A corn harvester energy-saving regulation method and system based on reinforcement learning

Abstract

The application belongs to the technical field of corn harvester management and control, and particularly relates to a corn harvester energy-saving regulation and control method and system based on reinforcement learning, wherein the system comprises a multi-source heterogeneous data acquisition module, a reinforcement learning decision module, an action execution regulation and control module, a safety constraint module and a running operation supervision terminal; the application collects relevant data of the corn harvester running process through the multi-source heterogeneous data acquisition module, designs a continuous action set based on the received data information by the reinforcement learning decision module, converts the action instruction into a specific control signal by the action execution regulation and control module, automatically and adaptively regulates the running of the corn harvester, and reduces the abnormal operation risk of the corn harvester by introducing a safety constraint mechanism, so as to realize the full-process intelligent control of the corn harvesting operation from environment perception to parameter optimization, which is beneficial to guarantee the safe, stable and efficient energy-saving running of the corn harvester, has high intelligence and automation level, and has small running management and control difficulty.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of corn harvester management and control, and particularly relates to a corn harvester energy-saving regulation and control method and system based on reinforcement learning. BACKGROUND

[0002] The corn harvester is a kind of agricultural machinery equipment specially used for corn harvesting operation, and its core function is to efficiently and accurately complete the harvesting, threshing, and cleaning of corn plants, so as to collect corn kernels from the field and prepare for subsequent storage, processing, or sales.

[0003] The traditional corn harvester usually adopts a fixed parameter control mode, which cannot adaptively adjust the operation state according to real-time changes in field conditions and crop density, resulting in serious energy waste. Moreover, it cannot reasonably analyze and timely warn the risk of the operation performance and constraint mechanism of the corn harvester, which is not conducive to ensuring the safe, stable, and energy-efficient operation of the corn harvester, and the operation supervision is difficult and the intelligent level is low.

[0004] In view of the above technical defects, a solution is proposed. SUMMARY

[0005] The present application aims to provide a corn harvester energy-saving regulation and control method and system based on reinforcement learning, which solves the problem that the prior art cannot adaptively adjust the operation state according to real-time changes in field conditions and crop density, and cannot reasonably analyze and timely warn the risk of the operation performance and constraint mechanism of the corn harvester, which is not conducive to ensuring the safe, stable, and energy-efficient operation of the corn harvester, and the operation supervision is difficult and the intelligent level is low.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical scheme:

[0007] A corn harvester energy-saving regulation and control system based on reinforcement learning includes a multi-source heterogeneous data acquisition module, a reinforcement learning decision module, an action execution and regulation module, a safety constraint module, and an operation control and supervision end. The multi-source heterogeneous data acquisition module collects the working environment and state parameters of the corn harvester in real time, performs spatio-temporal alignment and noise filtering on the original data, and generates a three-dimensional state vector containing crop density distribution, terrain slope, and current fuel consumption rate.

[0008] The reinforcement learning decision module maps the three-dimensional state vector output by the multi-source heterogeneous data acquisition module into a discrete state code, designs a continuous action set containing engine speed adjustment, header height adjustment, and threshing cylinder speed optimization, and the action execution and regulation module converts the action instructions output by the reinforcement learning decision module into specific control signals to automatically adaptively regulate the operation of the corn harvester.

[0009] The safety constraint module sets a plurality of groups of risk events to be monitored in advance, and if a corresponding risk event is detected during the operation of the corn harvester, a corresponding safety constraint operation is performed, and constraint alarm information is generated and sent to the operation control supervision end, and the operation control supervision end displays and warns the constraint alarm information.

[0010] Further, the multi-source heterogeneous data acquisition module scans the crop row spacing and density distribution by laser radar, the laser radar obtains crop canopy point cloud data in a 360° rotary scanning mode, removes the weed interference points by a voxel grid filtering algorithm, identifies the crop row spacing and local density distribution based on a density clustering algorithm, and generates a density heat map with a spatial resolution of 0.1 m x 0.1 m;

[0011] The body attitude of the corn harvester is monitored by an inertial measurement device, the body pitch angle, roll angle and acceleration change are monitored at a sampling frequency of 500 Hz, the Kalman filtering algorithm is used to fuse multi-sensor data, and a posture feature vector containing the terrain slope and body vibration amplitude is constructed;

[0012] The load change is obtained by an engine torque sensor, the instantaneous torque of the crankshaft is collected in real time, the load fluctuation coefficient is calculated in combination with the gear ring speed signal; and the work trajectory is recorded by a GPS positioning module.

