An ecological cycle management method based on the integration of agriculture and animal husbandry

Abstract

The present invention relates to the field of ecological cycle technology, and in particular to an ecological cycle management method based on the integration of agriculture and animal husbandry; the present invention adopts data mining technology, integrates soil sensor data, meteorological data, crop growth data and livestock health data, and constructs a dynamically coupled nutrient management model; utilizes algorithms such as deep learning to predict the nutrient requirements of crops and soil in real time, and adjusts the fertilization strategy according to the prediction results, thereby overcoming the limitations of static management and achieving precise fertilization; combines real-time monitoring technology to make real-time adjustments according to the actual environment and crop needs; continuously optimizes nutrient supply through a feedback mechanism system, avoids the problem of lag in existing management solutions, and ensures that crops obtain appropriate amounts of nutrients at different growth stages; utilizes data mining technology to deeply analyze historical data, digs out key factors affecting nutrient supply and demand, formulates scientific fertilization strategies, significantly improves data utilization, and reduces uncertainty in management decisions.

Description

Technical Field

[0001] The present invention relates to the field of ecological cycle technology, and in particular to an ecological cycle management method based on the integration of agriculture and animal husbandry. Background Art

[0002] Modern ecological circular agriculture is a comprehensive management mode that organically links planting, animal husbandry, fishery and other industries with processing industries. It uses microbial technology to form a virtuous cycle of the overall ecological chain among multiple modules of agriculture, forestry, animal husbandry, sideline production and fishery, and provides systematic solutions for solving agricultural pollution, optimizing industrial structure, saving agricultural resources, improving output effects, transforming agricultural ecology, and ensuring food safety.

[0003] In an ecosystem that integrates agriculture and animal husbandry, crop planting and livestock breeding achieve nutrient recycling through organic fertilizers, such as livestock manure and crop residues. This cycle helps reduce the use of chemical fertilizers and promotes soil health. However, current nutrient management is mostly based on static fertilization plans, ignoring dynamic factors such as crop growth cycles, climate change, and changes in livestock excrement, resulting in a mismatch between nutrient supply and crop demand, causing resource waste or nutrient deficiency. In addition, there is a large amount of potentially valuable data in the agriculture-animal husbandry system (such as soil nutrient levels, meteorological conditions, crop growth status, livestock health data, etc.), and existing methods lack effective data integration and analysis methods, and fail to make full use of these data to optimize management decisions. At the same time, current management methods lack real-time feedback mechanisms and cannot adjust fertilization strategies in a timely manner according to actual conditions. Lagging management methods will lead to long-term nutrient imbalances, affecting crop yields and soil health.

[0004] At present, regular soil testing is usually adopted to deal with such problems. Soil samples are collected regularly to analyze the soil nutrient content and adjust the fertilization plan accordingly. This method can reflect the nutrient status of the soil to a certain extent, but due to the low detection frequency, it is difficult to keep up with the changes in actual demand.

[0005] Some have adopted fixed crop rotation and rotational grazing systems to try to balance the supply and demand of soil nutrients; however, this system is based on long-term average effects and lacks the ability to respond to short-term changes; it has alleviated management problems to a certain extent, but there are still limitations, mainly manifested in insufficient response to dynamic changes, low data utilization and lagging feedback mechanisms; therefore, an ecological cycle management method based on the integration of agriculture and animal husbandry is urgently needed to solve such problems. Summary of the invention

[0006] To this end, the present invention provides an ecological cycle management method based on the integration of agriculture and animal husbandry, so as to overcome the problems in the prior art that the static fertilization plan is used, ignoring dynamic factors such as crop growth cycle, climate change, and changes in livestock excrement, resulting in a mismatch between nutrient supply and crop demand; lacking effective data integration and analysis methods, and failing to make full use of these data to optimize management decisions; lacking a real-time feedback mechanism, and being unable to adjust the fertilization strategy in time according to actual conditions.

