An intelligent agricultural machinery performance improvement method and system

Abstract

The present invention discloses a method and system for improving the performance of intelligent agricultural machinery. The method includes sensing data processing, virtual commissioning of agricultural machinery, optimization of agricultural machinery operations, and improvement of agricultural machinery performance. The present invention relates to the technical field of intelligent management of agricultural machinery, specifically referring to a method and system for improving the performance of intelligent agricultural machinery. This solution adopts sequential optimization in two dimensions of virtual commissioning of agricultural machinery and optimization of agricultural machinery operations, and deeply conducts digital modeling from virtual to actual, systematically solving the adaptability problem of agricultural machinery in complex farmland environments, improving operation accuracy and efficiency, while taking into account resource conservation and soil protection, meeting the actual needs of precision agriculture; adopts a dynamic digital twin mirror optimization method for virtual commissioning of agricultural machinery, and constructs a high-precision virtual farmland model through dynamic weight factors; adopts a multi-objective hierarchical adaptive reinforcement learning method for optimization of agricultural machinery operations.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent management of agricultural machinery, and specifically refers to an intelligent method and system for improving the performance of agricultural machinery. Background Art

[0002] The intelligent method and system for improving the performance of agricultural machinery is a comprehensive solution based on advanced sensing technology, digital twin, artificial intelligence, and the Internet of Things. It aims to significantly improve the operation efficiency, accuracy, and intelligence level of agricultural machinery through steps such as real-time data collection, virtual commissioning, operation optimization, and performance improvement. The system constructs a high-precision virtual farmland model and an agricultural machinery working condition model, combines multi-source data fusion and reinforcement learning algorithms, realizes autonomous decision-making and optimized operation of agricultural machinery in complex farmland environments, and at the same time takes into account resource conservation and soil protection, promotes the development of precision agriculture, reduces agricultural production costs, improves the quality and yield of agricultural products, and contributes to agricultural modernization and sustainable development.

[0003] However, in the existing methods for improving the performance of agricultural machinery, the environment for intelligent farmland operations is complex and changeable, resulting in problems of insufficient precision in the intelligent and automated processes of agricultural machinery operations. At the same time, the existing commissioning of agricultural machinery is often out of touch with the actual operating conditions, leading to the technical problem that the existing intelligent methods for improving the performance of agricultural machinery are difficult to balance efficiency and sustainability.

[0004] In the existing virtual commissioning methods for agricultural machinery, there are problems such as insufficient accuracy in virtual farmland modeling and a large deviation between the agricultural machinery working condition modeling and the actual situation. As a result, there is a lack of a feedback mechanism in the interaction process between the virtual model and the actual use, making it difficult to apply the virtual environment to the commissioning process of actual farmland operations. In the existing agricultural machinery operation optimization methods, the goals of agricultural machinery operations are usually single in the previous intelligent systems, which further exacerbates the problem of being difficult to balance efficiency and sustainability. At the same time, it also leads to low training efficiency in complex farmland environments and great difficulty in migrating from the virtual environment to the actual farmland. Summary of the Invention

[0005] In view of the above situation, to overcome the defects of the prior art, the technical solution adopted by the present invention is as follows: An intelligent method for improving the performance of agricultural machinery provided by the present invention includes the following steps:

[0006] Step S1: Sensing data processing;

[0007] Step S2: Virtual commissioning of agricultural machinery;

[0008] Step S3: Optimization of agricultural machinery operations;

[0009] Step S4: Improvement of agricultural machinery performance.

[0010] Further, in step S1, the sensing data processing is used to collect sensing data and integrate data resources. Specifically, it integrates a sensor combination to collect real-time agricultural machinery data information, and through data optimization and integrated storage, an original dataset for improving the performance of agricultural machinery is obtained;

[0011] The integrated sensor combination includes a positioning and attitude sensor, an environment perception sensor, a soil and crop sensor, and a working power monitoring sensor;

[0012] The original dataset for improving the performance of agricultural machinery includes agricultural machinery operation status data, operation environment data, crop growth condition data, soil status data, and agricultural machinery operation efficiency record data.

[0013] Further, in step S2, the virtual commissioning of agricultural machinery is used to build a digital twin model of agricultural machinery and conduct commissioning and optimization in a virtual environment. Specifically, based on the original dataset for improving the performance of agricultural machinery, through virtual farmland modeling and virtual agricultural machinery working condition modeling, and introducing a dynamic digital twin mirror optimization method, virtual commissioning of agricultural machinery is carried out to obtain virtual agricultural machinery environment simulation data, including the following steps:

[0014] Step S21: Virtual farmland modeling is used to build a virtual environment for farmland operations. Specifically, based on the operation environment data, crop growth condition data, and soil status data in the original dataset for improving the performance of agricultural machinery, a model of the distribution of soil moisture in the farmland is built to obtain data on the dynamic change modeling of the virtual farmland environment;

[0015] Step S22: Virtual agricultural machinery working condition modeling is used to conduct virtual modeling of the working conditions of agricultural machinery in combination with the operation data of agricultural machinery. Specifically, based on the agricultural machinery operation status data and agricultural machinery operation efficiency record data in the original dataset for improving the performance of agricultural machinery, a digital system modeling method combining physical modeling and data-driven modeling is adopted to conduct virtual agricultural machinery working condition modeling to obtain virtual agricultural machinery working condition modeling data;