[0013] Further, the action execution control module sends the engine speed adjustment instruction to the ECU through the CAN bus, realizes the target speed within 200 ms by using the feedforward-feedback composite control, adjusts the header height adjustment instruction through the PWM signal to drive the hydraulic proportional valve, realizes the closed loop control by cooperating with the displacement sensor, the positioning accuracy is ±2 mm, and the threshing cylinder torque limiting instruction is realized by changing the clutch solenoid valve current to realize the torque soft limit.

[0014] Further, the operation control supervision end is communicatively connected to the harvesting continuity detection module, the harvesting continuity detection module analyzes the harvesting continuity execution condition of the corn harvester per unit time, generates a continuity qualified signal or a continuity abnormal signal through analysis, and sends the continuity qualified signal or the continuity abnormal signal to the operation control supervision end, and the operation control supervision end issues a corresponding warning when receiving the continuity abnormal signal.

[0015] Further, the specific analysis process of the harvesting continuity detection module is as follows:

[0016] The engine speed, header height and cylinder roller speed of the corn harvester are acquired, the engine speed is subtracted from the standard engine speed value set at present and the absolute value is taken to obtain an engine detection value, and the header detection value and the cylinder roller detection value are obtained in the same way; the engine detection value, the header detection value and the cylinder roller detection value are compared with the preset engine detection threshold value, the preset header detection threshold value and the preset cylinder roller detection threshold value respectively, and if the engine detection value, the header detection value or the cylinder roller detection value exceeds the corresponding preset threshold value, it is judged that the corn harvester is in a non-ideal harvesting state;

[0017] The total time length of the corn harvester in the non-ideal harvesting state in a unit time is acquired and marked as a non-ideal harvesting characteristic value, and the non-ideal harvesting characteristic value is compared with the preset non-ideal harvesting characteristic threshold value, and if the non-ideal harvesting characteristic value exceeds the preset non-ideal harvesting characteristic threshold value, a continuity abnormal signal is generated.

[0018] Further, if the non-ideal harvesting characteristic value does not exceed the preset non-ideal harvesting characteristic threshold value, the engine detection value, the header detection value and the cylinder roller detection value are weighted and summed to obtain a harvester operation measurement value, all harvester operation measurement values in a unit time are averaged to obtain a harvester operation analysis value, and the largest harvester operation measurement value in a unit time is marked as a harvester amplitude value;

[0019] The non-ideal harvesting characteristic value, the harvester operation analysis value and the harvester amplitude value are weighted and summed to obtain a continuity evaluation value, and the continuity evaluation value is compared with the preset continuity evaluation threshold value, and if the continuity evaluation value exceeds the preset continuity evaluation threshold value, a continuity abnormal signal is generated; if the continuity evaluation value does not exceed the preset continuity evaluation threshold value, a continuity qualified signal is generated.

[0020] Further, the harvesting continuity detection module is communicatively connected to the constraint effectiveness evaluation module, the harvesting continuity detection module sends the continuity qualified signal to the constraint effectiveness evaluation module, the constraint effectiveness evaluation module evaluates and analyzes the safety constraint effectiveness of the corn harvester in a unit time when the continuity qualified signal is received, generates a constraint qualified signal or a constraint hidden danger signal through analysis, and sends the constraint qualified signal or the constraint hidden danger signal to the operation control supervision end, and the operation control supervision end sends an appropriate early warning when the constraint hidden danger signal is received.

