Artificial intelligence-based economic crop planting management method and system

Abstract

The invention discloses an artificial intelligence-based economic crop planting management method and system. The method comprises the steps of obtaining planting management original data, optimizing the original data, selecting target economic crops, predicting market prices of the economic crops, optimizing economic crop planting strategies and intelligently managing economic crop planting. The invention relates to the technical field of economic crop planting data processing, in particular to an artificial intelligence-based economic crop planting management method and system, which innovatively takes market price prediction as a guide and integrates a method system of target crop intelligent screening and full-cycle planting strategy optimization. The planting income stability of commercial crops is improved, price data signal decomposition processing is introduced, and the quality of commercial crop market price prediction input data is improved; a market price prediction model is innovatively designed, and the accuracy and stability of a model prediction result are improved; and an improved particle swarm optimization algorithm is introduced to realize efficient global optimization of the economic crop planting strategy combination.

Description

Technical Field

[0001] The present invention relates to the technical field of economic crop planting data processing, and specifically refers to an economic crop planting management method and system based on artificial intelligence. Background Art

[0002] An economic crop planting management method and system based on artificial intelligence refers to using artificial intelligence technology to intelligently manage and optimize economic crops throughout the planting cycle, thereby improving the economic benefits of economic crops and reducing production costs, and ultimately realizing digital, intelligent, and refined management of agricultural production, providing intelligent decision-making support for the whole process and the whole cycle for economic crop planting managers.

[0003] However, in traditional economic crop planting management methods, there are technical problems such as the lack of scientific basis for economic crop selection, lagging response to market prices, and difficulty in matching price high points, resulting in unstable returns in economic crop planting management and high market mismatch risks; there are technical problems in price time series data processing in traditional economic crop planting management methods, such as poor robustness to high-frequency noise in price series, instability in the mode decomposition process, and unsatisfactory trend extraction effects, resulting in strong volatility in the input data of subsequent price prediction models and easy overfitting, thus causing large deviations and poor reliability in price prediction results; there are technical problems in existing economic crop market price prediction models, such as insufficient learning of complex time series price fluctuation characteristics and lack of an effective identification mechanism for key time points and main influencing factors of price changes, thus affecting the obvious deficiencies in the accuracy and stability of economic crop price prediction results; in existing algorithms used for optimizing economic crop planting strategies, there are technical problems such as being easily trapped in local optimal solutions and insufficient search space exploration ability, which in turn leads to the lack of global optimality in the generated optimal planting strategy combinations. Summary of the Invention

[0004] In view of the above situation, to overcome the defects of the prior art, the present invention provides an economic crop planting management method and system based on artificial intelligence. Aiming at the technical problems existing in the traditional economic crop planting management method, such as the lack of scientific basis for the selection of economic crops, the lag in market price response, and the difficulty in matching the high price points, which lead to unstable benefits in economic crop planting management and high risks of market mismatch, this solution innovatively adopts a method system oriented by market price prediction, integrating intelligent screening of target crops and optimization of the whole-cycle planting strategy, effectively improving the scientificity and suitability of the target selection of economic crops, accurately identifying future high-price windows, realizing price-driven planting decisions, supporting personalized strategy recommendations according to local conditions, adapting to various crops and regional environments, and finally realizing price-based decision-making in the process of economic crop planting, improving the stability of the planting income of economic crops; aiming at the technical problems existing in the processing of price time-series data in the traditional economic crop planting management method, such as poor robustness to high-frequency noise in the price sequence, unstable modal decomposition process, and unsatisfactory trend extraction effect, which lead to strong volatility and overfitting in the input data of the subsequent price prediction model, resulting in large deviations and poor reliability in the price prediction results, this solution innovatively introduces a mechanism for constructing multiple groups of noise-added perturbation signals and a hierarchical progressive modal extraction strategy, significantly improving the stability and decomposition accuracy of the modal decomposition algorithm under high-frequency noise interference, realizing the fine extraction of predictable medium- and low-frequency trend signals in the economic crop price sequence, significantly improving the quality of the input data for economic crop market price prediction, and effectively improving the accuracy of economic crop market price prediction; aiming at the technical problems existing in the existing market price prediction model applicable to economic crops, such as insufficient learning of complex time-series price fluctuation characteristics and lack of an effective identification mechanism for key time points and main influencing factors of price changes, which affect the obvious deficiencies in the accuracy and stability of economic crop price prediction results, this solution innovatively constructs a multi-quantile price prediction model integrating improved ILSTM and introducing time attention and feature attention mechanisms. By introducing the improved ILSTM structure, the expression ability of the model for complex time-series dependence relationships and non-linear fluctuation trends in the economic crop price sequence is significantly improved. By introducing time attention and feature attention mechanisms, dual identification of key time points of price changes and dominant influencing factors is realized, effectively depicting the deep logic of price evolution over time and the dynamic effects of multi-source factors. Combining the multi-quantile output strategy, it can synchronously output the upper and lower limit estimates and central trend predictions of prices, improving the accuracy and stability of the model prediction results;In view of the technical problems in the existing algorithms for optimizing economic crop planting strategies, such as being prone to falling into local optimal solutions and having insufficient search space exploration ability, which lead to the lack of global optimality in the generated optimal planting strategy combinations, this solution innovatively introduces an improved particle swarm optimization algorithm with a hybrid cosine chaos mapping initialization and mutation intensity control mechanism to search for the optimal economic crop planting strategy, enhancing the diversity of the initial population and the search guidance, and enhancing the global search ability and the ability to jump out of local extreme values through dynamic mutation operations, so as to achieve the efficient global optimization of the economic crop planting strategy combination, and finally obtain an economic crop planting strategy plan that better fits the market price trend and has better revenue potential.

[0005] The technical solution adopted by the present invention is as follows: The present invention provides an economic crop planting management method and system based on artificial intelligence. The method includes the following steps:

[0006] Step S1: Obtain the original data of planting management;

[0007] Step S2: Optimize the original data;

[0008] Step S3: Select the target economic crop;

[0009] Step S4: Forecast the market price of economic crops;

[0010] Step S5: Optimize the economic crop planting strategy;

[0011] Step S6: Intelligently manage the economic crop planting.

[0012] Further, in step S1, the obtaining of the original data of planting management is specifically to obtain the original data of planting management through data collection; the original data of planting management includes reference economic crop selection data, target economic crop selection data, historical economic crop market price forecast data, and real-time economic crop market price forecast data; both the reference economic crop selection data and the target economic crop selection data include regional planting condition data, economic crop planting attribute data, and economic crop economic attribute data; the reference economic crop selection data further includes the reference economic crop selection result; both the historical economic crop price forecast data and the real-time economic crop price forecast data include various economic crop transaction data and external transaction impact data.

[0013] Further, in step S2, the optimization of the original data specifically includes the following steps:

[0014] Step S21: Perform data cleaning processing, specifically filling in missing values, removing outliers, and normalizing fields for the original data;

[0015] Step S22: Data normalization processing, specifically, normalizing the numerical data in the original data through the min-max normalization method;

[0016] Step S23: Data encoding processing, specifically, encoding the categorical fields in the original data using the one-hot encoding method to convert discrete text and label variables into sparse numerical vectors;

[0017] Step S24: Dynamic time window partitioning, specifically, dividing the time series data related to the target cash crop into an input window and a prediction window through a sliding window construction algorithm;

[0018] Step S25: Price data signal decomposition processing, specifically, improving the empirical mode decomposition algorithm by introducing a mechanism for constructing multiple groups of noisy perturbation signals and designing a hierarchical and progressive modal extraction strategy to achieve multi-level modal decomposition and high-frequency information removal of the cash crop price time series, obtaining the denoised cash crop price sequence data, including the following steps:

[0019] Step S251: Noisy signal construction processing, specifically, injecting Gaussian white noise into the original cash crop price sequence and modulating the perturbation in combination with the first-order modal decomposition operation function to construct multiple groups of noisy price signals, with the formula used as follows:

[0020] ;

[0021] In the formula, represents the value of the i-th group of perturbation signal sequences at time step t, represents the value of the original cash crop price sequence data at time step t, represents the noise adjustment coefficient used for the i-th perturbation, represents the first-layer modal decomposition operation function, represents the Gaussian white noise sample introduced at time step t in the i-th perturbation;

[0022] Step S252: Initial residual extraction, specifically, performing modal decomposition operations on each group of noisy signals to generate the modal components of each group, then performing local mean operations, and averaging the local mean results of all perturbation versions to form a stable first-layer residual signal , subtracting from to obtain the first-layer modal components;

[0023] Step S253: Multi-round modal iterative decomposition, specifically, continuously adding perturbation noise to each layer of residuals and performing modal decomposition operations to extract the current layer of modal components and residual signals, with the formula used as follows:

[0024] ;

[0025] ;

[0026] In the formula, represents the residual signal of the j-th layer, represents the residual component of the -th layer, represents the noise perturbation adjustment coefficient of the -th layer, represents the modal decomposition operation function of the j-th layer, represents the Gaussian white noise sample used in the j-th layer, represents the operation on the local mean, represents the operation of averaging all perturbed version IMFs;

[0027] Step S254: Reconstruct the denoised price sequence. Specifically, eliminate the high-frequency terms among all the decomposed modal components, retain the intermediate-frequency and low-frequency IMFs and the trend residual term, and perform weighted reconstruction on the retained components to form the denoised economic crop price sequence.