[0007] To achieve the above objectives, the present invention provides an ecological cycle management method based on the integration of agriculture and animal husbandry, comprising:

[0008] Step S1, dynamic nutrient demand prediction, using data mining technology to integrate multi-source data, build a real-time dynamic coupled nutrient management model, and predict the nutrient requirements of crops and soil;

[0009] Step S2, real-time feedback and adaptive adjustment, based on the output of the dynamic nutrient demand prediction model, combined with real-time monitoring technology, establish a feedback mechanism to perform adaptive adjustment of the fertilization strategy;

[0010] Step S3, data-driven decision support, identifies key factors affecting nutrient supply and demand based on the analysis of historical data, optimizes future fertilization strategies, and improves the scientificity and accuracy of management decisions.

[0011] Furthermore, in step S1, the dynamic nutrient demand prediction method includes:

[0012] Collect and integrate multidimensional data from soil sensors, weather stations, crop growth monitoring, and livestock health to build a complete ecosystem dataset;

[0013] Use deep learning algorithms to train the integrated data and build a dynamic coupling model;

[0014] Based on the constructed real-time dynamic coupled nutrient management model, the specific nutrient requirements of crops at different growth stages are predicted in real time.

[0015] Furthermore, in step S1, multi-source data collection is performed to construct a complete ecosystem dataset, and data from the following data sources are collected:

[0016] Soil Dataset D s : Contains soil moisture H s , soil temperature T s , soil organic matter content OM s , soil pH s ;

[0017] Meteorological Dataset D m : Including precipitation P m , sunshine duration S m, Temperature T m , wind speed W m ;

[0018] Crop Growth Dataset D c : Leaf Area Index LAI c , Crop nitrogen absorption N c , growth rate G c ;

[0019] Livestock Dataset D l :Including the nutrient content of livestock excrement N l , livestock feed consumption F l , excretion frequency R l ;

[0020] Assume that each data set D x The variable in is x i , for standardization: in is the standardized variable value, x i is the original data value, μ x For the dataset D x The mean of the variable in x For the dataset D x The standard deviation of the medium variable;

[0021] Integrate the standardized data into a multidimensional data matrix X: in, They are the standardized soil, meteorological, crop and livestock data respectively.

[0022] Furthermore, in step S1, a long short-term memory network LSTM is used to construct a dynamic coupling model: the LSTM network can effectively process time series data and is suitable for the coupling and dynamic prediction of multi-source data here;

[0023] The input layer is set to be the integrated multidimensional data matrix X, and the output is the predicted nutrient demand value N of crops at different growth stages. p , the LSTM network architecture includes: h t =o t *tanh(C t ), where f t is the forget gate, which controls the influence of the previous state; i t It is the input gate, which controls the input amount of new information; is a candidate memory unit, representing the candidate value of the new state; C t It is a memory unit that stores accumulated information; o t is the output gate; h tis the hidden layer state, i.e., the output; t is the time index of the current moment, σ is the Sigmoid activation function, W f , W i , W C , W o is a weight matrix, corresponding to the forget gate, input gate, candidate memory unit and output gate respectively. Each weight matrix is ​​used to connect the current input and the hidden state of the previous moment. f , b i , b C , b o is the bias term, corresponding to the forget gate, input gate, candidate memory unit and output gate respectively, h t-1 is the hidden state of the previous time step, including the information of the previous time step, X t The input data for the current time step includes all the collected standardized multidimensional data, i.e. soil, meteorological, crop growth and livestock data;

[0024] Using historical data and real-time data, the LSTM model is trained through the back propagation algorithm BPTT to optimize the model parameters θ={W f ,W i ,W C ,W o ,b f ,b i ,b C ,b o}, minimize the prediction error L(θ): Where N p,t is the actual nutrient requirement at the tth time step, For the model to predict nutrient requirements, N is the total number of time steps in the time series, and t is the current time step;

[0025] Input the standardized data collected in real time into the trained LSTM model to obtain the predicted value of nutrient demand at the current time step

[0026] According to the output of LSTM model Dynamically adjust fertilization strategies.

