Method for intelligent planning of crop planting, electronic device, and storage medium

Abstract

The present application relates to the technical field of agricultural planting, and provides a method for intelligent planning of crop planting, an electronic device, and a storage medium, said method comprising: determining a planned total planting area for various crops and additional subsidy values for subsidy crops among the various crops; abstracting a multi-farmer cooperative planting planning problem into a Markov decision process, and constructing a Markov decision process model; and, by using the planned total planting area for various crops as a constraint and using as the minimization of profit differences among farmers and maximization of crop rotation efficiency as a goal of optimization, finding a solution for the Markov decision process model, and obtaining a crop planting plan corresponding to each farmer, the profit differences among the farmers being determined on the basis of the additional subsidy values for the subsidy crops. According to the present application, farmer-level planning of crop planting is achieved while governmental requirements for agricultural production are met, and reference opinions for planting planning are provided for farmers in the implementation of agricultural planting policies.

Description

Crop intelligent planting planning method, electronic device and storage medium

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to Chinese patent application number 202410322288.8, filed on March 20, 2024, entitled “Intelligent Crop Planting Planning Method, Electronic Device and Storage Medium,” which is incorporated herein by reference in its entirety. Technical Field

[0003] The present application relates to the field of agricultural planting technology, and in particular to a method, electronic device and storage medium for intelligent crop planting planning. Background Art

[0004] The purpose of crop planting planning is to solve the problem of maximizing total revenue based on the goals and constraints in agricultural planting.

[0005] In the past, crop planting plans were generally based on the farmers' own perspectives, aiming to maximize their relevant interests. However, this often resulted in the overall planting plan failing to meet the government's agricultural production needs (such as the total crop planting area, the minimum crop planting area, and the maximum crop planting area).

[0006] Therefore, it is necessary to consider how to carry out crop planting planning at the farmer level while meeting the government's agricultural production needs.

[0007] Summary of the Invention

[0008] The present application provides an intelligent crop planting planning method, electronic device and storage medium to address the defect in the existing technology that the crop planting planning based on farmers themselves cannot meet the agricultural production needs of the government, and to realize farmer-level crop planting planning while meeting the agricultural production needs of the government.

[0009] This application provides a method for intelligent crop planting planning, including:

[0010] Determine the total planned planting area for various crops and the additional subsidy value for subsidized crops among the various crops;

[0011] The problem of multi-farmer cooperative planting planning is abstracted into a Markov decision process, and a Markov decision process model is constructed.

[0012] The Markov decision process model is solved using the planned total planting area of ​​various crops as a constraint, minimizing profit differences between farmers and maximizing crop rotation efficiency as optimization objectives, and obtaining a corresponding crop planting plan for each farmer. The profit differences between farmers are determined based on the additional subsidy value for subsidized crops. In some embodiments, the Markov decision process model includes: a state set, an action set, a reward function, a discount factor, and a state transition function; the discount factor and the state transition function are prior values.

[0013] The Markov decision process model is constructed, comprising:

[0014] Constructing the state set based on the current crop planting states of all plots and the current decision states of all plots;

[0015] Constructing the action set based on the type of crops planted;

[0016] The reward function is constructed based on one or more of different crop rotation methods, the decision completion status of each farmer, and the situation where the crop planting area does not meet the total planting plan area.

[0017] In some embodiments, determining the planned total planting area of ​​various crops and the additional subsidy value of the subsidized crops among the various crops includes:

[0018] Based on the predicted prices of various crops, a multi-objective constraint model is constructed, which includes multiple optimization objectives and multiple constraints; the multiple optimization objectives and the multiple constraints are determined based on the needs of planting structure adjustment, farmers' economic benefits, water and fertilizer resource utilization, and planting subsidy policies;

[0019] The optimal solution of the multi-objective constraint model is solved to obtain the planned planting areas of the various crops and the additional subsidy values ​​of the subsidized crops.

[0020] In some embodiments, the step of solving the multi-objective constraint model for an optimal solution to obtain the planned planting areas of the various crops and the additional subsidy values ​​of the subsidized crops includes:

[0021] Converting multiple optimization objectives in the multi-objective constraint model into a single optimization objective to obtain a single-objective constraint model;

[0022] The optimal solution of the single-objective constraint model is solved to obtain the planned planting areas of the various crops and the additional subsidy values ​​of the subsidized crops.

[0023] In some embodiments, converting multiple optimization objectives in the multi-objective constraint model into a single optimization objective to obtain a single-objective constraint model includes:

[0024] Determining the normalization factors corresponding to each optimization objective in the multi-objective constraint model, and the weight coefficients corresponding to each optimization objective;

[0025] Based on the normalization factors corresponding to the optimization objectives, each optimization objective is normalized to obtain each normalized objective;

[0026] Based on the weight coefficients corresponding to each optimization objective and each normalization objective, a single optimization objective is obtained;

[0027] Based on the single optimization objective and the multiple constraint conditions, the single objective constraint model is obtained.

[0028] In some embodiments, before constructing a multi-objective constraint model including multiple optimization objectives and multiple constraint conditions based on the predicted prices of various crops, the method further includes:

[0029] Taking any one of various crops as a target crop, obtaining historical price data of the target crop and historical price data of related crops; the prices of the related crops are Granger causes of the price of the target crop;

[0030] Inputting historical price data of the target crop and historical price data of related crops into a price prediction model to obtain a predicted price of the target crop;

[0031] The price prediction model includes an LSTM layer, an attention layer and a linear layer.

