Hydraulic tilling depth regulation and control method, device and system based on tilling depth prediction result

Abstract

The invention discloses a hydraulic tilling depth regulation and control method, device and system based on a tilling depth prediction result, belongs to the technical field of intelligent agricultural machinery, and particularly relates to self-adaptive regulation of the tilling depth. The problems that most of existing hydraulic tilling depth regulation and control systems depend on fixed control parameters, dynamic adjustment cannot be conducted according to real-time changing information such as the posture of a tractor body and the terrain, and therefore the performance and the effect of the hydraulic tilling depth regulation and control system in actual application are limited are solved. The method comprises the following steps: acquiring real farming depth data collected in real time; a control module is adopted to adjust a regulation and control instruction according to the deviation between the real tillage depth data and the given tillage depth, and self-adaptive adjustment of the given tillage depth is completed. The hydraulic tilling depth regulation and control method, device and system based on the tilling depth prediction result are suitable for predicting the tilling depth, dynamic adjustment is conducted according to the prediction result, and accurate control over the tilling depth is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent agricultural machinery, and particularly to the adaptive adjustment of tillage depth. Background Art

[0002] In modern agriculture, as the main tillage machinery, the tillage depth and uniformity of tractors have an important impact on the growth and yield of crops. However, most traditional tractor tillage depth controls rely on manual adjustment, which has many deficiencies. First, manual adjustment requires the operator to judge based on experience and manually adjust the tillage depth adjustment handle, which not only increases the labor intensity of the operator but also may lead to unstable and uneven tillage depth. Second, it is often difficult to achieve the ideal tillage depth manually, thus affecting the tillage quality and efficiency.

[0003] To solve the above problems, those skilled in the art have conducted a large number of studies and explorations. Among them, the tillage depth control technology based on the hydraulic system has attracted much attention due to its advantages such as fast response speed and high control accuracy. However, most existing hydraulic tillage depth control systems rely on fixed control parameters and cannot be dynamically adjusted according to real-time changing information such as the tractor body attitude and terrain, thus limiting their performance and effects in practical applications. Summary of the Invention

[0004] The present invention proposes a hydraulic tillage depth control method, device and system based on tillage depth prediction results, which solves the problem that most existing hydraulic tillage depth control systems rely on fixed control parameters and cannot be dynamically adjusted according to real-time changing information such as the tractor body attitude and terrain, thus limiting their performance and effects in practical applications.

[0005] The hydraulic tillage depth control method based on tillage depth prediction results according to the present invention includes the following steps: Step S1: Obtain the real-time collected tractor tillage data; the tractor tillage data includes tractor lower link angle data, tractor body attitude data and terrain data; preprocess the tractor tillage data to obtain multi-dimensional time series data; Step S2: Input the multi-dimensional time series data into the tillage depth prediction model to obtain the tillage depth prediction result; Step S3: Use the tillage depth prediction result as the given tillage depth, and calculate the control command by the control module; the control command is used to adjust the angle of the tillage implement and the lifting of the suspension system, so as to perform tillage operations according to the given tillage depth; Step S4: Obtain the real-time collected actual tillage depth data; use the control module to adjust the control command according to the deviation between the actual tillage depth data and the given tillage depth, and complete the adaptive adjustment of the given tillage depth.

[0006] Furthermore, a preferred embodiment is provided, where the tillage depth prediction model is the BOA-TCN-Transformer-LSTM model; The BOA-TCN-Transformer-LSTM model is obtained by using the Bayesian optimization algorithm to automatically tune the parameters of the TCN-Transformer-LSTM model and searching for the optimal hyperparameter configuration.

[0007] Furthermore, a preferred embodiment is provided, where the control module includes a PID control module that integrates an RL policy network and a BP neural network; The PID control module that integrates an RL policy network and a BP neural network has a two-layer control architecture, including a bottom control layer and an upper optimization layer; The bottom control layer is a PID controller based on a BP neural network, which is used to calculate the control quantity of the hydraulic system in real time and output a regulation instruction to the actuator of the hydraulic system to ensure a quick response to the tillage depth deviation; The upper layer is a reinforcement learning RL policy network, which is used to dynamically optimize the parameters of the PID controller and adjust the weights of the BP neural network by analyzing the tillage depth error of the hydraulic system to improve the global adaptability.

[0008] The present invention also proposes a hydraulic tillage depth regulation device based on the tillage depth prediction result. The device includes the following modules: Module S1: Obtain the tractor tillage data collected in real time; the tractor tillage data includes tractor lower link angle data, tractor body attitude data, and terrain data; preprocess the tractor tillage data to obtain multi-dimensional time series data; Module S2: Input the multi-dimensional time series data into the tillage depth prediction model to obtain the tillage depth prediction result; Module S3: Use the tillage depth prediction result as the given tillage depth, and calculate the regulation instruction by using the control module; the regulation instruction is used to adjust the angle of the tillage implement and the lifting of the suspension system, so as to perform the tillage operation according to the given tillage depth; Module S4: Obtain the real tillage depth data collected in real time; use the control module to adjust the regulation instruction according to the deviation between the real tillage depth data and the given tillage depth to complete the adaptive adjustment of the given tillage depth.