[0016] For the physical modeling, the ground friction coefficient and crop resistance are used as physical models, and the energy consumption and operation quality during operation are calculated;

[0017] For the data-driven modeling, historical operation data is used as the data source to train an operation optimization model to improve the operation accuracy prediction ability;

[0018] Step S23: Dynamic digital twin mirror optimization. Specifically, the operation parameters in the data on the dynamic change modeling of the virtual farmland environment and the virtual agricultural machinery working condition modeling data are optimized through a dynamic multi-modal data fusion method, and dynamic digital twin mirror optimization is carried out through a virtual-real interaction feedback mechanism;

[0019] The virtual-real interaction feedback mechanism simulates agricultural machinery operations in a virtual environment, calculates the operation indicators of agricultural machinery operations, compares them with the actual agricultural machinery operation data, calculates the virtual-real difference degree parameter, and realizes the automatic adjustment of the digital twin model parameters and dynamic optimization by setting a difference degree threshold and comparing it with the virtual-real difference degree parameter. The calculation formula of the virtual-real difference degree parameter is as follows:

[0020]

[0021] In the formula, D is the virtual-real difference degree parameter, N is the total number of agricultural machinery, i is the agricultural machinery index, V i is the virtual environment agricultural machinery operation index of the i-th agricultural machinery, and P i is the actual agricultural machinery operation index of the i-th agricultural machinery;

[0022] Step S24: Virtual commissioning of agricultural machinery, specifically, through the virtual farmland modeling, the virtual agricultural machinery working condition modeling, and the dynamic digital twin mirror optimization, virtual commissioning of agricultural machinery is carried out to obtain virtual environment simulation data of agricultural machinery;

[0023] The virtual environment simulation data of agricultural machinery includes agricultural machinery operation efficiency prediction data, agricultural machinery operation quality evaluation data, agricultural machinery operation path optimization data, and virtual environment adjustment parameters;

[0024] The agricultural machinery operation efficiency prediction data includes the farmland operation area per unit time and the energy consumption efficiency.

[0025] Furthermore, in step S3, the agricultural machinery operation optimization is used to control the agricultural machinery to autonomously learn and optimize strategies during the operation process. Specifically, based on the original agricultural machinery performance improvement dataset, a multi-objective hierarchical adaptive improved reinforcement learning method for agricultural machinery operation is adopted to carry out autonomous operation optimization of agricultural machinery to obtain autonomous operation optimization reference data for agricultural machinery, including the following steps:

[0026] Step S31: Construct a hierarchical decision-making architecture, including a global reinforcement learning layer and a local reinforcement learning layer; the global reinforcement learning layer is used for long-term operation task planning, and the local reinforcement learning layer is used for real-time agricultural machinery action adjustment;

[0027] The global reinforcement learning layer specifically uses Q-value learning for long-term operation task planning of agricultural machinery and generates a sequence of agricultural machinery operation sub-goals;

[0028] The sequence of agricultural machinery operation sub-goals includes path segmentation and operation mode switching;

[0029] The local reinforcement learning layer specifically uses the proximal policy optimization algorithm for real-time operation adjustment of agricultural machinery and generates a real-time adjustment plan for agricultural machinery;

[0030] The real-time adjustment scheme for agricultural machinery includes a speed control scheme, a steering angle scheme, and a seeding rate adjustment scheme;

[0031] Step S32: Construct multi-objective rewards. Specifically, design a multi-objective reward function and apply it to the hierarchical decision-making architecture for reinforcement learning training. The multi-objective reward function decomposes the agricultural machinery performance indicators into the dimensions of operation efficiency, resource consumption, and soil protection, and constructs the multi-objective reward function by dynamically adjusting the dimension weights. The calculation formula is:

[0032]

[0033] In the formula, R(t) is the multi-objective reward function, w1 is the operation efficiency weight, S(t) is the operation area calculation function, and S MAX is the maximum theoretical operation area, The whole is the operation efficiency reward term, w2 is the resource consumption weight, FC(t) is the fuel consumption calculation function, and FC p is the fuel consumption benchmark value, The whole is the resource consumption reward term, w3 is the soil protection weight, and FP(t) is the soil protection penalty term;

[0034] Step S33: Construct a soil protection penalty term to balance the performance of agricultural machinery and prevent the over-compaction of the land by agricultural machinery. Specifically, calculate the soil compaction degree parameter and construct a soil protection penalty term for constructing multi-objective rewards. The calculation formula is:

[0035]

[0036] In the formula, FP(t) is the soil protection penalty term, and F L (t) is the pressure calculation function exerted by the agricultural machinery tire on the soil, and F r is the critical pressure for soil compaction;

[0037] Step S34: Virtual environment migration. Specifically, by constructing a virtual migration loss function, the hierarchical decision-making architecture is first pre-trained in a digital twin virtual environment according to the virtual environment in the virtual environment simulation data of the agricultural machinery to obtain virtual policy parameters, and the virtual policy parameters are migrated and applied to the actual agricultural machinery operation environment for verification and fine-tuning to obtain virtual migration agricultural machinery operation optimization data;

[0038] Step S35: Strengthen agricultural machinery operation optimization. Specifically, through the hierarchical decision-making architecture, according to the multi-objective rewards and the soil protection penalty term, perform reinforcement learning training for virtual environment migration and agricultural machinery operation optimization to obtain agricultural machinery autonomous optimization reference data;

[0039] The reference data for the autonomous optimization of agricultural machinery includes operation efficiency optimization data, resource consumption optimization data, soil protection reference data, and optimization data for the state control of agricultural machinery equipment.

