG. The presence of the target network reduces fluctuations in network parameters, making the learning process more stable and the model easier to converge. The Actor network uses a six-layer neural network. The input is the system state s, with 8 nodes in the input layer; the first hidden layer uses the ReLU activation function and has 240 nodes; the second hidden layer also uses the ReLU activation function and has 200 nodes; the output layer is the action a, with 1 node. The learning rate is set to 0.001. The Critic network also uses a four-layer neural network. The input is the set of system state s and action a, hence the input layer contains 9 nodes; the two hidden layers have 200 and 150 nodes respectively. Both hidden layers use the ReLU activation function. The output layer is the value function Q(s,a), with 1 node, and the learning rate is set to 0.002.

[0133] Step S2.3: Set the reward function and evaluate the performance of the electronically controlled suspension system control strategy model trained in Step S2.2.

[0134] The reward is a feedback signal provided by the digital twin system. It shows the agent's performance under specific conditions when employing a specific strategy. By appropriately setting the reward function, the optimal suspension height can be determined. To achieve a control strategy with consistent tillage depth and low energy consumption, the reward variable R is used to evaluate the characteristics of the electronically controlled suspension system and to establish the reward function.

[0135] R = [lv] Formula 3

[0136] In Formula 3, R is the reward variable; l is the tillage depth in cm, which is an important constraint in suspension control; and v is the vehicle speed in m / s.

[0137] The reward variable R contains the two most important variables in the electronically controlled suspension system: tillage depth and vehicle speed. By solving an optimization problem with multiple control objectives, the coefficient matrix K in the reward function is dynamically adjusted so that different control objectives simultaneously achieve certain control effects. The instantaneous reward function is expressed as follows:

[0138] r = -(k1|l-l0| 2 +k2|v-v0| 2 ) Formula 4

[0139] In Formula 4, r is the immediate reward; k1 and k2 are the weighting coefficients for tillage depth and vehicle speed, respectively; l is the expected tillage depth in cm; l0 is the target tillage depth in cm; v is the expected vehicle speed in m / s; and v0 is the target vehicle speed in m / s.

[0140] The model is trained using the historical weather conditions, soil information, and tillage depth data collected and recorded in step S1. During the training process, the model learns the optimal suspension height control strategy by interacting with the digital twin system to maximize the cumulative reward of the reward function.

[0141] Step S2.4: Set the suspension height of the electric tractor's electronically controlled suspension system to the tillage depth of the electronically controlled suspension system control strategy model obtained after training in step S2.3, so as to achieve a control strategy with consistent tillage depth and low energy consumption.

[0142] Step S3: Apply the digital twin system in actual farming to determine the optimal tillage depth based on real-time environmental parameters and historical data; continuously collect actual farming data to iteratively improve the digital twin system and reinforcement learning model, thereby improving the accuracy of suspension height prediction and operational efficiency.

[0143] The iterative improvement steps include: data collection and analysis, model evaluation, model updating, and iterative cycles.

[0144] Data collection and analysis: Collect real-time tillage data, including tillage depth, tillage resistance, soil moisture, soil temperature, soil type, and vehicle speed; clean and preprocess the collected data to remove outliers and noise, and perform feature extraction and dimensionality reduction for subsequent analysis.

[0145] Model evaluation: Tillage depth consistency and energy consumption are used as indicators to evaluate model performance, assessing the prediction accuracy and control effect of digital twin systems and reinforcement learning models;

[0146] Model update: Based on the model evaluation results and problem diagnosis, adjust the parameters of the digital twin system model and the reinforcement learning model trained by the DDPG algorithm;

[0147] Iterative Loop: The updated model is trained using the DDPG algorithm, and new data is continuously collected for iterative training and optimization. Through data feedback loop, the control strategy of the reinforcement learning model is continuously optimized to improve the model's prediction accuracy and control effect.

[0148] Through the above iterative improvement steps, the digital twin system and the reinforcement learning model trained by the DDPG algorithm are continuously optimized to achieve more efficient suspension height control and operational efficiency.

[0149] Step S4: In actual farming, by monitoring environmental parameters and farming resistance in real time, and combining the digital twin system improved in step S3 with the reinforcement learning model trained by the DDPG algorithm, the optimal suspension height is predicted and determined.