[0028] Furthermore, in step S3, the selection of the target economic crop specifically includes the following steps:

[0029] Step S31: Construct an economic crop comprehensive evaluation model. Specifically, based on the multi-layer perceptron neural network structure, construct an economic crop comprehensive evaluation model; the economic crop comprehensive evaluation model has a dual-output structure, corresponding to two target economic crop selection decision indicators, namely the environmental adaptability score and the economic potential score respectively;

[0030] Step S32: Train the evaluation model. Specifically, use the reference economic crop selection data as the training data, input it into the economic crop comprehensive evaluation model, and perform model training on the model to obtain the trained economic crop comprehensive evaluation model;

[0031] Step S33: Calculate the comprehensive score. Specifically, input the target economic crop selection data of the current region into the trained economic crop comprehensive evaluation model to generate the environmental adaptability score and the economic potential score corresponding to the economic crop, and calculate the economic crop comprehensive score through a weighted fusion method ;

[0032] Step S34: Determine the target economic crop. Specifically, according to the economic crop comprehensive score, sort all candidate economic crops from high to low according to the score, and select the top k economic crops with the highest comprehensive score as the target economic crop set of the current region to obtain the target economic crop set.

[0033] Furthermore, in step S4, the cash crop market price forecast is used to predict the market price trend of the target cash crop in a specific future time period, identify the peak sale window of the cash crop, and provide a price driving basis for the planting strategy simulation; specifically, the following steps are included:

[0034] Step S41: Constructing a market price prediction model, specifically including the following steps:

[0035] Step S411: Establishing a time series coding structure for cash crop prices, specifically including the following steps:

[0036] Step S4111: Design an improved ILSTM network structure, specifically including forget gate calculation, input gate calculation, candidate state generation, state update, and output gate calculation; the input gate calculation is used to adjust the update amplitude of the current input information to the memory state, and the input gate output is translated by introducing an improved activation function to avoid signal collapse. The formula used is as follows:

[0037] ;

[0038] Where, represents the output of the input gate, represents the corresponding input gate weight matrix, represents the bias parameter of the input gate, Represents input data, Indicates the hidden state at the last moment. represents the sigmoid function, represents the hyperbolic tangent activation function, Represents the translation adjustment parameter of the nonlinear output of the input gate, and its value range is ;

[0039] Step S4112: Multi-layer time series coding output, specifically, multi-dimensional splicing of the denoised cash crop price series data, the corresponding cash crop transaction volume data, and the external transaction impact data to form the model input data tensor, which is sequentially input into the two-layer stacked improved ILSTM network structure for hierarchical coding processing to obtain the current time step hidden state of the second layer output ;

[0040] Step S412: Construct a price change key time point identification layer, specifically the hidden state of each time step Its previous state Splicing is performed, and through weighted calculation and nonlinear transformation, the time attention score is generated, and the scores of all time steps are normalized to generate the attention weights of each time position, and the hidden state of each time step is Perform weighted summation according to the corresponding weights to obtain the price change time identification vector;

[0041] Step S413: Construct a price influencing factor identification layer. Specifically, the feature hidden state at each time step is concatenated with the feature hidden state at the previous moment, and through weighted calculation and non-linear transformation, a feature attention score is generated. Then, the scores at all time steps are normalized to generate the attention weights of each feature. The feature hidden state at each time step is weighted and summed with the attention weights of the corresponding features to obtain a price influence identification vector;

[0042] Step S414: Construct an interval market price prediction output layer. Specifically, the price change time identification vector and the price influence identification vector are concatenated, and a fusion vector is generated through a non-linear mapping function. The fusion vector is input into a fully connected layer, and a multi-quantile price prediction strategy is introduced to perform interval prediction on the market price of the cash crop at future time points, obtaining the cash crop price prediction values corresponding to multiple quantiles;

[0043] Step S415: Design a price prediction loss function. Specifically, first, based on the quantile loss function, the deviation between each price prediction quantile and the actual observed price is calculated to generate the loss function values of each quantile. The loss function values in all the set quantile sets Q are weighted and integrated to generate a total loss function value, constituting the price prediction loss function;

[0044] Step S42: Train the price prediction model. Specifically, use the historical cash crop market price prediction data as training data, adopt the price prediction loss function as the supervised training objective function, and train the market price prediction model to obtain the trained market price prediction model;

[0045] Step S43: Predict the market price of the target cash crop. Specifically, select the transaction data and external transaction influence data corresponding to the target cash crop in the real-time cash crop price prediction data as real-time input data. Then, after optimizing and processing the original data, structured real-time price prediction data is generated. The structured real-time price prediction data is input into the trained market price prediction model in the preset input window format to obtain the multi-quantile price prediction results of the target cash crop within the future prediction time window. Based on the medium and high quantile prediction results, identify the time periods with upward potential in the future price trend, and label these time periods as the high-price interval windows of the target cash crop. Finally, take the multi-quantile price prediction results of the target cash crop and the corresponding high-price interval windows as the price prediction output results of the target cash crop.

[0046] Furthermore, in step S5, the optimization of the cash crop planting strategy specifically includes the following steps:

[0047] Step S51: Construct the planting strategy variable space. Specifically, by constructing the decision-making links in the whole planting process, the crop variety selection variable, sowing time variable, growth management intensity variable, and harvesting method variable are used to form the strategy variable optimization space S.

[0048] Step S52: Construct the strategy evaluation objective function. Specifically, for the strategy variable optimization space S, establish the revenue item function, risk item function, cost item function, and price matching degree item function respectively, and jointly construct the strategy evaluation objective function.

[0049] Step S53: Optimize the planting strategy search, which specifically includes the following steps:

[0050] Step S531: Initialize the search population. Specifically, take the strategy variable optimization space S as the search individual position coding space in the optimization algorithm, and perform particle population initialization operation on each variable vector using the hybrid cosine chaotic mapping function. The formula used is as follows:

[0051] ;

[0052] In the formula, represents the position of the th particle individual, represents the position of the th particle individual, and r represents a random number between 0 and 1.

[0053] Step S532: Calculate the individual fitness value. Specifically, calculate the fitness value f i of the particles in the particle swarm, and calculate the fitness value of the particle individual through the strategy evaluation objective function .

[0054] Step S533: Update the particles. Specifically, update the particle velocity and particle position.

[0055] Step S53: Particle mutation. Specifically, adopt a particle mutation mechanism based on the mutation intensity control factor to perform particle mutation on the selected M particles. The formula used is as follows:

[0056] ;

[0057] ;

[0058] In the formula, represents the mutation intensity coefficient of the e-th iteration, represents the parameter controlling the decline speed of the mutation intensity, represents the parameter controlling the decline shape of the mutation intensity, represents the fitness value of the worst particle in the current population, Denotes the fitness value of the optimal particle in the current population, Denotes a very small positive number to prevent the denominator from being zero, Denotes the number of particles to be mutated in this round, The position of the i-th particle after mutation, Denotes the random perturbation coefficient, which is a random number in the range of [0, 1], Denotes the i-th particle at the Position in the iteration, where e represents the current iteration number, Denotes the maximum number of iterations, Denotes the global optimal position of the particle, Denotes the mutation perturbation amplitude adjustment coefficient, which is used to control the proportional weight of the influence of the current particle individual fitness difference on the mutation intensity;

[0059] Step S535: Update the optimal position of the particle. Specifically, for all particles after particle update and mutation operations, re-evaluate their fitness values, and based on the fitness value of the current particle, compare it with its historical individual optimal value and the current population global optimal value respectively. If the current particle fitness is better, update the corresponding individual optimal position and global optimal position respectively; meanwhile, obtain the local optimal position of the particle individual for the next iteration and the global optimal position of the particle individual for the next iteration ;

[0060] Step S536: Terminate particle update. Specifically, when the particle fitness value f i is higher than the fitness threshold and when the maximum number of iterations is reached, terminate the search and obtain the global optimal position of the particle;

[0061] Step S54: Output the optimal planting strategy plan, which is used to determine the optimal full-cycle planting strategy configuration of cash crops. Specifically, extract the values of each dimension strategy variable from the global optimal position of the particle, and map them to the corresponding actual crop selection, sowing time, growth management intensity, and harvesting method respectively to obtain the optimal cash crop planting strategy combination.