[0021] Further, the specific analysis process of the constraint effectiveness evaluation module includes:

[0022] The risk items occurring in the operation of the corn harvester in a unit time are acquired, the occurrence time of the corresponding risk item is marked as a first time, the time of making the corresponding constraint operation is marked as a second time, the constraint delay value is calculated by calculating the time difference between the second time and the first time, the constraint delay value is compared with the corresponding preset constraint delay threshold value, if the constraint delay value exceeds the corresponding preset constraint delay threshold value, the constraint ineffective symbol TW-1 is assigned to the corresponding risk item;

[0023] The assignment times of the constraint ineffective symbol TW-1 in a unit time are acquired, and the constraint ineffective detection value is calculated by ratio calculation of the assignment times and the total number of occurrences of the risk items, the constraint ineffective detection value is compared with the preset constraint ineffective detection threshold value, if the constraint ineffective detection value exceeds the preset constraint ineffective detection threshold value, the constraint hidden danger signal is generated;

[0024] If the constraint ineffective detection value does not exceed the preset constraint ineffective detection threshold value, the constraint delay value is calculated by ratio calculation of the constraint delay value and the corresponding preset constraint delay threshold value, the constraint delay value in a unit time is calculated by mean value calculation, the constraint delay value with the maximum value in a unit time is marked as a constraint delay risk value;

[0025] The constraint effectiveness evaluation value is calculated by weighted summation calculation of the constraint ineffective detection value, the constraint delay analysis value and the constraint delay risk value, the constraint effectiveness evaluation value is compared with the preset constraint effectiveness evaluation threshold value, if the constraint effectiveness evaluation value exceeds the preset constraint effectiveness evaluation threshold value, the constraint hidden danger signal is generated; if the constraint effectiveness evaluation value does not exceed the preset constraint effectiveness evaluation threshold value, the constraint qualified signal is generated.

[0026] Further, the application also provides a corn harvester energy-saving regulation and control method based on reinforcement learning, which comprises the following steps:

[0027] Step one, collecting the working environment and state parameters of the corn harvester and performing data processing;

[0028] Step two, designing a continuous action set of the corn harvester;

[0029] Step three, generating specific control signals based on action instructions, and automatically adapting and regulating the operation of the corn harvester;

[0030] Step four, monitoring the risk items of the corn harvester, if the corresponding risk items are detected in the operation of the corn harvester, step five is performed;

[0031] Step five, making the corresponding safety constraint operation and generating the constraint alarm information.

[0032] Compared with the prior art, the application has the advantages that:

[0033] 1. In this invention, by collecting relevant data on the operation of the corn harvester and rationally designing a set of continuous actions, and converting the action commands into specific control signals, the operation of the corn harvester is automatically and adaptively regulated. Furthermore, by introducing a safety constraint mechanism to reduce the risk of abnormal operation of the corn harvester, the intelligent control of the entire process of corn harvesting operations from environmental perception to parameter optimization is realized, which is conducive to ensuring the safe, stable, efficient and energy-saving operation of the corn harvester.

[0034] 2. In this invention, the harvesting performance of the corn harvester is reasonably analyzed and comprehensively evaluated through the harvesting continuity detection module. When a continuity qualified signal is generated, the constraint effectiveness evaluation module accurately evaluates the timeliness of the constraint mechanism response of the corn harvester. This is conducive to managers conducting timely investigation and analysis of causes and making targeted improvement measures, further ensuring the efficient, stable and energy-saving operation of the corn harvester, significantly reducing the difficulty of operating and supervising the corn harvester, and demonstrating a high level of intelligence and automation. Attached Figure Description

[0035] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings;

[0036] Figure 1 This is a system block diagram of Embodiment 1 of the present invention;

[0037] Figure 2 This is a system block diagram of Embodiments 2 and 3 of the present invention;

[0038] Figure 3 This is a flowchart of the method in Embodiment 4 of the present invention. Detailed Implementation

[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] Example 1: As Figure 1 As shown, the present invention proposes an energy-saving control system for corn harvesters based on reinforcement learning, which includes a multi-source heterogeneous data acquisition module, a reinforcement learning decision-making module, an action execution control module, a safety constraint module, and an operation control monitoring terminal.