[0150] This process helps improve tillage efficiency, reduce energy consumption, and maintain uniform tillage depth. By continuously collecting actual tillage data for model training and optimization, the suspension height can be iteratively improved, resulting in higher operational efficiency and lower energy consumption in actual tillage.

Claims

1. An electric control suspension system for an electric tractor, characterized in that: The electronically controlled suspension system includes a low-voltage servo motor (1), a lower baffle plate (2), a frame (3), an upper baffle plate (4), a first cylindrical pin (5), an electric push rod (6), an "L"-shaped sliding frame (7), a tie rod (8), a hook pin (9), a suspension frame (10), a GPS locator (11), a main controller (12), a tillage depth controller (13), an electric push rod driver (14), a second cylindrical pin (15), a connecting pin (16), a pin shaft force sensor (17), and a reduction mechanism (18). The frame (3) is a rectangular frame structure, including left and right columns arranged on the left and right sides, and a top plate arranged between the upper ends of the left and right columns. The "L"-shaped sliding frame (7) includes a bottom rectangular frame, a left upright beam and a right upright beam. The bottom rectangular frame includes a front beam, a rear beam, a left side beam and a right side beam. The left and right upright beams are perpendicular to the bottom rectangular frame and extend upwards. The bottom of the left and right upright beams are connected to the rear ends of the left and right side beams. The front ends of the left and right side beams are connected to the lower end of the suspension frame (10). The upper ends of the two diagonal tie rods (8) are connected to the upper end of the suspension frame (10) by a hook pin (9), and the lower ends of the two diagonal tie rods (8) are respectively hinged to the rear of the left and right beams of the "L"-shaped sliding frame (7) by a connecting pin (16); the pin force sensor (17) is set at the connection between the left and right beams of the "L"-shaped sliding frame (7) and the suspension frame (10); The upper end of the electric push rod (6) is hinged to the top plate of the frame (3) through the first cylindrical pin (5), and the lower end of the electric push rod (6) is hinged to the rear beam of the "L"-shaped sliding frame (7) through the second cylindrical pin (15). The electric actuator (6) has a built-in magnetostrictive displacement sensor (23); The lower baffle (2) is located in front of the frame (3) and is connected to the left and right pillars of the frame (3) respectively; the upper baffle (4) is located in front of the frame (3) and is connected to the left and right pillars of the frame (3) respectively; the upper baffle (4) is located above the lower baffle (2); The left and right upright beams of the "L"-shaped sliding frame (7) are located between the left and right uprights of the frame (3) and the lower and upper baffles (2 and 4); The low-voltage servo motor (1), the reduction mechanism (18), and the frame (3) are connected in sequence; The GPS locator (11) is installed on the vehicle body and connected to the main controller (12) via a cable; The main controller (12) collects data from the GPS locator (11), the built-in magnetostrictive displacement sensor (23) of the electric push rod (6), and the pin force sensor (17) through a dedicated signal line to obtain information on the tillage position and the current status of the electric control suspension system; The data transceiver (24) is connected to the main controller (12); the data transceiver (24) remotely transmits and receives data collected by the main controller (12); The main controller (12), tillage depth controller (13), electric actuator driver (14) and electric actuator (6) are connected in sequence.

2. The electric tractor electronically controlled suspension system as described in claim 1, characterized in that: The electronically controlled suspension system also includes a long nylon pad (19) and a short nylon pad (22); the long nylon pad (19) is fixed to the front surface of the left and right pillars of the frame (3); the short nylon pad (22) is fixed to the rear surface of the lower baffle (2) and the upper baffle (4).

3. The electric tractor electronically controlled suspension system as described in claim 1, characterized in that: The electronically controlled suspension system also includes a bend plate (20) and a rolling bearing (21); the bend plate (20) is fixed on the left and right pillars of the frame (3), and the rolling bearing (21) is installed on the bend plate (20); the rolling bearing (21) contacts the left and right beams of the "L"-shaped sliding frame (7) to guide and limit, ensuring that the "L"-shaped sliding frame (7) rises and falls smoothly in the vertical direction.

