Intelligent scheduling method and system for agricultural machinery based on yield prediction

Abstract

The present invention discloses an intelligent dispatching method and system for agricultural machinery based on yield prediction, which belongs to the field of intelligent agricultural machinery technology. The method includes data collection, data preprocessing, crop yield prediction, intelligent dispatching of agricultural machinery and comprehensive dispatching of agricultural machinery. The present invention adopts a time convolution feature enhanced gated recurrent network for crop yield prediction, which has better long-term and short-term temporal dependency modeling capabilities, and can adaptively enhance key features, improve the stability and accuracy of crop yield prediction, and help improve the reliability of subsequent intelligent dispatching of agricultural machinery; the discrete particle swarm optimization algorithm is used for intelligent dispatching of agricultural machinery, which has strong global search capabilities, can adapt to different agricultural production environments and dispatching needs, has significant advantages in large-scale agricultural production, and improves the dispatching efficiency of agricultural machinery; the comprehensive intelligent dispatching of agricultural machinery based on yield prediction is adopted, which fully considers the differences in crop yields, reasonably arranges agricultural machinery operations, and effectively avoids waste of resources.

Description

Technical Field

[0001] The present invention belongs to the technical field of smart agricultural machinery, and specifically refers to an intelligent scheduling method and system for agricultural machinery based on yield prediction. Background Art

[0002] Intelligent scheduling of agricultural machinery based on yield forecast is a scheduling method that comprehensively considers crop yield forecast information and scientifically allocates agricultural machinery resources. It aims to improve the efficiency of agricultural machinery use, reduce unnecessary waste of resources, help reduce carbon emissions and environmental burdens in agricultural production, promote the development of green agriculture, and promote the transformation of agriculture towards intelligence.

[0003] However, in the existing crop yield prediction process, there are technical problems that traditional crop yield prediction usually relies on simple statistical models, which makes it difficult to fully consider the complex nonlinear relationship between multiple factors such as meteorology, images, and crop growth, and the crop growth process has a strong time-series dependence, and the traditional crop yield prediction has poor temporal feature modeling capabilities; in the existing intelligent scheduling of agricultural machinery, there are technical problems that traditional scheduling methods often rely on manual arrangements, are easily affected by human factors, and result in unreasonable operation arrangements, lack of global optimization, and are difficult to adapt to the needs of large-scale agricultural production; there are technical problems that traditional scheduling methods do not take into account crop yield differences, resulting in uneven distribution of agricultural machinery resources, excessive operation of agricultural machinery on plots, and unnecessary waste of resources. Summary of the invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an intelligent scheduling method and system for agricultural machinery based on yield prediction. In view of the technical problems in the existing crop yield prediction process, the traditional crop yield prediction usually relies on simple statistical models, which makes it difficult to fully consider the complex nonlinear relationship between multiple factors such as meteorology, images, crop growth, etc., and the crop growth process has a strong time series dependence, and the traditional crop yield prediction has a poor time series feature modeling ability. This scheme creatively uses a temporal convolution feature enhanced gated recurrent network to perform crop yield prediction, which has better long-term and short-term time series dependency modeling capabilities, and can adaptively enhance key features and suppress noise interference, thereby improving the stability and accuracy of crop yield prediction, and helping to improve the reliability of subsequent intelligent scheduling of agricultural machinery; in view of the technical problems in the existing intelligent scheduling of agricultural machinery, the traditional scheduling method The method often relies on manual arrangement and is easily affected by human factors, resulting in unreasonable operation arrangement, lack of global optimization, and difficult to adapt to the technical problems of large-scale agricultural production needs. This scheme creatively adopts discrete particle swarm optimization algorithm for intelligent scheduling of agricultural machinery, which has strong global search ability and can adapt to different agricultural production environments and scheduling needs. It has significant advantages in large-scale agricultural production, effectively avoids the high error problem of manual scheduling, and improves the efficiency of agricultural machinery scheduling. In the existing intelligent scheduling of agricultural machinery, the traditional scheduling method does not take into account the difference in crop yields, resulting in uneven allocation of agricultural machinery resources, excessive operation of agricultural machinery in the field, and unnecessary waste of resources. Technical problems, this scheme adopts comprehensive intelligent scheduling of agricultural machinery based on yield prediction, fully considers the difference in crop yields, reasonably arranges agricultural machinery operations, and effectively avoids waste of resources.

[0005] The technical solution adopted by the present invention is as follows: The present invention provides an intelligent dispatching method for agricultural machinery based on yield prediction, the method comprising the following steps:

[0006] Step S1: data collection;

[0007] Step S2: data preprocessing;

[0008] Step S3: crop yield prediction;

[0009] Step S4: Intelligent dispatching of agricultural machinery;

[0010] Step S5: Comprehensive agricultural machinery scheduling.

[0011] Furthermore, in step S1, the data collection is used to collect the original data required for the intelligent dispatching of agricultural machinery, specifically, by collecting the yield prediction related data and the agricultural machinery related data to obtain the original data for the dispatching of agricultural machinery;

[0012] The yield prediction related data includes plot information, meteorological data and crop growth data;

[0013] The agricultural machinery related data include agricultural machinery number, number of agricultural machinery, type of agricultural machinery, fuel consumption and maximum usage time;

[0014] The meteorological data include precipitation, temperature, humidity and wind speed;

[0015] The land plot information, including land plot area, location, soil type and crop type;

[0016] The crop growth data include vegetation index, crop height and crop coverage rate.

[0017] Furthermore, in step S2, the data preprocessing includes the following steps:

[0018] Step S21: data cleaning, specifically removing duplicate values ​​in the original data of agricultural machinery scheduling and interpolating missing values ​​to obtain cleaned data of agricultural machinery scheduling;

[0019] Step S22: data normalization, specifically, performing unique-hot encoding on the type data in the agricultural machinery scheduling cleaned data, and normalizing the numerical data in the agricultural machinery scheduling cleaned data, to obtain the agricultural machinery scheduling standard data.