[0041] The multi-source heterogeneous data acquisition module collects the working environment and state parameters of the corn harvester in real time (as the system perception layer, three groups of threads are started simultaneously for laser radar scanning, inertial navigation monitoring, and engine working condition acquisition), performs time and space alignment and noise filtering on the original data, and generates a three-dimensional state vector containing crop density distribution map, terrain slope, and current fuel consumption rate. That is, through time and space alignment of the three groups of original data, a three-dimensional state representation containing crop density encoding, terrain complexity score, and current fuel consumption rate is generated through a feature splicing network.

[0042] Specifically, the multi-source heterogeneous data acquisition module scans the crop row spacing and density distribution through laser radar. The laser radar obtains crop canopy point cloud data in a 360° rotary scanning mode, removes interference points from weeds through a voxel grid filtering algorithm, identifies crop row spacing and local density distribution based on a density clustering algorithm, and generates a density heat map with a spatial resolution of 0.1 m x 0.1 m.

[0043] The inertial measurement device (IMU) is used to monitor the attitude of the corn harvester. The six-axis IMU monitors the pitch angle, roll angle, and acceleration change of the machine body at a sampling frequency of 500 Hz. Kalman filtering algorithm is used to fuse multi-sensor data to construct an attitude feature vector containing terrain slope (accuracy ±0.5°) and machine body vibration amplitude (RMS value). The engine torque sensor is used to obtain load changes, and the instantaneous torque of the crankshaft is collected in real time. Combined with the gear ring speed signal, the load fluctuation coefficient is calculated. The GPS positioning module is used to record the working trajectory.

[0044] The reinforcement learning decision module maps the three-dimensional state vector output by the multi-source heterogeneous data acquisition module into a discrete state encoding, including crop density level (low / medium / high), terrain slope angle (-15° to 15°), and current fuel consumption rate interval (0-5 L / min). A continuous action set is designed, including engine speed regulation, header height adjustment, and threshing drum speed optimization. The action execution control module converts the action instructions output by the reinforcement learning decision module into specific control signals, and automatically adjusts and controls the operation of the corn harvester.

[0045] The action execution control module sends the engine speed regulation instruction to the ECU through the CAN bus, uses feedforward-feedback composite control to achieve the target speed within 200 ms, drives the hydraulic proportional valve through the PWM signal for the header height adjustment instruction, and cooperates with the displacement sensor to realize closed-loop control with a positioning accuracy of ±2 mm. The torque limiting instruction of the threshing drum is implemented by changing the current of the clutch solenoid valve to limit the torque.

[0046] The safety constraint module pre-sets several sets of risk items that need to be monitored (i.e., items that need to be monitored and constrained, such as constraints on operating speed and header height. Regarding the constraint on header height, it is necessary to ensure that the height is not less than 80% of the crop height to avoid missed harvesting). If a corresponding risk item is detected during the operation of the corn harvester, the corresponding safety constraint operation will be performed, and a constraint alarm message will be generated and sent to the operation control monitoring terminal. The operation control monitoring terminal displays and issues a warning on the constraint alarm message. The safety constraint mechanism reduces the risk of abnormal operation of the corn harvester.

[0047] Example 2: Figure 2 As shown, the difference between this embodiment and embodiment one is that the operation control and supervision terminal is connected to the harvesting continuity detection module. The harvesting continuity detection module analyzes the harvesting continuity execution status of the corn harvester within a unit of time, generates a continuity qualified signal or a continuity abnormal signal through analysis, and sends the continuity qualified signal or the continuity abnormal signal to the operation control and supervision terminal.

[0048] When the operation and control monitoring terminal receives continuous abnormal signals, it issues corresponding warnings. This allows for reasonable analysis and comprehensive evaluation of the corn harvester's harvesting performance, facilitating timely investigation and analysis of the causes by management personnel and enabling targeted improvement measures to ensure the efficient, stable, and energy-saving operation of the corn harvester. The specific analysis process of the harvesting continuity detection module is as follows:

[0049] The engine speed, header height, and threshing drum speed of the corn harvester are obtained. The difference between the engine speed and the currently set standard engine speed value is calculated and the absolute value is taken to obtain the engine detection value. The difference between the header height and the currently set standard header height value is calculated and the absolute value is taken to obtain the header detection value. The difference between the threshing drum speed and the currently set standard threshing drum speed value is calculated and the absolute value is taken to obtain the threshing drum detection value.