[0032] In some embodiments, inputting the historical price data of the target crop and the historical price data of related crops into a price prediction model to obtain the predicted price of the target crop includes:

[0033] Inputting the historical price data of the target crop and the historical price data of related crops into the LSTM layer of the price prediction model respectively to obtain the initial predicted price of the target crop and the initial predicted price of the related crops;

[0034] Inputting the initial predicted price of the target crop and the initial predicted prices of the related crops into the attention layer of the price prediction model to obtain a first weight coefficient corresponding to the initial predicted price of the target crop and a second weight coefficient corresponding to the initial predicted prices of the related crops;

[0035] The initial predicted price of the target crop, the initial predicted prices of the related crops, the first weight coefficient, and the second weight coefficient are input into a linear layer in the price prediction model to obtain the predicted price of the target crop.

[0036] In some embodiments, the weight coefficients corresponding to the respective optimization objectives are determined based on a month-on-month scoring method.

[0037] The present application also provides a crop intelligent planting planning device, comprising:

[0038] a determination module, configured to determine a total planned planting area of ​​various crops and an additional subsidy value for subsidized crops among the various crops;

[0039] A construction module is used to abstract the multi-farmer collaborative planting planning problem into a Markov decision process and construct a Markov decision process model;

[0040] The solution module is used to solve the Markov decision process model with the total planned planting area of ​​various crops as a constraint condition and minimizing the profit difference between farmers and maximizing the crop rotation efficiency as the optimization goal, so as to obtain the crop planting plan corresponding to each farmer; the profit difference between farmers is determined based on the additional subsidy value of the subsidized crops.

[0041] The present application also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for intelligent crop planting planning as described above is implemented.

[0042] The present application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described intelligent crop planting planning methods.

[0043] The present application also provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned intelligent crop planting planning methods.

[0044] The intelligent crop planting planning method, electronic device and storage medium provided in this application abstract the multi-farmer collaborative planting planning problem into a Markov decision process by determining the total planned planting area of ​​various crops and the additional subsidy value of subsidized crops, construct a Markov decision process model, and solve the Markov decision process model with the total planned planting area of ​​various crops as a constraint condition and minimizing the profit difference between farmers and maximizing crop rotation efficiency as the optimization goals to obtain the crop planting plan corresponding to each farmer, thereby implementing the government's agricultural production needs at the farmer level, enabling the overall policy of agricultural planting structure adjustment to be implemented at the actual operation level, realizing farmer-level crop planting planning while meeting the government's agricultural production needs, and providing farmers with reference opinions on planting planning in the implementation of agricultural planting policies. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the present application or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0046] FIG1 is a flow chart of a method for intelligent crop planting planning provided by the present application;

[0047] FIG2 is a schematic diagram of the optimal crop rotation pattern of soybean, corn and wheat;

[0048] FIG3 is a second flow chart of the crop intelligent planting planning method provided by the present application;

[0049] FIG4 is a third flow chart of the crop intelligent planting planning method provided by the present application;

[0050] FIG5 is a schematic diagram of the structure of the intelligent crop planting planning device provided by the present application;

[0051] FIG6 is a schematic structural diagram of the electronic device provided in this application. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are part of the embodiments of this application, rather than all the embodiments. It should be noted that, in the absence of conflict, the embodiments of this application and the features in the embodiments can be combined with each other. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0053] FIG1 is a flow chart of a method for intelligent crop planting planning provided by the present application. As shown in FIG1 , the present application provides a method for intelligent crop planting planning, comprising the following steps:

[0054] Step 110 , determining the planned total planting area of ​​various crops and the additional subsidy value of subsidized crops among the various crops.

[0055] Step 120 , abstracting the multi-farmer cooperative planting planning problem into a Markov decision process and constructing a Markov decision process model.

[0056] Step 130 , using the total planned planting area of ​​various crops as a constraint condition, minimizing the profit difference between farmers and maximizing the crop rotation efficiency as the optimization goal, solving the Markov decision process model to obtain the crop planting plan corresponding to each farmer; the profit difference between farmers is determined based on the additional subsidy value of the subsidized crops.

[0057] Specifically, when the total planned planting area of ​​various crops and the additional subsidy value of subsidized crops among various crops are known, the collaborative planting planning of multiple farmers is regarded as a sequential decision-making process. Through centralized decision-making, planting allocation is made to multiple farmers, and finally the planting planning calculation of multiple farmers is realized.

[0058] The total planned planting area of ​​various crops is used as a constraint condition so that the total planting area of ​​various crops finally allocated to multiple farmers should be equal to the total planned planting area of ​​various crops.

[0059] Taking the planting of three crops, corn, wheat, and soybeans, as an example, the expressions of the various constraints are as follows:

[0060] Where, F represents the total number of farmers, Area f represents the land area (mu) of the fth household, crop fi represents the crop of the i-th plot of land of the f-th farmer, 0 represents the crop is soybean, 1 represents the crop is corn, 2 represents the crop is wheat, A0 represents the total planned planting area of ​​soybean (mu), A1 represents the total planned planting area of ​​corn (mu), and A2 represents the total planned planting area of ​​wheat (mu).

[0061] The optimization objectives are to minimize profit differences between farmers and maximize crop rotation efficiency. Based on the principle of fairness, it is hoped that all farmers will receive the same profit. Therefore, the optimization goal is to minimize profit differences between farmers. The profit differences between farmers are determined based on the additional subsidy value of subsidized crops. Because different crop rotation methods bring different economic benefits, the optimization goal is to maximize crop rotation efficiency.

[0062] For example, the expression for minimizing the profit difference among farmers is as follows:

[0063] in,

[0064] in,

[0065] In the formula, F represents the total number of farmers, farmer_profitf represents the average profit of the fth household (yuan / mu), Avg_profit represents the average profit of all households (yuan / mu), Num_land f represents the total number of plots of the fth household, areaaf i represents the area (mu) of the i-th plot of land of the f-th farmer, cropf i represents the crop of the i-th plot of land of the f-th farmer, 0 represents the crop is soybean, 1 represents the crop is corn, 2 represents the crop is wheat, Area represents the crop profit of the i-th plot of land of the f-th farmer (yuan / mu), f represents the land area (mu) of the fth household, represents the crop yield per mu of the i-th plot of land of the f-th farmer (kg / mu), represents the predicted price of crops on the i-th plot of land of the f-th farmer, represents the crop production subsidy for the i-th plot of land of the f-th farmer, represents the crop production cost per mu of the i-th plot of land of the f-th farmer (yuan / mu), and extra_subsidy represents the additional subsidy value of soybeans (yuan / mu).