[0009] The present invention also proposes a hydraulic tillage depth regulation system based on the tillage depth prediction result. The system includes a collection device, a hydraulic system, and the above-mentioned hydraulic tillage depth regulation device based on the tillage depth prediction result; The collection device is used to collect tractor tillage data and real tillage depth data in real time; The hydraulic tillage depth control device based on the tillage depth prediction result is used to output a control command according to the tractor tillage data and the true tillage depth data collected in real time; The hydraulic system is used to adjust the angle of the tillage implement and the lifting of the suspension system according to the control command, so as to perform the tillage operation according to the given tillage depth.

[0010] Furthermore, a preferred embodiment is provided. The system further includes: a mode switching module, an alarm prompt module, and an operation interface module; The mode switching module is used to switch between the manual mode and the automatic mode; in the manual mode, the driver manually adjusts the tillage depth; in the automatic mode, the hydraulic tillage depth control device based on the tillage depth prediction result completes the adaptive adjustment of the given tillage depth; The alarm prompt module is used to send an alarm message when an abnormality occurs during the tractor tillage operation, reminding the driver to take corresponding measures; The operation interface module is used for human-computer interaction with the driver, including displaying the current tillage depth, displaying the alarm message, and setting the target value of the tillage depth.

[0011] The present invention also proposes a tractor, which includes the hydraulic tillage depth control system based on the tillage depth prediction result described in any one of the above.

[0012] The present invention also proposes a computer device, including: a processor and a memory. The memory is used to store the executable instructions of the processor, and the processor is configured to execute the hydraulic tillage depth control method described in any one of the above by executing the executable instructions.

[0013] The present invention also proposes a computer storage medium, in which a computer program is stored. When the computer program runs, it executes the hydraulic tillage depth control method described in any one of the above.

[0014] The present invention also proposes a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, the steps of the hydraulic tillage depth control method described in any one of the above are implemented.

[0015] The present invention has the following beneficial effects: 1. The hydraulic tillage depth control method based on the tillage depth prediction result of the present invention collects multi-dimensional information such as the angle of the tractor's lower link, the attitude of the tractor body, and the terrain, and uses an advanced deep learning algorithm to construct a high-precision tillage depth prediction model; the tillage depth prediction model is a BOA-TCN-Transformer-LSTM model, where the TCN_Transformer_LSTM model is a multi-level hybrid architecture that combines the advantages of a temporal convolutional network (TCN), a Transformer encoder, and a long short-term memory network (LSTM), aiming to efficiently process multi-dimensional time series data and improve the prediction performance; the core idea of the model is to rely on the TCN (Temporal Convolutional Network) to extract the local temporal features of the input data, use the Transformer encoder to capture the global dependencies, and finally perform multi-scale feature fusion through the LSTM (Long Short-Term Memory) to generate the prediction result; the key hyperparameters and the model structure of the model are crucial for optimizing the performance of the TCN-Transformer-LSTM model, and the fine-tuning of these parameters is often the key to achieving the best performance of the model; in view of the cumbersome and inefficient manual adjustment of these parameters, the Bayesian optimization algorithm (BOA) is innovatively introduced to intelligently search for the optimal parameter configuration.

[0016] 2. The hydraulic tillage depth control method based on the tillage depth prediction result of the present invention uses a PID control module that integrates an RL policy network and a BP neural network to receive and analyze the output result of the tillage depth prediction model, calculate the parameters required for adjusting the flow rate, pressure, etc., generate precise control instructions, and send them to the hydraulic system; the hydraulic system then responds by finely adjusting key components such as the flow valve, pressure valve, and electro-hydraulic proportional directional valve to accurately control the flow rate and pressure of the hydraulic oil and drive the lifting of the suspension system to reach the preset tillage depth; during the tillage process, the change in the tillage depth is monitored in real time, and once a deviation is found, a secondary adjustment is immediately made to ensure the accuracy of the tillage depth; as the operation progresses, the tillage depth prediction model continuously updates the prediction result, and the PID control module that integrates the RL policy network and the BP neural network continuously issues control instructions accordingly to achieve the dynamic adjustment and optimization of the tillage depth.

[0017] 3. The hydraulic tillage depth control method based on the tillage depth prediction result of the present invention integrates tillage depth prediction and hydraulic control to form a closed-loop intelligent control system, overcoming the deficiencies of traditional tillage depth control technologies and achieving precise and automatic control of the tillage depth.

[0018] 4. The hydraulic tillage depth control system based on the tillage depth prediction result of the present invention provides a user-friendly operation interface, enabling the driver to easily set the target tillage depth value, monitor the current tillage depth, and receive system alarm information.

[0019] 5. The hydraulic tillage depth control system based on the tillage depth prediction result of the present invention supports the switching between manual mode and automatic mode to adapt to different operating environments and requirements.