[0050] The engine detection value, header detection value, and threshing drum detection value are compared with the preset engine detection threshold, preset header detection threshold, and preset threshing drum detection threshold respectively. If the engine detection value, header detection value, or threshing drum detection value exceeds the corresponding preset threshold, the corn harvester is judged to be in a non-beneficial harvesting state.

[0051] The total duration of the corn harvester in a non-ideal harvesting state within a unit of time is obtained and marked as a non-ideal harvesting characteristic value. The non-ideal harvesting characteristic value is compared with a preset non-ideal harvesting characteristic threshold. If the non-ideal harvesting characteristic value exceeds the preset non-ideal harvesting characteristic threshold, it indicates that the corn harvester's harvesting performance is poor within a unit of time, which is not conducive to ensuring the efficient, stable and energy-saving operation of the harvester, and a continuous abnormal signal is generated.

[0052] Furthermore, if the non-good harvesting characteristic value does not exceed the preset non-good harvesting characteristic threshold, the harvester operation test value is obtained by weighted summation of the engine detection value, header detection value and threshing drum detection value. That is, the corresponding preset weight coefficients are assigned to the engine detection value, header detection value and threshing drum detection value respectively, and the engine detection value, header detection value and threshing drum detection value are multiplied by the corresponding preset weight coefficients respectively, and the sum of the three sets of product results is marked as the harvester operation test value.

[0053] It should be noted that the larger the value of the harvester's operational measurement, the worse the overall real-time performance of the corn harvester. The harvester's operational measurement value is obtained by averaging all the harvester's operational measurement values ​​within a unit of time, and the harvester's operational measurement value with the largest value within a unit of time is marked as the harvester's abnormal value.

[0054] The continuity assessment value is obtained by weighted summation of the harvesting defective characteristic value, harvester operation analysis value, and harvester aberration value. Specifically, the harvesting defective characteristic value, harvester operation analysis value, and harvester aberration value are each assigned a corresponding preset weight coefficient, and the harvesting defective characteristic value, harvester operation analysis value, and harvester aberration value are multiplied by the corresponding preset weight coefficient. The sum of the three sets of products is marked as the continuity assessment value.

[0055] It should be noted that the higher the continuity assessment value, the worse the overall harvesting performance of the corn harvester within a unit of time, which is less conducive to ensuring the efficient, stable, and energy-saving operation of the harvester. The continuity assessment value is compared with the preset continuity assessment threshold. If the continuity assessment value exceeds the preset continuity assessment threshold, it indicates that the overall harvesting performance of the corn harvester within a unit of time is poor, which is not conducive to ensuring the efficient, stable, and energy-saving operation of the harvester, and a continuity anomaly signal is generated. If the continuity assessment value does not exceed the preset continuity assessment threshold, it indicates that the overall harvesting performance of the corn harvester within a unit of time is good, and a continuity qualified signal is generated.

[0056] Example 3: Figure 2 As shown, the difference between this embodiment and Embodiments 1 and 2 is that the harvesting continuity detection module is connected to the constraint effectiveness evaluation module. The harvesting continuity detection module sends the continuity qualified signal to the constraint effectiveness evaluation module. When the constraint effectiveness evaluation module receives the continuity qualified signal, it evaluates and analyzes the safety constraint effectiveness of the corn harvester within a unit of time and generates a constraint qualified signal or constraint potential signal through analysis.