[0062] Furthermore, in step S6, the intelligent management of cash crop planting specifically formulates a full-cycle planting plan including cash crop selection, cash crop sowing time, cash crop growth management, and cash crop harvesting arrangement according to the optimal cash crop planting strategy combination, and converts it into executable agricultural production operation instructions to achieve the dynamic matching and management optimization of the cash crop planting plan and market trends.

[0063] The technical solution adopted by the present invention is as follows: The economic crop planting management system based on artificial intelligence provided by the present invention includes a module for acquiring original planting management data, an original data optimization module, a target economic crop selection module, an economic crop market price prediction module, an economic crop planting strategy optimization module, and an intelligent management module for economic crop planting;

[0064] The module for acquiring original planting management data obtains the original planting management data through data collection operations, and sends the original planting management data to the original data optimization module;

[0065] The original data optimization module receives the data sent by the module for acquiring original planting management data, performs data cleaning, data standardization, data encoding, time windowing, and price data signal decomposition processing on the original planting management data to obtain optimized planting management data, and sends the optimized planting management data to the target economic crop selection module;

[0066] The target economic crop selection module receives the data sent by the original data optimization module. By constructing an economic crop comprehensive evaluation model based on a multi-layer perceptron structure and performing model training, it inputs the target economic crop selection data of the current region into the model, generates the environmental adaptability score and economic potential score of the economic crop, forms a comprehensive score based on weight fusion, selects the economic crop with the highest score according to the score ranking to obtain a set of target economic crops, and sends the set of target economic crops to the economic crop market price prediction module;

[0067] The economic crop market price prediction module receives the data sent by the target economic crop selection module. First, it performs time series modeling by designing an improved ILSTM network structure, and then identifies the key time positions and main influencing features of price changes respectively; after fusing the identification results, it inputs them into the quantile prediction output layer to output multiple quantile price prediction results within a future time period, completes the construction of the market price prediction model, designs a price prediction loss function based on the quantile loss function, combines historical data to supervise and train the model, and finally inputs real-time data into the trained prediction model to output the multi-quantile price prediction results and their future high price interval windows of the target economic crop, obtains the economic crop price prediction results, and sends the economic crop price prediction results to the economic crop planting strategy optimization module;

[0068] The economic crop planting strategy optimization module receives the data sent by the economic crop market price prediction module. By constructing a planting strategy variable space that includes crop variety selection, sowing time, growth management intensity, and harvesting method, a multi-objective strategy evaluation function that takes into account expected revenue, market risk, planting cost, and price matching degree is established. An improved particle swarm optimization algorithm that combines a hybrid cosine chaotic mapping initialization and a mutation intensity control mechanism is used to search for the optimal strategy configuration, and the optimal economic crop planting strategy combination is obtained.

[0069] The economic crop intelligent management module receives the data from the economic crop planting strategy optimization module. Through the optimal economic crop planting strategy combination, corresponding agricultural operation instructions are generated to achieve the whole-process management configuration from crop variety selection to harvesting.

[0070] The beneficial effects achieved by the present invention using the above solution are as follows:

[0071] (1) Aiming at the technical problems in the traditional economic crop planting management method, such as the lack of scientific basis for economic crop selection, the lag in market price response, and the difficulty in matching price high points, which lead to unstable income and high market mismatch risk in economic crop planting management. This solution innovatively constructs a method system that takes market price prediction as the guide and integrates intelligent screening of target crops and optimization of the whole-cycle planting strategy. It effectively improves the scientificity and suitability of economic crop target selection, accurately identifies future high-price windows, realizes price-driven planting decisions, supports personalized strategy recommendations according to local conditions, adapts to various crops and regional environments, and finally realizes price-based decision-making in the process of economic crop planting, improving the stability of economic crop planting income.

[0072] (2) Aiming at the technical problems in the processing of price time-series data in the traditional economic crop planting management method, such as poor robustness to high-frequency noise in the price sequence, unstable modal decomposition process, and unsatisfactory trend extraction effect, which lead to strong volatility and overfitting in the input data of the subsequent price prediction model, resulting in large deviations and poor reliability in the price prediction results. This solution innovatively introduces a mechanism for constructing multiple groups of noise-added perturbation signals and a hierarchical progressive modal extraction strategy, significantly improving the stability and decomposition accuracy of the modal decomposition algorithm under high-frequency noise interference, realizing the fine extraction of predictable medium- and low-frequency trend signals in the economic crop price sequence, significantly improving the quality of the input data for economic crop market price prediction, and effectively improving the accuracy of economic crop market price prediction.

[0073] (3) Aiming at the technical problems existing in the existing market price prediction models applicable to cash crops, such as insufficient learning of complex time-series price fluctuation characteristics and lack of an effective identification mechanism for key time points and main influencing factors of price changes, which affects the obvious deficiencies in the accuracy and stability of cash crop price prediction results. This solution innovatively constructs a multi-quantile price prediction model that integrates improved ILSTM and introduces time attention and feature attention mechanisms. By introducing the improved ILSTM structure, the model's ability to express complex time-series dependencies and non-linear fluctuation trends in cash crop price sequences is significantly improved. By introducing time attention and feature attention mechanisms, dual identification of key time points of price changes and dominant influencing factors is achieved, effectively depicting the deep logic of price evolution over time and the dynamic effects of multi-source factors. Combining the multi-quantile output strategy, the upper and lower limit estimates and central trend predictions of prices can be output synchronously, improving the accuracy and stability of the model prediction results.

[0074] (4) Aiming at the technical problems existing in the algorithms used for optimizing cash crop planting strategies, such as being prone to falling into local optimal solutions and having insufficient search space exploration ability, which leads to the lack of global optimality in the generated optimal planting strategy combinations. This solution innovatively introduces an improved particle swarm optimization algorithm with a hybrid cosine chaos mapping initialization and mutation intensity control mechanism to search for the optimal cash crop planting strategy, enhancing the diversity of the initial population and search guidance, and enhancing the global search ability and the ability to jump out of local extreme values through dynamic mutation operations, thereby achieving efficient global optimization of cash crop planting strategy combinations and finally obtaining a cash crop planting strategy plan that better fits the market price trend and has better revenue potential. Brief Description of the Drawings

[0075] Figure 1 It is a schematic flowchart of the method for managing cash crop planting based on artificial intelligence provided by the present invention;

[0076] Figure 2 It is a schematic diagram of the modules of the system for managing cash crop planting based on artificial intelligence provided by the present invention;

[0077] Figure 3 It is a schematic flowchart of step S2;

[0078] Figure 4 It is a schematic flowchart of step S25;

[0079] Figure 5 It is a schematic flowchart of step S3;

[0080] Figure 6 It is a schematic flowchart of step S4;

[0081] Figure 7 It is a schematic flowchart of step S5;

[0082] Figure 8 is a flow chart of step S41;

[0083] 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

[0084] 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.

[0085] In the description of the present invention, it should be understood that terms such as "up", "down", "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. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the system or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they should not be understood as limiting the present invention.

[0086] Example 1, see Figure 1 The technical solution adopted by the present invention is as follows: The economic crop planting management method based on artificial intelligence provided by the present invention comprises the following steps:

[0087] Step S1: obtaining original planting management data, specifically by collecting data to obtain the original planting management data;

[0088] Step S2: optimizing the original data, specifically, performing data cleaning, data standardization, data coding, time windowing, and price data signal decomposition on the original planting management data to obtain planting management optimization data;

[0089] Step S3: Target cash crop selection, specifically by constructing a comprehensive cash crop evaluation model based on a multi-layer perceptron structure and conducting model training, inputting the target cash crop selection data for the current region into the model, generating environmental adaptability scores and economic potential scores for the cash crops, and forming a comprehensive score based on weight fusion. The cash crop with the highest score is selected based on the score sorting to obtain the target cash crop set;

[0090] Step S4: Prediction of the market price of cash crops. Specifically, first, perform time series modeling through the improved ILSTM network structure design, and then identify the key time positions and main influencing features of price changes respectively; after fusing the identification results, input them into the quantile prediction output layer to output multiple quantile price prediction results within the future time period, complete the construction of the market price prediction model, design the price prediction loss function based on the quantile loss function, and perform supervised training on the model in combination with historical data. Finally, input the real-time data into the trained prediction model to output the multi-quantile price prediction results of the target cash crop and its future high-price interval window, obtaining the cash crop price prediction results;

[0091] Step S5: Optimization of the cash crop planting strategy. Specifically, by constructing a planting strategy variable space including crop variety selection, sowing time, growth management intensity, and picking method, establish a multi-objective strategy evaluation function considering expected revenue, market risk, planting cost, and price matching degree, and use an improved particle swarm optimization algorithm integrating the initialization of the hybrid cosine chaotic mapping and the mutation intensity control mechanism to search for the optimal strategy configuration, obtaining the optimal cash crop planting strategy combination;

[0092] Step S6: Intelligent management of cash crop planting. Specifically, through the optimal cash crop planting strategy combination, generate corresponding agricultural operation instructions to realize the whole-process management configuration from crop variety selection to harvesting.