[0027] Furthermore, in step S2, the real-time feedback and adaptive adjustment method includes:

[0028] Continuously monitor soil conditions, crop status and environmental factors, feeding real-time data into nutrient demand prediction models;

[0029] By analyzing the deviation between real-time monitoring data and predicted results, the fertilization plan can be adjusted immediately to ensure that crops obtain the most suitable nutrients at each growth stage.

[0030] Furthermore, in step S2, real-time data monitoring is performed to collect data in real time:

[0031] Soil conditions: including soil moisture, soil temperature, and soil pH;

[0032] Crop status: crop leaf area index, nitrogen content, growth rate;

[0033] Environmental factors: temperature, precipitation, wind speed;

[0034] Let the real-time data matrix be X real (t);

[0035] Use real-time data to compare with the dynamic nutrient demand forecast model output and calculate the deviation:

[0036] Where Δt is the deviation matrix at time step t, is the output data matrix of the prediction model at time step t;

[0037] Using the deviation matrix Δt, an adaptive adjustment model of fertilization strategy is established. The amount of fertilizer in the facility is F(t), and the adjustment method of fertilization amount is:

[0038] F(t+1)=F(t)+η·W Δ Δ(t), where F(t) is the amount of fertilizer applied at time step t, η is the learning rate, and controls the adjustment amplitude, W Δ is the weight matrix, which is used to adjust the influence of different factors on the amount of fertilizer, and Δ(t) is the deviation matrix;

[0039] Optimization formula for adaptive adjustment strategy:

[0040] in is the weight matrix before and after the update, γ is the learning rate of weight update, L(W Δ ) is the loss function, which is the sum of square errors between the fertilization strategy and the actual demand;

[0041] Define the loss function: Where N p,real (t) is the actual nutrient requirement at time step t, and F(t) is the fertilizer application at time step t;

[0042] Through W Δ Iterative optimization gradually improves the accuracy of fertilization strategy;

[0043] Based on the adjusted fertilization strategy, the monitoring data of soil, crops and environment are updated in real time, the new deviation matrix Δt is continuously calculated and the adjustment process is repeated to form an adaptive closed-loop feedback system.

[0044] Furthermore, in step S3, the data-driven decision support method includes:

[0045] Use data mining technology to conduct in-depth analysis of accumulated historical data to extract key factors that have a significant impact on nutrient demand;

[0046] Based on the results of historical data analysis, the fertilization strategy is pre-optimized, and the effects of different fertilization plans are predicted through simulation to output fertilization decisions.

[0047] Furthermore, historical data analysis and feature selection are performed in step S3:

[0048] Suppose there is a historical data set X hist , which contains multidimensional data X at different time steps hist ={x1,x2,...,x T}, each data vector x t Contains n features, t refers to 1, 2, ..., T;

[0049] Principal component analysis (PCA) was used to extract the features that mainly affect nutrient requirements:

[0050] X pca =X hist W pca , where X pca is the feature matrix after dimensionality reduction, W pca is the eigenvector matrix of PCA, obtained by eigenvalue decomposition, X hist is the historical data matrix;

[0051] The principal component score PCS calculation formula is: PCS i is the score of the i-th principal component, w ij is the weight of feature j in the i-th principal component in principal component analysis, x j is the original data value of feature j.

[0052] Furthermore, the historical data analysis and feature selection method in step S3 also includes:

[0053] Through the ranking and contribution rate analysis of PCS, the top k key characteristics with the greatest impact on nutrient demand were determined;

[0054] For the selected k key features, a multivariate linear regression model was established to quantify the impact of the features on nutrient requirements:

[0055] Where N p,hist (t) is the historical nutrient demand at time step t, β0 is the intercept term, β i is the regression coefficient of feature i, x i(t) is the value of feature i at time step t, ε t is the error term;

[0056] The t-test method was used to screen out the most significant characteristics of nutrient requirements.

