An intelligent agricultural machinery control optimization method and system based on mathematical decoupling and deep learning

Abstract

The present invention discloses an intelligent agricultural machinery control optimization method and system based on mathematical decoupling and deep learning. The method includes constructing a simulation model of the intelligent agricultural machinery; collecting in-vehicle field operation data; based on the simulation model, restricting the excitation source to participate in the simulation analysis by changing parameters to obtain simulation data; qualitatively analyzing the influence of each excitation source on the operation quality based on the simulation data to obtain strong adverse excitation sources; performing quantitative analysis using the in-vehicle field operation data to establish a non-linear mapping model; based on the non-linear mapping model, obtaining the system characteristics corresponding to the agricultural machinery in the real operation environment; and optimizing the control parameters of the intelligent agricultural machinery based on the system characteristics and the determined strong adverse excitation sources. The present invention adopts a strategy formulation method combining theoretical simulation and processing of the collected in-vehicle data, which can effectively improve the rationality of the intelligent agricultural machinery control optimization.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic operation of intelligent agricultural machinery, and in particular to an intelligent agricultural machinery control optimization method and system based on mathematical decoupling and deep learning. Background Art

[0002] As smart agricultural machinery integrates more and more functions, such as rotary tillage, sowing, fertilization, etc., these devices usually use a variety of drive components, including traditional mechanical transmission systems and modern motors and hydraulic systems. Although this multifunctional integration improves work efficiency, it also brings new challenges. In actual operation, when the tractor needs to switch from one working state to another, due to the differences in response speed and characteristics between different actuators, it is easy to have operational inconsistencies, such as sudden acceleration or deceleration of the vehicle, and impact when shifting gears. These problems directly affect the driver's comfort experience. Especially in complex and changeable field conditions, this effect is particularly obvious when the speed and direction are frequently adjusted to adapt to different soil conditions or crop requirements. In addition, being in such a working environment for a long time not only increases the driver's physical fatigue, but may also reduce overall work efficiency. Therefore, in the design of modern smart agricultural machinery, how to improve the tractor's operating smoothness and reduce the discomfort caused by function switching by optimizing the control system has become the key to improving user experience.

[0003] In the prior art, the applicant's prior application CN 116476842A discloses a startup control method for intelligent agricultural machinery to improve driving comfort, but the method of the patent relies on a specially designed transmission system and has a high cost. In other prior arts, control strategies are generally formulated by establishing a simulation model of agricultural machinery for simulation. For example, patent CN103048148A discloses a semi-physical simulation test platform for a high-power tractor electro-hydraulic suspension system, which is used for the research and development, production and performance testing of key components of high-power tractors through simulation. This method lacks comprehensive consideration of the complex factors that may be encountered in actual use. Summary of the invention

[0004] Purpose of the invention: In order to overcome the deficiencies in the prior art, the present invention provides an intelligent agricultural machinery control optimization method and system based on mathematical decoupling and deep learning.

[0005] Technical solution: To achieve the above purpose, the intelligent agricultural machinery control optimization method based on mathematical decoupling and deep learning of the present invention comprises:

[0006] Construct a simulation model of intelligent agricultural machinery to describe the relationship between various excitation sources and transmission systems in intelligent agricultural machinery;

[0007] Collect real vehicle field operation data;

[0008] Based on the simulation model, the excitation source is restricted from participating in the simulation analysis by changing parameters to obtain simulation data;

[0009] Based on the simulation data, a qualitative analysis is performed on the influence of each excitation source on the operation quality to obtain strong adverse excitation sources;

[0010] Quantitative analysis is carried out using real vehicle field operation data to establish a non-linear mapping model;

[0011] Based on the non-linear mapping model, the system characteristics corresponding to the agricultural machinery in the real operation environment are obtained; the system characteristics include: the dynamic torque response speed of each working system and the influence relationship between them, and the mapping relationship of the control and delay response characteristics to the power output and the engine operating point;

[0012] Based on the system characteristics and the determined strong adverse excitation sources, the control parameters of the intelligent agricultural machinery are optimized.