[0066] Crop rotation efficiency is also considered. Figure 2 is a schematic diagram of the optimal rotation pattern for soybeans, corn, and wheat. As shown in Figure 2, if soybeans were planted last time, wheat should be planted next time; if wheat was planted last time, soybeans or corn can be planted next time; if corn was planted last time, soybeans should be planted next time.

[0067] The multi-farmer collaborative planting planning problem is modeled as a sequential decision model. Given the total planting area for each crop and the additional subsidy value for subsidized crops, centralized decisions are made for each farmer's plot, and the planting plan for each plot is determined sequentially. This process can be abstracted as a Markov decision process, and a Markov decision process model is constructed.

[0068] Solve the Markov decision process model to obtain the crop planting plan for each farmer. This can be solved using a reinforcement learning algorithm (such as the Double Deep Q Network (DDQN) algorithm). The solution to the Markov decision process model should satisfy the constraints (the total planned planting area for each crop) and the optimization objectives (minimizing profit differences between farmers and maximizing crop rotation efficiency).

[0069] Based on the determined total planting area of ​​various crops and the additional subsidy value of subsidized crops, the Double Deep Q Network (DDQN) algorithm was used to solve the crop planting planning schemes of 25 farmers when holding different numbers of plots. Table 2 shows the crop planting situation of the 25 farmers in the previous year, and Table 3 shows the crops planned to be planted by the 25 farmers in the next year and the expected average profit per mu.

[0070] Table 1 Crop planting situation of 25 farmers last year

[0071] Table 2 Crops planned to be planted by 25 farmers in the next year and their expected average profit per mu

[0072] The intelligent crop planting planning method provided in this application abstracts the multi-farmer collaborative planting planning problem into a Markov decision process while determining the total planned planting area of ​​various crops and the additional subsidy value of subsidized crops, constructs a Markov decision process model, takes the total planned planting area of ​​various crops as a constraint, and takes minimizing the profit difference between farmers and maximizing the crop rotation efficiency as the optimization goal, solves the Markov decision process model, and obtains the crop planting plan corresponding to each farmer, thereby implementing the government's agricultural production needs at the farmer level, enabling the overall policy of agricultural planting structure adjustment to be implemented at the actual operation level, realizing farmer-level crop planting planning while meeting the government's agricultural production needs, and providing farmers with reference opinions on planting planning in the implementation of agricultural planting policies.

[0073] In some embodiments, the Markov decision process model includes: a state set, an action set, a reward function, a discount factor, and a state transition function; the discount factor and the state transition function are prior values;

[0074] Construct a Markov decision process model, including:

[0075] Construct a state set based on the current crop planting status of all plots and the current decision status of all plots;

[0076] Construct an action set based on the type of crops planted;

[0077] A reward function is constructed based on different crop rotation patterns, the decision completion status of each farmer, and one or more situations in which the crop planting area does not meet the total planting plan area.

[0078] Specifically, the Markov decision process model consists of five tuples:<S,A,P,r,γ> The algorithm consists of: S, A, and r, where r(s, a) represents the reward function and r(s, a) represents the reward obtained by performing action a in the current state s; γ represents the discount factor, which takes a value between 0 and 1; and P(s′|s, a) represents the state transition function, which represents the probability of reaching state s′ after performing action a in the state s′. The state transition function P(s′|s, a) and the discount factor γ can be prior values.

[0079] State set S: The state of the crop distribution environment consists of parts s1 and s2. The state set S is composed of parts s1 and s2, and its size is 2*(Num_land+F-1), where Num_land is the total number of plots of land for all farmers and F is the total number of farmers.

[0080] 1) s1: represents the current crop status of all plots. Each number represents the crop planted in the current plot. The length of s1 is Num_land+F-1, and it is initialized to the current crop planted in all plots. For example, 0 means the current plot is planted with soybeans, 1 means the current plot is planted with corn, 2 means the current plot is planted with wheat, and -1 is a separator used to separate plots of different farmers.

[0081] 2) s2: Indicates the decision status of all current land parcels. 0 indicates a land parcel that has been decided or not yet decided, and 1 indicates a land parcel that is in the process of being decided. To facilitate integration with the s1 state, the size is also set to Num_land + F-1.

[0082] For example, Table 3 is a state set table showing the specific states of s1 and s2. As shown in Table 3, Farmer 1 owns two plots of land, with soybeans currently planted on the first and corn currently planted on the second. Farmer 2 owns three plots of land, with wheat currently planted on the first, soybeans currently planted on the second, and corn currently planted on the third. s1 indicates the current crop status of the plot, and s2 indicates whether a decision is being made on the plot, with 1 indicating a decision is being made and 0 indicating a decision has been made or not. Subsequent plantings on the remaining plots can be determined directly based on the crop rotation pattern in Figure 2.

[0083] Table 3 Status set table

[0084] Action set a: This determines the crops to be planted on the current plot. Based on the crop type, an action set is constructed. For example, if soybeans, corn, and wheat are planted, the action set has three actions: 0 for planting soybeans, 1 for planting corn, and 2 for planting wheat.

[0085] Reward function r: The design of the reward function r determines the learning skill and efficiency of the reinforcement learning algorithm. The calculation method of reward r is as follows:

[0086] Among them, reasonable crop rotation refers to the rotation method of planting soybeans, wheat and corn in sequence, soybean rotation refers to the rotation method of continuously planting soybeans, and other rotation methods refer to rotation methods other than the above two.