[0020] Further, in step S3, the crop yield prediction is used to predict the crop yield of each plot, specifically using a temporal convolution feature enhanced gated recurrent network to perform crop yield prediction and obtain crop yield prediction information;

[0021] The temporal convolutional feature enhanced gated recurrent network comprises a data input layer, a temporal convolutional feature enhanced subnet, a gated recurrent unit subnet and a prediction output layer;

[0022] The temporal convolutional feature enhancement subnet is used to model long-range temporal dependencies and enhance key features;

[0023] The gated recurrent unit subnetwork is used to capture short-term temporal dependencies;

[0024] The crop yield prediction comprises the following steps:

[0025] Step S31: constructing a data input layer, specifically, through the data input layer, performing time series and feature integration processing on the yield prediction related data in the agricultural machinery scheduling specification data, and performing tensor processing to obtain input data;

[0026] Step S32: constructing a temporal convolution feature enhancement subnet, specifically setting three temporal convolution feature enhancement blocks in the temporal convolution feature enhancement subnet, calculating the input data through the temporal convolution feature enhancement subnet, and obtaining the long-term enhanced features of the yield, including the following steps:

[0027] Step S321: construct a temporal convolution layer, specifically, set a batch normalization layer, a LeakyReLU activation function, a random dropout layer, and a dilated convolution layer in the temporal convolution layer, construct the temporal convolution layer, and obtain the yield temporal convolution feature. The calculation formula is:

[0028] ;

[0029] In the formula, H TCN is the yield temporal convolution feature, DilatedConv(·) is the dilated convolution operation, Dropout(·) is the random dropout operation, LeakyReLU(·) is the LeakyReLU activation function, and BN(·) is the batch normalization operation;

[0030] Step S322: constructing a feature enhancement layer, specifically setting a global average pooling layer, a fully connected combination layer and a dynamic soft threshold function in the feature enhancement layer, constructing the feature enhancement layer, and obtaining the production time enhancement feature, including the following steps:

[0031] Step S3221: Calculate the temporal convolutional features of the yield through a global average pooling layer to obtain the global average features of the yield;

[0032] Step S3222: Generate yield feature weights through a fully connected combination layer, where the fully connected combination layer includes a batch normalization layer, a ReLU activation function, a fully connected layer, and a Sigmoid activation function;

[0033] The calculation formula for generating the dynamic weight of yield characteristics through the fully connected combination layer is:

[0034] ;

[0035] Where w is the dynamic weight of the yield feature, Sigmoid(·) is the Sigmoid activation function, FC(·) is the fully connected layer, ReLU(·) is the ReLU activation function, BN(·) is the batch normalization operation, and H GAP is the global average characteristic of yield;

[0036] Step S3223: The production time enhancement feature is obtained by performing dynamic soft threshold processing on the production time convolution feature. The calculation formula is:

[0037] ;

[0038] ;

[0039] Where a is the dynamic soft threshold, is the element-by-element multiplication symbol, H enh It is a yield time-enhancing feature;

[0040] Step S323: Set residual connection, specifically, connect the temporal convolution layer and the feature enhancement layer through residual connection to obtain the long-term enhanced feature of the yield, and the calculation formula is:

[0041] ;

[0042] In the formula, H LT It is a characteristic of long-term enhancement of yield;

[0043] Step S324: constructing a temporal convolutional feature enhancement block, specifically, constructing a temporal convolutional feature enhancement block by constructing a temporal convolutional layer, constructing a feature enhancement layer, and setting a residual connection;

[0044] Step S325: Calculate the input data through three time convolution feature enhancement blocks to obtain the long-term comprehensive enhancement features of the output;

[0045] Step S33: constructing a gated recurrent unit subnet, specifically, setting two gated recurrent units in the gated recurrent unit subnet, calculating the comprehensive enhancement feature of the long time series of the yield through the gated recurrent unit subnet, and obtaining the enhancement feature of the long and short time series of the yield;

[0046] Step S34: constructing a prediction output layer, specifically setting a fully connected layer and a ReLU activation function in the prediction output layer, calculating the long-short sequence enhancement features of the output through the prediction output layer, and obtaining the model prediction result;

[0047] Step S35: constructing a crop yield prediction model, specifically, performing model training by constructing the data input layer, constructing the temporal convolutional feature enhancement subnet, constructing the gated recurrent unit subnet and constructing the prediction output layer to obtain a crop yield prediction model;

[0048] Step S36: predicting the crop yield of each plot, specifically predicting the crop yield of each plot by using a crop yield prediction model to obtain crop yield prediction information, wherein the crop yield prediction information includes a predicted crop yield value of each plot.

[0049] Further, in step S4, the agricultural machinery intelligent scheduling is used to generate an agricultural machinery scheduling plan according to the crop yield prediction information, specifically using a discrete particle swarm optimization algorithm to perform agricultural machinery intelligent scheduling to obtain an optimal agricultural machinery scheduling plan, including the following steps:

[0050] Step S41: constructing an agricultural machinery scheduling model, specifically modeling the agricultural machinery scheduling as a constrained optimization problem, including the following steps:

[0051] Step S411: Setting decision variables, specifically setting agricultural machinery allocation decision variables x ij, used to indicate whether the jth agricultural machine is operating on the i-th plot. If the jth agricultural machine is operating on the i-th plot, then x ij =1, otherwise, x ij =0;

[0052] Step S412: construct the agricultural machinery scheduling objective function, the calculation formula is:

[0053] ;

[0054] Where E is the objective function value of agricultural machinery scheduling, specifically the operation efficiency, i is the plot index, N is the number of plots, and Y is i is the predicted crop yield of the ith plot, A i is the completion degree of the i-th plot, j is the index of agricultural machinery, M is the number of agricultural machinery, C j is the use cost of the j-th agricultural machine, specifically the fuel consumption of the j-th agricultural machine;