[0057] Furthermore, the fertilization strategy optimization method in step S3 is:

[0058] Identify key characteristics and use them as inputs to optimize fertilization strategies;

[0059] Define fertilization strategy objectives: maximize crop yield, maintain or improve soil health, and minimize fertilization costs;

[0060] Through multi-objective optimization methods, find the fertilization plan that can achieve the best balance among multiple objectives;

[0061] Genetic algorithm and particle swarm optimization algorithm are used to input key features, objective functions and constraints to generate multiple possible fertilization strategies;

[0062] According to the optimization results, the fertilization plan is screened out.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] The present invention adopts data mining technology to integrate soil sensor data, meteorological data, crop growth data and livestock health data to construct a dynamically coupled nutrient management model; it uses deep learning and other algorithms to predict the nutrient requirements of crops and soil in real time, and adjusts the fertilization strategy according to the prediction results, overcoming the limitations of static management and achieving precise fertilization.

[0065] The present invention combines real-time monitoring technology to make real-time adjustments based on the actual environment and crop needs; the feedback mechanism system continuously optimizes nutrient supply, avoids the lag problem in existing management plans, and ensures that crops receive the right amount of nutrients at different growth stages.

[0066] The present invention uses data mining technology to deeply analyze historical data, dig out the key factors affecting nutrient supply and demand, formulate scientific fertilization strategies, significantly improve data utilization, and reduce uncertainty in management decisions.

[0067] The solution solves the problems that the existing technology is based on static fertilization plans, ignoring dynamic factors such as crop growth cycles, climate change, and changes in livestock excrement, resulting in a mismatch between nutrient supply and crop demand; lacks effective data integration and analysis methods, and fails to make full use of these data to optimize management decisions; lacks a real-time feedback mechanism and cannot adjust fertilization strategies in a timely manner according to actual conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1It is a schematic flow chart of the ecological cycle management method based on the integration of agriculture and animal husbandry of the present invention. DETAILED DESCRIPTION

[0069] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme 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, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0070] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0071] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0072] Example 1

[0073] See also Figure 1 The present invention provides an ecological cycle management method based on the integration of agriculture and animal husbandry, comprising:

[0074] Step S1, dynamic nutrient demand prediction, using data mining technology to integrate multi-source data, build a real-time dynamic coupled nutrient management model, and predict the nutrient requirements of crops and soil;

[0075] Dynamic nutrient demand prediction methods include:

[0076] Collect and integrate multidimensional data from soil sensors, weather stations, crop growth monitoring, and livestock health to build a complete ecosystem dataset;

[0077] Use deep learning algorithms to train the integrated data and build a dynamic coupling model;

[0078] Based on the constructed real-time dynamic coupled nutrient management model, the specific nutrient requirements of crops at different growth stages are predicted in real time;

[0079] Conduct multi-source data collection to build a complete ecosystem dataset, collecting data from the following data sources:

[0080] Soil Dataset D s : Contains soil moisture H s , soil temperature T s , soil organic matter content OM s , soil pH s ;

[0081] Meteorological Dataset D m : Including precipitation P m , sunshine duration S m , Temperature T m , wind speed W m ;

[0082] Crop Growth Dataset D c : Leaf Area Index LAI c , Crop nitrogen absorption N c , growth rate G c ;

[0083] Livestock Dataset D l :Including the nutrient content of livestock excrement N l , livestock feed consumption F l , excretion frequency R l ;

[0084] Assume that each data set D x The variable in is x i , for standardization: in is the standardized variable value, x i is the original data value, μ x For the dataset D x The mean of the variable in x For the dataset D x The standard deviation of the medium variable;

[0085] Integrate the standardized data into a multidimensional data matrix X: in, They are the standardized soil, meteorological, crop and livestock data;

[0086] The long short-term memory network LSTM is used to build a dynamic coupling model: the LSTM network can effectively process time series data and is suitable for the coupling and dynamic prediction of multi-source data here;