[0087] The crop intelligent planting planning method provided in this application realizes the construction of a Markov decision process model by constructing a state set, an action set and a reward function.

[0088] In some embodiments, when using reinforcement learning algorithms for simulation calculations, it is also necessary to determine whether the current round has ended. This round can end in two situations: 1) after providing a decision for the last farmer's last plot of land, the round ends, successfully providing the farmer with a crop planting decision; 2) when the execution of the action causes the crop area to exceed the total planned area, the round ends directly.

[0089] In some embodiments, reinforcement learning algorithms often produce illegal actions during the learning process. The following is a design for a valid action set:

[0090] Where avail_action[c] represents the set of valid actions for the current crop c, cur_num_crop c Indicates the currently allocated planting area (mu) of crop c. When the allocated planting area of ​​crop c is less than the total planting area of ​​crop c, the action of planting crop c is added to the valid action set, otherwise it is not added to the valid action set.

[0091] In some implementations, FIG3 is a second flow chart of the crop intelligent planting planning method provided by the present application. As shown in FIG3 , the aforementioned step 110 specifically includes:

[0092] Step 310, based on the predicted prices of various crops, construct a multi-objective constraint model including multiple optimization objectives and multiple constraints; the multiple optimization objectives and multiple constraints are determined based on agricultural production needs, farmers' economic benefits, water and fertilizer resource utilization, and planting subsidy policies.

[0093] Step 320 , finding the optimal solution for the multi-objective constraint model to obtain the planned planting areas of various crops and the additional subsidy values ​​for subsidized crops.

[0094] Specifically, farmers' economic benefits are a major objective function in crop planting area planning, and crop price is an important parameter in farmers' economic benefits.

[0095] While some existing technologies incorporate crop market price information into crop planting area planning, using historical market prices as a reference is subject to influences such as current market supply and demand, climate change, and national policies. Using current market prices as a reference can lead to a "price chasing" situation or a "cobweb dilemma." To avoid both of these situations, the projected prices of various crops are incorporated into crop planting area planning, constructing a multi-objective constraint model that includes multiple optimization objectives and constraints.

[0096] The multiple optimization objectives and multiple constraints in the multi-objective constraint model are determined by taking into account the needs of planting structure adjustment, farmers' economic benefits, water and fertilizer resource utilization, and planting subsidy policies, making the multi-objective constraint model more realistic and able to effectively provide policy support for relevant departments.

[0097] For example, consider the goal of expanding soybean planting by 10 million mu (approximately 1.5 acres). Specifically, four constraints are considered: total planted area (c1), maximum and minimum planted area (c2), maximum water use (c3), and maximum fertilizer use (c4). Furthermore, the model's four optimization objectives are maximizing farmer economic benefits (J1), maximizing ecological benefits (J2), achieving soybean planting targets (J3), and minimizing additional subsidies for subsidized crops (J4).

[0098] Constraint c1 indicates that the total planting area of ​​soybeans, corn, and wheat must be less than or equal to the total available planting area. The expression of constraint c1 is as follows:

[0099] In the formula, area k represents the planting area (mu) of the kth crop, N represents the total number of crops, and Area represents the total available planting area (mu).

[0100] The minimum planting area is to prevent crop shortages that would cause price increases, while the maximum planting area is to ensure that there is not too much crop redundancy, which would lead to storage problems and sales problems such as falling crop prices. Constraint c2 represents the maximum and minimum boundary constraints for the planting area of ​​each crop. The expression of constraint c2 is as follows:

[0101] Min_demand k ≤area k ≤Max_demand k

[0102] Where, Min_demand k Indicates the minimum planting area (mu) of the kth crop, area k Indicates the planting area of ​​the kth crop (mu), Max_demand krepresents the maximum planting area of ​​the kth crop (mu), and N represents the total number of crops.

[0103] Constraint c3 indicates that the total water consumption of all crops cannot exceed the given maximum total water consumption. The expression of constraint c3 is as follows:

[0104] In the formula, require_water k represents the water requirement per mu of the kth crop (cubic meters / mu), area k represents the planting area of ​​the kth crop (mu), Max_waterUse represents the maximum average water consumption of N crops (cubic meters / mu), and N represents the total number of crops.

[0105] To avoid damage to soil structure and environmental pollution caused by excessive use of fertilizers, constraint C4 states that the amount of fertilizer used for all crops cannot exceed the maximum amount of fertilizer used. The expression of constraint C4 is as follows:

[0106] In the formula, fertilizer k represents the total amount of fertilizer per mu for the k-th crop (yuan / mu), area k represents the planting area of ​​the kth crop (mu), Max_fertilizerUse represents the maximum amount of fertilizer used for N crops (yuan / mu), and N represents the total number of crops.

[0107] The expression of the optimization objective J1 is as follows:

[0108] in,

[0109] profit k =yield k *price k +subsidy k -cost k

[0110] In the formula, J1 represents the economic benefit of farmers (yuan / mu), profit1 represents the profit of soybeans (yuan / mu), extra_subsidy represents the additional subsidy value of soybeans (yuan / mu), area1 represents the planting area of ​​subsidized crops (mu), profit k (k=2,...,N) represents the profit of the kth crop except soybean, area k (k=2,...,N) represents the planting area (mu) of the kth crop except soybean, N represents the total number of crops, profit k represents the profit of the kth crop (yuan / mu), yieldk Indicates the yield per mu of the k-th crop (kg / mu), price k represents the predicted price of the k-th crop (yuan / kg), subsidy k represents the production subsidy for the kth crop (yuan / mu), cost k represents the production cost of the k-th crop (yuan / mu).

[0111] What needs to be understood is that production subsidies are basic subsidies, and additional subsidies are subsidies beyond production subsidies, which can be understood as local additional subsidies to encourage the production of certain crops.