[0055] Step S413: setting constraints, specifically setting operation allocation constraints, agricultural machinery operation capacity constraints and operation time window constraints;

[0056] The operation allocation constraint condition is that each plot of land can only be operated by one agricultural machine;

[0057] The agricultural machinery operation capacity constraint condition specifically includes that the operation time of each agricultural machinery cannot exceed the maximum operation time of each agricultural machinery;

[0058] The operation time window constraint condition is specifically to set the operation time of each plot to be completed within the specified time window;

[0059] Step S42: constructing a discrete particle swarm optimization algorithm, including the following steps:

[0060] Step S421: Initialize the particle swarm, specifically setting the particle swarm size, and randomly initializing the particle position in the solution space, and setting the current position of the particle to the initial value of the historical optimal solution;

[0061] The calculation formula of the particle position is:

[0062] ;

[0063] In the formula, pos k is the kth particle position, used to represent the agricultural machinery scheduling plan, k is the particle index, pos k1 is the position component of the kth particle in the first dimension, pos k2 is the position component of the kth particle in the second dimension, pos kdis the position component of the kth particle in the dth dimension, with a value range of {1,2,…,M}, which is used to represent the allocation relationship between agricultural machinery and the dth plot. d is the solution space dimension index, with a value range of {1,2,…,N};

[0064] Step S422: designing a fitness function, specifically, designing a fitness function according to the agricultural machinery scheduling objective function and constraint conditions, calculating the fitness value of each particle position through the fitness function, and taking the particle position with the best fitness value as the initial value of the global optimal solution;

[0065] The calculation formula of the fitness function is:

[0066] ;

[0067] Where f(·) is the fitness function, is the objective function weight, P con is a penalty term used to handle constraints, is the penalty term weight, P ass is the penalty for violating the job assignment constraint, P cap is the penalty for violating the constraints of agricultural machinery operation capacity, P time is the penalty for violating the job time window constraint;

[0068] Step S423: Particle position update, specifically using the probability transfer mechanism to update the particle position, the calculation formula is:

[0069] ;

[0070] In the formula, pos kd (t+1) is the position component of the kth particle in the dth dimension at the t+1th iteration, t is the iteration index, rand(·) is a random selection function used to randomly select a value from the set {1,2,…,M}, r is a random number uniformly distributed in the interval [0,1], and p r is the random search probability in the range [0,1], pb kd (t) is the position component of the optimal solution of the kth particle in the dth dimension at the tth iteration, p pb is the learning probability of the historical optimal solution, gb d (t) is the position component of the global optimal solution in the dth dimension at the tth iteration, p gb is the learning probability of the global optimal solution;

[0071] Step S424: Update the historical optimal solution and the global optimal solution. The calculation formula is:

[0072] ;

[0073] ;

[0074] In the formula, pb k (t+1) is the historical optimal solution of the kth particle at the t+1th iteration, pos k (t) is the position of the kth particle at the tth iteration, pb k (t) is the historical optimal solution of the kth particle at the tth iteration, gb(t+1) is the global optimal solution at the t+1th iteration, and gb(t) is the global optimal solution at the tth iteration;

[0075] Step S425: particle swarm iterative update, specifically, by repeatedly executing steps S423 to S424, performing particle swarm iterative update until the maximum number of iterations is reached and a global optimal solution is obtained;

[0076] Step S43: Generate an agricultural machinery scheduling plan, specifically use a discrete particle swarm optimization algorithm to perform intelligent scheduling of agricultural machinery and obtain the optimal agricultural machinery scheduling plan.

[0077] Further, in step S5, the comprehensive agricultural machinery scheduling is specifically to combine the crop yield prediction information and the optimal agricultural machinery scheduling plan, perform comprehensive agricultural machinery intelligent scheduling based on yield prediction, obtain comprehensive scheduling plan reference data, and execute agricultural machinery scheduling according to the comprehensive scheduling plan reference data;

[0078] The comprehensive scheduling scheme reference data includes crop yield prediction information and optimal agricultural machinery scheduling scheme.

[0079] The agricultural machinery intelligent scheduling system based on yield prediction provided by the present invention comprises: a data collection module, a data preprocessing module, a crop yield prediction module, an agricultural machinery intelligent scheduling module and a comprehensive agricultural machinery scheduling module;

[0080] The data collection module is used for data collection, obtains the original data of agricultural machinery scheduling through data collection, and sends the original data of agricultural machinery scheduling to the data preprocessing module;

[0081] The data preprocessing module is used for data preprocessing, and obtains agricultural machinery scheduling specification data through data preprocessing, and sends the agricultural machinery scheduling specification data to the crop yield prediction module and the agricultural machinery intelligent scheduling module;

[0082] The crop yield prediction module is used for crop yield prediction, obtains crop yield prediction information through crop yield prediction, and sends the crop yield prediction information to the agricultural machinery intelligent scheduling module and the comprehensive agricultural machinery scheduling module;

[0083] The agricultural machinery intelligent scheduling module is used for the intelligent scheduling of agricultural machinery, obtains the optimal agricultural machinery scheduling plan through the intelligent scheduling of agricultural machinery, and sends the optimal agricultural machinery scheduling plan to the comprehensive agricultural machinery scheduling module;

[0084] The comprehensive agricultural machinery scheduling module is used for comprehensive agricultural machinery scheduling, and combines the crop yield prediction information and the optimal agricultural machinery scheduling plan to perform comprehensive agricultural machinery intelligent scheduling based on yield prediction.