[0013] Furthermore, the construction of the simulation model of the intelligent agricultural machinery specifically includes:

[0014] An engine model, a hydraulic system model, and a PTO load model are established respectively;

[0015] The engine model, the hydraulic system model, and the PTO load model are coupled according to the torque balance equation based on the transmission system to form a comprehensive power transmission system model of mechanical-electrical-hydraulic multi-field coupling;

[0016] An environmental factor model applied to the transmission system model is established to obtain the complete simulation model.

[0017] Furthermore, the obtaining of the simulation data by restricting one or more excitations from participating in the simulation analysis based on the simulation model includes:

[0018] Based on the simulation model in sequence, a simulation analysis of the individual action of each excitation source is carried out to evaluate the dynamic response characteristics of each excitation source acting independently on the transmission system under different working conditions to obtain decoupling data;

[0019] Data acquisition of the coupled action of multiple excitations is carried out to obtain coupling data.

[0020] Furthermore, the analysis of the influence of each excitation source on the operation quality based on the simulation data to obtain strong adverse excitation sources includes the following steps:

[0021] Based on the decoupling data and the coupling data, a sensitivity analysis is performed on the influence of each excitation source on the dynamic characteristics of the transmission system output torque;

[0022] Based on the sensitivity analysis results, calculate the influence factor of each excitation source;

[0023] Arrange all the excitation sources in descending order of their influence factors to form a priority list;

[0024] According to the influence factor corresponding to the excitation source and specific torque fluctuation and response delay time indexes, construct a multi-dimensional feature vector corresponding to the excitation source;

[0025] Based on the multi-dimensional feature vector, perform clustering, and determine the strong adverse excitation sources through the analysis of the clustering results.

[0026] Furthermore, the use of actual vehicle field operation data for quantitative analysis and the establishment of a non-linear mapping model include:

[0027] Use the variable parameter One-Class-SVM algorithm to clean the actual vehicle field operation data and remove unreasonable outliers;

[0028] Organize the cleaned data into a time series database;

[0029] Use the LSTM algorithm to train the time series data and establish a non-linear mapping model;

[0030] Adjust the model parameters through the validation set.

[0031] Furthermore, based on the system characteristics and the determined strong adverse excitation sources, optimize the control parameters of the intelligent agricultural machinery, including:

[0032] For each identified strong adverse excitation source, according to its influence degree under different working conditions, adopt the feed-forward compensation algorithm to adjust the control parameters;

[0033] Combined with the dynamic torque response speed in the system characteristics, use the model predictive control algorithm to optimize the throttle opening θ;

[0034] According to the influence relationship between the dynamic torques of each working system in the system characteristics, use the adaptive shift logic algorithm to adjust the transmission gear g.

[0035] Adopt a delay compensation mechanism to offset the known response delay through the time lag term τ;

[0036] Continuously adjust the control parameters through real-time monitoring and iterative learning control algorithm to improve the operation efficiency and driving comfort.

[0037] The present invention also provides an intelligent agricultural machinery control optimization system based on mathematical decoupling and deep learning, which includes:

[0038] A building module, which is used to build a simulation model of an intelligent agricultural machine for describing the association between each excitation source and the transmission system in the intelligent agricultural machine;

[0039] An acquisition module, which is used to acquire on-vehicle field operation data;

[0040] A simulation module, which is used to limit the participation of the excitation source in the simulation analysis by changing parameters based on the simulation model to obtain simulation data;

[0041] A first analysis module, which is used to qualitatively analyze the influence of each excitation source on the operation quality based on the simulation data to obtain strong adverse excitation sources;

[0042] A second analysis module, which is used to perform quantitative analysis using on-vehicle field operation data to establish a non-linear mapping model;

[0043] An application module, which is used to obtain the system characteristics corresponding to agricultural machinery in the real operation environment based on the non-linear mapping model; the system characteristics include: the dynamic torque response speed of each working system and the influence relationship between them, and the mapping relationship between the control and delay response characteristics and the power output and engine operating point;

[0044] An optimization control module, which is used to optimize the control parameters of the intelligent agricultural machine based on the system characteristics and the determined strong adverse excitation sources.