[0087] The input layer is set to be the integrated multidimensional data matrix X, and the output is the predicted nutrient demand value N of crops at different growth stages. p , the LSTM network architecture includes: h t =o t *tanh(C t ), where f t is the forget gate, which controls the influence of the previous state; i t It is the input gate, which controls the input amount of new information; is a candidate memory unit, representing the candidate value of the new state; C t It is a memory unit that stores accumulated information; o t is the output gate; h t is the hidden layer state, i.e., the output; t is the time index of the current moment, σ is the Sigmoid activation function, W f , W i , W C , W o is a weight matrix, corresponding to the forget gate, input gate, candidate memory unit and output gate respectively. Each weight matrix is ​​used to connect the current input and the hidden state of the previous moment. f , b i , b C , b o is the bias term, corresponding to the forget gate, input gate, candidate memory unit and output gate respectively, h t-1 is the hidden state of the previous time step, including the information of the previous time step, X t The input data for the current time step includes all the collected standardized multidimensional data, i.e. soil, meteorological, crop growth and livestock data;

[0088] Using historical data and real-time data, the LSTM model is trained through the back propagation algorithm BPTT to optimize the model parameters θ={W f ,W i ,W C ,W o ,b f ,b i ,b C ,b o}, minimize the prediction error L(θ): Where N p,t is the actual nutrient requirement at the tth time step, For the model to predict nutrient requirements, N is the total number of time steps in the time series, and t is the current time step;

[0089] Input the standardized data collected in real time into the trained LSTM model to obtain the predicted value of nutrient demand at the current time step

[0090] According to the output of LSTM model Dynamically adjust fertilization strategies;

[0091] Specifically, through dynamic nutrient demand forecasting, accurate prediction of crop and soil nutrient demand can be achieved; based on integrated multi-dimensional data, including soil conditions, meteorological factors, crop growth status and livestock excrement, a dynamic coupling model built using a deep learning algorithm can process time series data in real time to generate accurate nutrient demand forecasts; real-time prediction ensures that fertilization strategies can be adjusted at any time to avoid resource waste and nutrient deficiency, which helps to increase crop yields and maintain soil health;

[0092] Step S2, real-time feedback and adaptive adjustment, based on the output of the dynamic nutrient demand prediction model, combined with real-time monitoring technology, establish a feedback mechanism to perform adaptive adjustment of the fertilization strategy;

[0093] Real-time feedback and adaptive adjustment methods include:

[0094] Continuously monitor soil conditions, crop status and environmental factors, feeding real-time data into nutrient demand prediction models;

[0095] By analyzing the deviation between real-time monitoring data and predicted results, the fertilization plan can be adjusted immediately to ensure that crops receive the most suitable nutrients at each growth stage;

[0096] Conduct real-time data monitoring and collect data in real time:

[0097] Soil conditions: including soil moisture, soil temperature, and soil pH;

[0098] Crop status: crop leaf area index, nitrogen content, growth rate;

[0099] Environmental factors: temperature, precipitation, wind speed;

[0100] Let the real-time data matrix be X real (t);

[0101] Use real-time data to compare with the dynamic nutrient demand forecast model output and calculate the deviation:

[0102] Where Δt is the deviation matrix at time step t, is the output data matrix of the prediction model at time step t;

[0103] Using the deviation matrix Δt, an adaptive adjustment model of fertilization strategy is established. The amount of fertilizer in the facility is F(t), and the adjustment method of fertilization amount is:

[0104] F(t+1)=F(t)+η·W Δ Δ(t), where F(t) is the amount of fertilizer applied at time step t, η is the learning rate, and controls the adjustment amplitude, W Δis the weight matrix, which is used to adjust the influence of different factors on the amount of fertilizer, and Δ(t) is the deviation matrix;

[0105] Optimization formula for adaptive adjustment strategy:

[0106] in is the weight matrix before and after the update, γ is the learning rate of weight update, L(W Δ ) is the loss function, which is the sum of square errors between the fertilization strategy and the actual demand;

[0107] Define the loss function: Where N p,real (t) is the actual nutrient requirement at time step t, and F(t) is the fertilizer application at time step t;

[0108] Through W Δ Iterative optimization gradually improves the accuracy of fertilization strategy;

[0109] Based on the adjusted fertilization strategy, the monitoring data of soil, crops and environment are updated in real time, the new deviation matrix Δt is calculated and the adjustment process is repeated to form an adaptive closed-loop feedback system.