[0085] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0086] (1) In view of the technical problems in the existing crop yield prediction process, the traditional crop yield prediction usually relies on simple statistical models, which makes it difficult to fully consider the complex nonlinear relationship between multiple factors such as meteorology, images, and crop growth. In addition, the crop growth process has a strong time series dependency, and the traditional crop yield prediction has a poor time series feature modeling capability. This scheme creatively uses the temporal convolution feature enhanced gated recurrent network for crop yield prediction, which has better long-term and short-term time series dependency modeling capabilities, and can adaptively enhance key features and suppress noise interference, thereby improving the stability and accuracy of crop yield prediction, which helps to improve the reliability of subsequent intelligent scheduling of agricultural machinery;

[0087] (2) In view of the technical problems in the existing intelligent scheduling of agricultural machinery, the traditional scheduling methods often rely on manual arrangements, are easily affected by human factors, and have unreasonable operation arrangements. In addition, they lack global optimization and are difficult to adapt to the needs of large-scale agricultural production. This solution creatively uses discrete particle swarm optimization algorithm for intelligent scheduling of agricultural machinery. It has strong global search capabilities and can adapt to different agricultural production environments and scheduling needs. It has significant advantages in large-scale agricultural production, effectively avoids the high error problem of manual scheduling, and improves the efficiency of agricultural machinery scheduling.

[0088] (3) In view of the technical problem that the traditional scheduling method does not take into account the difference in crop yields in the existing intelligent scheduling of agricultural machinery, resulting in uneven allocation of agricultural machinery resources, excessive operation of agricultural machinery on the land, and unnecessary waste of resources, this scheme adopts a comprehensive intelligent scheduling of agricultural machinery based on yield prediction, fully considers the difference in crop yields, reasonably arranges agricultural machinery operations, and effectively avoids waste of resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] Figure 1 A schematic diagram of the flow of the intelligent dispatching method for agricultural machinery based on yield prediction provided by the present invention;

[0090] Figure 2 A schematic diagram of an intelligent agricultural machinery scheduling system based on yield prediction provided by the present invention;

[0091] Figure 3is a schematic diagram of the process of step S3;

[0092] Figure 4 It is a schematic diagram of the process of step S4.

[0093] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0094] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0095] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0096] Example 1, see Figure 1 The invention provides an intelligent dispatching method for agricultural machinery based on yield prediction, the method comprising the following steps:

[0097] Step S1: data collection;

[0098] Step S2: data preprocessing;

[0099] Step S3: crop yield prediction;

[0100] Step S4: Intelligent dispatching of agricultural machinery;

[0101] Step S5: Comprehensive agricultural machinery scheduling.

[0102] Example 2, see Figure 1 , this embodiment is based on the above embodiment. In step S1, the data collection is used to collect the original data required for intelligent scheduling of agricultural machinery, specifically, by collecting yield prediction related data and agricultural machinery related data to obtain the original data for agricultural machinery scheduling;

[0103] The yield prediction related data includes plot information, meteorological data and crop growth data;

[0104] The agricultural machinery related data include agricultural machinery number, number of agricultural machinery, type of agricultural machinery, fuel consumption and maximum usage time;

[0105] The meteorological data include precipitation, temperature, humidity and wind speed;

[0106] The land plot information, including land plot area, location, soil type and crop type;

[0107] The crop growth data include vegetation index, crop height and crop coverage rate.

[0108] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, the data preprocessing includes the following steps:

[0109] Step S21: data cleaning, specifically removing duplicate values ​​in the original data of agricultural machinery scheduling and interpolating missing values ​​to obtain cleaned data of agricultural machinery scheduling;

[0110] Step S22: data normalization, specifically, performing unique-hot encoding on the type data in the agricultural machinery scheduling cleaned data, and normalizing the numerical data in the agricultural machinery scheduling cleaned data, to obtain the agricultural machinery scheduling standard data.

[0111] Example 4, see Figure 1 and Figure 3 , this embodiment is based on the above embodiment. In step S3, the crop yield prediction is used to predict the crop yield of each plot, specifically using a temporal convolution feature enhanced gated recurrent network to perform crop yield prediction and obtain crop yield prediction information;

[0112] The temporal convolutional feature enhanced gated recurrent network comprises a data input layer, a temporal convolutional feature enhanced subnet, a gated recurrent unit subnet and a prediction output layer;

[0113] The temporal convolutional feature enhancement subnet is used to model long-range temporal dependencies and enhance key features;

[0114] The gated recurrent unit subnetwork is used to capture short-term temporal dependencies;

[0115] The crop yield prediction comprises the following steps:

[0116] Step S31: constructing a data input layer, specifically, through the data input layer, performing time series and feature integration processing on the yield prediction related data in the agricultural machinery scheduling specification data, and performing tensor processing to obtain input data;

[0117] Step S32: constructing a temporal convolution feature enhancement subnet, specifically setting three temporal convolution feature enhancement blocks in the temporal convolution feature enhancement subnet, calculating the input data through the temporal convolution feature enhancement subnet, and obtaining the long-term enhanced features of the yield, including the following steps:

[0118] Step S321: construct a temporal convolution layer, specifically, set a batch normalization layer, a LeakyReLU activation function, a random dropout layer, and a dilated convolution layer in the temporal convolution layer, construct the temporal convolution layer, and obtain the yield temporal convolution feature. The calculation formula is:

[0119] ;

[0120] In the formula, H TCN is the yield temporal convolution feature, DilatedConv(·) is the dilated convolution operation, Dropout(·) is the random dropout operation, LeakyReLU(·) is the LeakyReLU activation function, and BN(·) is the batch normalization operation;

[0121] Step S322: constructing a feature enhancement layer, specifically setting a global average pooling layer, a fully connected combination layer and a dynamic soft threshold function in the feature enhancement layer, constructing the feature enhancement layer, and obtaining the production time enhancement feature, including the following steps:

[0122] Step S3221: Calculate the temporal convolutional features of the yield through a global average pooling layer to obtain the global average features of the yield;