[0045] Beneficial effects: The intelligent agricultural machine control optimization method and system based on mathematical decoupling and deep learning adopt a strategy formulation method that combines theoretical simulation and processing of on-vehicle collected data. On the one hand, based on the theoretical structure of agricultural machinery, simulation is carried out and qualitatively analyzed to determine strong adverse excitation sources that have a greater impact on driving comfort. On the other hand, quantitative analysis is carried out based on on-vehicle operation collected data to obtain the system characteristics in the real operation environment, which can make up for the limitations of the simulation model. Targeted control of the intelligent agricultural machine based on strong adverse excitation sources and system characteristics can effectively improve the rationality of the control of the intelligent agricultural machine during operation and enhance driving comfort. Description of the Drawings

[0046] Figure 1 It is a schematic flow chart of the intelligent agricultural machine control optimization method based on mathematical decoupling and deep learning;

[0047] Figure 2 It is a schematic diagram of the intelligent agricultural machine data acquisition system;

[0048] Figure 3 It is a schematic diagram of the composition of the intelligent agricultural machine control optimization device based on mathematical decoupling and deep learning. Detailed Embodiments

[0049] The present invention will be further described in conjunction with the accompanying drawings.

[0050] As Figure 1 shown, the intelligent agricultural machinery control optimization method based on mathematical decoupling and deep learning of the present invention includes the following steps S101 - S107:

[0051] Step S101, construct a simulation model of the intelligent agricultural machinery for describing the association between various excitation sources and the transmission system in the intelligent agricultural machinery;

[0052] In this step, the simulation model is constructed based on the mechanical structure of the intelligent agricultural machinery. The simulation model describes the connection between each component in the intelligent agricultural machinery and the transmission system. Each component can generate an excitation acting on the transmission system, that is, acting as an excitation source. The connection between each part and the transmission system includes mechanical transmission relationship, hydraulic coupling relationship, electrical connection relationship, traction / hanging connection relationship (the component is towed behind the tractor or can be lifted up and down); the transmission system includes the engine and the gearbox. The excitation sources include each component that can affect the engine harmonic torque, hydraulic working dynamic torque, and PTO working torque, such as: seeding mechanism, fertilizing mechanism, workbench hydraulic lifting mechanism, rotary tillage mechanism, fan mechanism, etc. The state switching of some components can instantaneously increase the load on the transmission system and have an obvious impact on driving comfort, while some components have a smaller impact on driving comfort. The parameter data ranges corresponding to the above - mentioned components can be obtained by analyzing the field operation data of the actual vehicle. For example, when the workbench is in the lifted state and the lowered state, the hydraulic working dynamic torque can be measured.

[0053] Step S102, collect the field operation data of the actual vehicle;

[0054] In this step, as Figure 2 shown, torque sensors are installed on the front and rear drive shafts and the PTO shaft of the tractor in the intelligent agricultural machinery, and a sound level meter is installed at the gearbox housing. The torque and noise data are collected by using a multi - channel data acquisition system; the real - time parameters of the engine, throttle position, gear, and hydraulic system in the agricultural machinery CAN bus are read by using D2P rapid prototyping; all the data are aggregated by D2P and unified in time scale, and uploaded to the upper computer for recording and storage through Kvaser. Among them, the noise spectrum is collected by using the sound level meter for auxiliary analysis.

[0055] Step S103, based on the simulation model, limit the excitation sources to participate in the simulation analysis by changing parameters to obtain simulation data;

[0056] In this step, the simulation data is collected from the simulation model. After inputting the input data corresponding to the excitation source participating in the simulation, the simulation data is the output data obtained from a specific position of the simulation model. These output data can be used as indicators to measure the operation quality of the intelligent agricultural machinery. In this embodiment, the torque fluctuation and response delay time of the PST output shaft are used as the above indicators. The torque fluctuation can effectively reflect the indicators affecting driving comfort such as jitter, jerks, and noise during the operation of the intelligent agricultural machinery. The response delay can reflect the time delay between the input data and the generation of the response of the indicators.

[0057] Step S104, based on the simulation data, qualitatively analyze the influence of each excitation source on the operation quality to obtain strong adverse excitation sources. Here, a strong adverse excitation source refers to an excitation source that can have an obvious impact on driving comfort.