[0040] Furthermore, in step S4, the performance improvement of agricultural machinery is used to optimize and enhance the performance of agricultural machinery by combining the virtual environment and optimization strategies. Specifically, based on the simulated data of the agricultural machinery virtual environment and the reference data for the autonomous operation optimization of agricultural machinery, the comparison and weighted evaluation of the simulated data and the actual operation data are carried out. And through the agricultural machinery control decision-making that combines the simulated data and the agricultural machinery control decision-making of the autonomous optimization data of agricultural machinery, the collaborative optimization of the simulation level and the real level of the performance of agricultural machinery is carried out to obtain the reference data for the comprehensive performance of agricultural machinery.

[0041] An intelligent agricultural machinery performance improvement system provided by the present invention includes a data perception module, a network transmission module, an intelligent application module, and a performance improvement module;

[0042] The data perception module is used for sensing data processing. Through sensing data processing, the original data set for the performance improvement of agricultural machinery is obtained, and the original data set for the performance improvement of agricultural machinery is sent to the network transmission module and the intelligent application module;

[0043] The network transmission module is used for virtual commissioning of agricultural machinery. Through virtual commissioning of agricultural machinery, the simulated data of the agricultural machinery virtual environment is obtained, and the simulated data of the agricultural machinery virtual environment is sent to the intelligent application module and the performance improvement module;

[0044] The intelligent application module is used for optimizing the operation of agricultural machinery. Through optimizing the operation of agricultural machinery, the reference data for the autonomous operation optimization of agricultural machinery is obtained, and the reference data for the autonomous operation optimization of agricultural machinery is sent to the performance improvement module;

[0045] The performance improvement module is used for the performance improvement of agricultural machinery. Through the performance improvement of agricultural machinery, the reference data for the comprehensive performance of agricultural machinery is obtained.

[0046] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0047] (1) Aiming at the problems existing in the existing methods for improving the performance of agricultural machinery, where the environment of intelligent farm operations is complex and changeable, resulting in insufficient accuracy in the intelligent and automated process of agricultural machinery operations. At the same time, the existing agricultural machinery commissioning is often easily disconnected from the actual operating conditions, thus leading to the technical problem that the existing methods for intelligent improvement of agricultural machinery performance are difficult to balance efficiency and sustainability. This scheme creatively adopts sequential optimization in two dimensions of virtual agricultural machinery commissioning and agricultural machinery operation optimization, and deeply conducts digital modeling from virtual to actual, systematically solving the adaptability problem of agricultural machinery in complex farm environments, improving operation accuracy and efficiency, while taking into account resource conservation and soil protection, and meeting the actual needs of precision agriculture;

[0048] (2) In the existing virtual commissioning methods for agricultural machinery, there are problems such as insufficient accuracy in virtual farmland modeling and significant deviations between the modeled agricultural machinery working conditions and the actual ones. As a result, there is a lack of a feedback mechanism in the interaction process between the virtual model and the actual use, making it difficult to apply the virtual environment to the commissioning process of actual farmland operations. This solution creatively uses the dynamic digital twin mirror optimization method for virtual commissioning of agricultural machinery. By constructing a high-precision virtual farmland model through dynamic weight factors and designing a feedback mechanism for the difference degree between the virtual and the real, it achieves a high degree of consistency between the virtual commissioning results and the actual operations, shortens the commissioning cycle, improves the reliability of commissioning, and enhances the overall usability of the virtual digital model.

[0049] (3) In the existing agricultural machinery operation optimization methods, the goals of agricultural machinery operations are usually single in the previous intelligent systems. This not only further exacerbates the problem of being difficult to balance efficiency and sustainability but also leads to low training efficiency in complex farmland environments and great difficulty in migrating from the virtual environment to the actual farmland. This solution creatively uses the multi-objective hierarchical adaptive reinforcement learning method for agricultural machinery operation optimization. Through a hierarchical decision-making architecture and a multi-objective reward function, it realizes the multi-objective collaborative optimization of operation efficiency, resource consumption, and soil protection, and introduces a virtual migration loss function to accelerate the strategy migration from the virtual environment to the actual farmland, significantly enhancing the comprehensive performance of autonomous operation of agricultural machinery and better meeting the actual needs of precision agriculture. Description of the Drawings

[0050] Figure 1 It is a schematic flow chart of an intelligent agricultural machinery performance improvement method provided by the present invention;

[0051] Figure 2 It is a schematic diagram of an intelligent agricultural machinery performance improvement system provided by the present invention;

[0052] Figure 3 It is a schematic flow chart of the virtual commissioning of agricultural machinery in step S2;

[0053] Figure 4 It is a schematic flow chart of the agricultural machinery operation optimization in step S3.

[0054] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation to the present invention. Detailed Embodiments

[0055] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0056] In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc. indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention.

[0057] Embodiment 1. Refer to Figure 1 , a method for improving the performance of intelligent agricultural machinery provided by the present invention, the method includes the following steps:

[0058] Step S1: Sensing data processing;

[0059] Step S2: Virtual commissioning of agricultural machinery;

[0060] Step S3: Optimization of agricultural machinery operation;

[0061] Step S4: Improvement of agricultural machinery performance.