[0110] Specifically, through real-time feedback and adaptive adjustment mechanisms, timely responses can be made to environmental changes and fluctuations in crop demand; soil, crop and environmental data can be continuously monitored and compared with the output of the prediction model, deviations can be calculated and fertilization strategies can be dynamically adjusted; the adaptive adjustment mechanism effectively forms a closed-loop feedback system, so that the fertilization strategy always maintains the optimal state at different growth stages of the crop, further improving the flexibility and accuracy of management;

[0111] Step S3, data-driven decision support, based on the analysis of historical data, identifies key factors affecting nutrient supply and demand, optimizes future fertilization strategies, and improves the scientificity and accuracy of management decisions;

[0112] Data-driven decision support methods include:

[0113] Use data mining technology to conduct in-depth analysis of accumulated historical data to extract key factors that have a significant impact on nutrient demand;

[0114] Based on the results of historical data analysis, the fertilization strategy is optimized in advance, and the effects of different fertilization schemes are predicted through simulation to output fertilization decisions;

[0115] Historical data analysis and feature selection are performed in step S3:

[0116] Suppose there is a historical data set X hist , which contains multidimensional data X at different time steps hist={x1,x2,...,x T}, each data vector x t Contains n features, t refers to 1, 2, ..., T;

[0117] Principal component analysis (PCA) was used to extract the features that mainly affect nutrient requirements:

[0118] X pca =X hist W pca , where X pca is the feature matrix after dimensionality reduction, W pca is the eigenvector matrix of PCA, obtained by eigenvalue decomposition, X hist is the historical data matrix;

[0119] The principal component score PCS calculation formula is: PCS i is the score of the i-th principal component, w ij is the weight of feature j in the i-th principal component in principal component analysis, x j is the original data value of feature j;

[0120] Through the ranking and contribution rate analysis of PCS, the top k key characteristics with the greatest impact on nutrient demand were determined;

[0121] For the selected k key features, a multivariate linear regression model was established to quantify the impact of the features on nutrient requirements:

[0122] Where N p,hist (t) is the historical nutrient demand at time step t, β0 is the intercept term, β i is the regression coefficient of feature i, x i (t) is the value of feature i at time step t, ε t is the error term;

[0123] The t-test method was used to screen out the most significant characteristics of nutrient requirements;

[0124] The optimization method of fertilization strategy is:

[0125] Identify key characteristics and use them as inputs to optimize fertilization strategies;

[0126] Define fertilization strategy objectives: maximize crop yield, maintain or improve soil health, and minimize fertilization costs;

[0127] Through multi-objective optimization methods, find the fertilization plan that can achieve the best balance among multiple objectives;

[0128] Genetic algorithm and particle swarm optimization algorithm are used to input key features, objective functions and constraints to generate multiple possible fertilization strategies;

[0129] According to the optimization results, the fertilization plan is screened out;

[0130] Specifically, through in-depth analysis of historical data, the key factors affecting nutrient supply and demand are extracted, and the system can perform multi-objective optimization to balance crop yield, soil health and fertilization costs; using genetic algorithms and particle swarm optimization algorithms, it can find the best balance between multiple objectives and generate the optimal fertilization plan; data-driven optimization strategies not only improve the scientific nature of fertilization decisions, but also reduce uncertainties in management, laying the foundation for the long-term sustainable development of agricultural and animal husbandry systems.

