Seeding line spacing self-adaptive adjusting method and system

Abstract

The invention discloses a seeding row spacing adaptive adjustment method and system, and relates to the technical field of agricultural machinery control, and the method comprises the steps: training a regression neural network based on meteorological disturbance constraint, and obtaining a seeding machine slip prediction model; calling a probability analysis rule set according to the number of times of sliding offset, calculating and generating a slip rate, and performing anti-slip critical analysis on the slip rate and a predicted seeding machine running offset value output by the seeding machine slip prediction model to obtain a running speed reference value; inputting the driving speed reference value and a set seeding spacing into a seeding rate adjustment function with optimization so as to generate a seeding rate adjustment value; the meteorological data and the operation data of the sowing machine are fused and modeled, the running deviation value and the slip rate of the sowing machine at the future moment are predicted, the running speed and the sowing rate of the sowing machine are adjusted in time, crop sowing row line deviation caused by slipping of the sowing machine is avoided, and the stability and the consistency of the sowing operation row spacing are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of agricultural machinery control, in particular to a seeding row spacing self-adaptive adjustment method and system. BACKGROUND

[0002] In the field of agricultural machinery operation, the consistency of seeding row spacing is an important basis for improving seeding uniformity and achieving high yield and high quality of crops. The accuracy and consistency of seeding row spacing not only relate to the growth space of crops, but also affect subsequent field management and mechanized harvesting efficiency. Traditional seeding row spacing adjustment relies on fixed speed and preset spacing parameters, but in variable weather conditions and complex field soil environment, this method often cannot adapt to external disturbance factors, resulting in uneven crop spacing, accumulated row spacing deviation, and reduced resource utilization.

[0003] Currently, during the operation of the seeding machine, due to the straw or residue in the field, poor soil quality, uneven soil, and variables such as air moisture content, air pressure, and wind speed in weather conditions often interact with operating data such as soil moisture and wheel speed, which can easily cause the seeding machine to slip during driving. This slip not only increases the lateral difference between the driving path and the planned path, but also significantly increases the sensitivity of the seeding machine to driving deviation under rainfall trends. Due to the lack of deep integration of weather disturbance factors and seeding machine operating data, the prediction result lacks stability in complex environments, making it difficult to accurately identify potential deviation risks during seeding.

[0004] Secondly, during the seeding operation, the frequency of lateral deviation of the seeding machine and its cumulative effect often cause row spacing deviation. Although the number of deviations and deviation thresholds within a unit time window can be determined, the maximum allowed lateral deviation of different crops is quite different, making it difficult to obtain a slip rate that matches the actual crop requirements. This results in the inability to reasonably constrain the driving speed of the seeding machine under different crop conditions, resulting in excessive speed, causing skidding or low speed affecting operation efficiency.

[0005] In addition, during the seeding process of different crops, the target seeding spacing is different. If only fixed speed and seeding rate relationship are relied on for adjustment, it is easy to lose row spacing consistency in complex environments. When the slip rate changes and the seeding rate is not corrected in time, it will cause the amplification of crop row spacing deviation. SUMMARY

[0006] To solve the above technical problems, the present application provides a seeding row spacing self-adaptive adjustment method, which comprises: S11, training a regression neural network based on weather disturbance constraints to obtain a seeding machine slip prediction model; S12, calling a probability analysis rule set according to the number of sliding offset times, calculating a generated slip rate, and performing anti-slip critical analysis on the slip rate and a predicted seeder travel offset value output by the seeder slip prediction model to obtain a travel speed reference value; S13, inputting the travel speed reference value and a set seeding distance into a seeder speed regulation function with accompanying optimization to generate a seeding speed adjustment value.

[0007] Further, the step of training the regression neural network based on the meteorological disturbance constraint is: S111, collecting meteorological data, seeder operation data and seeder travel offset values at the same time of the same seeder in history , as training samples, and constructing a slip prediction training set, a slip prediction verification set and a slip prediction test set based on multiple groups of training samples; S112, inputting the slip prediction training set into the regression neural network, setting the input layer nodes to correspond to air moisture content data, air pressure data, wind speed data, wheel speed and soil humidity data, and setting the output layer nodes to be the predicted seeder travel offset value; S113, introducing a meteorological disturbance constraint in the training process of the regression neural network, and generating a meteorological penalty factor based on the meteorological disturbance constraint; S114, optimizing the loss function according to the meteorological penalty factor, and updating the regression neural network parameters based on the optimized loss function; S115, repeating the training iteration until the loss function converges to a preset convergence threshold or reaches a maximum iteration number, and outputting the converged regression neural network as an initial slip prediction network; S116, inputting the slip prediction test set into the initial slip prediction network, calculating the test error between the predicted seeder travel offset value and the true seeder travel offset value, and outputting the initial slip prediction network as the final seeder slip prediction model when the test error is less than a preset test threshold.

[0008] Further, the meteorological data includes air moisture content data, air pressure data and wind speed data; the seeder operation data includes wheel speed and soil humidity data.