[0093] By performing the above operations, aiming at the technical problems of lack of scientific basis for cash crop selection, lag in market price response, and difficulty in matching price highs in the traditional cash crop planting management method, resulting in unstable planting management benefits and high market mismatch risks of cash crops, this solution innovatively constructs a method system oriented by market price prediction, integrating intelligent screening of target crops and optimization of the whole-cycle planting strategy, effectively improving the scientificity and suitability of cash crop target selection, accurately identifying future high-price windows, realizing price-driven planting decisions, supporting personalized strategy recommendations according to local conditions, adapting to multiple crops and regional environments, and finally realizing price-based decision-making in the cash crop planting process, improving the stability of the planting benefits of cash crops.

[0094] Embodiment 2, refer to Figure 1 and Figure 2, this embodiment is based on the above embodiment. In step S1, the acquisition of the original planting management data is used to obtain the original data set for the intelligent planting management of cash crops. Specifically, data collection is carried out through various methods of the agricultural big data platform and the agricultural product trading platform to obtain the original planting management data. The original planting management data includes reference cash crop selection data, target cash crop selection data, historical cash crop market price prediction data, and real-time cash crop market price prediction data. The reference cash crop selection data and the target cash crop selection data both include regional planting condition data, cash crop planting attribute data, and cash crop economic attribute data. The reference cash crop selection data also includes the reference cash crop selection result. The historical cash crop price prediction data and the real-time cash crop price prediction data both include various cash crop trading data and external trading impact data. The regional planting condition data includes soil type, temperature, precipitation, humidity, sowing time, altitude, and light. The cash crop planting attribute data includes cash crop variety, growth cycle, regulation ability, and water requirement. The cash crop economic attribute data includes crop planting cost, sales volume, price volatility, average price, and price seasonal fluctuation degree. The reference cash crop selection result includes cash crop yield, cash crop quality, planting failure rate, average cash crop sales price, and evaluated cash crop sales volume. The various cash crop trading data includes cash crop price sequence data and cash crop trading volume data. The external trading impact data includes holiday data, meteorological disaster record data, cash crop market policies, and market supply and demand data.

[0095] Embodiment 3, refer to Figure 1 , Figure 3 and Figure 4 , this embodiment is based on the above embodiment. In step S2, the optimization of the original data is used to perform optimization processing on the original data. Specifically, data cleaning processing, data standardization processing, data encoding processing, time windowing, and price data signal decomposition processing are performed on the original data to obtain the optimized planting management data. It includes the following steps:

[0096] Step S21: Data cleaning processing is used to ensure the integrity and consistency of the original data and remove invalid and incorrect data. Specifically, missing value filling, outlier removal, and field normalization processing are performed on the original data. The missing value filling is used to achieve the integrity of the data. Specifically, the numerical values of the missing data items in the original data are complemented by the mean filling method. The outlier removal is used to control the rationality of the data distribution. Specifically, the extreme values and logical outliers in the original data are detected and removed through the Z-Score algorithm. The field normalization processing is used to ensure the consistency of the model input. Specifically, different formats of data are standardized through a unified unit conversion rule to ensure the consistency of the model input.

[0097] Step S22: Data normalization processing, which is used to unify the numerical scales of numerical features. Specifically, the numerical data in the original data is normalized by the min-max normalization method;

[0098] Step S23: Data encoding processing, which is used to convert categorical and symbolic data into numerical vectors. Specifically, the one-hot encoding method is used to encode the categorical fields in the original data, and the discrete text and label variables are converted into sparse numerical vectors;

[0099] Step S24: Dynamic time window partitioning, which is used to convert the original time series data into a sliding time window structure required for prediction, and to achieve market price prediction after a complete growth cycle. Specifically, through the sliding window construction algorithm, the time series data related to the target cash crop is divided into an input window and a prediction window;

[0100] The input window is used to construct the input time series data of the model. Specifically, based on the current time point, the price prediction data of the cash crop within a continuous number of time steps is selected and slid forward to form a complete input time data series. The time step is in units of natural weeks, and the sliding time length forward can be set between 2 and 4 weeks, which is set according to the growth cycle of the cash crop variety;

[0101] The prediction window is used to predict the market price trend within a specific time period before and after the crop matures. Specifically, it is a bidirectional sliding time window formed by sliding forward and backward a set number of time steps centered on the standard maturity time point of the cash crop, covering the price change trend range within the adjustable range; The time step is in units of natural weeks, and the window range is set to between 1 and 6 weeks before and after according to the controllable maturity cycle of the cash crop;

[0102] Step S25: Price data signal decomposition processing, which is used for multi-scale modal decomposition and high-frequency denoising reconstruction processing of the cash crop price sequence data, and extracting mid-low frequency trend signals with predictable value; Specifically, by introducing a mechanism for constructing multiple groups of noisy perturbation signals and designing a hierarchical and progressive modal extraction strategy to improve the empirical mode decomposition algorithm, multi-level modal decomposition and high-frequency information elimination of the cash crop price time series are realized, and the denoised cash crop price sequence data is obtained, including the following steps:

[0103] Step S251: Noisy signal construction processing, which is used to construct multiple groups of perturbed versions of the price signal sequence to enhance the robustness and stability of subsequent modal decomposition to high-frequency noise; Specifically, Gaussian white noise is injected into the original cash crop price sequence, and combined with the first-order modal decomposition operation function for perturbation modulation to construct multiple groups of noisy price signals. The formula used is as follows:

[0104] ;

[0105] Where, represents the value of the i-th group of disturbance signal sequence at time step t, represents the value of the original cash crop price series data at time step t, represents the noise adjustment coefficient used for the ith disturbance, represents the first-layer modal decomposition operation function, represents the Gaussian white noise sample introduced at time step t in the i-th disturbance, which conforms to the standard normal distribution with mean zero and variance 1;

[0106] Step S252: Initial residual extraction, specifically performing a modal decomposition operation on each group of noisy signals to generate modal components of each group, and then performing a local mean operation to average the local mean results of all perturbation versions to form a stable first-layer residual signal. ,Will and Perform difference operation to obtain the first layer modal component; the formula used is as follows

[0107] ;

[0108] ;

[0109] Where, Represents the local mean operation, represents the modal decomposition function, Indicates the average operation of all perturbation versions of IMF, represents the first layer modal component, represents the first layer residual signal;

[0110] Step S253: Multiple rounds of modal iterative decomposition are used to continue extracting deeper mid- and low-frequency modal signals based on the initial residual. Specifically, disturbance noise is added to each layer of residual and modal decomposition is performed to extract the modal components and residual signals of the current layer. The formula used is as follows:

[0111] ;

[0112] ;

[0113] Where, represents the residual signal of the jth layer, Indicates the The residual component of the layer, Indicates the The noise perturbation adjustment coefficient of the layer, represents the modal decomposition function of the j-th layer, denotes the Gaussian white noise samples used in the j-th layer, denotes the modal component of the j-th layer;

[0114] Step S254: Reconstruct the denoised price sequence to remove high-frequency modal terms and retain medium-frequency, low-frequency, and trend terms; specifically, remove the high-frequency terms from all the decomposed modal components, retain the medium-frequency and low-frequency IMFs and the trend residual term, and perform weighted reconstruction on the retained components to form the denoised economic crop price sequence. The formula used is as follows:

[0115] ;

[0116] In the formula, denotes the denoised economic crop price sequence, denotes the final residual signal, k denotes the cut-off layer number of the high-frequency mode, k = 3, and n denotes the total number of extracted IMF layers.

[0117] By performing the above operations, aiming at the technical problems existing in the processing of price time series data in the traditional economic crop planting management method, such as poor robustness to high-frequency noise in the price sequence, instability in the modal decomposition process, and unsatisfactory trend extraction effect, which lead to strong volatility and overfitting in the input data of the subsequent price prediction model, resulting in large deviations and poor reliability in the price prediction results. This solution innovatively introduces a mechanism for constructing multiple groups of noise-added perturbation signals and a hierarchical progressive modal extraction strategy, significantly improving the stability and decomposition accuracy of the modal decomposition algorithm under high-frequency noise interference, realizing the fine extraction of predictable medium- and low-frequency trend signals in the economic crop price sequence, significantly improving the quality of the input data for economic crop market price prediction, and effectively improving the accuracy of economic crop market price prediction.