4. The electric tractor electronically controlled suspension system as described in claim 1, characterized in that: The extension and retraction of the electric actuator (6) can drive the "L"-shaped sliding frame (7) to achieve lifting and lowering within a range of 37cm.

5. A method for controlling the suspension height of an electric tractor's electronically controlled suspension system according to any one of claims 1 to 4, characterized in that: The method includes the following steps: Step S1: Establish a digital twin system, which includes an electronically controlled suspension system, farmland monitoring equipment, a digital twin, twin data, a digital twin service platform, and the connections between the various parts of the digital twin system; The electric suspension system includes a magnetostrictive displacement sensor (23), a GPS locator (11), a pin force sensor (17) built into the electric actuator (6), the electric actuator (6) as a suspension height actuator, and a data transceiver (24); The GPS locator (11) can measure the tractor's location information and tillage position; the pin force sensor (17) can measure tillage resistance; the electric push rod (6) can perform lifting and lowering actions; the data transceiver (24) can complete the remote transmission and reception of information. Farmland monitoring equipment collects weather and soil information in real time and transmits the weather and soil information to a digital twin service platform; The weather conditions information includes rain / snow information and temperature; The soil information includes soil moisture, soil temperature, soil type, and soil hardness; A digital twin is a dynamic simulation model of the plowing process built on a server; Based on the real-time status information of the electric tractor and its suspension system, the collected weather conditions and soil information, the digital twin runs synchronously in the virtual world. The digital twin simulates and generates dynamic and soil model data, and supports data query and status review. Twin data is a collection of historical and real-time data generated by the electronically controlled suspension system, digital twin, and digital twin services: including tillage depth data measured by the magnetostrictive displacement sensor (23) built into the electric actuator (6), geographical location information provided by the GPS locator (11), force data obtained by the pin force sensor (17), and dynamics and soil model data simulated by the digital twin. These data are fused and stored in the server for subsequent analysis and decision-making. Among them, the pin pressure measured by the pin force sensor (17) is converted into the tillage resistance experienced by the electric tractor in real time; the displacement signal of the magnetostrictive displacement sensor (23) built into the electric push rod (6) is converted into the tillage depth in real time. The digital twin service platform runs on a server and can provide real-time monitoring and data acquisition functions, prediction and simulation functions, and support remote monitoring and management; The connections between the various parts of a digital twin system are accomplished through TCP / IP and MQTT protocols; Step S2: Suspension height control: Based on the historical weather conditions, soil information, and tillage depth data collected and recorded in Step S1, establish an online control problem model for the electric suspension system. Use the deep deterministic strategy gradient algorithm to train and obtain the optimal tillage depth. Set the suspension height of the electric tractor's electric suspension system to the tillage depth to achieve a control strategy with consistent tillage depth and low energy consumption. Step S2.1: Establish the control strategy model for the electronically controlled suspension system. The online control problem model for the electronically controlled suspension system is established as follows: s t = [x y Fr θ T S l v] Equation 1 a t =[h] Formula 2 In Formula 1 and Formula 2, s t Let a be the state observation at time t; t t represents the amount of motion at time t; x and y represent the horizontal and vertical coordinates of the current tillage position in the geodetic coordinate system, in meters; Fr represents the current tillage resistance, in N; θ represents soil moisture, in %mass; T represents soil temperature, in °C; S represents soil type; l represents the current tillage depth, in cm; v represents the current vehicle speed, in m / s; h represents the current suspension lifting height, in cm. Step S2.2: Train the control strategy model of the electronically controlled suspension system established in step S2.1 using the deep deterministic policy gradient algorithm. The deep deterministic policy gradient algorithm adopts a dual-network mode, with both the Actor network and the Critic network equipped with a target network of the same structure. Therefore, there are four networks in the deep deterministic policy gradient algorithm. Step S2.3: Set the reward function and evaluate the performance of the electronically controlled suspension system control strategy model trained in Step S2.