[0123] Step S3222: Generate yield feature weights through a fully connected combination layer, where the fully connected combination layer includes a batch normalization layer, a ReLU activation function, a fully connected layer, and a Sigmoid activation function;

[0124] The calculation formula for generating the dynamic weight of yield characteristics through the fully connected combination layer is:

[0125] ;

[0126] Where w is the dynamic weight of the yield feature, Sigmoid(·) is the Sigmoid activation function, FC(·) is the fully connected layer, ReLU(·) is the ReLU activation function, BN(·) is the batch normalization operation, and H GAP is the global average characteristic of yield;

[0127] Step S3223: The production time enhancement feature is obtained by performing dynamic soft threshold processing on the production time convolution feature. The calculation formula is:

[0128] ;

[0129] ;

[0130] Where a is the dynamic soft threshold, is the element-by-element multiplication symbol, H enh It is a yield time-enhancing feature;

[0131] Step S323: Set residual connection, specifically, connect the temporal convolution layer and the feature enhancement layer through residual connection to obtain the long-term enhanced feature of the yield, and the calculation formula is:

[0132] ;

[0133] In the formula, H LT It is a characteristic of long-term enhancement of yield;

[0134] Step S324: constructing a temporal convolutional feature enhancement block, specifically, constructing a temporal convolutional feature enhancement block by constructing a temporal convolutional layer, constructing a feature enhancement layer, and setting a residual connection;

[0135] Step S325: Calculate the input data through three time convolution feature enhancement blocks to obtain the long-term comprehensive enhancement features of the output;

[0136] Step S33: constructing a gated recurrent unit subnet, specifically, setting two gated recurrent units in the gated recurrent unit subnet, calculating the comprehensive enhancement feature of the long time series of the yield through the gated recurrent unit subnet, and obtaining the enhancement feature of the long and short time series of the yield;

[0137] Step S34: constructing a prediction output layer, specifically setting a fully connected layer and a ReLU activation function in the prediction output layer, calculating the long-short sequence enhancement features of the output through the prediction output layer, and obtaining the model prediction result;

[0138] Step S35: constructing a crop yield prediction model, specifically, performing model training by constructing the data input layer, constructing the temporal convolutional feature enhancement subnet, constructing the gated recurrent unit subnet and constructing the prediction output layer to obtain a crop yield prediction model;

[0139] Step S36: predicting the crop yield of each plot, specifically predicting the crop yield of each plot by using a crop yield prediction model to obtain crop yield prediction information, wherein the crop yield prediction information includes a crop yield prediction value of each plot;

[0140] By performing the above operations, in view of the technical problems in the existing crop yield prediction process, the traditional crop yield prediction usually relies on simple statistical models, which makes it difficult to fully consider the complex nonlinear relationship between multiple factors such as meteorology, images, crop growth, etc., and the crop growth process has strong time series dependence, and the traditional crop yield prediction has poor time series feature modeling ability. This scheme creatively uses the time convolution feature enhanced gated recurrent network for crop yield prediction, which has better long-term and short-term time series dependency modeling capabilities, and can adaptively enhance key features and suppress noise interference, thereby improving the stability and accuracy of crop yield prediction, and helping to improve the reliability of subsequent intelligent scheduling of agricultural machinery.

[0141] Example 5, see Figure 1 and Figure 4 This embodiment is based on the above embodiment. In step S4, the agricultural machinery intelligent scheduling is used to generate an agricultural machinery scheduling plan according to the crop yield prediction information. Specifically, the agricultural machinery intelligent scheduling is performed using a discrete particle swarm optimization algorithm to obtain an optimal agricultural machinery scheduling plan, including the following steps:

[0142] Step S41: constructing an agricultural machinery scheduling model, specifically modeling the agricultural machinery scheduling as a constrained optimization problem, including the following steps:

[0143] Step S411: Setting decision variables, specifically setting agricultural machinery allocation decision variables x ij , used to indicate whether the jth agricultural machine is operating on the i-th plot. If the jth agricultural machine is operating on the i-th plot, then x ij =1, otherwise, x ij =0;

[0144] Step S412: construct the agricultural machinery scheduling objective function, the calculation formula is:

[0145] ;

[0146] Where E is the objective function value of agricultural machinery scheduling, specifically the operation efficiency, i is the plot index, N is the number of plots, and Y is i is the predicted crop yield of the ith plot, A i is the completion degree of the i-th plot, j is the index of agricultural machinery, M is the number of agricultural machinery, C j is the use cost of the j-th agricultural machine, specifically the fuel consumption of the j-th agricultural machine;

[0147] Step S413: setting constraints, specifically setting operation allocation constraints, agricultural machinery operation capacity constraints and operation time window constraints;

[0148] The operation allocation constraint condition is that each plot of land can only be operated by one agricultural machine;

[0149] The agricultural machinery operation capacity constraint condition specifically includes that the operation time of each agricultural machinery cannot exceed the maximum operation time of each agricultural machinery;

[0150] The operation time window constraint condition is specifically to set the operation time of each plot to be completed within the specified time window;

[0151] Step S42: constructing a discrete particle swarm optimization algorithm, including the following steps:

[0152] Step S421: Initialize the particle swarm, specifically setting the particle swarm size, and randomly initializing the particle position in the solution space, and setting the current position of the particle to the initial value of the historical optimal solution;

[0153] The calculation formula of the particle position is:

[0154] ;

[0155] In the formula, pos k is the kth particle position, used to represent the agricultural machinery scheduling plan, k is the particle index, pos k1 is the position component of the kth particle in the first dimension, pos k2 is the position component of the kth particle in the second dimension, pos kd is the position component of the kth particle in the dth dimension, with a value range of {1,2,…,M}, which is used to represent the allocation relationship between agricultural machinery and the dth plot. d is the solution space dimension index, with a value range of {1,2,…,N};