[0062] By performing the above operations, in the existing methods for improving the performance of agricultural machinery, there are problems that the environment of intelligent farm operations is complex and changeable, resulting in insufficient accuracy in the intelligent and automated process of agricultural machinery operations. At the same time, the existing agricultural machinery commissioning is often disconnected from the actual operating conditions, and thus the existing methods for intelligent improvement of agricultural machinery performance are difficult to balance efficiency and sustainability. This solution creatively adopts sequential optimization in two dimensions of virtual agricultural machinery commissioning and optimization of agricultural machinery operation, and deeply conducts digital modeling from virtual to actual, systematically solving the adaptability problems of agricultural machinery in complex farm environments, improving operation accuracy and efficiency, while taking into account resource conservation and soil protection, and meeting the actual needs of precision agriculture.

[0063] Embodiment 2. Refer to Figure 1 and Figure 2 , in step S1, the sensing data processing is used to collect sensing data and integrate data resources, specifically by integrating a sensor combination, collecting real-time agricultural machinery data information, and obtaining the original data set for improving the performance of agricultural machinery through data optimization and integrated storage;

[0064] The integrated sensor combination includes a positioning attitude sensor, an environmental perception sensor, a soil and crop sensor, and an operation power monitoring sensor;

[0065] The agricultural machinery performance improvement original data set includes agricultural machinery operation status data, operation environment data, crop growth data, soil status data and agricultural machinery operation efficiency record data;

[0066] The positioning attitude sensor is used for agricultural machinery positioning, attitude detection and trajectory tracking, including a GNSS sensor and an IMU sensor;

[0067] The environmental perception sensor is used to identify farmland terrain and farmland obstacles, including a lidar sensor, a camera, and an air temperature and humidity sensor;

[0068] The soil crop sensor is used to monitor soil moisture, temperature and nutrition, including a soil moisture sensor, a temperature and humidity sensor and a conductivity sensor;

[0069] The operating power monitoring sensor is used to monitor the power system status, fuel consumption status and mechanical load status of the agricultural machinery, including an engine sensor, a fuel consumption sensor and a torque sensor;

[0070] The agricultural machinery operation status data includes the agricultural machinery position, speed, acceleration, steering angle, engine speed and torque parameter data;

[0071] The working environment data includes terrain type, obstacle distribution type, weather type, and air temperature and humidity data;

[0072] The crop growth data include farmland crop growth cycle and health status type data;

[0073] The soil state data includes soil moisture, temperature, pH value and conductivity data;

[0074] The agricultural machinery operation efficiency record data includes operation area, operation speed and fuel consumption reference data.

[0075] Example 3, see Figure 1 , Figure 2 and Figure 3 This embodiment is based on the above embodiment. In step S2, the agricultural machinery virtual debugging is used to build a digital twin model of agricultural machinery and debug and optimize it in a virtual environment. Specifically, the original data set of agricultural machinery performance improvement is used to perform virtual farmland modeling and virtual agricultural machinery working condition modeling, and introduce a dynamic digital twin image optimization method to perform virtual debugging of agricultural machinery and obtain agricultural machinery virtual environment simulation data, including the following steps:

[0076] Step S21: Virtual farmland modeling, which is used to construct a virtual environment for farmland operations. Specifically, based on the operation environment data, crop growth data, and soil condition data in the original dataset for improving agricultural machinery performance, a model of the farmland soil moisture distribution is constructed to obtain the modeling data of the dynamic changes in the virtual farmland environment. The calculation formula is as follows:

[0077]

[0078] In the formula, SoilMap(x, y, t) is the function for modeling the farmland soil moisture distribution, x is the horizontal index of the sensing position, y is the vertical index of the sensing position, t is the current time index, m is the total number of sensors, k is the sensor index, γ k is the sensor weight, Sensor k (x, y) is the value-taking function of the k-th sensor, exp(-λ|t - t k |) is the natural exponential function, λ is the time decay factor used to control the timeliness of soil moisture, t k is the time when the sensing data of the k-th sensor is collected;

[0079] Step S22: Virtual agricultural machinery working condition modeling, which is used to conduct virtual modeling of the working conditions of agricultural machinery in combination with the operation data of agricultural machinery. Specifically, based on the operation state data and the working efficiency record data of agricultural machinery in the original dataset for improving agricultural machinery performance, a digital system modeling method combining physical modeling and data-driven modeling is adopted to conduct virtual agricultural machinery working condition modeling to obtain the modeling data of virtual agricultural machinery working conditions;

[0080] For the physical modeling, the ground friction coefficient and the crop resistance are used as the physical model, and the energy consumption and working quality during the operation are calculated;

[0081] For the data-driven modeling, historical operation data is used as the data source to train the operation optimization model to improve the operation accuracy prediction ability;

[0082] Step S23: Dynamic digital twin mirror optimization. Specifically, the operation parameters in the modeling data of the dynamic changes in the virtual farmland environment and the modeling data of the virtual agricultural machinery working conditions are optimized through the dynamic multi-modal data fusion method, and dynamic digital twin mirror optimization is carried out through the virtual-real interaction feedback mechanism;

[0083] The calculation formula for the dynamic multi-modal data fusion is as follows:

[0084]