2. Use the reward variable R to evaluate the characteristics of the electronically controlled suspension system and establish the reward function: R = [lv] Formula 3 In Formula 3, R is the reward variable; l is the tillage depth in cm; and v is the vehicle speed in m / s. The instant reward function is expressed as follows: r = -(k1|l-l0| 2 +k2|v-v0| 2 ) Formula 4 In Formula 4, r is the immediate reward; k1 and k2 are the weighting coefficients of tillage depth and vehicle speed, respectively; l is the expected tillage depth in cm; l0 is the target tillage depth in cm; v is the expected vehicle speed in m / s; v0 is the target vehicle speed in m / s. The model is trained using the historical weather conditions, soil information, and tillage depth data collected and recorded in step S1. During the training process, the model learns the optimal suspension height control strategy by interacting with the digital twin system in order to maximize the cumulative reward of the reward function. Step S2.4: Set the suspension height of the electric tractor's electronically controlled suspension system to the tillage depth of the electronically controlled suspension system control strategy model obtained after training in step S2.3, so as to achieve a control strategy with consistent tillage depth and low energy consumption. Step S3: Apply the digital twin system in actual farming to determine the optimal tillage depth based on real-time environmental parameters and historical data; continuously collect actual farming data to iteratively improve the digital twin system and reinforcement learning model in order to improve the accuracy of suspension height prediction and operational efficiency. Step S4: In actual farming, by monitoring environmental parameters and farming resistance in real time, and combining the digital twin system improved in step S3 with the reinforcement learning model trained by the deep deterministic policy gradient algorithm, the optimal suspension height is predicted and determined.

6. The suspension height control method for the electric tractor's electronically controlled suspension system as described in claim 5, characterized in that: in, In step S1, the digital twin is constructed as follows: Step S1.1: Use MATLAB to construct a dynamic model of the electric suspension actuator, including the motor model, transmission system model and load model of the electric actuator (6); Step S1.2: Using the soil information data obtained from farmland monitoring equipment, establish a soil mechanical property model in MATLAB, including the soil shear stress-strain relationship; Step S1.3: Use Unity3D to build a visualization model to present the real-time status and operation of the electric tractor and its suspension system. The visualization model is synchronized with the plowing process and updated in real time to ensure that the model's predictions are consistent with the actual operation.

7. The suspension height control method for an electric tractor's electronically controlled suspension system as described in claim 5, characterized in that: wherein, In step S2.2, the Actor network uses a six-layer neural network; the input is the system state s, and the input layer has 8 nodes; the activation function of the first hidden layer uses the ReLU function, and the number of nodes is 240; the activation function of the second hidden layer also uses the ReLU function, and the number of nodes is 200; the output layer is action a, and it has 1 node; the learning rate of the network is set to 0.

001. The Critic network also uses a four-layer neural network; the input is the set of system state s and action a, so the input layer contains 9 nodes; the number of nodes in the two hidden layers are 200 and 150 respectively; the activation function of the hidden layers is the ReLU function; the output layer is the value function Q(s,a), and it has 1 node; the learning rate of the network is set to 0.

002.

8. The suspension height control method for the electric tractor's electronically controlled suspension system as described in claim 5, characterized in that: in, In step S3, the iterative improvement steps include: data collection and analysis, model evaluation, model updating, and iterative loop; Data collection and analysis: Collect real-time tillage data, including tillage depth, tillage resistance, soil moisture, soil temperature, soil type, and vehicle speed; clean and preprocess the collected data to remove outliers and noise, and perform feature extraction and dimensionality reduction for subsequent analysis. Model evaluation: Tillage depth consistency and energy consumption are used as indicators to evaluate model performance, assessing the prediction accuracy and control effect of digital twin systems and reinforcement learning models; Model Update: Based on model evaluation results and problem diagnosis, adjust the parameters of the digital twin system model and the reinforcement learning model trained by the deep deterministic policy gradient algorithm; Iterative loop: The updated model is trained using a deep deterministic policy gradient algorithm, and new data is continuously collected for iterative training and optimization; through data feedback loop, the control policy of the reinforcement learning model is continuously optimized to improve the model's prediction accuracy and control effect. Through the above iterative improvement steps, the digital twin system and the reinforcement learning model trained by the deep deterministic policy gradient algorithm are continuously optimized to achieve more efficient suspension height control and operational efficiency.

Images