[0156] Step S422: designing a fitness function, specifically, designing a fitness function according to the agricultural machinery scheduling objective function and constraint conditions, calculating the fitness value of each particle position through the fitness function, and taking the particle position with the best fitness value as the initial value of the global optimal solution;

[0157] The calculation formula of the fitness function is:

[0158] ;

[0159] Where f(·) is the fitness function, is the objective function weight, P con is a penalty term used to handle constraints, is the penalty term weight, P ass is the penalty for violating the job assignment constraint, P cap is the penalty for violating the constraints of agricultural machinery operation capacity, P time is the penalty for violating the job time window constraint;

[0160] Step S423: Particle position update, specifically using the probability transfer mechanism to update the particle position, the calculation formula is:

[0161] ;

[0162] In the formula, pos kd (t+1) is the position component of the kth particle in the dth dimension at the t+1th iteration, t is the iteration index, rand(·) is a random selection function used to randomly select a value from the set {1,2,…,M}, r is a random number uniformly distributed in the interval [0,1], and p r is the random search probability in the range [0,1], pb kd (t) is the position component of the optimal solution of the kth particle in the dth dimension at the tth iteration, p pb is the learning probability of the historical optimal solution, gb d (t) is the position component of the global optimal solution in the dth dimension at the tth iteration, p gb is the learning probability of the global optimal solution;

[0163] The calculation formulas for the historical optimal solution learning probability and the global optimal solution learning probability are:

[0164] ;

[0165] Step S424: Update the historical optimal solution and the global optimal solution. The calculation formula is:

[0166] ;

[0167] ;

[0168] In the formula, pb k (t+1) is the historical optimal solution of the kth particle at the t+1th iteration, pos k (t) is the position of the kth particle at the tth iteration, pb k (t) is the historical optimal solution of the kth particle at the tth iteration, gb(t+1) is the global optimal solution at the t+1th iteration, and gb(t) is the global optimal solution at the tth iteration;

[0169] Step S425: particle swarm iterative update, specifically, by repeatedly executing steps S423 to S424, performing particle swarm iterative update until the maximum number of iterations is reached and a global optimal solution is obtained;

[0170] Step S43: generating an agricultural machinery scheduling plan, specifically using a discrete particle swarm optimization algorithm to perform intelligent scheduling of agricultural machinery and obtain an optimal agricultural machinery scheduling plan;

[0171] By performing the above operations, in view of the technical problems in the existing intelligent scheduling of agricultural machinery, the traditional scheduling methods often rely on manual arrangements, are easily affected by human factors, have unreasonable job arrangements, lack of global optimization, and are difficult to adapt to the needs of large-scale agricultural production. This solution creatively adopts the discrete particle swarm optimization algorithm for intelligent scheduling of agricultural machinery. It has strong global search capabilities, can adapt to different agricultural production environments and scheduling needs, has significant advantages in large-scale agricultural production, effectively avoids the high error problem of manual scheduling, and improves the efficiency of agricultural machinery scheduling.

[0172] Example 6, see Figure 1 This embodiment is based on the above embodiment. In step S5, the comprehensive agricultural machinery scheduling is specifically to combine the crop yield prediction information and the optimal agricultural machinery scheduling plan to perform comprehensive agricultural machinery intelligent scheduling based on yield prediction, obtain comprehensive scheduling plan reference data, and execute agricultural machinery scheduling according to the comprehensive scheduling plan reference data;

[0173] The comprehensive scheduling scheme reference data includes crop yield forecast information and optimal agricultural machinery scheduling scheme;

[0174] By executing the above operations, in order to address the technical problem that in the existing intelligent scheduling of agricultural machinery, the traditional scheduling method does not take into account the difference in crop yields, resulting in uneven allocation of agricultural machinery resources, excessive operation of agricultural machinery on the land, and unnecessary waste of resources, this solution adopts comprehensive intelligent scheduling of agricultural machinery based on yield prediction, fully considers the difference in crop yields, reasonably arranges agricultural machinery operations, and effectively avoids waste of resources.

[0175] Embodiment 7, see Figure 2 , this embodiment is based on the above embodiment, and the agricultural machinery intelligent scheduling system based on yield prediction provided by the present invention includes: a data collection module, a data preprocessing module, a crop yield prediction module, an agricultural machinery intelligent scheduling module and a comprehensive agricultural machinery scheduling module;

[0176] The data collection module is used for data collection, obtains the original data of agricultural machinery scheduling through data collection, and sends the original data of agricultural machinery scheduling to the data preprocessing module;

[0177] The data preprocessing module is used for data preprocessing, and obtains agricultural machinery scheduling specification data through data preprocessing, and sends the agricultural machinery scheduling specification data to the crop yield prediction module and the agricultural machinery intelligent scheduling module;

[0178] The crop yield prediction module is used for crop yield prediction, obtains crop yield prediction information through crop yield prediction, and sends the crop yield prediction information to the agricultural machinery intelligent scheduling module and the comprehensive agricultural machinery scheduling module;

[0179] The agricultural machinery intelligent scheduling module is used for the intelligent scheduling of agricultural machinery, obtains the optimal agricultural machinery scheduling plan through the intelligent scheduling of agricultural machinery, and sends the optimal agricultural machinery scheduling plan to the comprehensive agricultural machinery scheduling module;

[0180] The comprehensive agricultural machinery scheduling module is used for comprehensive agricultural machinery scheduling, and combines the crop yield prediction information and the optimal agricultural machinery scheduling plan to perform comprehensive agricultural machinery intelligent scheduling based on yield prediction.