[0112] The expression of the optimization objective J2 is as follows:

[0113] Where, J2 represents ecological benefit (yuan / mu), require_water k represents the water requirement per mu of the kth crop (cubic meters / mu), area k represents the planting area of ​​the kth crop / (mu), and N represents the total number of crops.

[0114] The expression of the optimization objective J3 is as follows:

[0115] J3=area1

[0116] Where J3 represents the expanded soybean planting area (mu); area1 represents the soybean planting area (mu).

[0117] The expression of the optimization target J4 is as follows:

[0118] J4 = extra_subsidy

[0119] In the formula, J4 represents the additional subsidy for soybeans (yuan / mu), and extra_subsidy represents the additional subsidy for soybeans (yuan / mu).

[0120] The intelligent crop planting planning method provided in this application constructs a multi-objective constraint model based on the predicted prices of various crops to avoid the situation of "chasing prices" or falling into the "spider web dilemma". In addition, it comprehensively considers key factors such as the demand for planting structure adjustment, farmers' economic benefits, water and fertilizer resource utilization, and planting subsidy policies. The multi-objective constraint model can not only guide the adjustment of agricultural planting structure more scientifically, but also achieve more sustainable and economically maximizing results in practice, providing reliable support for the decision-making of relevant departments.

[0121] In some embodiments, the optimal solution of the multi-objective constraint model is solved to obtain the planned planting area of ​​various crops and the subsidy value of the subsidized crops, including:

[0122] Convert multiple optimization objectives in the multi-objective constraint model into a single optimization objective to obtain a single-objective constraint model;

[0123] The optimal solution of the single-objective constraint model is solved to obtain the planned planting area of ​​various crops and the additional subsidy value of subsidized crops.

[0124] Specifically, multiple optimization objectives in the multi-objective constraint model are converted into a single optimization objective to obtain a single-objective constraint model in order to simplify the subsequent solution process.

[0125] An intelligent optimization algorithm (e.g., particle swarm optimization (PSO)) was used to find the optimal solution for the single-objective constraint model, determining the planned planting area for each crop and the additional subsidy value for subsidized crops. The specific parameter settings for the PSO algorithm are shown in Table 4. The results for the planned planting area for each crop and the additional subsidy value for subsidized crops are shown in Table 5.

[0126] Table 4 PSO algorithm parameter setting table

[0127] Table 5 Results of calculation of planned planting areas of various crops and additional subsidy values ​​for subsidized crops

[0128] The intelligent crop planting planning method provided in this application first converts multiple optimization objectives in the multi-objective constraint model into a single optimization objective to obtain a single-objective constraint model; then the single-objective constraint model is solved for the optimal solution to obtain the planned planting area of ​​various crops and the additional subsidy value of subsidized crops, thereby simplifying the solution process.

[0129] In some embodiments, multiple optimization objectives in a multi-objective constraint model are converted into a single optimization objective to obtain a single-objective constraint model, including:

[0130] Determine the normalization factors corresponding to each optimization objective in the multi-objective constraint model, as well as the weight coefficients corresponding to each optimization objective;

[0131] Based on the normalization factors corresponding to the optimization objectives, each optimization objective is normalized to obtain each normalized objective;

[0132] Based on the weight coefficients corresponding to each optimization objective and each normalization objective, a single optimization objective is obtained;

[0133] Based on a single optimization objective and multiple constraints, a single-objective constraint model is obtained.

[0134] Specifically, the method adopted is to normalize each optimization objective separately, and then use linear weighting to convert multiple optimization objectives into a single optimization objective for solution.

[0135] For example, the expression for converting the above optimization objectives J1, J2, J3, and J4 into a single optimization objective J is as follows:

[0136] Wherein, norm1 represents the normalization factor corresponding to the optimization target J1, norm2 represents the normalization factor corresponding to the optimization target J2, norm3 represents the normalization factor corresponding to the optimization target J3, norm4 represents the normalization factor corresponding to the optimization target J4, w1 represents the weight coefficient corresponding to the optimization target J1, w2 represents the weight coefficient corresponding to the optimization target J2, w3 represents the weight coefficient corresponding to the optimization target J3, and w4 represents the weight coefficient corresponding to the optimization target J4.

[0137] in,

[0138] Where norm1 represents the normalization factor corresponding to the optimization target J1, prefit k Indicates the profit of the k-th crop (yuan / mu); Max_demand k represents the maximum planting area of ​​the kth crop; N represents the total number of crops.

[0139] in,

[0140] Where norm2 represents the normalization factor corresponding to the optimization target J2, N represents the total number of crops, and require_water k Indicates the water requirement per mu of the k-th crop (cubic meters / mu), Max_demand k Indicates the maximum planting area of ​​the k-th crop.

[0141] in,

[0142] norm3=Area

[0143] Where norm3 represents the normalization factor corresponding to the optimization target J3, and Area represents the total available planting area (mu).

[0144] in,

[0145] norm4=Max_extra_subsidy

[0146] Where norm4 represents the normalization factor corresponding to the optimization target J4, and Max_extra_subsidy represents the maximum value of soybean additional subsidy.

[0147] The intelligent crop planting planning method provided in this application normalizes each optimization objective separately and then uses linear weighting to convert multiple optimization objectives into a single optimization objective, which is conducive to simplifying the subsequent solution process.

[0148] In some embodiments, the weight coefficient corresponding to each optimization objective is determined based on a month-on-month scoring method.

[0149] Specifically, according to the month-on-month scoring method, the weight coefficients corresponding to each optimization objective can be obtained.

[0150] For example, the weight coefficients corresponding to the optimization targets J1, J2, J3 and J4 obtained based on the month-on-month scoring method are shown in Table 6.