[0058] In this step, mathematical modeling and simulation analysis methods are used to qualitatively find the dynamic torque response law of the transmission system under coupled excitation. Since it is based on a mathematical model, many system parameters are assumed and it is difficult to be accurate. Therefore, the results cannot fully conform to the actual situation, but the law trend can be referenced. Therefore, it is a qualitative analysis. The strong adverse excitation sources determined based on this qualitative analysis have great reference significance.

[0059] Step S105, use the actual vehicle field operation data for quantitative analysis to establish a non - linear mapping model.

[0060] Step S106, based on the non - linear mapping model, obtain the system characteristics corresponding to the agricultural machinery in the real operation environment. The system characteristics include: the dynamic torque response speed of each working system and the influence relationship between them, and the mapping relationship between the control and delay response characteristics and the power output and engine operating point. The above - mentioned dynamic torque of the working system includes the PST output shaft torque, PTO torque, engine torque, and hydraulic pump shaft torque.

[0061] Step S107, based on the system characteristics and the determined strong adverse excitation sources, optimize the control parameters of the intelligent agricultural machinery. The control parameters include throttle opening, transmission gear, and delay compensation.

[0062] The above - mentioned intelligent agricultural machinery automatic operation strategy optimization method adopts a strategy formulation method that combines theoretical simulation and processing of actual vehicle collected data. On the one hand, based on the theoretical structure of the agricultural machinery, simulation is carried out and qualitatively analyzed to determine strong adverse excitation sources that have a greater impact on driving comfort. On the other hand, quantitative analysis is carried out based on the actual vehicle operation collected data to obtain the system characteristics in the real operation environment, which can make up for the limitations of the simulation model. Targeted control of the intelligent agricultural machinery based on strong adverse excitation sources and system characteristics can effectively improve the rationality of the control optimization of the intelligent agricultural machinery and enhance driving comfort.

[0063] The construction of the simulation model of the intelligent agricultural machinery described in the above step S101 specifically includes the following steps S201 - S202:

[0064] Step S201, establish an engine model, a hydraulic system model, and a PTO load model respectively;

[0065] Step S202, couple the engine model, the hydraulic system model, and the PTO load model according to the torque balance equation based on the transmission system to form a comprehensive power transmission system model with multi - field coupling of machine - electricity - hydraulics;

[0066] Step S203, establish an environmental factor model applied to the transmission system model to obtain the complete simulation model. The environmental factor model includes soil resistance, wind resistance, ground slope, etc.

[0067] In the above simulation system, the harmonic torque output by the engine model is coupled and transmitted to the input shaft of the transmission system to simulate the excitation effect of the engine on the transmission system; the hydraulic pump in the hydraulic system model is connected to the transmission system through a hydraulic pipeline to form a dynamic coupling relationship between the hydraulic pump torque and the transmission system; the output torque of the PTO model is directly transmitted to the PTO shaft of the transmission system, and the PTO shaft in the coupling model rotates synchronously with other output shafts of the transmission system to simulate the working conditions of various working tools under the action of the PTO.

[0068] In one embodiment, the simulation data includes decoupled data obtained by each of the excitation sources participating in the simulation alone and coupled data obtained by multiple excitation sources participating in the simulation; in the above step S103, based on the simulation model, by changing parameters to limit one or more excitations to participate in the simulation analysis to obtain simulation data, which includes the following steps S301 - S302:

[0069] Step S301, sequentially perform simulation analysis on each of the excitation sources based on the simulation model for individual actions, evaluate the dynamic response characteristics of each of the excitation sources acting independently on the transmission system under different working conditions, and obtain decoupled data;

[0070] Step S302, perform data acquisition for the coupled action of multiple excitations to obtain coupled data.

[0071] In another embodiment, the simulation data includes decoupled data obtained by each of the excitation sources participating in the simulation alone; in addition, coupled data is obtained by analyzing the actual vehicle operation data.