[0009] Further, the construction formula of the meteorological penalty factor is:

[0010] In the formula, is the meteorological penalty factor of the i-th sample in the slip prediction training set at the j-th iteration training, , , , ​respectively, are the mean and standard deviation of all air moisture content data in the slip prediction training set, , respectively, are the mean and standard deviation of all air pressure data in the slip prediction training set, is the air moisture content data corresponding to the th sample in the slip prediction training set, is the air pressure data corresponding to the th sample in the slip prediction training set, is the constraint strength coefficient of the th iteration training, is the maximum value of the data in the parentheses.

[0011] Further, the generation logic of the constraint strength coefficient of the th iteration training is as follows: a1, obtain the mean and standard deviation of all air moisture content data in the slip prediction training set, and the mean and standard deviation of all air pressure data, and based on the air moisture content data and air pressure data in the slip prediction validation set, divide the meteorological disturbance sample set and the normal meteorological sample set through the established meteorological disturbance condition and normal meteorological condition; a2, according to the number of the divided meteorological disturbance sample set and the normal meteorological sample set, and the real seeder travel deviation value and the predicted seeder travel deviation value corresponding to the samples in the two sets, obtain the validation set prediction mean square error of the th iteration training under the two conditions; a3, according to the validation set prediction mean square error of the th iteration training under the two conditions, retrieve the preset sensitivity parameter to generate the constraint strength coefficient.

[0012] Further, the generation logic of the test threshold is as follows: Obtain the maximum allowed lateral deviation of the seeder under the agronomic requirements, square the maximum allowed lateral deviation to generate a maximum test error that can exist, and output the maximum test error as the test threshold.

[0013] Further, according to the probability analysis rule set according to the number of slip deviation times, the step of calculating and generating the slip rate is as follows: S121, in the process of seeder operation, the number of times that the real-time seeder travel deviation value exceeds the preset deviation threshold in the unit time window is counted, which is recorded as the number of slip deviation times; S122, based on the seeder travel deviation value, retrieve the established maximum allowed lateral deviation to generate a deviation weight; S123, according to the number of slip deviation times in the unit time window and the deviation weight, obtain the slip rate at moment. S124, based on the real-time slip ratio and offset weight, obtain The slip ratio at any given moment.

[0014] Furthermore, the steps to obtain the reference value for driving speed are as follows: S125, in real-time operation of the seeder At any given moment, acquire meteorological data and seeder operation data, and input them into the trained seeder slippage prediction model to generate... Predicted seeder travel offset at any given time; S126, based on The slip ratio and predicted seeder travel offset at any given time are compared with a pre-stored anti-slip critical database in the cloud to obtain the results. The slip ratio at any given time, the reference slip ratio, and the reference driving offset value are the same as the predicted driving offset value of the seeder. Based on the reference slip ratio and the reference driving offset value, the corresponding driving speed reference value is retrieved. S127: Calculate the slip ratio within a continuous time window and determine whether a correction alarm is triggered. If no correction alarm is triggered, output the driving speed reference value.

[0015] Furthermore, the generation logic for the anti-slip critical database is as follows: b1 sets the initial travel speed of the seeder. , This is an initial empirical value lower than the slippage speed; b2, based on meteorological and soil moisture data, sets up multiple sets of simulated environments. In one of these simulated environments, the seeder is set to operate at its initial travel value. The vehicle will reciprocate in a predetermined number of cycles. b3. After the preset number of reciprocating cycles has been reached, if no slippage occurs, proceed to step b4; if slippage occurs, proceed to step b5. b4. If no slippage occurs, update the initial driving value according to the predetermined speed step size, input the updated initial driving value into step b1, and repeat steps b1 to b3. b5, when skidding occurs, discard the initial speed value and stop the test. Subtract the speed step size from the initial speed value to generate a critical speed, and output the critical speed as a reference value for the driving speed. The slip ratio and travel offset of the seeder during its reciprocating cycles with an initial travel value are obtained. The slip ratio and travel offset are used as reference slip ratio and reference travel offset, and the reference slip ratio and reference travel offset are bound to a travel speed reference value. b6, based on multiple sets of reference slip rates, reference travel offset values and bound travel speed reference values, a skid prevention critical database is constructed and stored in the cloud.

[0016] An automatic seeding row distance adjustment system for performing any of the automatic seeding row distance adjustment methods, the system comprising: A model training module S21 for training a regression neural network based on meteorological disturbance constraints to obtain a seeder slip prediction model; A speed analysis module S22 for calling probability analysis rules set according to the number of slip offsets, calculating a slip rate, and performing skid prevention critical analysis on the slip rate and the predicted seeder travel offset value output by the seeder slip prediction model to obtain a travel speed reference value; A seeding adjustment module S23 for inputting the travel speed reference value and the predetermined seeding distance into a seeding rate adjustment function with accompanying optimization to generate a seeding rate adjustment value.