[0151] Table 6 Weight coefficients corresponding to optimization objectives J1, J2, J3 and J4 determined based on the ring-comparison scoring method

[0152] In some embodiments, FIG4 is a third flow chart of the crop intelligent planting planning method provided by the present application. As shown in FIG4 , before the aforementioned step 310, the method further includes:

[0153] Step 410 , taking any one of various crops as a target crop, obtaining historical price data of the target crop and historical price data of related crops; the prices of the related crops are Granger causes of the price of the target crop;

[0154] Step 420 , inputting the historical price data of the target crop and the historical price data of related crops into a price prediction model to obtain a predicted price of the target crop;

[0155] Among them, the price prediction model includes LSTM layer, attention layer and linear layer.

[0156] Specifically, traditional crop price forecasting often uses only historical data of the crop being forecasted as input to a regression model, which is then used to fit and predict future crop prices. However, crop price fluctuations are actually caused by a variety of natural and social factors, such as socioeconomic conditions, natural disasters, and related crop prices. When price forecasting solely based on historical data, when there is a significant time span between training and test data, it often fails to accurately fit the crop price forecast curve.

[0157] When constructing a multi-objective constraint model, predicted prices for various crops are required. Therefore, any one of these crops is chosen as the target crop. There is a certain correlation between the prices of different crops, which is driven by specific social factors. For example, the prices of corn and soybeans have a strong positive correlation because they are substitutes for each other in production and have similar primary production areas and planting and harvesting seasons. Therefore, by incorporating historical price data for these crops, we can enrich the input data for the prediction model, thereby improving the accuracy of the target crop price forecast.

[0158] First, a correlation analysis is conducted on the prices of various crops to determine the related crops corresponding to the target crop, that is, a Granger causality test is conducted on the prices of various crops, and the prices of related crops are the Granger causes of the prices of the target crop.

[0159] For example, by performing a Granger causality test on the monthly price of corn and the monthly price of soybeans, we can find out that the monthly price of corn is the Granger cause of the monthly price of soybeans, but the monthly price of soybeans is not the Granger cause of the monthly price of corn. Therefore, the soybean price can be predicted in combination with the corn price.

[0160] Subsequently, historical price data for the target crop and related crops are obtained and aligned in time. The historical price data can be monthly data to be closer to the actual price, thereby improving the accuracy of price prediction.

[0161] Then, the historical price data of the target crop and related crops are input into the price prediction model to obtain the predicted price of the target crop. The price prediction model includes a long short-term memory (LSTM) layer, an attention layer, and a linear layer.

[0162] The LSTM layer performs price predictions based on the input time series of historical crop prices. The attention layer learns attention parameters and assigns weight coefficients to the price predictions obtained by the LSTM layer, representing the degree of attention paid to the input crop's historical prices. The linear layer maps the output of the attention layer to obtain the final price prediction results.

[0163] In addition, the price prediction model can calculate the loss in combination with the real price of the crop, and implement network updates in the price prediction model through back propagation.

[0164] Taking the prediction of soybean prices based on corn prices as an example, considering that corn and soybeans are planted around May and harvested and put on the market around early October, the planting cycle is about 5 months.

[0165] Therefore, when constructing training data, we chose to use the monthly prices of crops for the previous 10 months as historical price data and predict the market price for the 16th month to avoid cumulative errors caused by multiple predictions. When constructing training data, we used a "sliding window" approach to reasonably predict the prices of various crops. The window size is w, and the price prediction model input is a price sequence of length w, with the w+6th price as the label. Alternatively, when the amount of data is limited, the first 90% of the data can be used as the training set, and the last 10% as the test set.

[0166] For example, Table 7 is a table showing the training data and label data construction of the price prediction model. As shown in Table 7, part X is the training data, part Y is the label data, the sliding window size is 10, and the data size is m. n The labeled data can be used to verify the accuracy of the prediction.

[0167] Table 7: Training data and label data construction table of price prediction model

[0168] In some embodiments, historical price data of a target crop and historical price data of related crops are input into a price prediction model to obtain a predicted price of the target crop, including:

[0169] The historical price data of the target crop and the historical price data of related crops are respectively input into the LSTM layer of the price prediction model to obtain the initial predicted price of the target crop and the initial predicted price of related crops;

[0170] Inputting the initial predicted price of the target crop and the initial predicted prices of related crops into the attention layer of the price prediction model to obtain a first weight coefficient corresponding to the initial predicted price of the target crop and a second weight coefficient corresponding to the initial predicted prices of the related crops;

[0171] The initial predicted price of the target crop, the initial predicted prices of related crops, the first weight coefficient and the second weight coefficient are input into the linear layer of the price prediction model to obtain the predicted price of the target crop.

[0172] Specifically, taking the prediction of soybean price based on corn price as an example, the historical price data of corn and soybean are aligned in time and input into the LSTM layer of the price prediction model. The two LSTM models in the LSTM layer respectively predict the historical price data of corn [x1, x2, ..., x 10 ] and the historical price data of soybeans [y1,y2,…,y 10 ] to make price predictions and get the initial predicted price of corn h 10 and the initial forecast price of soybeans h ′ 10 .

[0173] The initial predicted price of corn h 10 and the initial forecast price of soybeans h ′ 10 , which is input into the attention layer of the price prediction model. The attention layer is the initial predicted price of corn h 10 Assign weight coefficient w1 to the initial predicted price h of soybeans ′ 10 Assign weight coefficient w2.

[0174] The initial predicted price of corn h 10 and the initial forecast price of soybeans h ′ 10 , and the initial predicted price of corn h 10 Assign weight coefficient w1 to the initial predicted price h of soybeans ′ 10 Assign the weight coefficient w2 and input it into the linear layer in the price prediction model. The linear layer performs weighted summation to obtain h 10 *w1+h ′ 10 *w2 is the predicted price of soybeans.