[0020] The hydraulic tillage depth control method, device, and system based on the tillage depth prediction result of the present invention are applicable to predicting the tillage depth by real-time monitoring and analyzing information such as the angle of the tractor's lower link, the attitude of the tractor body, and the terrain, and dynamically adjusting the control parameters of the hydraulic system according to the prediction result to achieve precise control of the tillage depth. Its application scenarios cover multi-dimensional complex tillage environments and are applicable to multi-scale slope terrains. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0022] Figure 1 It is a schematic flow chart of the hydraulic tillage depth control method based on the tillage depth prediction result in an embodiment of the present invention; Figure 2 It is a schematic diagram of the tractor tillage data in an embodiment of the present invention; Figure 3 It is a schematic structural diagram of the BOA-TCN-Transformer-LSTM model in an embodiment of the present invention; Figure 4 It is a simulation schematic diagram of the PID controller based on the BP neural network in an embodiment of the present invention; In the figure, zeros(s): represents the zeros of the system, that is, the s value when the numerator of the transfer function is zero. The zeros have an important impact on the dynamic response of the system, and they can affect the stability and response speed of the system; poles(s): represents the poles of the system, that is, the s value when the denominator of the transfer function is zero. The poles determine the stability and dynamic characteristics of the system, such as oscillation, damping ratio, etc. The distribution of the poles in the complex plane directly affects the stability and response characteristics of the system; 1 / z: In a discrete-time system, z is a complex frequency domain variable, and 1 / z represents a delay of one sampling period. This module is used to simulate the transmission delay of the signal in the system and is a common delay link in discrete-time systems; BPPID: represents the PID controller based on the BP neural network; Figure 5 In an embodiment of the present invention, it is a schematic diagram of the simulation result of a PID controller based on a BP neural network; Figure 6 In an embodiment of the present invention, it is a schematic structural diagram of a PID controller based on a BP neural network; Figure 7 In an embodiment of the present invention, it is a schematic structural diagram of a PID control module integrating an RL policy network and a BP neural network. Detailed implementation manners

[0023] To make the technical solutions and advantages of the present invention more clearly expressed, the following will further describe in detail and completely the specific implementation manners of the present invention in conjunction with the accompanying drawings. The following described implementation manners are only some preferred solutions of the present invention, rather than all implementation solutions; the following described implementation manners are intended to explain the present invention and should not be construed as a limitation to the present invention; the reasonable combination of the technical features defined in each implementation manner of the present invention, and all other implementation manners obtained by those of ordinary skill in the art without creative efforts based on the implementation manners of the present invention belong to the scope of protection of the present invention.

[0024] Embodiment 1 provides a hydraulic tillage depth regulation method based on the tillage depth prediction result. The method includes the following steps: Step S1: Obtain the tractor tillage data collected in real time; the tractor tillage data includes the tractor lower link angle data, the tractor body attitude data, and the terrain data; preprocess the tractor tillage data to obtain multi-dimensional time series data; Step S2: Input the multi-dimensional time series data into the tillage depth prediction model to obtain the tillage depth prediction result; Step S3: Use the tillage depth prediction result as the given tillage depth, and calculate the regulation instruction by using the control module; the regulation instruction is used to adjust the angle of the tillage implement and the lifting of the suspension system, so as to perform the tillage operation according to the given tillage depth; Step S4: Obtain the real tillage depth data collected in real time; use the control module to adjust the regulation instruction according to the deviation between the real tillage depth data and the given tillage depth to complete the adaptive adjustment of the given tillage depth.

[0025] In this embodiment, the tractor tillage data is collected in real time by sensors and drones.

[0026] In this embodiment, the control module calculates the regulation instruction by using PID regulation or fuzzy control.

[0027] In this embodiment, the terrain data is DEM data, which contains slope information at different scales.

[0028] DEM data, namely Digital Elevation Model, is a raster dataset that expresses the surface elevation through an ordered numerical array. Its core is to construct a continuous terrain surface through discretized elevation points (such as at intervals of 3 meters or 1 meter per point), reflecting the terrain undulation characteristics of the region. In this embodiment, after the data acquisition is completed, the obtained DEM data is precisely processed to extract the slope information of the study area at (five) different scales.

[0029] Five different scales: flight altitude of 25 meters, image resolution of 0.68 cm; flight altitude of 30 meters, image resolution of 0.82 cm; flight altitude of 35 meters, image resolution of 0.96 cm; flight altitude of 40 meters, image resolution of 1.12 cm; flight altitude of 45 meters, image resolution of 1.23 cm.

[0030] In this embodiment, the tractor body attitude data includes the horizontal tilt angle and pitch angle of the tractor body during deep plowing operation.

[0031] In this embodiment, the tractor lower link angle data is the lifting angle of the link connecting the hydraulic reversible plow of the tractor relative to the horizontal plane.

[0032] In this embodiment, the real tillage depth data can be collected by using the "Online Tillage Depth Monitoring System and Method for a Towed Tillage and Soil Preparation Implement" described in the patent document with the publication number "CN119289888A" and the name "Online Tillage Depth Monitoring System and Method for a Towed Tillage and Soil Preparation Implement".

[0033] In addition, in one embodiment, in step S3, the angle of the tillage implement is adjusted: Based on the regulation strategy, the displacement of the hydraulic cylinder, the flow valve and the pressure valve (of the hydraulic system) are adjusted to optimize the (tillage) angle of the tillage implement.

[0034] In addition, in one embodiment, in step S3, the lifting of the suspension system is adjusted: Electro-hydraulic multi-objective optimization regulation is adopted to adjust the lifting of the suspension system: The electro-hydraulic proportional directional valve is adjusted by using a control module (such as PID control), and then the hydraulic oil flow rate is adjusted to realize the adjustment of the lifting of the suspension system.