[0085] In the formula, w k(t) is the dynamic weight factor of the k-th sensor, which is used as the dynamic weight for multimodal data fusion. k is the sensor index, t is the current time index, α is the confidence weight of sensor data, Conf k (t) is the confidence value of the k-th sensor. The confidence value is specifically evaluated and calculated based on errors and noise. β is the weight of the correlation degree of agricultural machinery operation conditions, Relev k (t) is the correlation degree value of the k-th sensor with the agricultural machinery operation conditions. The correlation degree value of the agricultural machinery operation conditions is used to represent the correlation between the sensor data and the current agricultural machinery operation task. n is the total number of fused sensor data, which represents the total number of sensor data that needs to be fused for multimodal data fusion. j is the fused sensor index;

[0086] The virtual-real interaction feedback mechanism simulates agricultural machinery operations in a virtual environment and calculates the operation indicators of agricultural machinery operations. It is used to compare with the actual agricultural machinery operation data and calculate the virtual-real difference parameter. By setting a difference threshold and comparing it with the virtual-real difference parameter, the parameters of the digital twin model are automatically adjusted and dynamically optimized. The calculation formula of the virtual-real difference parameter is:

[0087]

[0088] In the formula, D is the virtual-real difference parameter, N is the total number of agricultural machinery, i is the agricultural machinery index, V i is the virtual environment agricultural machinery operation indicator of the i-th agricultural machinery, P i is the actual agricultural machinery operation indicator of the i-th agricultural machinery;

[0089] Step S24: Virtual commissioning of agricultural machinery. Specifically, through the virtual farmland modeling, the virtual agricultural machinery working condition modeling, and the dynamic digital twin mirror optimization, virtual commissioning of agricultural machinery is carried out to obtain virtual environment simulation data of agricultural machinery;

[0090] The virtual environment simulation data of agricultural machinery includes agricultural machinery operation efficiency prediction data, agricultural machinery operation quality evaluation data, agricultural machinery operation path optimization data, and virtual environment adjustment parameters;

[0091] The agricultural machinery operation efficiency prediction data includes the farmland operation area per unit time and the energy consumption efficiency.

[0092] By performing the above operations, in the existing virtual commissioning method for agricultural machinery, there are problems such as insufficient accuracy in virtual farmland modeling and often large deviations between the modeling of agricultural machinery working conditions and the actual situation. As a result, there is a lack of a feedback mechanism in the interaction process between the virtual model and the actual use, making it difficult to apply the virtual environment to the commissioning process of actual farmland operations. In this solution, a dynamic digital twin mirror optimization method is creatively adopted for virtual commissioning of agricultural machinery. A high-precision virtual farmland model is constructed through a dynamic weight factor, and a feedback mechanism for the difference between the virtual and the real is designed to achieve a high degree of consistency between the virtual commissioning results and the actual operation, shorten the commissioning cycle, and improve the reliability of commissioning and the overall usability of the virtual digital model.

[0093] Example 4, refer to Figure 1 、 Figure 2 and Figure 4 Based on the above example, in step S3, the optimization of agricultural machinery operations is used to control the agricultural machinery to autonomously learn and optimize strategies during the operation process. Specifically, according to the original dataset for improving the performance of agricultural machinery, a multi-objective hierarchical adaptive improved reinforcement learning method for agricultural machinery operations is adopted to perform autonomous operation optimization of agricultural machinery, and reference data for autonomous operation optimization of agricultural machinery is obtained, including the following steps:

[0094] Step S31: Construct a hierarchical decision-making architecture, including a global reinforcement learning layer and a local reinforcement learning layer; the global reinforcement learning layer is used for long-term operation task planning, and the local reinforcement learning layer is used for real-time adjustment of agricultural machinery actions;

[0095] The global reinforcement learning layer specifically uses Q-value learning for long-term operation task planning of agricultural machinery and generates a sequence of sub-goals for agricultural machinery operations;

[0096] The sequence of sub-goals for agricultural machinery operations includes path segmentation and operation mode switching;

[0097] The local reinforcement learning layer specifically uses the proximal policy optimization algorithm to perform real-time operation adjustment of agricultural machinery and generates a real-time adjustment plan for agricultural machinery;

[0098] The real-time adjustment plan for agricultural machinery includes a speed control plan, a steering angle plan, and a seeding rate adjustment plan;

[0099] Step S32: Construct multi-objective rewards. Specifically, design a multi-objective reward function and apply it to the hierarchical decision-making architecture for reinforcement learning training; the multi-objective reward function decomposes the performance indicators of agricultural machinery into dimensions of operation efficiency, resource consumption, and soil protection, and constructs the multi-objective reward function by dynamically adjusting the dimension weights. The calculation formula is:

[0100]

[0101] Wherein, R(t) is the multi-objective reward function, w1 is the operation efficiency weight, S(t) is the operation area calculation function, and S MAX is the maximum theoretical operation area, as a whole is the operation efficiency reward term, w2 is the resource consumption weight, FC(t) is the fuel consumption calculation function, and FC p is the fuel consumption reference value, as a whole is the resource consumption reward term, w3 is the soil protection weight, and FP(t) is the soil protection penalty term;

[0102] Step S33: Construct a soil protection penalty term to balance the performance of agricultural machinery and prevent excessive compaction of the land by agricultural machinery. Specifically, calculate the soil compaction degree parameter and construct the soil protection penalty term for constructing the multi-objective reward. The calculation formula is:

[0103]