[0118] Example 4, refer to Figure 1 , Figure 2 and Figure 5 Based on the above example, in step S3, the selection of the target economic crop is used to screen out the target economic crop that is suitable for planting and has good market potential under the current regional environmental conditions from multiple alternative economic crops. The specific steps are as follows:

[0119] Step S31: Construct an economic crop comprehensive evaluation model to establish a scoring model that can simultaneously evaluate the environmental adaptability and market economic potential of each alternative economic crop. Specifically, construct an economic crop comprehensive evaluation model based on the multi-layer perceptron neural network structure; the economic crop comprehensive evaluation model is a dual-output structure, corresponding to two target economic crop selection decision indicators, namely the environmental adaptability score and the economic potential score respectively.

[0120] Step S32: Evaluation model training, which is used to supervise and train the comprehensive evaluation model of cash crops to obtain a scoring model with reliable output ability. Specifically, the reference cash crop selection data is used as training data and input into the comprehensive evaluation model of cash crops for model training to obtain the trained comprehensive evaluation model of cash crops;

[0121] The model training is specifically to construct a supervision label, use the supervision label as the target output of the model, and use the mean square error loss function as the optimization objective function, and combine the backpropagation algorithm to adjust the network parameters;

[0122] The supervision label is jointly composed of an environmental adaptability scoring label and an economic potential scoring label;

[0123] The environmental adaptability scoring label is specifically composed by weighted fusion according to the cash crop yield, cash crop quality and planting failure rate, and the economic potential scoring label is weighted and composed according to the average selling price of cash crops and the evaluated sales volume of cash crops;

[0124] Step S33: Calculate the comprehensive score. Specifically, the target cash crop selection data of the current region is input into the trained comprehensive evaluation model of cash crops to generate the environmental adaptability score and economic potential score corresponding to the cash crops, and the comprehensive score of the cash crops is calculated through weighted fusion ; The formula used is as follows:

[0125] ;

[0126] In the formula, represents the comprehensive score of the cash crop , represents the weight parameter of the environmental adaptability of the cash crop, represents the weight parameter of the economic potential of the cash crop, represents the cash crop 's environmental adaptability score, represents the cash crop 's economic potential score;

[0127] Step S34: Target cash crop determination, which is used as the input basis for subsequent price prediction and planting strategy optimization. Specifically, according to the comprehensive score of the cash crops, all candidate cash crops are sorted from high to low according to the score, and the top k cash crops with the highest comprehensive score are selected as the target cash crop set in the current region to obtain the target cash crop set.

[0128] Example Five, refer to Figure 1 、 Figure 2 、 Figure 6 and Figure 8, this embodiment is based on the above embodiment. In step S4, the economic crop market price prediction is used to predict the market price trend of the determined target economic crop within a specific future time interval, identify the peak window for the sale of economic crops, and provide a price-driven basis for the planting strategy simulation; specifically, it includes the following steps:

[0129] Step S41: Construct a market price prediction model for performing time series modeling on the economic crop market price and its related influencing factor sequences, and constructing a neural network coding structure with long-term dependence memory ability and non-linear modeling ability; specifically, it includes the following steps:

[0130] Step S411: Establish an economic crop price time series coding structure, specifically including the following steps:

[0131] Step S4111: Design an improved ILSTM network structure for constructing an improved long short-term memory network suitable for predicting economic crop price sequences, specifically including forget gate calculation, input gate calculation, candidate state generation, state update, and output gate calculation; the forget gate calculation, candidate state generation, state update, and output gate calculation all follow the calculation methods of the standard LSTM network structure, and only the input gate calculation part is structurally improved; the input gate calculation is used to adjust the update amplitude of the current input information to the memory state, and by introducing an improved activation function, a translation transformation is performed on the output of the input gate to avoid signal collapse. The formula used is as follows:

[0132] ;

[0133] In the formula, represents the output of the input gate, represents the corresponding input gate weight matrix, represents the bias term parameter of the input gate, represents the input data, represents the hidden state at the previous moment, represents the sigmoid function, represents the hyperbolic tangent activation function, represents the translation adjustment parameter of the non-linear output of the input gate, and its value range is ;

[0134] Step S4112: Multilayer time series coding output, specifically, the denoised economic crop price sequence data, the corresponding economic crop trading volume data, and the external trading influence data are multi-dimensionally spliced to form a model input data tensor, which is sequentially input into two stacked improved ILSTM network structures for hierarchical coding processing to obtain the hidden state of the current time step output by the second layer ; The formula used is as follows:

[0135] ;

[0136] ;

[0137] In the formula, represents the hidden state of the current time step of the first-layer output, and extracts the short-term change pattern of the economic crop price, represents the hidden state of the current time step of the second-layer output, and is further integrated into the medium- and long-term trend, represents the model input data composed of the economic crop price series data, economic crop trading volume data, and external trading impact data after denoising;

[0138] Step S412: Construct a key time point identification layer for price change to identify the time positions in the economic crop price series that have a key impact on future price prediction, specifically the hidden state at each time step is concatenated with its previous moment state and through weighted calculation and non-linear transformation, generate a time attention score, and normalize the scores of all time steps to generate the attention weights of each time position. Weighted sum the hidden state of each time step according to the corresponding weights to obtain a price change time identification vector; the formula used is as follows:

[0139] ;

[0140] ;

[0141] ;

[0142] In the formula, represents the time attention score at the t-th time step, represents the attention weight at the t-th time step, T represents the time step length, represents the weight vector used to project the intermediate layer into a single score value, represents the weight matrix, represents the bias term parameter, represents the concatenation operation, represents the price change time identification vector at the current time step;

[0143] Step S413: Construct a price influencing factor identification layer. Specifically, the feature hidden state at each time step is concatenated with its previous moment feature hidden state, and through weighted calculation and non-linear transformation, generate a feature attention score, and normalize the scores of all time steps to generate the attention weights of each feature. Weighted sum the feature hidden state of each time step with the attention weights of the corresponding features to obtain a price influence identification vector; the formula used is as follows:

[0144] ;

[0145] ;

[0146] ;

[0147] In the formula, represents the i-th feature attention score at the t-th time step, represents the feature attention weight vector, represents the feature attention weight matrix, represents the bias term parameter for feature attention calculation, represents the i-th feature attention weight at the t-th time step, F represents the number of features, represents the price impact identification vector, represents the previous moment state of the i-th feature, represents the current moment state of the i-th feature;

[0148] Step S414: Construct an interval-based market price prediction output layer for realizing the deep fusion modeling of when the economic crop price changes and why the economic crop price changes. Specifically, splice the price change time identification vector and the price impact identification vector, generate a fusion vector through a non-linear mapping function, input the fusion vector into a fully connected layer, introduce a multi-quantile price prediction strategy, and perform interval prediction on the market price of the economic crop at a future time point to obtain the economic crop price prediction values corresponding to multiple quantiles; the formula used is as follows:

[0149] ;

[0150] In the formula, represents the economic crop price prediction result at the current time step, represents the fully connected layer, represents the weight matrix for vector fusion, represents the bias term parameter for vector fusion, represents that the output result is a real vector with a dimension of Q, corresponding to the price estimation values under different confidence levels, Q represents the set of quantiles, which is , the lower quantile represents the pessimistic price estimate, representing the price lower limit; the middle quantile represents the neutral price estimate, representing the central prediction trend of the price; the upper quantile represents the optimistic price estimate, representing the price upper limit;

[0151] Step S415: Design a price prediction loss function to achieve interval modeling for price prediction and better adapt to the volatility and risk of economic crop prices. Specifically, first calculate the error of the deviation between each price prediction quantile and the actual observed price based on the quantile loss function to generate the loss function values for each quantile. Then, weight and integrate the loss function values in all the set quantile sets Q to generate the total loss function value, thus constituting the price prediction loss function. The formula used is as follows:

[0152] ;

[0153] ;

[0154] In the formula, represents the loss function value of this quantile, represents the predicted value of the economic crop price at this quantile, represents the true value of the economic crop price, represents the quantile index, represents the loss weight coefficient for each quantile;

[0155] Step S42: Train the price prediction model. Specifically, use the historical economic crop market price prediction data as the training data, adopt the price prediction loss function as the supervised training objective function, and train the market price prediction model to obtain the trained market price prediction model. The model training is specifically to perform backpropagation calculation based on the total loss function value, gradually adjust the internal parameters of the model, and use an optimizer to iteratively update the parameters in the model. When the loss function converges or reaches the set number of training rounds, the model training process is completed;

[0156] Step S43: Predict the market price of the target economic crop, which is used to perform real-time interval prediction on the market price trend of the current target economic crop and identify its future high-price time periods. Specifically, select the transaction data and external transaction impact data corresponding to the target economic crop in the real-time economic crop price prediction data as the real-time input data. Then, after optimizing and processing the original data, generate structured real-time price prediction data. Input the structured real-time price prediction data into the trained market price prediction model in the preset input window format to obtain the multi-quantile price prediction results of the target economic crop within the future prediction time window. Based on the medium and high quantile prediction results, identify the time periods with upward potential in the future price trend, and label this time period as the high-price interval window of the target economic crop. Finally, use the multi-quantile price prediction results of the target economic crop and the corresponding high-price interval window as the price prediction output results of the target economic crop.