[0181] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0182] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0183] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. An intelligent dispatching method for agricultural machinery based on yield prediction, characterized by: The method comprises the following steps: Step S1: data collection; Step S2: data preprocessing; Step S3: Crop yield prediction, which is used to predict the crop yield of each plot, specifically using a temporal convolution feature enhanced gated recurrent network to perform crop yield prediction and obtain crop yield prediction information; The temporal convolutional feature enhanced gated recurrent network comprises a data input layer, a temporal convolutional feature enhanced subnet, a gated recurrent unit subnet and a prediction output layer; The temporal convolutional feature enhancement subnet is used to model long-range temporal dependencies and enhance key features; The gated recurrent unit subnetwork is used to capture short-term temporal dependencies; Step S4: Intelligent dispatching of agricultural machinery, used to generate an agricultural machinery dispatching plan according to crop yield prediction information, including the following steps: Step S41: constructing an agricultural machinery dispatching model; Step S42: constructing a discrete particle swarm optimization algorithm; Step S43: generating an agricultural machinery dispatching plan; Step S5: Comprehensive agricultural machinery scheduling.

2. The method for intelligent dispatching of agricultural machinery based on yield prediction according to claim 1 is characterized in that: In step S3, the crop yield prediction includes the following steps: Step S31: constructing a data input layer, specifically, through the data input layer, performing time series and feature integration processing on the yield prediction related data in the agricultural machinery scheduling specification data, and performing tensor processing to obtain input data; Step S32: constructing a temporal convolution feature enhancement subnet, specifically setting three temporal convolution feature enhancement blocks in the temporal convolution feature enhancement subnet, calculating the input data through the temporal convolution feature enhancement subnet, and obtaining the long-term enhanced features of the yield, including the following steps: Step S321: construct a temporal convolution layer, specifically, set a batch normalization layer, a LeakyReLU activation function, a random dropout layer, and a dilated convolution layer in the temporal convolution layer, construct the temporal convolution layer, and obtain the yield temporal convolution feature. The calculation formula is: ； In the formula, H TCN is the yield temporal convolution feature, DilatedConv(·) is the dilated convolution operation, Dropout(·) is the random dropout operation, LeakyReLU(·) is the LeakyReLU activation function, and BN(·) is the batch normalization operation; Step S322: constructing a feature enhancement layer, specifically setting a global average pooling layer, a fully connected combination layer and a dynamic soft threshold function in the feature enhancement layer, constructing the feature enhancement layer, and obtaining the production time enhancement feature, including the following steps: Step S3221: Calculate the temporal convolutional features of the yield through a global average pooling layer to obtain the global average features of the yield; Step S3222: Generate yield feature weights through a fully connected combination layer, where the fully connected combination layer includes a batch normalization layer, a ReLU activation function, a fully connected layer, and a Sigmoid activation function; The calculation formula for generating the dynamic weight of yield characteristics through the fully connected combination layer is: ； Where w is the dynamic weight of the yield feature, Sigmoid(·) is the Sigmoid activation function, FC(·) is the fully connected layer, ReLU(·) is the ReLU activation function, BN(·) is the batch normalization operation, and H GAP is the global average characteristic of yield; Step S3223: The production time enhancement feature is obtained by performing dynamic soft threshold processing on the production time convolution feature. The calculation formula is: ； ； Where a is the dynamic soft threshold, is the element-by-element multiplication symbol, H enh It is a yield time-enhancing feature; Step S323: Set residual connection, specifically, connect the temporal convolution layer and the feature enhancement layer through residual connection to obtain the long-term enhanced feature of the yield, and the calculation formula is: ； In the formula, H LT It is a characteristic of long-term enhancement of yield; Step S324: constructing a temporal convolutional feature enhancement block, specifically, constructing a temporal convolutional feature enhancement block by constructing a temporal convolutional layer, constructing a feature enhancement layer, and setting a residual connection; Step S325: Calculate the input data through three time convolution feature enhancement blocks to obtain the long-term comprehensive enhancement features of the output; Step S33: constructing a gated recurrent unit subnet, specifically, setting two gated recurrent units in the gated recurrent unit subnet, calculating the comprehensive enhancement feature of the long time series of the yield through the gated recurrent unit subnet, and obtaining the enhancement feature of the long and short time series of the yield; Step S34: constructing a prediction output layer, specifically setting a fully connected layer and a ReLU activation function in the prediction output layer, calculating the long-short sequence enhancement features of the output through the prediction output layer, and obtaining the model prediction result; Step S35: constructing a crop yield prediction model, specifically, performing model training by constructing the data input layer, constructing the temporal convolutional feature enhancement subnet, constructing the gated recurrent unit subnet and constructing the prediction output layer to obtain a crop yield prediction model; Step S36: predicting the crop yield of each plot, specifically predicting the crop yield of each plot through a crop yield prediction model to obtain crop yield prediction information.