[0175] To verify that incorporating historical price data of related crops into target crop price predictions can significantly improve crop price prediction effectiveness, this example compared the performance of the models with and without incorporating related crop prices. The results are shown in Table 8. MAE, RMSE, and MAPE represent mean absolute error, root mean square error, and mean absolute percentage error, respectively.

[0176] Table 8 Performance comparison of price prediction

[0177] The intelligent crop planting planning method provided in this application obtains the predicted price of the target crop by inputting the historical price data of the target crop and the historical price data of related crops into a price prediction model, which includes an LSTM layer, an attention layer and a linear layer, thereby improving the accuracy of the predicted price of the target crop.

[0178] The crop intelligent planting planning device provided in this application is described below. The crop intelligent planting planning device described below and the crop intelligent planting planning method described above can be referenced to each other.

[0179] FIG5 is a schematic diagram of the structure of the intelligent crop planting planning device provided by the present application. As shown in FIG5 , the present application provides an intelligent crop planting planning device, including:

[0180] A determination module 510 is configured to determine a total planned planting area for various crops and an additional subsidy value for subsidized crops among the various crops;

[0181] A construction module 520 is used to abstract the multi-farmer cooperative planting planning problem into a Markov decision process and construct a Markov decision process model;

[0182] Solution module 530 is used to solve the Markov decision process model with the total planned planting area of ​​various crops as a constraint condition and minimizing the profit difference between farmers and maximizing the crop rotation efficiency as the optimization goal, so as to obtain the crop planting plan corresponding to each farmer; the profit difference between farmers is determined based on the additional subsidy value of the subsidized crop.

[0183] In some embodiments, the Markov decision process model includes: a state set, an action set, a reward function, a discount factor, and a state transition function; the discount factor and the state transition function are prior values;

[0184] The construction module 520 is specifically used to: construct the state set based on the current crop planting status of all plots and the current decision status of all plots; construct the action set based on the type of planted crops; and construct the reward function based on one or more situations of different crop rotation methods, the decision completion status of each farmer, and the crop planting area not meeting the total planting plan area.

[0185] In some embodiments, the determining module 510 includes:

[0186] Constructing a submodule for constructing a multi-objective constraint model including multiple optimization objectives and multiple constraints based on the predicted prices of various crops; the multiple optimization objectives and the multiple constraints are determined based on the needs of planting structure adjustment, farmers' economic benefits, water and fertilizer resource utilization, and planting subsidy policies;

[0187] The solution submodule is used to solve the optimal solution of the multi-objective constraint model to obtain the planned planting area of ​​the various crops and the additional subsidy value of the subsidized crops.

[0188] In some embodiments, the solution submodule includes:

[0189] a conversion unit, configured to convert multiple optimization objectives in the multi-objective constraint model into a single optimization objective to obtain a single-objective constraint model;

[0190] A solving unit is used to solve the optimal solution of the single-objective constraint model to obtain the planned planting areas of the various crops and the additional subsidy values ​​of the subsidized crops.

[0191] In some embodiments, the conversion unit is specifically used to: determine the normalization factors corresponding to each optimization objective in the multi-objective constraint model, and the weight coefficients corresponding to each optimization objective; normalize each optimization objective based on the normalization factors corresponding to each optimization objective to obtain each normalized objective; obtain a single optimization objective based on the weight coefficients corresponding to each optimization objective and each normalized objective; obtain the single-objective constraint model based on the single optimization objective and the multiple constraints.

[0192] In some embodiments, the apparatus further comprises:

[0193] an acquisition module, configured to take any one of various crops as a target crop and acquire historical price data of the target crop and historical price data of related crops; the price of the related crops being the Granger cause of the price of the target crop;

[0194] a prediction module, configured to input historical price data of the target crop and historical price data of related crops into a price prediction model to obtain a predicted price of the target crop;

[0195] The price prediction model includes an LSTM layer, an attention layer and a linear layer.

[0196] In some embodiments, the prediction module is specifically configured to:

[0197] Inputting the historical price data of the target crop and the historical price data of related crops into the LSTM layer of the price prediction model respectively to obtain the initial predicted price of the target crop and the initial predicted price of the related crops;

[0198] Inputting the initial predicted price of the target crop and the initial predicted prices of the related crops into the attention layer of the price prediction model to obtain a first weight coefficient corresponding to the initial predicted price of the target crop and a second weight coefficient corresponding to the initial predicted prices of the related crops;

[0199] The initial predicted price of the target crop, the initial predicted prices of the related crops, the first weight coefficient, and the second weight coefficient are input into a linear layer in the price prediction model to obtain the predicted price of the target crop.

[0200] It should be noted here that the above-mentioned crop intelligent planting planning device provided in this application can implement all the method steps implemented in the above-mentioned method embodiment and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as the method embodiment will not be described in detail here.

[0201] FIG6 is a schematic diagram of the structure of an electronic device provided by the present application. As shown in FIG6 , the electronic device may include: a processor 610, a communication interface 620, a memory 630, and a communication bus 640, wherein the processor 610, the communication interface 620, and the memory 630 communicate with each other via the communication bus 640. The processor 610 may call logic instructions in the memory 630 to execute a method for intelligent crop planting planning, which includes: determining the total planned planting area of ​​various crops and the additional subsidy value of subsidized crops among the various crops; abstracting the multi-farmer collaborative planting planning problem into a Markov decision process and constructing a Markov decision process model; solving the Markov decision process model with the total planned planting area of ​​various crops as a constraint condition and minimizing the profit difference between farmers and maximizing crop rotation efficiency as the optimization goal to obtain the crop planting plan corresponding to each farmer; the profit difference between farmers is determined based on the additional subsidy value of the subsidized crop.