[0072] The analysis of the influence of each of the excitation sources on the operation quality based on the simulation data in the above step S104 to obtain strong adverse excitation sources includes the following steps S401 - S405:

[0073] Step S401: Based on the decoupled data and the coupled data, perform a sensitivity analysis on the influence of each excitation source on the dynamic characteristics of the transmission system output torque, so as to evaluate the influence degree of each excitation source on the torque fluctuation and response delay time of the PST output shaft under different working conditions; this process identifies which excitation sources have a significant impact on torque fluctuation and response delay time under different working conditions;

[0074] Step S402: Based on the sensitivity analysis results, calculate the influence factor of each excitation source, which reflects the contribution degree of the excitation source to the change of the transmission system output characteristics;

[0075] Step S403: Arrange all the excitation sources in descending order of their influence factors to form a priority list, where the excitation source with a larger influence factor indicates a greater influence on the transmission system;

[0076] Step S404: According to the influence factor corresponding to the excitation source and the specific torque fluctuation and response delay time indexes, construct a multi-dimensional feature vector corresponding to the excitation source; the feature vector contains the key attribute information of the excitation source;

[0077] Step S405: Based on the multi-dimensional feature vector, perform clustering, and through the analysis of the clustering results, determine the strong adverse excitation sources.

[0078] In the above steps, the purpose of clustering is to find groups of excitation sources with similar behavior patterns, so as to better understand their common influence on the operation quality of intelligent agricultural machinery. Through the analysis of the clustering results, especially those categories containing excitation sources with high influence factors, it is possible to further confirm which excitation sources are the main factors causing phenomena such as gearshift jitter and jerks, and define them as strong adverse excitation sources.

[0079] The quantitative analysis using the actual vehicle field operation data in the above step S105 to establish a non-linear mapping model includes the following steps S501 - S504:

[0080] Step S501: Use the variable parameter One-Class-SVM algorithm to clean the actual vehicle field operation data and remove unreasonable outliers;

[0081] Step S502: Organize the cleaned data into a time series database;

[0082] Step S503: Use the LSTM algorithm to train the time series data and establish a non-linear mapping model;

[0083] Step S504: Adjust the model parameters through the validation set to ensure that the model has good generalization ability.

[0084] In step S106 above, based on the non-linear mapping model, the system characteristics corresponding to the agricultural machinery in the actual operating environment are obtained, including the following steps S601 - S603:

[0085] Step S601: Use the non-linear mapping model to predict the dynamic torque response speed of each working system in the actual operating environment;

[0086] Step S602: Analyze the mutual influence relationship of dynamic torques between different working systems;

[0087] Step S603: Determine the mapping relationship between the control and delay response characteristics and the power output and engine operating point.

[0088] In step S107 above, based on the system characteristics and the determined strong adverse excitation sources, the control parameters of the intelligent agricultural machinery are optimized, including the following steps S701 - S705:

[0089] Step S701: For each identified strong adverse excitation source, according to its influence degree under different working conditions, use the feedforward compensation algorithm to adjust the control parameters;

[0090] In this step, assuming that the influence of a certain strong adverse excitation source I can be expressed as a function f(t), then the feedforward compensation amount C f can be calculated by the following formula: C f (t) = -K f ·f(t) where K f is the feedforward gain coefficient, which is determined by experimental or simulation tuning.

[0091] Step S702: Combine the dynamic torque response speed in the system characteristics and use the model predictive control algorithm to optimize the throttle opening θ;

[0092] In this step, the objective function J of the model predictive control algorithm can be defined as: where e(k) is the tracking error, u(k) is the control input, w 1 and w 2 are the weight coefficients, and N is the prediction horizon length;

[0093] Step S703: According to the mutual influence relationship of the dynamic torques of each working system in the system characteristics, use the adaptive shifting logic algorithm to adjust the transmission gear g.

[0094] In this step, the adaptive shifting logic algorithm determines the optimal gear based on the current engine speed n, vehicle speed v, and load torque TL. The decision rule can be expressed as: g * = argmin g (|n - n +∣+α·∣v - v + ∣+β·∣T , -T ,+ ∣) where n + , v + , and T ,+ are the target engine speed, target vehicle speed, and target load torque respectively, and α and β are adjustment factors.