[0131] The present invention first forms a dynamic prediction of crop and soil nutrient requirements through data integration and deep learning model construction; based on the prediction, real-time monitoring data is continuously updated, and the prediction results output by the model are directly used to guide immediate fertilization adjustments; through continuous data accumulation and analysis, models and strategies are continuously optimized to achieve short-term to long-term fertilization management optimization and ensure the sustainability of the entire agricultural and animal husbandry ecosystem.

[0132] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. An ecological cycle management method based on the integration of agriculture and animal husbandry, characterized in that: include: Step S1, dynamic nutrient demand prediction, using data mining technology to integrate multi-source data, build a real-time dynamic coupled nutrient management model, and predict the nutrient requirements of crops and soil; Step S2, real-time feedback and adaptive adjustment, based on the output of the dynamic nutrient demand prediction model, combined with real-time monitoring technology, establish a feedback mechanism to perform adaptive adjustment of the fertilization strategy; Step S3, data-driven decision support, identifies key factors affecting nutrient supply and demand based on the analysis of historical data and optimizes future fertilization strategies; In step S1, the dynamic nutrient demand prediction method includes: Collect and integrate multidimensional data from soil sensors, weather stations, crop growth monitoring, and livestock health to build a complete ecosystem dataset; Use deep learning algorithms to train the integrated data and build a dynamic coupling model; Based on the constructed real-time dynamic coupled nutrient management model, the specific nutrient requirements of crops at different growth stages are predicted in real time; In step S1, multi-source data collection is performed to build a complete ecosystem dataset, collecting data from the following data sources: Soil Dataset D s : Contains soil moisture H s , soil temperature T s , soil organic matter content OM s , soil pH s ; Meteorological Dataset D m : Including precipitation P m , sunshine duration S m , Temperature T m , wind speed W m ; Crop Growth Dataset D c : Leaf Area Index (LAI) c , Crop nitrogen absorption N c , growth rate G c ; Livestock Dataset D l :Including the nutrient content of livestock excrement N l , livestock feed consumption F l , excretion frequency R l ; Assume that each data set D x The variable in is x i , for standardization: in is the standardized variable value, x i is the original data value, μ x For the dataset D x The mean of the variable in x For the dataset D x The standard deviation of the medium variable; Integrate the standardized data into a multidimensional data matrix X: in, They are the standardized soil, meteorological, crop and livestock data; In step S1, a long short-term memory network LSTM is used to build a dynamic coupling model: The input layer is set to be the integrated multidimensional data matrix X, and the output is the predicted nutrient demand value N of crops at different growth stages. p , the LSTM network architecture includes: h t =o t *tanh(C t ), where f t is the forget gate, which controls the influence of the previous state; i t It is the input gate, which controls the input amount of new information; is a candidate memory unit, representing the candidate value of the new state; C t It is a memory unit that stores accumulated information; o t is the output gate; h t is the hidden layer state, i.e., the output; t is the time index of the current moment, σ is the Sigmoid activation function, W f , W i , W C , W o is a weight matrix, corresponding to the forget gate, input gate, candidate memory unit and output gate respectively. Each weight matrix is ​​used to connect the current input and the hidden state of the previous moment. f , b i , b C , b o is the bias term, corresponding to the forget gate, input gate, candidate memory unit and output gate respectively, h t-1 is the hidden state of the previous time step, including the information of the previous time step, X t The input data for the current time step includes all the collected standardized multidimensional data, i.e. soil, meteorological, crop growth and livestock data; Using historical data and real-time data, the LSTM model is trained through the back propagation algorithm BPTT to optimize the model parameters θ={W f ,W i ,W C ,W o ,b f ,b i ,b C ,b o }, minimize the prediction error L(θ): Where N p,t is the actual nutrient requirement at the tth time step, For the model to predict nutrient requirements, N is the total number of time steps in the time series, and t is the current time step; Input the standardized data collected in real time into the trained LSTM model to obtain the predicted value of nutrient demand at the current time step According to the output of LSTM model Dynamically adjust fertilization strategies; In step S2, the real-time feedback and adaptive adjustment method includes: Continuously monitor soil conditions, crop status and environmental factors, feeding real-time data into nutrient demand prediction models; Immediately adjust the fertilization plan by analyzing the deviation between real-time monitoring data and predicted results; In step S2, real-time data monitoring and real-time data collection are performed: Soil conditions: including soil moisture, soil temperature, and soil pH; Crop status: crop leaf area index, nitrogen content, growth rate; Environmental factors: temperature, precipitation, wind speed; Let the real-time data matrix be X real (t); Use real-time data to compare with the dynamic nutrient demand forecast model output and calculate the deviation: Where Δt is the deviation matrix at time step t, is the output data matrix of the prediction model at time step t; Using the deviation matrix Δt, an adaptive adjustment model of fertilization strategy is established. The amount of fertilizer in the facility is F(t), and the adjustment method of fertilization amount is: F(t+1)=F(t)+η·W Δ Δ(t), where F(t) is the amount of fertilizer applied at time step t, η is the learning rate, and controls the adjustment amplitude, W Δ is the weight matrix, which is used to adjust the influence of different factors on the amount of fertilizer, and Δ(t) is the deviation matrix; Optimization formula for adaptive adjustment strategy: in is the weight matrix before and after the update, γ is the learning rate of weight update, L(W Δ ) is the loss function, which is the sum of square errors between the fertilization strategy and the actual demand; Define the loss function: Where N p,real (t) is the actual nutrient requirement at time step t, and F(t) is the amount of fertilizer applied at time step t.