[0202] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0203] On the other hand, the present application also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the crop intelligent planting planning method provided by the above methods, which includes: determining the total planned planting area of ​​various crops and the additional subsidy value of the subsidized crops among the various crops; abstracting the multi-farmer collaborative planting planning problem into a Markov decision process, and constructing a Markov decision process model; using the total planned planting area of ​​various crops as a constraint condition, and minimizing the profit difference between farmers and maximizing crop rotation efficiency as optimization goals, solving the Markov decision process model to obtain the crop planting plan corresponding to each farmer; the profit difference between farmers is determined based on the additional subsidy value of the subsidized crops.

[0204] On the other hand, the present application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the intelligent crop planting planning method provided by the above-mentioned methods, the method including: determining the total planned planting area of ​​various crops and the additional subsidy value of the subsidized crops among the various crops; abstracting the multi-farmer collaborative planting planning problem into a Markov decision process, and constructing a Markov decision process model; using the total planned planting area of ​​various crops as a constraint condition, minimizing the profit difference between farmers and maximizing crop rotation efficiency as optimization goals, solving the Markov decision process model to obtain the crop planting plan corresponding to each farmer; the profit difference between farmers is determined based on the additional subsidy value of the subsidized crops.

[0205] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0206] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiments.

[0207] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application pertains. The terms used herein in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application.

[0208] It should be further noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not preclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0209] In this application, "at least one" means one or more, and "a plurality" means two or more than two. The terms "first," "second," "third," "fourth," etc. (if any) in this application are used to distinguish similar objects, rather than to describe a particular order or sequence.

[0210] In the embodiments of this application, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0211] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for intelligent crop planting planning, comprising: Determine the total planned planting area for various crops and the additional subsidy value for subsidized crops among the various crops; The problem of multi-farmer cooperative planting planning is abstracted into a Markov decision process, and a Markov decision process model is constructed. With the total planned planting area of ​​various crops as a constraint condition, and minimizing the profit difference between farmers and maximizing the crop rotation efficiency as the optimization goal, the Markov decision process model is solved to obtain the crop planting plan corresponding to each farmer; the profit difference between farmers is determined based on the additional subsidy value of the subsidized crops.

2. The method for intelligent crop planting planning according to claim 1, wherein: The Markov decision process model includes: a state set, an action set, a reward function, a discount factor and a state transition function; the discount factor and the state transition function are prior values; The Markov decision process model is constructed, comprising: Constructing the state set based on the current crop planting states of all plots and the current decision states of all plots; Constructing the action set based on the type of crops planted; The reward function is constructed based on one or more of different crop rotation methods, the decision completion status of each farmer, and the situation where the crop planting area does not meet the total planting plan area.

3. The method for intelligent crop planting planning according to claim 1, wherein: The determination of the planned total planting area of ​​various crops and the additional subsidy value of subsidized crops among the various crops includes: Based on the predicted prices of various crops, a multi-objective constraint model is constructed, which includes multiple optimization objectives and multiple constraints; the multiple optimization objectives and the multiple constraints are determined based on the needs of planting structure adjustment, farmers' economic benefits, water and fertilizer resource utilization, and planting subsidy policies; The optimal solution of the multi-objective constraint model is solved to obtain the planned planting areas of the various crops and the additional subsidy values ​​of the subsidized crops.

4. The method for intelligent crop planting planning according to claim 3, wherein: Solving the multi-objective constraint model for an optimal solution to obtain the planned planting areas of the various crops and the additional subsidy values ​​of the subsidized crops includes: Converting multiple optimization objectives in the multi-objective constraint model into a single optimization objective to obtain a single-objective constraint model; Solve the optimal solution of the single-objective constraint model to obtain the planting planning area of ​​the various crops and the Additional subsidy value for subsidized crops.

5. The method for intelligent crop planting planning according to claim 4, wherein: The step of converting multiple optimization objectives in the multi-objective constraint model into a single optimization objective to obtain a single-objective constraint model includes: Determining the normalization factors corresponding to each optimization objective in the multi-objective constraint model, and the weight coefficients corresponding to each optimization objective; Based on the normalization factors corresponding to the optimization objectives, each optimization objective is normalized to obtain each normalized objective; Based on the weight coefficients corresponding to each optimization objective and each normalization objective, a single optimization objective is obtained; Based on the single optimization objective and the multiple constraint conditions, the single objective constraint model is obtained.

6. The method for intelligent crop planting planning according to claim 3, wherein: Before constructing a multi-objective constraint model including multiple optimization objectives and multiple constraints based on the predicted prices of various crops, the following steps are also included: Taking any one of various crops as a target crop, obtaining historical price data of the target crop and historical price data of related crops; the prices of the related crops are Granger causes of the price of the target crop; Inputting historical price data of the target crop and historical price data of related crops into a price prediction model to obtain a predicted price of the target crop; The price prediction model includes an LSTM layer, an attention layer and a linear layer.

7. The method for intelligent crop planting planning according to claim 6, wherein: Inputting the historical price data of the target crop and the historical price data of related crops into a price prediction model to obtain a predicted price of the target crop includes: Inputting the historical price data of the target crop and the historical price data of related crops into the LSTM layer of the price prediction model respectively to obtain the initial predicted price of the target crop and the initial predicted price of the related crops; Inputting the initial predicted price of the target crop and the initial predicted prices of the related crops into the attention layer of the price prediction model to obtain a first weight coefficient corresponding to the initial predicted price of the target crop and a second weight coefficient corresponding to the initial predicted prices of the related crops; The initial predicted price of the target crop, the initial predicted prices of the related crops, the first weight coefficient, and the second weight coefficient are input into a linear layer in the price prediction model to obtain the predicted price of the target crop.

8. The method for intelligent crop planting planning according to claim 5, wherein: The weight coefficients corresponding to the various optimization objectives are determined based on the month-on-month scoring method.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the crop intelligent planting planning method according to any one of claims 1 to 8 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, wherein: When the computer program is executed by a processor, the crop intelligent planting planning method according to any one of claims 1 to 8 is implemented.

Images