[0057] and send the constraint qualified signal or the constraint hidden danger signal to the operation control supervision end, when the operation control supervision end receives the constraint hidden danger signal, issue corresponding early warning, can analyze and evaluate the timeliness of the constraint mechanism reaction of the corn harvester, is beneficial to the management personnel to investigate and analyze the cause in time and make targeted improvement measures, further guarantee the efficient and stable and energy-saving operation of the corn harvester, significantly reduce the operation supervision difficulty of the corn harvester, the intelligent and automatic level is high;The specific analysis process of the constraint effectiveness evaluation module includes:

[0058] All risk items of the corn harvester in the running process in unit time are obtained, the occurrence time of the corresponding risk item is marked as the first time, the time of making the corresponding constraint operation is marked as the second time, the constraint delay value is calculated by the time difference between the second time and the first time;The greater the value of constraint delay value, the more timely the constraint for the corresponding risk item;

[0059] The constraint delay value is compared with the corresponding preset constraint delay threshold value, if the constraint delay value exceeds the corresponding preset constraint delay threshold value, it is indicated that the constraint for the corresponding risk item is not timely, then the constraint ineffective symbol TW-1 is given to the corresponding risk item;

[0060] The number of times of giving constraint ineffective symbol TW-1 in unit time is obtained, and the ratio of the number of times of giving constraint ineffective symbol TW-1 in unit time to the total number of occurrence of risk items is calculated to obtain constraint ineffective detection value, and the constraint ineffective detection value is compared with the preset constraint ineffective detection threshold value, if the constraint ineffective detection value exceeds the preset constraint ineffective detection threshold value, it is indicated that the constraint risk of the safety constraint for the corn harvester is high, which is not conducive to ensure the safe and stable operation of the corn harvester, then the constraint hidden danger signal is generated;

[0061] If the constraint ineffective detection value does not exceed the preset constraint ineffective detection threshold value, the constraint delay value is compared with the corresponding preset constraint delay threshold value to obtain the constraint delay value, and the average value of all constraint delay values in unit time is calculated to obtain the constraint delay value, and the maximum constraint delay value in unit time is marked as the constraint delay risk value;

[0062] The constraint effectiveness evaluation value is calculated by weighted summation of constraint ineffective detection value, constraint delay value and constraint delay risk value;That is, the corresponding preset weight coefficient is given to constraint ineffective detection value, constraint delay value and constraint delay risk value respectively, and constraint ineffective detection value, constraint delay value and constraint delay risk value are multiplied by the corresponding preset weight coefficient respectively, and the sum of the three groups of product results is marked as the constraint effectiveness evaluation value;

[0063] It should be noted that the greater the value of the constraint effectiveness evaluation value, the higher the constraint risk of the safety constraint module for the corn harvester; the constraint effectiveness evaluation value is compared with the preset constraint effectiveness evaluation threshold value, if the constraint effectiveness evaluation value exceeds the preset constraint effectiveness evaluation threshold value, it is indicated that the constraint risk of the safety constraint for the corn harvester is high, and a constraint hidden danger signal is generated; if the constraint effectiveness evaluation value does not exceed the preset constraint effectiveness evaluation threshold value, it is indicated that the constraint risk of the safety constraint for the corn harvester is low, and a constraint qualified signal is generated.

[0064] Embodiment four: as shown in the embodiment and embodiment one, embodiment two, embodiment three, the difference lies in that the corn harvester energy-saving regulation and control method based on reinforcement learning comprises the following steps: Figure 3

[0065] Step one, collect the working environment and state parameters of the corn harvester and perform data processing;

[0066] Step two, design a continuous action set of the corn harvester;

[0067] Step three, generate specific control signals based on action instructions to automatically adaptively regulate and control the operation of the corn harvester;

[0068] Step four, monitor the risk items of the corn harvester, if the corresponding risk items of the corn harvester in the operation process are detected, step five is performed;

[0069] Step five, make corresponding safety constraint operation and generate constraint alarm information.

[0070] The working principle of the present application: in use, the related data of the corn harvester operation process is collected through the multi-source heterogeneous data acquisition module, the reinforcement learning decision module designs a continuous action set based on the received data information, the action execution control module converts the action instruction into a specific control signal to automatically adaptively regulate and control the operation of the corn harvester, and the safety constraint mechanism is introduced to reduce the abnormal operation risk of the corn harvester, realizing the full-process intelligent control of corn harvesting operation from environment perception to parameter optimization, which is conducive to ensuring the safe, stable and efficient and energy-saving operation of the corn harvester, significantly reducing the operation control difficulty of the corn harvester and improving its intelligent and automation level.