[0017] Compared with the prior art, the present application has the following advantages: The present application introduces meteorological disturbance constraints in the training of the seeder slip prediction model, fuses air moisture content data, air pressure data and wind speed data with seeder operation data to improve the prediction stability of the model for seeder travel offset values under complex weather conditions, thereby enabling early identification of seeder skid risk under rainfall trend environment, thereby enhancing the model's adaptability to environmental changes. In addition, the present application generates a real-time slip rate based on the number of slip offsets and offset weights, and compares the prediction result of the seeder slip prediction model with the skid prevention critical database to obtain a travel speed reference value within the maximum allowed lateral offset range corresponding to different crops, thereby achieving dynamic constraint of the seeder speed while maintaining job continuity and effectively avoiding offset accumulation caused by increased soil humidity or deteriorating weather conditions. Further, the present application inputs the travel speed reference value and the target seeding distance into the seeding rate adjustment function with accompanying optimization, and dynamically corrects the seeding rate in combination with the real-time slip rate influence factor, thereby maintaining the consistency of the seeding distance under different crop types and row distance requirements, and thereby achieving stable control of the seeding process. In summary, the present application fuses meteorological data with seeder operation data for modeling, predicts the travel offset value and slip rate of the seeder at future time under the influence of meteorological data and seeder operation data, realizes timely adjustment of the seeder travel speed and seeding speed, and thereby improves the stability of the seeding operation. BRIEF DESCRIPTION OF DRAWINGS

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1 This is a flowchart of a method for adaptive adjustment of sowing row spacing provided in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the lateral difference and sowing spacing in Embodiment 1 of the present invention; Figure 3 This is a block diagram of an automatic row spacing adjustment system for sowing provided in Embodiment 2 of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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. Example

[0021] Please see Figure 1 As shown in the figure, this embodiment discloses a method for adaptive adjustment of sowing row spacing, the method comprising: S11, The regression neural network based on meteorological disturbance constraints is trained to obtain the seeder slippage prediction model; Specifically, the steps for training a regression neural network based on meteorological disturbance constraints are as follows: S111, Collecting data from the same seeder in history Real-time weather data, seeder operation data, and The driving offset value of the seeder at each moment is used as a training sample. Based on multiple sets of training samples, a slip prediction training set, a slip prediction validation set, and a slip prediction test set are constructed. It should be noted that the seeder travel offset value is the lateral difference between the seeder and the planned travel path during the sowing process; During the sowing process, the seeder sows along a pre-planned straight path, denoted as the planned travel path. However, sowing deviation is more likely to occur during or after rain. This means that during sowing, the seeder may slip, and when the operator corrects the machine, it resumes a straight, forward motion, but this time it has already slipped. The difference between the two ends of the current travel path and the two ends of the planned travel path is called the lateral difference. Figure 3 The Lac shown is the lateral difference. The dashed line represents the trajectory of the two ends of the current driving path, and the solid line represents the trajectory of the two ends of the planned driving path. The lateral difference can be calculated by using GPS or RTK positioning to obtain the actual path coordinates, comparing them with the preset work trajectory coordinates, and then generating the lateral difference. The slip prediction training set, slip prediction validation set, and slip prediction test set all include multiple sets of data. Real-time meteorological data, seeder operation data, and corresponding The offset value of the seeder at any given time.

[0022] Specifically, the meteorological data includes air moisture content data, air pressure data, and wind speed data; the seeder operation data includes wheel speed and soil moisture data. It should be noted that the air moisture content data is collected by a capacitive humidity sensor installed at the front of the seeder or at the meteorological monitoring point in the work area. The higher the air moisture content, the more saturated the environment is, and the greater the possibility of future rainfall. Rainfall will increase soil moisture and reduce the coefficient of friction, making the seeder more prone to slippage during operation and increasing the probability of the seeder deviating from its course. Air pressure data is collected by a silicon piezoresistive air pressure sensor installed on the seeder. A continuous drop in air pressure usually indicates the arrival of a low-pressure system and is accompanied by a rainfall trend. Wind speed data is collected by an ultrasonic anemometer installed on the top of the seeder. When the wind speed is high or changes suddenly, it is often accompanied by the passage of a front and rainfall. The wheel speed is collected by a Hall effect wheel speed sensor installed at the drive wheel axle. Soil moisture data is collected by a capacitive soil moisture sensor installed under the seeder or in the shallow layer of the working area. When the soil moisture increases, the ground friction coefficient decreases, the drive wheel adhesion decreases, and the seeder is more likely to slip in the soil and deviate from the intended trajectory.

[0023] S112, Input the slip prediction training set into the regression neural network, set the input layer nodes to correspond to air moisture content data, air pressure data, wind speed data, wheel speed and soil moisture data, and set the output layer nodes to predict the seeder travel offset value; It should be noted that before being used as input and output layer nodes, the data in the slip prediction training set needs to be preprocessed. The preprocessing includes, but is not limited to, timestamp alignment, missing value completion, outlier removal, and numerical normalization of air moisture content data, air pressure data, wind speed data, wheel speed data, and soil moisture data. This is to ensure that data from different sources participate in training on a uniform scale and to improve the prediction stability when the seeder's driving offset value is used as the output. ReLU is used as the activation function for the hidden layer, and a linear function is used as the activation function for the output layer, which will not be elaborated further.