[0035] In Embodiment 2, the tillage depth prediction model is the BOA-TCN-Transformer-LSTM model; The BOA-TCN-Transformer-LSTM model is obtained by using the Bayesian optimization algorithm to automatically tune the parameters of the TCN-Transformer-LSTM model and searching for the optimal hyperparameter configuration.

[0036] In this embodiment, the Bayesian optimization algorithm, whose full English name is "Bayesian Optimization Algorithm" and is abbreviated as "BOA".

[0037] In this embodiment, the Bayesian optimization algorithm (BOA) is used to intelligently search for the optimal hyperparameter configuration (automatically adjust hyperparameters) of the TCN-Transformer-LSTM model to improve the prediction performance.

[0038] Deep learning models (such as BOA-TCN-Transformer-LSTM) usually contain a large number of hyperparameters (such as learning rate, the convolutional kernel size of TCN, the number of attention heads of Transformer, the hidden layer dimension of LSTM, etc.). Manually adjusting these parameters is time-consuming and inefficient.

[0039] The Bayesian optimization algorithm automatically finds the hyperparameter combination that optimizes the model performance (such as minimizing the prediction error) through mathematical modeling and iterative sampling, avoiding the inefficiency of exhaustive search (such as grid search) or random search.

[0040] The Bayesian optimization process: Surrogate Model: Use Gaussian Process or Random Forest to model the relationship between hyperparameters and model performance.

[0041] Acquisition Function: According to the "potential gain" predicted by the surrogate model, select the next set of hyperparameters to be evaluated (such as Expected Improvement EI, Upper Confidence Bound UCB).

[0042] Iterative optimization: Repeat "sample hyperparameters → train the model → evaluate the performance → update the surrogate model" until the preset number of iterations or convergence is reached.

[0043] The advantages of Bayesian optimization: High efficiency: Through the surrogate model and the acquisition function, preferentially explore the hyperparameter regions with better performance and reduce ineffective attempts.

[0044] Adapt to complex objective functions: Suitable for complex relationships between model performance and hyperparameters that are non-linear and non-convex.

[0045] Automation: No manual intervention is required, which is suitable for large-scale hyperparameter tuning.

[0046] Bayesian optimization does not modify the model structure but only adjusts the hyperparameters. By using the optimized parameters, the feature extraction capabilities of the TCN, Transformer, and LSTM modules are maximally coordinated to improve the final prediction accuracy.

[0047] In summary, Bayesian optimization is an automated hyperparameter tuning tool that, through mathematical modeling and intelligent sampling, finds the optimal hyperparameter configuration for the TCN-Transformer-LSTM combined model in the tillage depth prediction task. The introduction of this method significantly reduces the cost of manual hyperparameter tuning and improves the generalization ability and prediction accuracy of the model.

[0048] The optimized poor parameters are shown in the following table:

[0049] In this embodiment, the BOA-TCN-Transformer-LSTM model combines the advantages of the Temporal Convolutional Network (TCN), Transformer encoder, and Long Short-Term Memory network (LSTM) to process multi-dimensional time series data.

[0050] The BOA-TCN-Transformer-LSTM model relies on the TCN (Temporal Convolutional Network) to extract local temporal features of the input data, uses the Transformer encoder to capture global dependencies, and finally performs multi-scale feature fusion through the LSTM (Long Short-Term Memory) to generate the tillage depth prediction result.

[0051] In this embodiment, the TCN-Transformer-LSTM model includes a TCN module, a Transformer encoder, and an LSTM module; The TCN module, the Transformer encoder, and the LSTM module are connected in series in sequence; The TCN module is used to extract local temporal features of multi-dimensional time series data; The Transformer encoder is used to capture global dependencies of multi-dimensional time series data; The LSTM module is used to perform multi-scale feature fusion to obtain the tillage depth prediction result.

[0052] In this embodiment, the TCN module, Temporal Convolutional Network, is a time convolutional network: The TCN module consists of multiple stacked temporal convolutional residual blocks (Temporal Block); Each residual block extracts local temporal patterns through causal convolution and dilated convolution to enhance the perception of short-term dynamic changes.

[0053] The output of this residual block is added to the output of the previous residual block and used as the input of the next residual block.

[0054] Each residual block includes two repeated tandem structures; each tandem structure consists of a dilated causal convolution layer, a weight normalization layer, a ReLU layer, and a Dropout layer in series.

[0055] In this embodiment, the Transformer encoder includes a multi-head attention layer (i.e., multi-head self-attention mechanism, Self-Attention) and a feed-forward layer.

[0056] Through the multi-head self-attention mechanism, it is possible to model long-term dependencies in the time series and improve the expression ability of global features.

[0057] In this embodiment, the LSTM module, Long Short-Term Memory, long short-term memory network: Further fuse time series information to capture long-term temporal dependencies and generate the final tillage depth prediction result.

[0058] In this embodiment, the LSTM module includes an LSTM layer and a fully connected layer (Fully Connected Layer); the fully connected layer is used to map the hidden state output by the LSTM layer to the target monitoring space to complete the final regression task.

[0059] In this embodiment, the training method of the BOA-TCN-Transformer-LSTM model: The dataset is divided into a training set and a test set with a division ratio of 4:1. The training set is used for training, and the test set is used to verify the model.

[0060] In this embodiment, the BOA-TCN-Transformer-LSTM model uses the Huber loss function to fine-tune the predicted output values.