[0071] ​The threshold or the preset value, the preset range and the like in the technical scheme of the present application are set for result comparison and analysis, so as to determine whether it is good or not, and the size of the value is set according to the large model analysis of sample data and the combination of artificial experience, and is also stored, and can be appropriately adjusted through seasonal or reasonable influence conditions. The preset weight coefficient and the influence factor are set according to the influence of each parameter on the result, and the specific value is allocated to finally reflect the influence on the result, and the large model analysis of sample data and the combination of artificial experience are also used to set the storage, and appropriate adjustment can be made through seasonal or reasonable influence conditions.

[0072] The preferred embodiments of the present application disclosed above are only used to help explain the present application, and the preferred embodiments do not describe all the details and limit the present application to the specific embodiments. Obviously, many modifications and changes can be made according to the content of the present application. The present application selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present application, so that the skilled in the art can well understand and utilize the present application. The present application is limited by the claims and the entire scope and equivalents thereof.

Claims

1. A reinforcement learning based energy saving regulation system for corn harvester, characterized in that, The system comprises a multi-source heterogeneous data acquisition module, a reinforcement learning decision module, an action execution control module, a safety constraint module, and an operation control supervision terminal. The multi-source heterogeneous data acquisition module collects the working environment and state parameters of the corn harvester in real time, performs time-space alignment and noise filtering on the original data, and generates a three-dimensional state vector. The reinforcement learning decision module maps the three-dimensional state vector output by the multi-source heterogeneous data acquisition module into a discrete state code, designs a continuous action set, and the action execution control module converts the action instruction output by the reinforcement learning decision module into a specific control signal to automatically adaptively control the operation of the corn harvester. The safety constraint module sets a number of risk events to be monitored in advance, and if a corresponding risk event is detected during the operation of the corn harvester, a corresponding safety constraint operation is performed, a constraint alarm information is generated and sent to the operation control supervision terminal, and the operation control supervision terminal displays and warns the constraint alarm information. The operation control supervision terminal is communicatively connected to the harvesting continuity detection module, which analyzes the harvesting continuity execution condition of the corn harvester within a unit time, generates a continuity qualified signal or a continuity abnormal signal through analysis, and sends the continuity qualified signal or the continuity abnormal signal to the operation control supervision terminal. When the operation control supervision terminal receives the continuity abnormal signal, it issues a corresponding warning. The specific analysis process of the harvesting continuity detection module is as follows: The engine speed, header height, and cylinder drum speed of the corn harvester are obtained, and the engine speed is compared with the standard engine speed value set at present to obtain the engine detection value, and similarly, the header detection value and the cylinder drum detection value are obtained. The engine detection value, header detection value, and cylinder drum detection value are compared with the preset engine detection threshold, preset header detection threshold, and preset cylinder drum detection threshold, respectively. If the engine detection value, header detection value, or cylinder drum detection value exceeds the corresponding preset threshold, it is determined that the corn harvester is in a non-optimal harvesting state. The total duration of the corn harvester in a non-optimal harvesting state within a unit time is obtained and marked as a non-optimal harvesting characteristic value. If the non-optimal harvesting characteristic value does not exceed the preset non-optimal harvesting characteristic threshold, the engine detection value, header detection value, and cylinder drum detection value are weighted and summed to obtain a harvester operation measurement value. The maximum harvester operation measurement value within a unit time is marked as a harvester amplitude value. The non-optimal harvesting characteristic value, harvester operation analysis value, and harvester amplitude value are weighted and summed to obtain a continuity evaluation value. If the continuity evaluation value exceeds the preset continuity evaluation threshold, a continuity abnormal signal is generated; if the continuity evaluation value does not exceed the preset continuity evaluation threshold, a continuity qualified signal is generated. The harvesting continuity detection module is communicatively connected to the constraint validity evaluation module. The harvesting continuity detection module sends a continuity qualified signal to the constraint validity evaluation module. When the continuity qualified signal is received, the constraint validity evaluation module evaluates the safety constraint validity of the corn harvester in unit time, generates a constraint qualified signal or a constraint hidden danger signal through analysis, and sends the constraint qualified signal or the constraint hidden danger signal to the operation control supervision end. When the constraint hidden danger signal is received, the operation control supervision end issues a corresponding early warning. The specific analysis process of the constraint validity evaluation module includes: All risk events occurring in the running process of the corn harvester in unit time are obtained. The occurrence time of the corresponding risk event is marked as the first time, the time of making the corresponding constraint operation is marked as the second time, and the constraint delay value is calculated by calculating the time difference between the second time and the first time. The constraint delay value is compared with the corresponding preset constraint delay threshold value. If the constraint delay value exceeds the corresponding preset constraint delay threshold value, the constraint non-effective symbol TW-1 is assigned to the corresponding risk event. The number of times of assigning the constraint non-effective symbol TW-1 in unit time is obtained, and a constraint non-effective detection value is calculated by ratio calculation of the number of times and the total number of occurrences of the risk event. The constraint non-effective detection value is compared with the preset constraint non-effective detection threshold value. If the constraint non-effective detection value exceeds the preset constraint non-effective detection threshold value, a constraint hidden danger signal is generated. If the constraint non-effective detection value does not exceed the preset constraint non-effective detection threshold value, a constraint delay value is calculated by ratio calculation of the constraint delay value and the corresponding preset constraint delay threshold value. The mean value of all constraint delay values in unit time is calculated to obtain a constraint delay analysis value, and the maximum constraint delay value in unit time is marked as a constraint delay risk value. The constraint validity evaluation value is calculated by weighted summation of the constraint non-effective detection value, the constraint delay analysis value and the constraint delay risk value. The constraint validity evaluation value is compared with the preset constraint validity evaluation threshold value. If the constraint validity evaluation value exceeds the preset constraint validity evaluation threshold value, a constraint hidden danger signal is generated. If the constraint validity evaluation value does not exceed the preset constraint validity evaluation threshold value, a constraint qualified signal is generated.