[0157] By performing the above operations, aiming at the technical problems existing in the existing market price prediction model for cash crops, such as insufficient learning of complex time-series price fluctuation characteristics and lack of an effective identification mechanism for key time points and main influencing factors of price changes, which affect the obvious deficiencies in the accuracy and stability of cash crop price prediction results. This solution innovatively constructs a multi-quantile price prediction model that integrates improved ILSTM and introduces time attention and feature attention mechanisms. By introducing the improved ILSTM structure, the model's ability to express complex time-series dependencies and non-linear fluctuation trends in the cash crop price sequence is significantly improved. By introducing time attention and feature attention mechanisms, dual identification of key time points of price changes and dominant influencing factors is achieved, effectively depicting the deep logic of price evolution over time and the dynamic effects of multi-source factors. Combining with the multi-quantile output strategy, it can synchronously output the upper and lower limit estimates and central trend predictions of prices, improving the accuracy and stability of the model prediction results.

[0158] Example 6, refer to Figure 1 、 Figure 2 and Figure 7 Based on the above embodiments, further, in step S5, the optimization of the cash crop planting strategy is used to generate a full-cycle cash crop planting plan for future high-price windows, specifically including the following steps:

[0159] Step S51: Construct a planting strategy variable space for defining various key controllable strategy variables in the full-cycle production process of cash crops, and construct an optimization search space for planting strategy combinations. Specifically, by constructing the decision-making links in the whole planting process, the crop variety selection variable, sowing time variable, growth management intensity variable, and picking method variable are used to form the strategy variable optimization space S. The formula is as follows:

[0160] ;

[0161] In the formula, represents the strategy variable optimization space, represents the crop variety selection variable, which is used to determine the optimal planting variety in the target cash crop set. The value of the crop variety selection variable is selected from the target cash crop set; represents the sowing time variable, which is used to determine the actual starting week number of crop sowing to achieve the maximum overlap between the maturity period and the market high-price interval. The value of the sowing time variable is the time period from 3 weeks before to 3 weeks after the regular sowing period of the crop, and the time unit is in weeks; represents the growth management intensity variable, which is used to adjust the input level of field management. Different intensities affect crop quality and input costs. The value of the field management variable includes low intensity, normal intensity, and high intensity. High-intensity management can improve crop quality and stability as well as high cost, and at the same time can accelerate the maturity time of cash crops; Indicates the picking method variable, which is used to determine the picking method and time arrangement, and hits the predicted high-price range. The picking method variable takes values of normal picking, delayed picking, and early picking, and each picking method corresponds to a different picking ratio;

[0162] Step S52: Construct a strategy evaluation objective function, which is used to comprehensively evaluate the value of the economic crop full-cycle planting strategy combination, and improve the high-price window matching degree and the robustness of the actual income. Specifically, for the strategy variable optimization space S, establish a revenue term function, a risk term function, a cost term function, and a price matching degree term function respectively, and jointly construct a strategy evaluation objective function; the formula used is as follows:

[0163] ;

[0164] ;

[0165] ;

[0166] ;

[0167] In the formula, Indicates the revenue objective function, Indicates the revenue term function, Indicates the risk term function, Indicates the cost term function, Indicates the price matching degree term function, Indicates the picking cycle, Indicates the Predicted value of the high-price economic crop price in the Indicates the Predicted value of the medium-price economic crop price in the Indicates the quantile fusion weight coefficient, Indicates the Picking quantity of the economic crop in the Indicates the Predicted value of the low-price economic crop price in the Indicates the risk weight coefficient, Indicates the misalignment distance between the picking period and the high-price prediction interval, Indicates the high-price interval window of the economic crop market price prediction;

[0168] Step S53: Optimize the planting strategy search, which is used to search for the optimal economic crop full-cycle planting strategy combination in the constructed multi-dimensional planting strategy variable space based on the set strategy evaluation objective function; specifically includes the following steps:

[0169] Step S531: Initialize the search population. Specifically, use the strategy variable optimization space S as the search individual position coding space in the optimization algorithm, and for each variable vector Perform particle population initialization using the hybrid cosine chaotic mapping function; the formula used is as follows:

[0170] ;

[0171] In the formula, represents the position of the particle individual, represents the position of the particle individual, r represents a random number between 0 and 1;

[0172] Step S532: Calculate the individual fitness value. Specifically, calculate the fitness value f i of the particles in the particle swarm, and calculate the fitness value of the particle individual through the strategy evaluation objective function ;

[0173] Step S533: Update the particles. Specifically, update the particle velocity and particle position; the formula used is as follows:

[0174] ;

[0175] In the formula, represents the velocity of the i-th particle in the e-th iteration, represents the velocity of the i-th particle in the e-th iteration, represents the position of the i-th particle in the e-th iteration, represents the local optimal position of the particle individual, represents the global optimal position of the particle, and represent random numbers in the range [0, 1], represents the individual learning factor, which is used to control the speed of the particle moving towards the individual optimal position, represents the group learning factor, which is used to control the speed of the particle moving towards the global optimal position, represents the position of the i-th particle in the e-th iteration, e represents the current iteration number, represents the inertia factor in the e-th iteration;

[0176] Step S534: Particle mutation, which is used to solve the problem that the particle swarm falls into a local optimal solution and oscillates near the global optimum. Specifically, adopt a particle mutation mechanism based on the mutation intensity control factor to perform particle mutation on the selected M particles; the formula used is as follows:

[0177] ;

[0178] ;

[0179] ;

[0180] In the formula, represents the mutation intensity coefficient of the e-th iteration, represents the parameter for controlling the decline rate of the mutation intensity, represents the parameter for controlling the decline shape of the mutation intensity, represents the fitness value of the worst particle in the current population, represents the fitness value of the best particle in the current population, represents a very small positive number to prevent the denominator from being zero; represents the number of particles to be mutated in this round, represents the total number of particles in the population, represents the maximum mutation rate, the position of the i-th particle after mutation, represents the random perturbation coefficient, which is a random number within the range of [0, 1], represents the mutation perturbation amplitude adjustment coefficient, which is used to control the proportional weight of the influence of the fitness difference of the current particle individual on the mutation intensity;

[0181] Step S535: Update the optimal position of the particle. Specifically, for all particles after particle update and mutation operations, re-evaluate their fitness values, and based on the fitness value of the current particle, compare it with its historical individual optimal value and the current population global optimal value respectively. If the fitness of the current particle is better, update the corresponding individual optimal position and global optimal position respectively; meanwhile, obtain the local optimal position of the particle individual for the next iteration and the global optimal position of the particle individual for the next iteration ;

[0182] Step S536: Terminate the particle update. Specifically, when the particle fitness value f i is higher than the fitness threshold and when the maximum number of iterations is reached, terminate the search and obtain the global optimal position of the particle;

[0183] Step S54: Output the optimal planting strategy plan for determining the optimal full-cycle planting strategy configuration of cash crops. Specifically, extract the values of each dimension of the strategy variables from the global optimal position of the particle, and map them to the corresponding actual crop selection, sowing time, growth management intensity, and picking method respectively to obtain the optimal cash crop planting strategy combination.

[0184] By performing the above operations, in the algorithms currently used for optimizing economic crop planting strategies, there are technical problems such as being prone to falling into local optimal solutions and having insufficient search space exploration ability, which in turn leads to the lack of global optimality in the generated optimal planting strategy combinations. This solution innovatively introduces an improved particle swarm optimization algorithm with a hybrid cosine chaos mapping initialization and mutation intensity control mechanism to search for the optimal economic crop planting strategy, enhancing the diversity of the initial population and the search guidance, and enhancing the global search ability and the ability to jump out of local extreme values through dynamic mutation operations, thereby achieving efficient global optimization of the economic crop planting strategy combination and finally obtaining an economic crop planting strategy plan that better fits the market price trend and has better revenue potential.

[0185] Example VII. Refer to Figure 1 and Figure 2 In this example, based on the above example, in step S6, for the intelligent management of economic crop planting, specifically according to the optimal economic crop planting strategy combination, a full-cycle planting plan including the selection of economic crops, the sowing time of economic crops, the growth management of economic crops, and the picking arrangement of economic crops is formulated, and it is converted into executable agricultural production operation instructions to achieve the dynamic matching and management optimization of the economic crop planting plan and the market trend.