[0024] S113 introduces meteorological perturbation constraints during the training of the regression neural network and generates a meteorological penalty factor based on the meteorological perturbation constraints; In one specific embodiment, the meteorological disturbance constraint is constructed based on the joint trend of air moisture content data and air pressure data. When air moisture content increases and air pressure decreases, a meteorological penalty factor is added to the predicted offset value during weight updates of the regression neural network to simulate the sensitivity of the seeder's operating state to the driving offset value under rainfall conditions. The formula for constructing the meteorological penalty factor is as follows:

[0025] In the formula, For the slip prediction training set One sample in Weather penalty factor during subsequent training iterations , These represent the mean and standard deviation of all air moisture content data in the slip prediction training set, respectively. , These represent the mean and standard deviation of all air pressure data in the slip prediction training set, respectively. For the slip prediction training set Air moisture content data for each sample For the slip prediction training set Air pressure data corresponding to each sample For the first The constraint strength coefficient of the next iteration of training To retrieve the maximum value of the data within the parentheses; It should be noted that: Used to measure how high the humidity of the sample is relative to the average level, when When the value is greater than zero, it indicates that the air moisture content of the sample is higher than the average value, and the larger the value, the closer the humidity is to saturation, thus increasing the possibility of rainfall. When the calculation result is less than or equal to zero, zero is taken as the constraint input, indicating that the air humidity does not exceed the average level and does not trigger the risk signal of rainfall trend. This is used to measure how low the air pressure of the sample is relative to the average level. When the calculation result is greater than zero, it means that the air pressure of the sample is lower than the average value, and the larger the value, the stronger the low pressure and the greater the probability of rainfall. When the calculation result is less than or equal to zero, zero is taken as the constraint input, indicating that the air pressure of the sample is not significantly lower than the average level and does not trigger the risk signal of rainfall trend.

[0026] Specifically, the first Constraint strength coefficients for the next iteration of training The generation logic is as follows: a1. Obtain the mean and standard deviation of all air moisture content data and all air pressure data in the slip prediction training set. Based on the air moisture content data and air pressure data in the slip prediction validation set, divide the meteorological disturbance sample set and the normal meteorological sample set according to the given meteorological disturbance conditions and normal meteorological conditions. In a specific embodiment, the meteorological disturbance condition is expressed as:

[0027] in, For the slip prediction validation set, the first Air moisture content data for each sample , These represent the mean and standard deviation of all air moisture content data in the slip prediction training set, respectively. For the first in the slip pre-verification set Air pressure data for each sample, , These are the mean and standard deviation of all air pressure data in the slip prediction training set, respectively. It should be noted that the validation set of samples selected from meteorological disturbance conditions indicates that the air moisture content is significantly higher than normal, the air pressure is significantly lower than normal, and the air humidity is close to saturation, which is a typical combination of precursors to rainfall.

[0028] Normal weather conditions are expressed as follows:

[0029] It should be noted that the validation set samples selected under normal meteorological conditions indicate that the air moisture content and air pressure are within the normal fluctuation range and there are no significant abnormalities. These samples represent working environments with sunny days or stable climate.

[0030] a2, based on the number of meteorological disturbance sample sets and regular meteorological sample sets, and the actual and predicted seeder travel offset values ​​corresponding to the samples in the two sets, the results are obtained under the two types of conditions. Mean squared error of the validation set predictions during the next iteration of training; It should be noted that the real sowing machine running deviation value is the sowing machine running deviation value when the training sample is constructed, and the predicted sowing machine running deviation value is the sowing machine running deviation value predicted by the regression neural network in the current iteration training process; in a specific embodiment, when the meteorological disturbance condition is met, the logical formula of the verification set prediction mean square error of the first iteration training is:

[0031] In the formula, is the verification set prediction mean square error of the first iteration training under the meteorological disturbance condition, is the number of meteorological disturbance samples, is the real sowing machine running deviation value of the i-th sample in the meteorological disturbance sample set, is the predicted sowing machine running deviation value of the i-th sample in the first iteration training, is the predicted sowing machine running deviation value of the i-th sample in the first iteration training; When the regular meteorological condition is met, the logical formula of the verification set prediction mean square error of the first iteration training is:

[0032] In the formula, is the verification set prediction mean square error calculated by the first iteration training under the regular meteorological condition, is the number of regular meteorological samples, is the real sowing machine running deviation value of the i-th sample in the regular meteorological sample set, is the predicted sowing machine running deviation value of the i-th sample in the regular meteorological sample set, is the predicted sowing machine running deviation value of the i-th sample in the first iteration training, is the predicted sowing machine running deviation value of the i-th sample in the first iteration training; a3, according to the verification set prediction mean square error of the first iteration training under the two conditions, calls the preset sensitivity parameter to generate the constraint strength coefficient; is expressed as:

[0033] In the formula, is the constraint strength coefficient of the first iteration training, is the initial value set based on historical meteorological sample data, is the sensitivity parameter;For example, when historical meteorological sample data shows that under a strong rainfall trend, the actual seeder driving deviation value increases by 30%–50% compared to normal conditions, the initial value can be... The sensitivity parameter is set to [0.3, 0.5] to ensure that the model has sufficient constraints on perturbation conditions in the early stages of training; the sensitivity parameter is then gradually adjusted on the slip prediction validation set. The value of is taken, and observation is made. The update rate and the final model's test error on the test set, when the sensitivity parameter Setting the sensitivity parameter too low results in a slow constraint response, potentially leading to distorted predictions under perturbed samples; conversely, setting it too high may cause the model to distort its predictions under perturbed samples. Setting a large sensitivity parameter may lead to overfitting of normal samples. Based on these two critical conditions, the optimal sensitivity parameter is selected. Preferred The value range is [0.5, 2].

[0034] S114. Optimize the loss function based on the meteorological penalty factor, and update the parameters of the regression neural network based on the optimized loss function; It should be noted that the parameters of the regression neural network include the weights and biases of the regression neural network; In a specific embodiment, the optimization formula for the loss function is:

[0035] In the formula, For the first The loss function for the next iteration of training. The total number of samples in the training set for slip prediction. For the first The actual driving offset value of each sample seeder. For the first The sample at the th Predicted seeder travel offset values ​​from the next iteration of training. For the slip prediction training set One sample in Weather penalty factor during subsequent training iterations; Based on the optimized loss function, the weights and biases of the regression neural network are updated, as follows:

[0036]

[0037] In the formula, For the first The weights of a regressive neural network trained iteratively. For the first The weights of the regression neural network trained in the next iteration. the bias of the regression neural network trained for the first iteration, the bias of the regression neural network trained for the first iteration, is a predetermined learning rate; It should be noted that the learning rate is used to control the step size of the neural network parameters in the process of updating each iteration. It can be understood that in the regression neural network of the embodiment, the learning rate is a predetermined value, and the step size shows the update speed of each iteration training. The larger the learning rate is, the faster the update speed is, and vice versa. Preferably, the value range of the learning rate is .

[0038] S115, repeating the training iteration until the loss function converges to a preset convergence threshold or reaches a maximum iteration number, and outputting the converged regression neural network as an initial slip prediction network; It should be noted that the convergence threshold is obtained by: In the initial experiment, different candidate thresholds are set, for example , the error change trend on the validation set is observed, and the threshold that can reach a smaller error within a reasonable training time and does not cause overfitting is selected as the convergence threshold.

[0039] S116, inputting the slip prediction test set into the initial slip prediction network, calculating the test error between the predicted seeder travel offset value and the true seeder travel offset value, and when the test error is less than a preset test threshold, outputting the initial slip prediction network as a final seeder slip prediction model; If the test error is greater than the preset test threshold, return to step S111 to readjust the training parameters and perform training and verification again until a final seeder slip prediction model that meets the requirements is obtained; In a specific embodiment, the calculation formula of the test error is:

[0040] is the test error, is the total number of samples in the slip prediction test set, is the true seeder travel offset value of the first sample, is the predicted seeder travel offset value of the first sample; It should be noted that the test threshold is set by experimenters based on the maximum lateral offset allowed in actual seeder operation; Specifically, the generation logic of the test threshold is: acquire the maximum allowed lateral deviation of the seeding machine under the agricultural requirements, square the maximum allowed lateral deviation to generate a maximum test error that can exist, and output the maximum test error as a test threshold; is expressed as:

[0041] In the formula, is the maximum allowed lateral deviation, is the maximum test error. It should be noted that the dimension of the lateral deviation can be changed to ensure that the numerical range of the lateral deviation is between 0 and 1.

[0042] S12, calling a probability analysis rule set according to the number of sliding deviations, calculating a sliding rate, and performing anti-slip critical analysis on the sliding rate and the predicted seeding machine travel deviation value output by the seeding machine sliding prediction model to obtain a travel speed reference value; It should be noted that the purpose of this step is to obtain a travel speed reference value that can ensure that the seeding machine does not slip and deviate, and can travel as efficiently as possible; Specifically, according to the probability analysis rule set according to the number of sliding deviations, the step of calculating the sliding rate is: S121, in the process of seeding machine operation, the number of times that the real-time seeding machine travel deviation value in a unit time window exceeds a preset deviation threshold is counted, which is recorded as the number of sliding deviations; It should be noted that: In a specific embodiment, the length of the time window is set to The number of times that the deviation threshold is exceeded in the time window is counted as The total number of samples in the window is recorded as ; It should be noted that the deviation threshold is a seeding machine travel deviation value set by a person according to the type of crop and the agricultural requirements, and the deviation threshold corresponding to different crops is different: For example, the deviation threshold for wheat, rape and other row crops can be set to 6-8 cm, and the deviation threshold for soybean, corn and other hill planting crops can be set to 12-18 cm.

[0043] The deviation threshold is determined by agricultural personnel before seeding by combining crop characteristics, planting density and mechanical operation conditions, and is used as a judgment basis when counting the number of sliding deviations.