2. The energy saving regulation system for corn harvester based on reinforcement learning according to claim 1, characterized in that, The multi-source heterogeneous data acquisition module scans the crop row spacing and density distribution by laser radar. The laser radar obtains crop canopy point cloud data in a 360° rotary scanning mode, removes weed interference points by a voxel grid filtering algorithm, identifies crop row spacing and local density distribution based on a density clustering algorithm, and generates a density heat map with a spatial resolution of 0.1m×0.1m. The body attitude of the corn harvester is monitored by an inertial measurement device. The body pitch angle, roll angle and acceleration change are monitored at a sampling frequency of 500Hz. A Kalman filtering algorithm is used to fuse multi-sensor data to construct an attitude feature vector including terrain slope and body vibration amplitude. The load change is obtained by an engine torque sensor. The instantaneous torque of the crankshaft is collected in real time, and the load fluctuation coefficient is calculated in combination with the gear ring speed signal. The working track is recorded by a GPS positioning module.

3. The energy saving regulation system for corn harvester based on reinforcement learning according to claim 1, characterized in that, The action execution control module sends the engine speed regulation instruction to the ECU through the CAN bus, uses feedforward-feedback compound control to achieve the target speed within 200 ms, drives the hydraulic proportional valve through the PWM signal to adjust the header height, cooperates with the displacement sensor to realize closed-loop control, and has a positioning accuracy of ±2 mm, and changes the clutch solenoid valve current to realize torque soft limit through the threshing cylinder torque limit instruction.

4. A control method for an energy-saving control system for a corn harvester based on reinforcement learning as described in any one of claims 1-3, characterized in that, The method comprises the following steps: Step one, collecting the working environment and state parameters of the corn harvester and processing the data; Step two, designing a continuous action set of the corn harvester; Step three, generating specific control signals based on action instructions; Step four, monitoring risk items of the corn harvester, and if corresponding risk items are detected during the operation of the corn harvester, step five is performed; Step five, making corresponding safety constraint operations and generating constraint alarm information.

Images