[0061] Huber loss function: 。

[0063] In Embodiment 3, the control module includes a PID controller based on a BP neural network: The PID controller based on a BP neural network includes a BP neural network and a PID controller; The PID controller is used to perform closed-loop feedback control on the hydraulic system, calculate the control quantity of the hydraulic system in real time, and output a regulation instruction to the actuator of the hydraulic system; The BP neural network takes the system state (i.e., the state of the hydraulic system) as the input and the system performance index (i.e., the performance index of the hydraulic system) as the optimization goal, and adjusts the neural network weights and biases through the backpropagation algorithm to adaptively optimize the parameters of the PID controller.

[0064] In this embodiment, the BP neural network (Back Propagation Neural Network, that is, the backpropagation neural network): "BP" is an abbreviation of "Back Propagation", referring to a multi-layer feedforward neural network trained by the backpropagation algorithm. Its core structure includes an input layer, a hidden layer, and an output layer, and iteratively adjusts the weights and biases to minimize the prediction error.

[0065] In this embodiment, the PID controller is used to perform closed-loop feedback control on the controlled object (i.e., the hydraulic system) and regulate the flow rate and pressure of the hydraulic oil in the hydraulic system.

[0066] In this embodiment, the PID controller based on the BP neural network can be abbreviated as the BP-PID regulation system, which is a control strategy that combines the backpropagation (BP) neural network and the PID controller. It dynamically optimizes the PID parameters (proportional, integral, and differential coefficients) through the self-learning ability of the neural network, making the control algorithm have both the real-time performance of the PID and the adaptability of the neural network, and solving the problems of fixed traditional PID control parameters, poor adaptability, and insufficient stability of pure neural network control.

[0067] In this embodiment, the system performance indexes include overshoot, adjustment time, and steady-state error.

[0068] In this embodiment, the parameters of the PID controller include proportional, integral, and differential coefficients.

[0069] The proportional, integral, and differential parameters of PID control have the characteristics of cooperating with each other and restricting each other.

[0070] In this embodiment, by virtue of the mapping ability of the BP neural network for nonlinear functions, the optimal solutions of various nonlinear combinations of the three parameters in the PID controller can be obtained through self-learning.

[0071] In this embodiment, the BP neural network dynamically optimizes the parameters of the PID controller according to the system state and the learning algorithm, where: (1) System state: Refers to the real-time feedback variables of the input layer of the BP neural network, including: error (e) and rate of change of error (Δe). Error (e): The deviation between the reference input r (the given tillage depth) and the system output y (the real tillage depth data), i.e., e = r - y; Rate of change of error (Δe): The change in error at adjacent moments, usually expressed as Δe(k) = e(k) - e(k - 1). These state variables constitute the input feature vector of the neural network, reflecting the current dynamic characteristics of the system. For example, when there is a large error in the system, the network will first adjust the proportional term Kp; when the error persists (the integral characteristic is obvious), the Ki parameter will be optimized emphatically.

[0072] (2) Learning algorithm: Refers to the backpropagation algorithm (Backpropagation) that drives the optimization of neural network parameters, and its operation mechanism is as follows: Forward propagation: Input layer → Hidden layer → Output layer, generating the PID parameter adjustment amounts (ΔKp, ΔKi, ΔKd); Error backpropagation: According to the error function of the output layer (such as E = ½(r - y)²), calculate the gradient layer by layer through the chain rule of differentiation; Weight update: Use the gradient descent method to correct the network weights and biases, and the update formula is:

[0073] where η is the learning rate, Calculated through error backpropagation.

[0074] (3) The process of the BP neural network dynamically optimizing the PID parameters according to the system state and learning algorithm includes: Initialize the PID parameters and neural network weights → Collect the system output y and reference input r → Calculate the output of the PID controller → Adjust the parameters through the neural network → Iteratively update until the system is stable.

[0075] In the implementation process, the input layer of the BP neural network receives the system state variables (error and rate of change of error), the hidden layer processes the signals through the activation function, and the output layer generates the adjustment amounts of the PID parameters; through the error backpropagation algorithm, the neural network continuously optimizes the weights and biases, thereby realizing the dynamic adjustment of the PID parameters.

[0076] The combination of the backpropagation (BP) neural network and the PID controller breaks through the limitation of the linear combination of the traditional PID and can effectively cope with complex working conditions such as time-varying and non-linear.

[0077] In this embodiment, the advantages of the PID controller based on the BP neural network compared with the traditional PID control module (or traditional PID controller) are shown in the following table:

[0078] It should be noted that in the prior art, the "real-time acquisition of tillage depth data and traditional PID control" combination method is generally used for real-time control of tillage depth. In this embodiment, the "tillage depth prediction model and the PID controller based on the BP neural network" combination method is used for real-time control of tillage depth.

[0079] Compared with the traditional "real-time acquisition of tillage depth data and traditional PID control" combination method, the "tillage depth prediction model and the PID controller based on the BP neural network" combination method realizes dynamic adaptive decision-making through deep learning, can fuse multi-source data to predict the change trend of tillage depth, and actively optimize control parameters (such as hydraulic oil pressure), and dynamically adjust according to real-time changing information such as the tractor body attitude and terrain, significantly reducing the lag effect and mechanical loss. Its advantages are as follows: (1) Adapt to complex environments (such as soft and hard alternating soils) without manual parameter adjustment; (2) Automatically optimize strategies during long-term operations, improve energy efficiency, and ultimately achieve a tillage process with higher precision and lower energy consumption through intelligent decision-making.