2. The ecological cycle management method based on the integration of agriculture and animal husbandry according to claim 1 is characterized in that: In step S3, the data-driven decision support method includes: Use data mining technology to conduct in-depth analysis of accumulated historical data to extract key factors that have a significant impact on nutrient demand; Based on the results of historical data analysis, the fertilization strategy is pre-optimized, and the effects of different fertilization plans are predicted through simulation to output fertilization decisions.

3. The ecological cycle management method based on the integration of agriculture and animal husbandry according to claim 2 is characterized in that: Historical data analysis and feature selection are performed in step S3: Suppose there is a historical data set X hist , which contains multidimensional data X at different time steps hist ={x1,x2,...,x T }, each data vector x t Contains n features, t refers to 1, 2, ..., T; Principal component analysis (PCA) was used to extract the features that mainly affect nutrient requirements: X pca =X hist W pca , where X pca is the feature matrix after dimensionality reduction, W pca is the eigenvector matrix of PCA, obtained by eigenvalue decomposition, X hist is the historical data matrix; The principal component score PCS calculation formula is: PCS i is the score of the i-th principal component, w ij is the weight of feature j in the i-th principal component in principal component analysis, x j is the original data value of feature j.

4. The ecological cycle management method based on the integration of agriculture and animal husbandry according to claim 3 is characterized in that: The historical data analysis and feature selection method in step S3 also includes: Through the ranking and contribution rate analysis of PCS, the top k key characteristics with the greatest impact on nutrient demand were determined; For the selected k key features, a multivariate linear regression model was established to quantify the impact of the features on nutrient requirements: Where N p,hist (t) is the historical nutrient demand at time step t, β0 is the intercept term, β i is the regression coefficient of feature i, x i (t) is the value of feature i at time step t, ε t is the error term; The t-test method was used to screen out the significant characteristics of nutrient requirements.

5. The ecological cycle management method based on the integration of agriculture and animal husbandry according to claim 4 is characterized in that: The fertilization strategy optimization method in step S3 is: Identify key characteristics and use them as inputs to optimize fertilization strategies; Define fertilization strategy objectives: maximize crop yield, maintain or improve soil health, and minimize fertilization costs; Through multi-objective optimization methods, find the fertilization plan that can achieve the best balance among multiple objectives; Genetic algorithm and particle swarm optimization algorithm are used to input key features, objective functions and constraints to generate multiple possible fertilization strategies; According to the optimization results, the fertilization plan is screened out.

Images