[0186] Example VIII. Refer to Figure 1 and Figure 2 In this example, based on the above example, the technical solution adopted by the present invention is as follows: The economic crop planting management system based on artificial intelligence provided by the present invention includes a module for obtaining original planting management data, an original data optimization module, a target economic crop selection module, an economic crop market price prediction module, an economic crop planting strategy optimization module, and an economic crop planting intelligent management module;

[0187] The module for obtaining original planting management data obtains original planting management data through data collection operations and sends the original planting management data to the original data optimization module;

[0188] The original data optimization module receives the data sent by the module for obtaining original planting management data, performs data cleaning, data standardization, data encoding, time windowing, and price data signal decomposition processing on the original planting management data to obtain optimized planting management data, and sends the optimized planting management data to the target economic crop selection module;

[0189] The target cash crop selection module receives the data sent by the original data optimization module. By constructing a comprehensive evaluation model of cash crops based on the multi-layer perceptron structure and conducting model training, it inputs the target cash crop selection data of the current region into the model, generates the environmental adaptability scores and economic potential scores corresponding to the cash crops, forms a comprehensive score by weight fusion, selects the cash crop with the highest score according to the score ranking to obtain the target cash crop set, and sends the target cash crop set to the cash crop market price prediction module;

[0190] The cash crop market price prediction module receives the data sent by the target cash crop selection module. First, it conducts time series modeling processing by designing an improved ILSTM network structure, and then respectively identifies the key time positions and main influencing features of price changes; after fusing the identification results, it inputs them into the quantile prediction output layer to output multiple quantile price prediction results within the future time period, completes the construction of the market price prediction model, designs a price prediction loss function based on the quantile loss function, combines historical data to supervise and train the model, and finally inputs real-time data into the trained prediction model to output the multi-quantile price prediction results of the target cash crops and their future high price interval windows, obtains the cash crop price prediction results, and sends the cash crop price prediction results to the cash crop planting strategy optimization module;

[0191] The cash crop planting strategy optimization module receives the data sent by the cash crop market price prediction module. By constructing a planting strategy variable space including crop selection, sowing time, growth management intensity, and harvesting methods, it establishes a multi-objective strategy evaluation function considering expected revenue, market risk, planting cost, and price matching degree, and uses an improved particle swarm optimization algorithm integrating the initialization of the hybrid cosine chaotic mapping and the mutation intensity control mechanism to search for the optimal strategy configuration to obtain the optimal cash crop planting strategy combination;

[0192] The cash crop planting intelligent management module receives the data of the cash crop planting strategy optimization module. Through the optimal cash crop planting strategy combination, it generates corresponding agricultural operation instructions to realize the whole process management configuration from crop selection to harvesting.

[0193] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0194] Although embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention.

[0195] The present invention and its embodiments have been described above. Such description is not restrictive. What is shown in the drawings is only one of the embodiments of the present invention, and the actual structure is not limited thereto. Generally speaking, if those of ordinary skill in the art are inspired by it and, without departing from the purpose of the present invention, design similar structural modes and embodiments to this technical solution without creative efforts, they shall fall within the protection scope of the present invention.

Claims

1. An economic crop planting management method based on artificial intelligence, characterized in that: The method includes the following steps: Step S1: Obtain the original planting management data. Through data collection operations, obtain the original planting management data; Step S2: Optimize the original data. Successively perform data cleaning, standardization processing, coding conversion, and windowing on the original planting management data, and introduce multiple groups of noise-adding perturbation signal construction mechanisms and hierarchical progressive mode extraction strategies to perform price data signal decomposition processing on the price data to obtain optimized planting management data; Step S3: Select target cash crops. By constructing a comprehensive evaluation model for cash crops and conducting model training, input the data of the current region into the model, output the environmental adaptability score and economic potential score, and form a comprehensive score based on weight fusion, and then sort and select the target cash crops accordingly to obtain a set of target cash crops; Step S4: Forecast the market price of cash crops. Design an improved ILSTM network structure, and introduce time attention and feature attention mechanisms to identify the key time points of price changes and the main influencing factors respectively. Combine the multi-quantile prediction output layer to construct a market price prediction model. Design a price prediction loss function based on the quantile loss function, conduct model training, input real-time data into the trained model, and output the multi-quantile price prediction results of the target cash crops and their future high-price intervals to obtain the cash crop price prediction results; Step S5: Optimize the planting strategy of cash crops. By constructing a planting strategy variable space, establish a multi-objective strategy evaluation function, and use an improved particle swarm optimization algorithm that combines a hybrid cosine chaotic mapping initialization and mutation intensity control mechanism to search for the optimal strategy configuration to obtain the optimal cash crop planting strategy combination; Step S6: Intelligent management of cash crop planting. Generate corresponding agricultural operation instructions through the optimal cash crop planting strategy combination to achieve the whole-process management of cash crops.

2. The method for managing the cultivation of cash crops based on artificial intelligence according to claim 1, characterized in that: In step S2, the optimization of the original data specifically includes the following steps: Step S21: Data cleaning process, specifically filling missing values, removing outliers, and normalizing fields in the original data; Step S22: Data standardization process, specifically normalizing the numerical data in the original data through the min-max normalization method; Step S23: Data coding process, specifically encoding the categorical fields in the original data using the one-hot encoding method to convert discrete text and label variables into sparse numerical vectors; Step S24: Dynamic time window division, specifically dividing the time series data related to the target cash crop into an input window and a prediction window through a sliding window construction algorithm; Step S25: Price data signal decomposition process, specifically improving the empirical mode decomposition algorithm to achieve multi-level mode decomposition and high-frequency information removal of the cash crop price time series to obtain the denoised cash crop price series data.

3. The economic crop planting management method based on artificial intelligence according to claim 1, characterized in that: In step S25, the price data signal decomposition process specifically includes the following steps: Step S251: Construction and processing of the noisy signal. Specifically, Gaussian white noise is injected into the original economic crop price sequence, and perturbation modulation is performed in combination with the first-order mode decomposition operation function to construct multiple groups of noisy price signals. The formula used is as follows: ； Wherein, represents the value of the i-th group of disturbance signal sequences at time step t, represents the value of the original cash crop price sequence data at time step t, represents the noise adjustment coefficient used for the i-th secondary disturbance, represents the first-layer mode decomposition operation function, represents the Gaussian white noise sample introduced at time step t in the i-th disturbance; Step S252: Initial residual extraction, specifically, perform a modal decomposition operation on each group of noisy signals to generate modal components for each group, then perform a local mean operation, and average the local mean results of all perturbed versions to form a stable first-layer residual signal , subtract from to obtain the first-layer modal components; Step S253: Multi-round modal iterative decomposition. Specifically, perturbation noise is continuously added to the residual of each layer and modal decomposition operations are performed to extract the modal components and residual signals of the current layer. The formula used is as follows: ； ； wherein, represents the residual signal of the j-th layer, represents the residual component of the -th layer, represents the noise perturbation adjustment coefficient of the -th layer, represents the modal decomposition operation function of the j-th layer, represents the Gaussian white noise sample used in the j-th layer, represents the operation on the local mean, represents the operation of averaging all the perturbed version IMFs; Step S254: Reconstruct the denoised price sequence. Specifically, the high-frequency terms among all the decomposed modal components are removed, the intermediate-frequency and low-frequency IMFs as well as the trend residual terms are retained, and the retained components are weighted and reconstructed to form the denoised economic crop price sequence.

4. The method for planting and managing cash crops based on artificial intelligence according to claim 1, characterized in that: In step S3, the selection of the target economic crop specifically includes the following steps: Step S31: Construct an economic crop comprehensive evaluation model. Specifically, an economic crop comprehensive evaluation model is constructed based on the multi-layer perceptron neural network structure; the economic crop comprehensive evaluation model has a dual-output structure, corresponding to two target economic crop selection decision indicators, namely the environmental adaptability score and the economic potential score respectively; Step S32: Training of the evaluation model. Specifically, the reference economic crop selection data is used as the training data and input into the economic crop comprehensive evaluation model to train the model and obtain the trained economic crop comprehensive evaluation model; Step S33: Calculate the comprehensive score. Specifically, input the target cash crop selection data of the current region into the trained comprehensive evaluation model for cash crops to generate the environmental adaptability score and economic potential score corresponding to the cash crops, and calculate the comprehensive score of the cash crops through a weighted fusion method ; Step S34: Determination of the target economic crop. Specifically, according to the economic crop comprehensive score, all candidate economic crops are sorted from high to low according to the economic crop comprehensive score, and the top k economic crops with the highest comprehensive score are selected as the target economic crop set for the current region to obtain the target economic crop set.

5. The method for managing the cultivation of cash crops based on artificial intelligence according to claim 1, characterized in that: In step S4, the prediction of the market price of economic crops specifically includes the following steps: Step S41: Construct a market price prediction model; Step S42: Training of the price prediction model. Specifically, historical economic crop market price prediction data is used as the training data, and the price prediction loss function is used as the supervised training objective function to train the market price prediction model and obtain the trained market price prediction model; Step S43: Prediction of the market price of the target economic crop. Specifically, the transaction data and external transaction impact data corresponding to the target economic crop in the real-time economic crop price prediction data are selected as the real-time input data, and then after the original data optimization process, structured real-time price prediction data is generated. The structured real-time price prediction data is input into the trained market price prediction model in the preset input window format to obtain the multi-quantile price prediction results of the target economic crop within the future prediction time window. Based on the medium and high quantile prediction results, the time periods with upward potential in the future price trend are identified, and these time periods are marked as the high-price interval windows of the target economic crop. Finally, the multi-quantile price prediction results of the target economic crop and the corresponding high-price interval windows are used as the price prediction output results of the target economic crop.