[0044] S122, based on the seeding machine travel deviation value, the established maximum allowed lateral deviation is called to generate a deviation weight; is expressed as:

[0045] wherein, is the offset weight of the time point , is the offset weight of the time point is the seeder driving offset value of the time point is the maximum allowed lateral offset; It should be noted that the offset weight of the time point in the model is a historical time point, while the time point in step S12 is a real-time operation time point; S123, according to the sliding offset number and the offset weight in the unit time window, the sliding rate of the time point is obtained; is represented as:

[0046] wherein, is the real-time sliding rate of the time point , is the offset weight of the time point , is the indication variable of the time point ; if the seeder driving offset value of the time point exceeds the preset offset threshold, the indication variable is 1, otherwise, the indication variable is 0; is the time window with the target reference time point , is the total number of sampling points in the time window with the target reference time point . It should be noted that the target reference time point is the final time point of the time window.

[0047] S124, according to the real-time sliding rate and the offset weight, the sliding rate of the time point is obtained.

[0048] wherein, is the sliding rate of the time point , is the time balance coefficient based on experience , is the sliding offset number change amount of adjacent time windows, is the real-time sliding rate of the time point , is the total sampling number in the time window.

[0049] Specifically, the step of obtaining the driving speed reference value is: S125, in real-time operation of the seeder At any given moment, acquire meteorological data and seeder operation data, and input them into the trained seeder slippage prediction model to generate... Predicted seeder travel offset at any given time; S126, based on The slip ratio and predicted seeder travel offset at any given time are compared with a pre-stored anti-slip critical database in the cloud to obtain the results. The slip ratio at any given time, the reference slip ratio, and the reference driving offset value are the same as the predicted driving offset value of the seeder. Based on the reference slip ratio and the reference driving offset value, the corresponding driving speed reference value is retrieved. Specifically, the generation logic for the anti-skid critical database is as follows: b1, setting the initial travel speed value of the seeder. This is an initial empirical value lower than the slippage speed; It should be noted that: The minimum speed at which slippage did not occur after multiple experiments under different soil moisture and weather conditions was determined, and this minimum speed was used as a unified empirical initial value. b2, based on meteorological and soil moisture data, sets up multiple sets of simulated environments. In one of these simulated environments, the seeder is set to operate at its initial travel value. The vehicle will reciprocate in a predetermined number of cycles. It should be noted that the reciprocating motion is carried out on the same terrain of the farm road. The larger the number of reciprocating motions, the higher the accuracy of the critical speed determination. When computing resources are limited, it can be set to 10 motions, while under high-precision computing conditions, the number of reciprocating motions must be greater than or equal to 100. b3. After the preset number of reciprocating cycles has been reached, if no slippage occurs, proceed to step b4; if slippage occurs, proceed to step b5. b4, When no skidding occurs, proceed according to the predetermined speed step size. Update the initial driving value, represented as: Input the updated initial driving value into step b1, and repeat steps b1 to b3; b5, When skidding occurs, the initial driving value is discarded. And stop the test, and set the initial driving value. Subtract speed step size Generate critical velocity It is output as a reference value for driving speed, expressed as: ;as well as, Obtain the seeder with initial travel value. The slip ratio and seeder travel offset value during the reciprocating travel cycles are calculated. These slip ratio and seeder travel offset values ​​are used as reference slip ratio and reference travel offset values, respectively. These reference slip ratio and reference travel offset values ​​are then compared with the travel speed reference value. Bind; It should be noted that the slip ratio calculation formula for this step is:

[0050] The number of times the offset threshold is exceeded within the time window, serving as the reference slip ratio, is denoted as... The total number of samples within the window is denoted as ; The logic for generating the reference driving offset value is the same as the logic for generating the seeder driving offset value in step S111, see step S111 for details; b6, based on multiple sets of reference slip ratios, reference driving offset values ​​and bound driving speed reference values, constructs an anti-skid critical database and stores it in the cloud.

[0051] S127: Calculate the slip ratio within a continuous time window and determine whether a correction alarm is triggered. If no correction alarm is triggered, output the driving speed reference value. Statistical continuity The slip ratio within a time window, when the slip ratio of consecutive time windows exceeds a preset slip ratio threshold. When this happens, a correction alarm signal is output to prompt the driver to perform direction correction and trajectory reset operations; The determination formula is expressed as follows:

[0052] In the formula, For the first Slip rate within a time window To preset the slip ratio threshold, This is an indicator function that takes the value 1 when the condition is true and 0 otherwise. This represents the number of consecutive time windows. It should be noted that: slip ratio threshold Based on the agronomic requirements of different crops, the spacing can be set to 0.3 to 0.4 for crops with medium row spacing such as corn and rapeseed, and to 0.15 to 0.25 for crops with high requirements for row spacing consistency such as rice and vegetables. Number of consecutive time windows The preferred value is 3 to 5; when multiple consecutive windows exceed the threshold, it indicates that the current sliding state has entered the continuous offset stage, and the accuracy cannot be restored by speed adjustment alone, so direction correction needs to be triggered.