3. The method for intelligent dispatching of agricultural machinery based on yield prediction according to claim 2 is characterized in that: In step S41, the agricultural machinery scheduling model is constructed, specifically, the agricultural machinery scheduling is modeled as a constrained optimization problem, including the following steps: Step S411: Setting decision variables, specifically setting agricultural machinery allocation decision variables x ij , used to indicate whether the jth agricultural machine is operating on the i-th plot. If the jth agricultural machine is operating on the i-th plot, then x ij =1, otherwise, x ij =0; Step S412: construct the agricultural machinery scheduling objective function, the calculation formula is: ； Where E is the objective function value of agricultural machinery scheduling, specifically the operation efficiency, i is the plot index, N is the number of plots, and Y is i is the predicted crop yield of the ith plot, A i is the completion degree of the i-th plot, j is the index of agricultural machinery, M is the number of agricultural machinery, C j is the use cost of the jth agricultural machine; Step S413: setting constraints, specifically setting operation allocation constraints, agricultural machinery operation capacity constraints and operation time window constraints; The operation allocation constraint condition is that each plot of land can only be operated by one agricultural machine; The agricultural machinery operation capacity constraint condition specifically includes that the operation time of each agricultural machinery cannot exceed the maximum operation time of each agricultural machinery; The operation time window constraint condition is specifically to set the operation time of each plot to be completed within the specified time window; In step S42, the discrete particle swarm optimization algorithm is constructed, including the following steps: Step S421: Initialize the particle swarm, specifically setting the particle swarm size, and randomly initializing the particle position in the solution space, and setting the current position of the particle to the initial value of the historical optimal solution; The calculation formula of the particle position is: ； In the formula, pos k is the kth particle position, used to represent the agricultural machinery scheduling plan, k is the particle index, pos k1 is the position component of the kth particle in the first dimension, pos k2 is the position component of the kth particle in the second dimension, pos kd is the position component of the kth particle in the dth dimension, with a value range of {1,2,…,M}, which is used to represent the allocation relationship between agricultural machinery and the dth plot. d is the solution space dimension index, with a value range of {1,2,…,N}; Step S422: designing a fitness function, specifically, designing a fitness function according to the agricultural machinery scheduling objective function and constraint conditions, calculating the fitness value of each particle position through the fitness function, and taking the particle position with the best fitness value as the initial value of the global optimal solution; The calculation formula of the fitness function is: ； Where f(·) is the fitness function, is the objective function weight, P con is a penalty term used to handle constraints, is the penalty term weight, P ass is the penalty for violating the job assignment constraint, P cap is the penalty for violating the constraints of agricultural machinery operation capacity, P time is the penalty for violating the job time window constraint; Step S423: Particle position update, specifically using the probability transfer mechanism to update the particle position, the calculation formula is: ； In the formula, pos kd (t+1) is the position component of the kth particle in the dth dimension at the t+1th iteration, t is the iteration index, rand(·) is a random selection function used to randomly select a value from the set {1,2,…,M}, r is a random number uniformly distributed in the interval [0,1], and p r is the random search probability in the range [0,1], pb kd (t) is the position component of the optimal solution of the kth particle in the dth dimension at the tth iteration, p pb is the learning probability of the historical optimal solution, gb d (t) is the position component of the global optimal solution in the dth dimension at the tth iteration, p gb is the learning probability of the global optimal solution; Step S424: Update the historical optimal solution and the global optimal solution. The calculation formula is: ； ； In the formula, pb k (t+1) is the historical optimal solution of the kth particle at the t+1th iteration, pos k (t) is the position of the kth particle at the tth iteration, pb k (t) is the historical optimal solution of the kth particle at the tth iteration, gb(t+1) is the global optimal solution at the t+1th iteration, and gb(t) is the global optimal solution at the tth iteration; Step S425: particle swarm iterative update, specifically, by repeatedly executing steps S423 to S424, performing particle swarm iterative update until the maximum number of iterations is reached and a global optimal solution is obtained; In step S43, the agricultural machinery scheduling plan is generated by using a discrete particle swarm optimization algorithm to perform intelligent scheduling of agricultural machinery to obtain the optimal agricultural machinery scheduling plan.

4. The method for intelligent dispatching of agricultural machinery based on yield prediction according to claim 3 is characterized in that: In step S5, the comprehensive agricultural machinery scheduling is specifically to combine the crop yield prediction information and the optimal agricultural machinery scheduling plan to perform comprehensive agricultural machinery intelligent scheduling based on yield prediction, obtain comprehensive scheduling plan reference data, and execute agricultural machinery scheduling based on the comprehensive scheduling plan reference data.

5. The method for intelligent dispatching of agricultural machinery based on yield prediction according to claim 4 is characterized in that: In step S2, the data preprocessing includes the following steps: Step S21: data cleaning, specifically removing duplicate values ​​in the original data of agricultural machinery scheduling and interpolating missing values ​​to obtain cleaned data of agricultural machinery scheduling; Step S22: data normalization, specifically, performing unique-hot encoding on the type data in the agricultural machinery scheduling cleaned data, and normalizing the numerical data in the agricultural machinery scheduling cleaned data, to obtain the agricultural machinery scheduling standard data.

6. The method for intelligent dispatching of agricultural machinery based on yield prediction according to claim 5 is characterized in that: In step S1, the data collection is used to collect the original data required for the intelligent scheduling of agricultural machinery, specifically, by collecting yield prediction related data and agricultural machinery related data to obtain the original data for agricultural machinery scheduling.

7. An intelligent dispatching system for agricultural machinery based on yield prediction, used to implement the intelligent dispatching method for agricultural machinery based on yield prediction as claimed in any one of claims 1 to 6, characterized in that: It includes data collection module, data preprocessing module, crop yield prediction module, agricultural machinery intelligent scheduling module and comprehensive agricultural machinery scheduling module.

8. The intelligent agricultural machinery scheduling system based on yield prediction according to claim 7 is characterized in that: The data collection module is used for data collection, obtains the original data of agricultural machinery scheduling through data collection, and sends the original data of agricultural machinery scheduling to the data preprocessing module; The data preprocessing module is used for data preprocessing, and obtains agricultural machinery scheduling specification data through data preprocessing, and sends the agricultural machinery scheduling specification data to the crop yield prediction module and the agricultural machinery intelligent scheduling module; The crop yield prediction module is used for crop yield prediction, obtains crop yield prediction information through crop yield prediction, and sends the crop yield prediction information to the agricultural machinery intelligent scheduling module and the comprehensive agricultural machinery scheduling module; The agricultural machinery intelligent scheduling module is used for the intelligent scheduling of agricultural machinery, obtains the optimal agricultural machinery scheduling plan through the intelligent scheduling of agricultural machinery, and sends the optimal agricultural machinery scheduling plan to the comprehensive agricultural machinery scheduling module; The comprehensive agricultural machinery scheduling module is used for comprehensive agricultural machinery scheduling, and combines the crop yield prediction information and the optimal agricultural machinery scheduling plan to perform comprehensive agricultural machinery intelligent scheduling based on yield prediction.

Images