6. The method for managing the cultivation of cash crops based on artificial intelligence according to claim 1, characterized in that: In step S41, the construction of the market price prediction model specifically includes the following steps: Step S411: Establish an economic crop price time series coding structure, specifically including the following steps: Step S4111: Design an improved ILSTM network structure, specifically including forget gate calculation, input gate calculation, candidate state generation, state update, and output gate calculation; the input gate calculation is used to adjust the update amplitude of the current input information to the memory state, and a translation transformation is performed on the output of the input gate by introducing an improved activation function to avoid signal collapse. The formula is as follows: ； In the formula, represents the output of the input gate, represents the weight matrix corresponding to the input gate, represents the bias term parameter of the input gate, represents the input data, represents the hidden state at the previous moment, represents the sigmoid function, represents the hyperbolic tangent activation function, represents the translation adjustment parameter of the non-linear output of the input gate; Step S4112: Multilayer time series coding output. Specifically, the denoised economic crop price sequence data, the corresponding economic crop trading volume data, and the external trading impact data are multi-dimensionally concatenated to form a model input data tensor, which is sequentially input into two stacked improved ILSTM network structures for hierarchical coding processing to obtain the hidden state of the current time step output by the second layer ; Step S412: Construct a key time point identification layer for price changes, specifically the hidden state at each time step is concatenated with its previous moment state to generate time attention scores through weighted calculation and non-linear transformation, and normalize the scores of all time steps to generate attention weights for each time position. The hidden state of each time step is weighted and summed according to the corresponding weights to obtain a price change time identification vector; Step S413: Construct a price influence factor identification layer, specifically by concatenating the feature hidden states of each time step with the feature hidden state of the previous time step, and generating a feature attention score through weighted calculation and non-linear transformation. The scores of all time steps are normalized to generate the attention weights of each feature, and the feature hidden state of each time step is weighted and summed with the attention weights of the corresponding features to obtain a price influence identification vector; Step S414: Construct an interval market price prediction output layer, specifically by concatenating the price change time identification vector and the price influence identification vector, and generating a fusion vector through a non-linear mapping function. The fusion vector is input into a fully connected layer, and a multi-quantile price prediction strategy is introduced to perform interval prediction on the market price of economic crops at future time points, and obtain the economic crop price prediction values corresponding to multiple quantiles; Step S415: Design a price prediction loss function, specifically by first calculating the error of the deviation between each price prediction quantile and the actual observed price based on the quantile loss function, generating the loss function values of each quantile, and performing weighted integration on the loss function values in all the set quantile sets Q to generate the total loss function value, thereby constituting the price prediction loss function.

7. The method for managing the cultivation of cash crops based on artificial intelligence according to claim 1, characterized in that: In step S5, the optimization of the economic crop planting strategy specifically includes the following steps: Step S51: Construct a planting strategy variable space, specifically by constructing the decision-making links in the whole planting process, and forming a strategy variable optimization space S with crop variety selection variables, sowing time variables, growth management intensity variables, and harvesting method variables; Step S52: Construct a strategy evaluation objective function, specifically by establishing a revenue item function, a risk item function, a cost item function, and a price matching degree item function for the strategy variable optimization space S respectively, and jointly constructing a strategy evaluation objective function; Step S53: Optimize the search for the planting strategy, specifically including the following steps: Step S531: Initialize the search population. Specifically, take the strategy variable optimization space S as the search individual position encoding space in the optimization algorithm, and for each variable vector perform the initialization operation of the particle population using the hybrid cosine chaotic mapping function; the formula used is as follows: ； In the formula, represents the position of the th particle individual, represents the position of the th particle individual, and r represents a random number between 0 and 1; Step S532: Calculate the individual fitness value, specifically, calculate the fitness value f of the particles in the particle swarm i , and calculate the fitness value of the individual particle through the policy evaluation objective function ; Step S533: Particle update, specifically by updating the particle velocity and particle position; Step S534: Particle mutation, specifically by adopting a particle mutation mechanism based on a mutation intensity control factor to perform particle mutation on the selected M particles. The formula is as follows: ； ； Wherein, represents the mutation intensity coefficient of the e-th iteration, represents the parameter for controlling the decline rate of the mutation intensity, represents the parameter for controlling the decline shape of the mutation intensity, represents the fitness value of the worst particle in the current population, represents the fitness value of the best particle in the current population, represents a very small positive number to prevent the denominator from being zero, represents the number of particles to be mutated in this round, The position of the i-th particle after mutation, represents the random perturbation coefficient, which is a random number within the range of [0, 1], represents the position of the i-th particle in the iteration, and e represents the current iteration number, represents the maximum number of iterations, represents the global optimal position of the particle, represents the mutation perturbation amplitude adjustment coefficient, which is used to control the proportional weight of the influence of the individual fitness difference of the current particle on the mutation intensity; Step S535: Update the optimal position of the particles. Specifically, for all particles after particle update and mutation operations, re-evaluate their fitness values, and based on the fitness value of the current particle, compare it with its historical individual optimal value and the current population global optimal value respectively. If the fitness of the current particle is better, update the corresponding individual optimal position and global optimal position respectively; meanwhile, obtain the local optimal position of the particle individuals for the next iteration and the global optimal position of the particle individuals for the next iteration ; Step S536: Particle update termination, specifically, when the fitness value f of the particle i is higher than the fitness threshold and when the maximum number of iterations is reached, terminate the search and obtain the global optimal position of the particle; Step S54: Output the optimal planting strategy plan, specifically by extracting the values of each dimension of the strategy variables from the global optimal position of the particles, and mapping them to the corresponding actual crop variety selection, sowing time, growth management intensity, and harvesting method respectively to obtain the optimal economic crop planting strategy combination.

8. The method for planting and managing cash crops based on artificial intelligence according to claim 1, characterized in that: In step S1, the original planting management data includes reference economic crop selection data, target economic crop selection data, historical economic crop market price prediction data, and real-time economic crop market price prediction data; The reference cash crop selection data and the target cash crop selection data both include regional planting condition data, cash crop planting attribute data, and cash crop economic attribute data; the reference cash crop selection data further includes reference cash crop selection results; the historical cash crop price prediction data and the real-time cash crop price prediction data both include various cash crop transaction data and external transaction impact data.

9. The economic crop planting management system based on artificial intelligence according to claim 1, which is used to implement the economic crop planting management method based on artificial intelligence described in any one of claims 1-8, is characterized in that: A planting management raw data acquisition module, a raw data optimization module, a target cash crop selection module, a cash crop market price prediction module, a cash crop planting strategy optimization module, and a cash crop planting intelligent management module; The planting management raw data acquisition module obtains planting management raw data through data collection operations, and sends the planting management raw data to the raw data optimization module; The raw data optimization module receives the data sent by the planting management raw data acquisition module, performs data optimization processing on the planting management raw data to obtain optimized planting management data, and sends the optimized planting management data to the target cash crop selection module; The target cash crop selection module receives the data sent by the raw data optimization module. By constructing a cash crop comprehensive evaluation model based on a multi-layer perceptron structure and performing model training, it inputs the target cash crop selection data of the current region into the model, generates the environmental adaptability score and economic potential score of the cash crop, and forms a comprehensive score based on weight fusion. It selects the cash crop with the highest score according to the score ranking to obtain a set of target cash crops, and sends the set of target cash crops to the cash crop market price prediction module; The cash crop market price prediction module receives the data sent by the target cash crop selection module. First, it completes the construction of a market price prediction model, designs a price prediction loss function based on the quantile loss function, combines historical data for price prediction model training, and finally predicts the market price of the target cash crop to obtain the cash crop price prediction result, and sends the cash crop price prediction result to the cash crop planting strategy optimization module; The cash crop planting strategy optimization module receives the data sent by the cash crop market price prediction module. By constructing a planting strategy variable space that includes crop selection, sowing time, growth management intensity, and harvesting method, it establishes a multi-objective strategy evaluation function that takes into account expected revenue, market risk, planting cost, and price matching degree, and uses an improved particle swarm optimization algorithm to search for the optimal strategy configuration to obtain the optimal cash crop planting strategy combination; The cash crop planting intelligent management module receives the data of the cash crop planting strategy optimization module, and generates corresponding agricultural operation instructions through the optimal cash crop planting strategy combination to achieve the whole-process management configuration from crop selection to harvesting.

Images