[0053] S13, input the driving speed reference value and the predetermined seeding distance into the seeding rate adjustment function with optimization to generate a seeding rate adjustment value; Specifically, the seeding rate adjustment function is represented as:

[0054] In the formula, is the seeding rate adjustment value, is the theoretical seeding rate, is the slip rate influence factor; The theoretical seeding rate is calculated by the formula:

[0055] In the formula, is the driving speed reference value, is the target seeding distance; The logical formula of the slip rate influence factor is:

[0056] In the formula, is the slip rate at the moment.

[0057] It should be noted that the predetermined seeding distance is the seeding distance in the driving direction of the seeding machine, such as Lz shown in Figure 3 The seeding distance is set based on the crop growth manual; For example, soybeans can be set to 20-25m, and corn can be set to 10-16cm.

[0058] Embodiment 2 Referring to Figure 2 Based on the unified invention concept, the embodiment discloses a seeding row distance automatic adjustment system, which comprises a model training module S21, a speed analysis module S22 and a seeding adjustment module S23, specifically: The model training module S21 is used for training a regression neural network based on meteorological disturbance constraints to obtain a seeding machine slip prediction model; the model training module speed analysis module seeding adjustment module The speed analysis module S22 is used for calling a probability analysis rule set according to the number of slip offsets, calculating and generating a slip rate, and performing anti-slip critical analysis on the slip rate and the predicted seeding machine driving offset value output by the seeding machine slip prediction model to obtain a driving speed reference value; The seeding adjustment module S23 is used for inputting the driving speed reference value and the predetermined seeding distance into the seeding rate adjustment function with optimization to generate a seeding rate adjustment value.

[0059] The above-described embodiments can be implemented in part or in whole through software implemented by a processor. The above-described embodiments can be implemented by a special-purpose computer, a programmable computer or other programmable devices. The software implemented by the processor or the other programmable device can be stored in one or more computer-readable storage media such as one or more of the system memories, the portable storage unit, and the storage devices shown in the drawings, or can be stored in another computer-readable storage medium for use by the computer or the other programmable devices.

[0060] In several embodiments of the present application, it should be understood that the disclosed system, apparatus, and method can be implemented in other ways. For example, the apparatus embodiments described above are merely schematic, and the division of the units is merely one of the possible divisions. In actual implementation, another division manner can be used, for example, a plurality of units or components can be combined or integrated into another system, or some features can be omitted or not implemented. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0061] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, can be located in one place or can be distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0062] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically, or two or more units can be integrated into one unit.

[0063] Part of data in the above formula is calculated by removing dimension, and the formula is obtained by software simulation of a large amount of collected data to be closest to the real situation; the preset parameters and the preset threshold in the formula are set by the person skilled in the art according to the actual situation or obtained by a large amount of data simulation.

[0064] The above examples are only used to illustrate the technical method of the present application but not limit the present application. Although the present application is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.

Claims

1. A method for adaptive adjustment of seeding row spacing, characterized in that, The method comprises: S11, training a regression neural network based on meteorological disturbance constraints to obtain a seeder slip prediction model; S12, calling a probability analysis rule set according to the number of slip offset, calculating the slip rate, and performing anti-slip critical analysis on the slip rate and the predicted seeder travel offset value output by the seeder slip prediction model to obtain a travel speed reference value; S13, inputting the travel speed reference value and the predetermined seeding spacing into the seeding rate adjustment function with accompanying optimization to generate a seeding rate adjustment value.

2. The method according to claim 1, wherein, The step of training the regression neural network based on meteorological disturbance constraints is: S111, Collecting data from the same seeder in history Real-time weather data, seeder operation data, and The driving offset value of the seeder at each moment is used as a training sample. Based on multiple sets of training samples, a slip prediction training set, a slip prediction validation set, and a slip prediction test set are constructed. S112, inputting the slip prediction training set into the regression neural network, setting the input layer nodes to correspond to air moisture content data, air pressure data, wind speed data, wheel speed and soil humidity data, and setting the output layer nodes to be the predicted seeder travel offset value; S113, introducing meteorological disturbance constraints during the training of the regression neural network, and generating a meteorological penalty factor based on the meteorological disturbance constraints; S114, optimizing the loss function according to the meteorological penalty factor, and updating the regression neural network parameters based on the optimized loss function; S115, repeating the training iteration until the loss function converges to a preset convergence threshold or reaches a maximum number of iterations, and outputting the converged regression neural network as an initial slip prediction network; S116, inputting the slip prediction test set into the initial slip prediction network, calculating the test error between the predicted seeder travel offset value and the true seeder travel offset value, and outputting the initial slip prediction network as the final seeder slip prediction model when the test error is less than a preset test threshold.

3. The method of claim 2, wherein, The meteorological data includes air moisture content data, air pressure data and wind speed data; and the seeder operation data includes wheel speed and soil humidity data.