[0080] (3) The method of combining the "tillage depth prediction model and the PID controller based on the BP neural network" has certain robustness and fault tolerance capabilities, and can resist data noise and interference to a certain extent. This means that even if there are certain errors or fluctuations in the real tillage depth data collected in real time, the neural network model can still output relatively accurate results. The traditional "real-time acquisition of tillage depth data and traditional PID control" combination method completely depends on the real-time collected data. If the data has errors or fluctuations, it may lead to a decline in the control effect.

[0081] In this embodiment, as Figure 5 shown, it is a schematic diagram of the simulation results of the PID controller based on the BP neural network; among them, the controller output is the output of the PID controller based on the BP neural network. It can be seen from the figure that: The large fluctuations in the controller output at the initial stage indicate that the controller has high flexibility. It can quickly adjust its output according to the real-time state and requirements of the system to more effectively control the behavior of the system. The system output is more affected by the controller output and has relatively low flexibility, mainly following the changes in the controller output. In the stable stage, the controller output gradually becomes stable, the fluctuation range significantly decreases, and finally stabilizes within a small range. This stability helps the system output to remain near the desired stable value, avoiding unnecessary oscillations or overshoots in the system output. The stability of the controller output has an important impact on the stability of the system output and is one of the key factors to ensure the stable operation of the system.

[0082] Embodiment 4, the control module includes a PID control module that integrates an RL policy network and a BP neural network; The PID control module that integrates an RL policy network and a BP neural network has a two-layer control architecture, including a bottom control layer and an upper optimization layer; The bottom control layer is a PID controller based on a BP neural network, which is used to calculate the control quantity of the hydraulic system in real time and output a control command to the actuator of the hydraulic system to ensure a quick response to the tillage depth deviation; The upper layer is a reinforcement learning RL policy network, which is used to dynamically optimize the PID controller parameters and adjust the BP neural network weights by analyzing the tillage depth error of the hydraulic system to improve the global adaptability.

[0083] In this embodiment, the control quantity of the hydraulic system, such as the displacement or flow rate of the hydraulic cylinder.

[0084] Further, in one embodiment, the PID controller based on a BP neural network as the bottom control layer includes a BP neural network and a PID controller; The PID controller is used to perform closed-loop feedback control on the hydraulic system, calculate the control quantity of the hydraulic system in real time, and output a control command to the actuator of the hydraulic system; The BP neural network uses the system state (i.e., the state of the hydraulic system) as the input and the system performance index (i.e., the performance index of the hydraulic system) as the optimization goal, and adjusts the neural network weights and biases through the backpropagation algorithm to adaptively optimize the parameters of the PID controller.

[0085] Further, in one embodiment, the reinforcement learning RL policy network (or RL algorithm) as the upper optimization layer adopts an Actor-Critic structure and a multi-objective optimization mechanism to keep the control strategy stable while exploring the optimal parameters. Among them: Multi-objective optimization mechanism: Design a reward function that takes into account the tillage depth tracking error, response speed, and energy consumption of the hydraulic system.

[0086] In this embodiment, aiming at the problems of non-linearity, time-variation and insufficient anti-interference ability of the farmland tillage depth hydraulic (control) system, a hierarchical intelligent control framework integrating reinforcement learning (RL policy network) and a PID controller based on BP neural network is proposed; the adaptive adjustment of tillage depth is realized through a double-layer optimization structure to meet the precise control requirements under complex farmland environments.

[0087] It should be noted that traditional PID control relies on manual parameter tuning in tillage depth adjustment and is difficult to adapt to changes in terrain and landform; although BP neural network PID has the ability of parameter self-tuning, it is prone to falling into local optimum under complex disturbances.

[0088] In this embodiment, to solve the above problems, a double-layer control architecture is constructed: The bottom layer is a PID controller based on BP neural network, which outputs control quantities (regulation instructions) to the hydraulic actuator (hydraulic system) in real time; the RL policy network is introduced in the upper layer, and the parameters of the PID controller are optimized online or the weights of the neural network are adjusted by analyzing the system states (such as pressure, flow rate, error signal).

[0089] In this embodiment, the double-layer control architecture is adopted to achieve high-precision and adaptive control of tillage depth in farmland operations, adapt to different terrain and landform changes, and improve the intelligent level of agricultural machinery operations.

[0090] In addition, in one embodiment, in step S3, the lifting of the suspension system is adjusted: Electro-hydraulic multi-objective optimization regulation is adopted to adjust the lifting of the suspension system: A PID controller based on BP neural network is used to adjust the electro-hydraulic proportional reversing valve, and then the hydraulic oil flow rate is adjusted to realize the adjustment of the lifting of the suspension system.

[0091] In addition, in one embodiment, a field experiment is carried out to verify the accuracy and robustness of the method (the BOA-TCN-Transformer-LSTM model outputs the tillage depth prediction result) in this embodiment, and it can meet the requirements of the consistency of fine finishing depth.

[0092] Specifically, the field experiment is as follows: The total area of the experimental plot is 178 mu, the terrain is relatively flat, and the soil type is typical black soil.