[0095] Step S704, adopt a delay compensation mechanism to offset the known response delay through the time lag term τ;

[0096] In this step, for a control system with delay, compensation can be performed according to the following formula: G - (s) = G . (s) / (1 + τ / ); where G . (S) is the transfer function of the controlled object, and G - (s) is the transfer function of the controller including delay compensation;

[0097] Step S705, continuously adjust the control parameters through real-time monitoring and iterative learning control algorithm to improve the operation efficiency and driving comfort. After each operation, update the controller parameter ΔK, and its update formula can be expressed as: K 012 = K 34+ + η·ΔK, where η is the learning rate, K new is the new control parameter, and K 34+ is the old control parameter.

[0098] The present invention also discloses an intelligent agricultural machinery control optimization system based on mathematical decoupling and deep learning. The optimization system can include or be divided into one or more program modules. One or more program modules are stored in a storage medium and executed by one or more processors to complete the present invention and implement the above optimization method. The program modules referred to in the embodiments of the present invention refer to a series of computer program instruction segments that can complete specific functions, and are more suitable for describing the execution process of the optimization method in the storage medium than the program itself. The following description will specifically introduce the functions of each program module in this embodiment, as Figure 3 shown, which includes:

[0099] A construction module 801, which is used to construct a simulation model of the intelligent agricultural machinery to describe the association between each excitation source and the transmission system in the intelligent agricultural machinery;

[0100] An acquisition module 802, which is used to acquire real vehicle field operation data;

[0101] A simulation module 803, which is used to limit the participation of the excitation source in the simulation analysis by changing parameters based on the simulation model to obtain simulation data;

[0102] A first analysis module 804, which is configured to perform qualitative analysis on each of the excitation source pairs based on the simulation data to obtain strong adverse excitation sources;

[0103] A second analysis module 805, which is configured to perform quantitative analysis using the actual vehicle field operation data to establish a non - linear mapping model;

[0104] An application module 806, which is configured to obtain system characteristics corresponding to the agricultural machinery in the real operation environment based on the non - linear mapping model; the system characteristics include: the dynamic torque response speeds of each working system and the influence relationship between them, and the mapping relationship between the control and delay response characteristics and the power output and the engine operating point;

[0105] An optimization control module 807, which is configured to optimize the control parameters of the intelligent agricultural machinery based on the system characteristics and the determined strong adverse excitation sources.

[0106] The content of other implementations of the above - mentioned intelligent agricultural machinery control optimization method based on the intelligent agricultural machinery control optimization system has been introduced in detail in the previous embodiments. For corresponding content in the previous embodiments, reference can be made, and details are not repeated here.

[0107] The above - mentioned are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art in this technical field, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. An intelligent agricultural machinery control optimization method based on mathematical decoupling and deep learning, characterized in that: The method comprises: Construct a simulation model of intelligent agricultural machinery to describe the relationship between various excitation sources and transmission systems in intelligent agricultural machinery; Collect real vehicle field operation data; Based on the simulation model, one or more excitations are restricted from participating in the simulation analysis by changing parameters to obtain simulation data; Based on the simulation data, a qualitative analysis is performed on the influence of each of the excitation sources on the operation quality to obtain a strong adverse excitation source; Use the actual vehicle field operation data for quantitative analysis and establish a nonlinear mapping model; Based on the nonlinear mapping model, the system characteristics corresponding to the agricultural implement in the real working environment are obtained; the system characteristics include: the dynamic torque response speed of each working system and the influence relationship between them, and the mapping relationship between the control and delayed response characteristics and the power output and the engine working point; Based on the system characteristics and the determined strong adverse incentive source, the control parameters of the intelligent agricultural machinery are optimized; the control parameters include: throttle opening, gearbox gear and delay compensation; The simulation model is based on which one or more excitations are restricted from participating in the simulation analysis by changing parameters to obtain simulation data, including: Based on the simulation model, simulation analysis is performed on each of the excitation sources acting alone in turn, and the dynamic response characteristics of each of the excitation sources acting independently on the transmission system under different working conditions are evaluated to obtain decoupling data; Carry out data collection of multi-channel excitation coupling to obtain coupling data; The method of analyzing the influence of each of the excitation sources on the operation quality based on the simulation data to obtain a strong unfavorable excitation source comprises the following steps: Based on the decoupling data and the coupling data, a sensitivity analysis is performed on the influence of each excitation source on the dynamic characteristics of the output torque of the transmission system; Based on the sensitivity analysis results, calculate the impact factor of each excitation source; Arrange all motivation sources in descending order of their impact factors to form a priority list; Constructing a multidimensional feature vector corresponding to the excitation source according to the influencing factors corresponding to the excitation source and specific torque fluctuation and response delay time indicators; Performing clustering based on the multi-dimensional feature vector, and determining the strong adverse incentive source by analyzing the clustering result; The method of using the actual vehicle field operation data for quantitative analysis and establishing a nonlinear mapping model includes: Use variable parameter One-Class-SVM algorithm to clean the field operation data of real vehicles and remove unreasonable outliers; Organize the cleaned data into a time series database; Use LSTM algorithm to train time series data and establish nonlinear mapping model; Tune model parameters using the validation set.