[0104] Wherein, FP(t) is the soil protection penalty term, and F L (t) is the pressure calculation function exerted by the agricultural machinery tire on the soil, and F r is the critical pressure for soil compaction;

[0105] Step S34: Virtual environment migration. Specifically, by constructing a virtual migration loss function, pre-train the hierarchical decision-making architecture according to the virtual environment in the virtual environment simulation data of the agricultural machinery to obtain virtual policy parameters, and apply the virtual policy parameters to the actual agricultural machinery operation environment for verification and fine-tuning to obtain virtual migration agricultural machinery operation optimization data;

[0106] The calculation formula of the virtual migration loss function is:

[0107] L adapt = ||μ V - μ R || 2 + λ C · KL(P V || P R ) ;

[0108] Wherein, L adapt is the virtual migration loss function, μ V is the average value of the agricultural machinery state in the virtual environment, μ R is the average value of the agricultural machinery state in the actual environment, λ C is the divergence adjustment coefficient, KL(P V || P R ) is the KL divergence calculation function, and P V is the probability distribution of the agricultural machinery state distribution in the virtual environment, and P Ris the probability distribution of the agricultural machinery state distribution in the actual environment;

[0109] Step S35: Strengthen the optimization of agricultural machinery operations. Specifically, through the hierarchical decision-making architecture, based on the multi-objective reward and the soil protection penalty term, perform reinforcement learning training for virtual environment migration and agricultural machinery operation optimization to obtain agricultural machinery autonomous optimization reference data;

[0110] The agricultural machinery autonomous optimization reference data includes operation efficiency optimization data, resource consumption optimization data, soil protection reference data, and agricultural machinery equipment state control optimization data.

[0111] By performing the above operations, in the existing agricultural machinery operation optimization methods, the goals of agricultural machinery operations are usually single in the previous intelligent systems, which further exacerbates the problem of being difficult to balance efficiency and sustainability. At the same time, it also leads to low training efficiency in complex farmland environments and great difficulty in migrating from the virtual environment to the actual farmland. This solution creatively adopts a multi-objective hierarchical adaptive reinforcement learning method for agricultural machinery operation optimization, realizes multi-objective collaborative optimization of operation efficiency, resource consumption, and soil protection through a hierarchical decision-making architecture and a multi-objective reward function, and introduces a virtual migration loss function to accelerate the strategy migration from the virtual environment to the actual farmland, significantly improving the comprehensive performance of agricultural machinery autonomous operations and better meeting the actual needs of precision agriculture.

[0112] Example Five, refer to Figure 1 、 Figure 2 and Figure 4 Based on the above embodiment, in step S3, Table 1 is a reference example table of the content of the agricultural machinery autonomous optimization reference data. As shown in the table, the operation efficiency optimization data includes optimized operation area, optimized operation speed, path planning deviation reference, predicted time for completing farmland operations, and predicted time for operation mode switching; the resource consumption optimization data includes predicted fuel consumption value, predicted battery current consumption value, predicted total operation energy consumption value, and predicted agricultural input consumption value; the soil protection reference data includes reference value of soil compaction change, reference value of tire contact pressure change, reference value of soil moisture change, and reference value of soil organic matter content; the agricultural machinery equipment state control optimization data includes engine load rate, transmission working state, and equipment vibration frequency.

[0113] Table 1 Reference Example Table of the Content of Agricultural Machinery Autonomous Optimization Reference Data

[0114]

[0115] Example Six, refer to Figure 1 and Figure 2, based on the above embodiment, in step S4, the agricultural machinery performance improvement is used to optimize and improve the agricultural machinery performance by combining the virtual environment and the optimization strategy. Specifically, according to the simulated data of the agricultural machinery virtual environment and the reference data for the autonomous operation optimization of the agricultural machinery, the comparison and weighted evaluation of the simulated data and the actual operation data are carried out, and through the agricultural machinery control decision combining the simulated data and the agricultural machinery control decision of the agricultural machinery autonomous optimization data, the collaborative optimization of the simulation level and the real level of the agricultural machinery performance is carried out to obtain the reference data for the comprehensive performance of the agricultural machinery.

[0116] Embodiment Seven, refer to Figure 1 and Figure 2 , based on the above embodiment, an intelligent agricultural machinery performance improvement system provided by the present invention includes a data perception module, a network transmission module, an intelligent application module, and a performance improvement module;

[0117] The data perception module is used for sensing data processing. Through sensing data processing, the original data set for agricultural machinery performance improvement is obtained, and the original data set for agricultural machinery performance improvement is sent to the network transmission module and the intelligent application module;

[0118] The network transmission module is used for virtual commissioning of agricultural machinery. Through virtual commissioning of agricultural machinery, the simulated data of the agricultural machinery virtual environment is obtained, and the simulated data of the agricultural machinery virtual environment is sent to the intelligent application module and the performance improvement module;

[0119] The intelligent application module is used for optimizing the operation of agricultural machinery. Through optimizing the operation of agricultural machinery, the reference data for the autonomous operation optimization of agricultural machinery is obtained, and the reference data for the autonomous operation optimization of agricultural machinery is sent to the performance improvement module;

[0120] The performance improvement module is used for improving the performance of agricultural machinery. Through improving the performance of agricultural machinery, the reference data for the comprehensive performance of agricultural machinery is obtained.

[0121] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0122] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention.