[0093] The agricultural machinery and equipment used in the experiment include Deutz-Fahr CD2004(G4) wheeled tractor and Lovol 1LFTT-4455R hydraulic reversible plow.

[0094] Record the tractor attitude, GPS position, and hitch angle information at a frequency of every 5 seconds, and match it with the terrain elevation data (collected by RTK drones) as the input of the model (tillage depth prediction model) to output the tillage depth prediction result.

[0095] To evaluate the model performance, four comparison models were set up, including: linear regression, K-nearest neighbor regression, random forest regression, and support vector regression.

[0096] The comparison results are as follows: Traditional models: The range of the coefficient of determination (R²) is between 0.53 - 0.65; the range of the root mean square error (RMSE) is 3.71 - 4.02 cm; the range of the mean absolute error (MAE) is 2.85 - 3.24 cm.

[0097] BOA-TCN-Transformer-LSTM model: The coefficient of determination (R²) is 0.765, the root mean square error (RMSE) is 1.978 cm, and the mean absolute error (MAE) is 1.489 cm.

[0098] In summary, compared with the traditional models, the proposed BOA-TCN-Transformer-LSTM model achieves the highest monitoring accuracy.

[0099] Embodiment 5 provides a hydraulic tillage depth control device based on the tillage depth prediction result. The device includes the following modules: Module S1: Obtain the real-time collected tractor tillage data; the tractor tillage data includes tractor lower hitch angle data, tractor body attitude data, and terrain data; preprocess the tractor tillage data to obtain multi-dimensional time series data; Module S2: Input the multi-dimensional time series data into the tillage depth prediction model to obtain the tillage depth prediction result; Module S3: Use the tillage depth prediction result as the given tillage depth, and calculate the control command by the control module; the control command is used to adjust the angle of the tillage implement and the lifting of the suspension system, so as to perform the tillage operation according to the given tillage depth; Module S4: Obtain the real-time collected actual tillage depth data; use the control module to adjust the control command according to the deviation between the actual tillage depth data and the given tillage depth to complete the adaptive adjustment of the given tillage depth.

[0100] Embodiment 6 provides a hydraulic tillage depth control system based on the tillage depth prediction result. The system includes a collection device, a hydraulic system, and the hydraulic tillage depth control device based on the tillage depth prediction result described in Embodiment 4; The collection device is used to collect the tractor tillage data and the actual tillage depth data in real time; The hydraulic tillage depth control device based on the tillage depth prediction result is used to output a control command according to the tractor tillage data and the true tillage depth data collected in real time; The hydraulic system is used to adjust the angle of the tillage implement and the lifting of the suspension system according to the control command, so as to perform the tillage operation according to the given tillage depth.

[0101] Embodiment Seven, the system further includes: a mode switching module, an alarm prompt module, and an operation interface module; The mode switching module is used to switch between the manual mode and the automatic mode; in the manual mode, the driver manually adjusts the tillage depth; in the automatic mode, the hydraulic tillage depth control device based on the tillage depth prediction result completes the adaptive adjustment of the given tillage depth; The alarm prompt module is used to send an alarm message when an abnormality occurs during the tractor tillage operation, reminding the driver to take corresponding measures; The operation interface module is used for human-computer interaction with the driver, including displaying the current tillage depth, displaying the alarm message, and setting the target value of the tillage depth.

[0102] In this embodiment, the occurrence of an abnormality during the tractor tillage operation includes that the tillage depth deviates from the predetermined range.

[0103] Embodiment Eight, provides a tractor, and the tractor includes the hydraulic tillage depth control system based on the tillage depth prediction result described in any one of the above.

[0104] Embodiment Nine, provides a computer device, including: a processor and a memory, the memory is used to store the executable instructions of the processor, and the processor is configured to execute the hydraulic tillage depth control method based on the tillage depth prediction result described in any one of the above by executing the executable instructions.

[0105] Embodiment Ten, provides a computer storage medium, and a computer program is stored in the storage medium. When the computer program runs, it executes the hydraulic tillage depth control method based on the tillage depth prediction result described in any one of the above.

[0106] Embodiment Eleven, provides a computer program product, including a computer program / instructions, and when the computer program / instructions are executed by a processor, the steps of the hydraulic tillage depth control method based on the tillage depth prediction result described in any one of the above are implemented.

[0107] A computer device or system provided by this embodiment, the hardware device of this part is a general model and is not shown in the form of a diagram. The system includes a processor and a memory. The processor and the memory can be connected through a bus or other means. The memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs and modules, as well as corresponding program instructions / modules. The processor executes various functional applications and data processing of the processor by running the non-transitory software programs, instructions and modules stored in the memory, so as to implement the data space entity parsing data quality enhancement method in the above method embodiments.

[0108] The memory may include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created by the processor and the like. In addition, the memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory may optionally include a memory remotely set relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above networks include but are not limited to the Internet, enterprise intranets, enterprise internal networks, mobile communication networks and their combinations.

[0109] One or more modules are stored in the memory. When the processor executes, the method steps in the embodiments are executed. In this way, through the method, device and process of the present invention, the invention purpose of the present invention can be achieved. The specific details of the above computer device can be understood by referring to the corresponding relevant descriptions and effects in the embodiments, and will not be elaborated here.

[0110] Those skilled in the art can understand that to implement all or part of the processes in the above method embodiments, it can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above method embodiments. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above types of memories.