2. The intelligent agricultural machinery control optimization method based on mathematical decoupling and deep learning according to claim 1 is characterized in that: The construction of the simulation model of intelligent agricultural machinery specifically includes: Establish engine model, hydraulic system model and PTO load model respectively; The engine model, the hydraulic system model and the PTO load model are coupled according to the torque balance equation based on the transmission system to form a comprehensive power transmission system model of mechanical-electrical-hydraulic multi-field coupling; An environmental factor model applied to the transmission system model is established to obtain the complete simulation model.

3. The intelligent agricultural machinery control optimization method based on mathematical decoupling and deep learning according to claim 1 is characterized in that: The optimizing the control parameters of the intelligent agricultural machinery based on the system characteristics and the determined strong adverse excitation source includes: For each identified strong adverse excitation source, the control parameters are adjusted using the feedforward compensation algorithm according to its influence under different working conditions. The feedforward compensation amount corresponding to the feedforward compensation algorithm is Calculated by the following formula: ,in is the feedforward gain coefficient, function Strong adverse incentive I The impact of Combined with the dynamic torque response speed in the system characteristics, the model predictive control algorithm is used to optimize the throttle opening θ ; The objective function J of the model predictive control algorithm is defined as: ; in e ( k ) is the tracking error, u ( k ) is the control input, w 1 and w 2 is the weight coefficient, N is the prediction time domain length; According to the influence relationship between the dynamic torque of each working system in the system characteristics, the adaptive shifting logic algorithm is used to adjust the gearbox gear g ; The adaptive shift logic algorithm is based on the current engine speed n , vehicle speed v and load torque T L To determine the optimal gear, the decision rule is expressed as: ,in , , and They are target engine speed, target vehicle speed and target load torque. α and β is the regulating factor; Using the delay compensation mechanism, the time lag term To offset the known response delay; for control systems with delay, compensation is performed using the following formula: in is the transfer function of the controlled object, is the controller transfer function including delay compensation; Continuously adjust control parameters through real-time monitoring and iterative learning control algorithms to improve work efficiency and driving comfort; update controller parameters after each operation ΔK , the update formula is expressed as: ,in η is the learning rate, is the new control parameter, is the old control parameter.

4. An intelligent agricultural machinery control optimization system based on mathematical decoupling and deep learning, used to implement the intelligent agricultural machinery control optimization method based on mathematical decoupling and deep learning described in claim 1, characterized in that: It includes: A construction module, which is used to construct a simulation model of the intelligent agricultural machinery and to describe the relationship between each excitation source and the transmission system in the intelligent agricultural machinery; A collection module, which is used to collect field operation data of real vehicles; A simulation module, which is used to obtain simulation data by changing parameters to limit one or more excitations to participate in simulation analysis based on the simulation model; A first analysis module, which is used to analyze the influence of each of the excitation sources on the operation quality based on the simulation data to obtain a strong adverse excitation source; The second analysis module is used to perform quantitative analysis using actual vehicle field operation data and establish a nonlinear mapping model; An application module, which is used to obtain system characteristics corresponding to agricultural machinery in a real working environment based on the nonlinear mapping model; the system characteristics include: dynamic torque response speed of each working system and the influence relationship between them, and the mapping relationship between control and delayed response characteristics and power output and engine working point; An optimization control module is used to optimize the control parameters of the intelligent agricultural machinery based on the system characteristics and the determined strong adverse excitation source; the control parameters include: throttle opening, gearbox gear and delay compensation.

Images