4. The method of claim 3, wherein, The construction formula of the meteorological penalty factor is: In the formula, For the slip prediction training set One sample in Weather penalty factor during subsequent training iterations , These represent the mean and standard deviation of all air moisture content data in the slip prediction training set, respectively. , These represent the mean and standard deviation of all air pressure data in the slip prediction training set, respectively. For the slip prediction training set Air moisture content data for each sample For the slip prediction training set Air pressure data corresponding to each sample For the first The constraint strength coefficient of the next iteration of training To retrieve the maximum value of the data within the parentheses.

5. The method of claim 4, wherein, The first The constraint strength coefficient generation logic for the second iteration is: a1, obtaining the mean and standard deviation of all air moisture content data in the slip prediction training set, and the mean and standard deviation of all air pressure data, and dividing the air moisture content data and air pressure data in the slip prediction validation set into a meteorological disturbance sample set and a normal meteorological sample set based on the predetermined meteorological disturbance conditions and normal meteorological conditions; a2, according to the number of the divided weather disturbance sample set and the conventional weather sample set, and the true seeding machine travel deviation value and the predicted seeding machine travel deviation value corresponding to the sample in the two sets, the first iteration training verification set prediction mean square error; a3, according to the first iteration training under two conditions Validation set prediction mean square error of the second iteration training, call the preset sensitivity parameter, generate constraint strength coefficient.

6. The method of claim 5, wherein, The generation logic of the test threshold is: Obtaining the maximum allowed lateral offset of the seeder under the requirements of agronomy, squaring the maximum allowed lateral offset to generate a maximum test error that can exist, and outputting the maximum test error as the test threshold.

7. The method of claim 6, wherein, The step of calculating the slip rate according to the probability analysis rule set according to the number of slip offset is: S121, in the process of seeder operation, counting the number of times that the real-time seeder travel offset value exceeds a preset offset threshold in a unit time window, which is recorded as the number of slip offset; S122, based on the seeder travel offset value, calling the maximum allowed lateral offset to generate an offset weight; S123, according to the sliding offset times and offset weight within the unit time window, obtain slip rate at the moment; S124, obtaining, according to the real-time slip ratio and the offset weight, a slip ratio at the moment.

8. The method of claim 7, wherein, The step of obtaining the travel speed reference value is: S125, at the moment when the seeding machine is running in real time , acquire the meteorological data and the seeding machine running data at the moment, and input them to the trained seeding machine slip prediction model to generate the predicted seeding machine driving deviation value at the moment; S126, based on the slip rate and the predicted seeder travel offset value at the time, and a pre-stored anti-slip critical database in the cloud to obtain a reference slip rate and a reference travel offset value corresponding to the slip rate and the predicted seeder travel offset value at the time. the slip rate and the predicted seeder travel offset value at the time, and a pre-stored anti-slip critical database in the cloud to obtain a reference slip rate and a reference travel offset value corresponding to the slip rate and the predicted seeder travel offset value at the time. S127, counting the slip rate in a continuous time window, judging whether a correction alarm is triggered, and outputting the travel speed reference value if the correction alarm is not triggered.

9. The method of claim 8, wherein, The generation logic of the anti-slip critical database is: b1, initial driving value of the sowing machine driving speed , is an empirical initial value below the slip speed b2, based on the weather data and the soil moisture data, set a plurality of groups of simulated environments, under one of the groups of simulated environments, let the seeding machine run at an initial driving value reciprocally drive according to the preset reciprocally driving wheel times; b3, if no driving slip occurs after the preset number of reciprocating driving rounds, step b4 is performed; If slip occurs, step b5 is performed; b4, when no driving slip occurs, the initial driving value is updated according to the predetermined speed step, and the updated initial driving value is input into step b1, and steps b1-b3 are repeated; b5, when driving slip occurs, the initial driving value is removed and the test is stopped, the initial driving value is reduced by the speed step, the critical speed is generated, and the critical speed is output as the driving speed reference value, and The slip rate and the driving offset value of the seeding machine in the initial driving value are obtained in the reciprocating driving rounds, and the slip rate and the driving offset value of the seeding machine are taken as the reference slip rate and the reference driving offset value, and the reference slip rate and the reference driving offset value are bound with the driving speed reference value; b6, based on the plurality of reference slip rates, reference driving offset values and bound driving speed reference values, a slip prevention critical database is constructed and stored in the cloud.

10. A seeding row distance automatic adjustment system for performing the seeding row distance automatic adjustment method according to any one of claims 1 to 9, characterized by The system comprises: A model training module S21 is configured to train a regression neural network based on meteorological disturbance constraints to obtain a seeding machine slip prediction model; A speed analysis module S22 is configured to call a probability analysis rule set according to the number of slip offset times, calculate a slip rate, and perform slip prevention critical analysis on the slip rate and the predicted seeding machine driving offset value output by the seeding machine slip prediction model to obtain a driving speed reference value; A seeding adjustment module S23 is configured to input the driving speed reference value and the predetermined seeding interval into a seeding rate adjustment function with accompanying optimization to generate a seeding rate adjustment value.

Images