[0111] The above further describes the technical solutions provided by the present invention through several specific embodiments to highlight the advantages and beneficial effects of the technical solutions provided by the present invention. However, the several specific embodiments described above are not used as a limitation to the present invention. Any reasonable modifications and improvements to the present invention, reasonable combinations of implementation manners, equivalent replacements, etc. within the spirit and principle of the present invention should be included within the protection scope of the present invention.

Claims

1. A hydraulic tillage depth control method based on tillage depth prediction results, characterized in that, The method includes the following steps: Step S1: Obtain the tractor tillage data collected in real time; the tractor tillage data includes the tractor lower link angle data, the tractor body attitude data, and the terrain data; preprocess the tractor tillage data to obtain multi-dimensional time series data; Step S2: Input the multi-dimensional time series data into the tillage depth prediction model to obtain the tillage depth prediction result; Step S3: Use the tillage depth prediction result as the given tillage depth, and use the control module to calculate the control command; the control command is used to adjust the angle of the tillage implement and the lifting of the suspension system, so as to perform the tillage operation according to the given tillage depth; Step S4: Obtain the real tillage depth data collected in real time; use the control module to adjust the control command according to the deviation between the real tillage depth data and the given tillage depth, and complete the adaptive adjustment of the given tillage depth.

2. The hydraulic tillage depth control method based on the tillage depth prediction result according to claim 1, wherein The tillage depth prediction model is the BOA-TCN-Transformer-LSTM model; The BOA-TCN-Transformer-LSTM model is obtained by using the Bayesian optimization algorithm to automatically tune the parameters of the TCN-Transformer-LSTM model and searching for the optimal hyperparameter configuration.

3. The hydraulic tillage depth control method based on the tillage depth prediction result according to claim 1, wherein The control module includes a PID control module that integrates the RL policy network and the BP neural network; The PID control module that integrates the RL policy network and the BP neural network is a two-layer control architecture, including a bottom control layer and an upper optimization layer; The bottom control layer is a PID controller based on the BP neural network, which is used to calculate the control quantity of the hydraulic system in real time and output the control command to the actuator of the hydraulic system to ensure a quick response to the tillage depth deviation; The upper layer is the reinforcement learning RL policy network, which is used to dynamically optimize the PID controller parameters and adjust the BP neural network weights by analyzing the tillage depth error of the hydraulic system to improve the global adaptability.

4. The hydraulic tillage depth control device based on the tillage depth prediction result is characterized in that The device includes the following modules: Module S1: Obtain the tractor tillage data collected in real time; the tractor tillage data includes the tractor lower link angle data, the tractor body attitude data, and the terrain data; preprocess the tractor tillage data to obtain multi-dimensional time series data; Module S2: Input the multi-dimensional time series data into the tillage depth prediction model to obtain the tillage depth prediction result; Module S3: Use the tillage depth prediction result as the given tillage depth, and use the control module to calculate the control command; the control command is used to adjust the angle of the tillage implement and the lifting of the suspension system, so as to perform the tillage operation according to the given tillage depth; Module S4: Obtain the real tillage depth data collected in real time; Use the control module to adjust the control command according to the deviation between the real tillage depth data and the given tillage depth, and complete the adaptive adjustment of the given tillage depth.

5. The hydraulic tillage depth control system based on the tillage depth prediction result is characterized in that, The system includes a collection device, a hydraulic system, and the hydraulic tillage depth control device based on the tillage depth prediction result as claimed in claim 4; The collection device is used to collect the tractor tillage data and the real tillage depth data in real time; The hydraulic tillage depth control device based on the tillage depth prediction result is used to output the control command according to the tractor tillage data and the real tillage depth data collected in real time; The hydraulic system is used to adjust the angle of the tillage implement and the lifting of the suspension system according to the control instructions, so as to perform tillage operations according to the given tillage depth.

6. The hydraulic tillage depth control system based on the tillage depth prediction result according to claim 5, wherein The system further includes: a mode switching module, an alarm prompting module, and an operation interface module; The mode switching module is used to switch between manual mode and automatic mode; in manual mode, the driver manually adjusts the tillage depth; in automatic mode, the hydraulic tillage depth control device based on the tillage depth prediction result completes the adaptive adjustment of the given tillage depth; The alarm prompting module is used to send an alarm message when an abnormality occurs during the tillage operation of the tractor, reminding the driver to take corresponding measures; The operation interface module is used for human-machine interaction with the driver, including displaying the current tillage depth, displaying the alarm message, and setting the target value of the tillage depth.

7. A tractor, characterized in that, The tractor includes the hydraulic tillage depth control system based on the tillage depth prediction result described in claim 5 or 6.

8. A computer device, comprising: A processor and a memory, characterized in that the memory is used to store executable instructions of the processor, and the processor is configured to execute the hydraulic tillage depth control method based on the tillage depth prediction result described in any one of claims 1-3 by executing the executable instructions.

9. A computer storage medium, characterized in that, A computer program is stored in the storage medium, and when the computer program runs, it executes the hydraulic tillage depth control method based on the tillage depth prediction result described in any one of claims 1-3.

10. A computer program product, comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, the steps of the hydraulic tillage depth control method based on the tillage depth prediction result described in any one of claims 1-3 are implemented.

Images