[0123] The above description of the present invention and its embodiments is not restrictive. What is shown in the drawings is only one of the embodiments of the present invention, and the actual structure is not limited thereto. Generally speaking, if those of ordinary skill in the art are inspired by it and, without departing from the gist of the present invention, design similar structural modes and embodiments to this technical solution without creative efforts, they shall fall within the protection scope of the present invention.

Claims

1. A method for improving the performance of intelligent agricultural machinery, characterized in that: The method comprises the following steps: Step S1: sensor data processing to obtain the original data set for agricultural machinery performance improvement; Step S2: virtual debugging of agricultural machinery, through virtual farmland modeling and virtual agricultural machinery working condition modeling, and introducing dynamic digital twin image optimization method, virtual debugging of agricultural machinery is carried out to obtain agricultural machinery virtual environment simulation data, including the following steps: step S21: virtual farmland modeling; step S22: virtual agricultural machinery working condition modeling; step S23: dynamic digital twin image optimization; step S24: virtual debugging of agricultural machinery; In step S23, the dynamic digital twin image optimization is specifically to optimize the operating parameters in the dynamic change modeling data of the virtual farmland environment and the virtual agricultural machinery working condition modeling data through a dynamic multimodal data fusion method, and perform dynamic digital twin image optimization through a virtual-real interactive feedback mechanism; The virtual-reality interactive feedback mechanism simulates agricultural machinery operations in a virtual environment and calculates the operation indicators of agricultural machinery operations for comparison with actual agricultural machinery operation data, and calculates virtual-reality difference parameters. By setting a difference threshold and comparing it with the virtual-reality difference parameters, the parameters of the digital twin model are automatically adjusted and dynamically optimized. The calculation formula of the virtual-reality difference parameters is: Where D is the virtual-real difference parameter, N is the total number of agricultural machinery, i is the index of agricultural machinery, V i is the virtual environment agricultural machinery operation index of the i-th agricultural machinery, P i is the actual agricultural machinery operation index of the i-th agricultural machinery; Step S3: agricultural machinery operation optimization, adopting a multi-objective hierarchical adaptively improved enhanced agricultural machinery operation learning method to optimize the autonomous operation of agricultural machinery and obtain reference data for the optimization of autonomous operation of agricultural machinery, including the following steps: step S31: constructing a hierarchical decision-making framework; step S32: constructing a multi-objective reward; step S33: constructing a soil protection penalty item; step S34: virtual environment migration; step S35: strengthening the optimization of agricultural machinery operations; Step S4: Improve the performance of agricultural machinery by coordinating the simulation and reality levels of agricultural machinery performance to obtain reference data on the comprehensive performance of agricultural machinery.

2. The method for improving the performance of intelligent agricultural machinery according to claim 1, characterized in that: In step S1, the sensor data processing is used to collect sensor data and integrate data resources, specifically integrating sensor combinations to collect real-time agricultural machinery data information, and obtain the original data set for agricultural machinery performance improvement through data optimization and integrated storage; The integrated sensor combination includes a positioning attitude sensor, an environmental perception sensor, a soil and crop sensor, and an operation power monitoring sensor; The agricultural machinery performance improvement original data set includes agricultural machinery operation status data, operation environment data, crop growth data, soil status data and agricultural machinery operation efficiency record data.

3. The method for improving the performance of intelligent agricultural machinery according to claim 2, characterized in that: In step S2, the agricultural machinery virtual debugging is used to build a digital twin model of agricultural machinery and debug and optimize it in a virtual environment. Specifically, based on the original data set of agricultural machinery performance improvement, virtual farmland modeling and virtual agricultural machinery working condition modeling are performed, and a dynamic digital twin image optimization method is introduced to perform virtual debugging of agricultural machinery to obtain agricultural machinery virtual environment simulation data, including the following steps: Step S21: virtual farmland modeling, which is used to construct a virtual environment for farmland operations, specifically, based on the operating environment data, crop growth data and soil state data in the original data set for improving the performance of agricultural machinery, farmland soil moisture distribution modeling is performed to obtain virtual farmland environment dynamic change modeling data; Step S22: virtual agricultural machinery working condition modeling, which is used to perform virtual modeling of agricultural machinery working conditions in combination with agricultural machinery operation data, specifically, based on the agricultural machinery operation status data and agricultural machinery operation efficiency record data in the original data set of agricultural machinery performance improvement, a digital system modeling method combining physical modeling and data-driven modeling is used to perform virtual agricultural machinery working condition modeling to obtain virtual agricultural machinery working condition modeling data; The physical modeling uses the ground friction coefficient and crop resistance as the physical model, and calculates the energy consumption and operation quality during the operation; The data-driven modeling uses historical operation data as a data source to train the operation optimization model to improve the operation accuracy prediction capability; Step S23: dynamic digital twin image optimization; Step S24: virtual debugging of agricultural machinery, specifically, performing virtual debugging of agricultural machinery through the virtual farmland modeling, the virtual agricultural machinery working condition modeling and the dynamic digital twin image optimization to obtain agricultural machinery virtual environment simulation data.

4. The method for improving the performance of intelligent agricultural machinery according to claim 3, characterized in that: In step S2, the agricultural machinery virtual environment simulation data includes agricultural machinery operation efficiency prediction data, agricultural machinery operation quality assessment data, agricultural machinery operation path optimization data and virtual environment adjustment parameters; The agricultural machinery operation efficiency prediction data includes the farmland operation area and energy consumption efficiency per unit time.

5. The method for improving the performance of intelligent agricultural machinery according to claim 4, characterized in that: In step S3, the agricultural machinery operation optimization is used to control the agricultural machinery to autonomously learn the optimization strategy during the operation process, specifically, based on the original data set of agricultural machinery performance improvement, a multi-objective hierarchical adaptive improved enhanced agricultural machinery operation learning method is used to optimize the autonomous operation of the agricultural machinery, and obtain the reference data for the autonomous operation optimization of the agricultural machinery, including the following steps: Step S31: constructing a hierarchical decision-making architecture, including a global reinforcement learning layer and a local reinforcement learning layer; the global reinforcement learning layer is used for long-term operation task planning, and the local reinforcement learning layer is used for real-time agricultural machinery action adjustment; The global reinforcement learning layer specifically uses Q-value learning to plan long-term operation tasks of agricultural machinery and generates a sequence of agricultural machinery operation sub-goals; The agricultural machinery operation sub-goal sequence includes path segmentation and operation mode switching; The local reinforcement learning layer specifically uses a proximal strategy optimization algorithm to adjust the real-time operation of the agricultural machinery and generate a real-time adjustment plan for the agricultural machinery; The real-time adjustment scheme of the agricultural machinery includes a speed control scheme, a steering angle scheme and a seeding amount adjustment scheme; Step S32: constructing a multi-objective reward, specifically designing a multi-objective reward function and applying it to the hierarchical decision-making framework to perform reinforcement learning training; the multi-objective reward function decomposes the agricultural machinery performance indicators into the operating efficiency dimension, the resource consumption dimension and the soil protection dimension, and constructs the multi-objective reward function by dynamically adjusting the dimension weights. The calculation formula is: Where R(t) is the multi-objective reward function, w1 is the operation efficiency weight, S(t) is the operation area calculation function, S MAX is the maximum theoretical operating area, The whole is the work efficiency reward item, w2 is the resource consumption weight, FC(t) is the fuel consumption calculation function, FC p is the fuel consumption benchmark value, The whole is the resource consumption reward term, w3 is the soil protection weight, and FP(t) is the soil protection penalty term; Step S33: construct a soil protection penalty term to balance the performance of agricultural machinery and prevent excessive compaction of the land by agricultural machinery. Specifically, the soil compaction degree parameter is calculated and a soil protection penalty term is constructed to construct a multi-objective reward. The calculation formula is: Where FP(t) is the soil protection penalty term, F L (t) is the calculation function of the pressure applied by the tire of the agricultural machinery to the soil, F r is the critical pressure for soil compaction; Step S34: virtual environment migration, specifically, by constructing a virtual migration loss function, pre-training the hierarchical decision architecture in a digital twin virtual environment according to the virtual environment in the agricultural machinery virtual environment simulation data to obtain virtual strategy parameters, and migrating the virtual strategy parameters to the actual agricultural machinery operation environment for verification and fine-tuning to obtain virtual migration agricultural machinery operation optimization data; Step S35: Strengthening the optimization of agricultural machinery operations, specifically, through the hierarchical decision-making framework, based on the multi-objective rewards and the soil protection penalty items, performing reinforcement learning training for virtual environment migration and agricultural machinery operation optimization, and obtaining reference data for autonomous optimization of agricultural machinery.

6. The method for improving the performance of intelligent agricultural machinery according to claim 5, characterized in that: In step S3, the agricultural machinery autonomous optimization reference data includes operation efficiency optimization data, resource consumption optimization data, soil protection reference data and agricultural machinery equipment status control optimization data.

7. The method for improving the performance of intelligent agricultural machinery according to claim 6, characterized in that: In step S4, the agricultural machinery performance improvement is used to optimize and improve the agricultural machinery performance in combination with the virtual environment and the optimization strategy. Specifically, based on the agricultural machinery virtual environment simulation data and the agricultural machinery autonomous operation optimization reference data, the simulation data and the actual operation data are compared and weighted evaluated. By combining the agricultural machinery control decisions of the simulation data and the agricultural machinery autonomous optimization data, the simulation level and the actual level of the agricultural machinery performance are coordinated optimized to obtain the comprehensive performance reference data of the agricultural machinery.

8. An intelligent agricultural machinery performance improvement system, used to implement an intelligent agricultural machinery performance improvement method as described in any one of claims 1 to 7, characterized in that: It includes data perception module, network transmission module, intelligent application module and performance improvement module.

9. The intelligent agricultural machinery performance improvement system according to claim 8, characterized in that: The data perception module is used for sensor data processing, and obtains an original data set for improving agricultural machinery performance through sensor data processing, and sends the original data set for improving agricultural machinery performance to the network transmission module and the intelligent application module; The network transmission module is used for virtual debugging of agricultural machinery, obtains agricultural machinery virtual environment simulation data through virtual debugging of agricultural machinery, and sends the agricultural machinery virtual environment simulation data to the intelligent application module and the performance improvement module; The intelligent application module is used for optimizing agricultural machinery operations, and obtains reference data for optimizing autonomous agricultural machinery operations through optimizing agricultural machinery operations, and sends the reference data for optimizing autonomous agricultural machinery operations to the performance improvement module; The performance improvement module is used for improving the performance of agricultural machinery, and through improving the performance of agricultural machinery, comprehensive performance reference data of agricultural machinery is obtained.

Images