Greenhouse water, fertilizer and pesticide optimization control system and method based on environment-crop-input multi-element coupling

Abstract

The invention provides a greenhouse water, fertilizer and pesticide optimization control system and method based on environment-crop-input multi-element coupling, and the method comprises the steps: deploying a sensor network, collecting greenhouse data, processing the greenhouse data, generating a space-time continuous data set, and carrying out the optimization control of the greenhouse water, fertilizer and pesticide on the basis of the generated space-time continuous data set. And quantifying a dynamic causal chain of the environment, the crops and the input, generating a decision rule base, fusing the decision rule base with a real-time simulation result of the three-dimensional digital twinborn model of the root zone, and generating a water, pesticide and fertilizer regulation and control instruction. Space-time continuous acquisition and fusion processing of greenhouse environment-crop-input product data are realized through a multi-modal sensor network, a decision rule base containing a dynamic causal chain is constructed, real-time simulation of a root zone three-dimensional digital twinborn model and Bayesian causal reasoning are combined, an accurate water, fertilizer and pesticide regulation and control instruction is generated, and the accuracy of the water, fertilizer and pesticide regulation and control is improved. The nonlinear causal relationship among multiple elements is effectively quantified, and the decision scientificity is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent control technology for facility agriculture, in particular to a greenhouse water-fertilizer-pesticide optimization control system and method based on environment-crop-input multi-element coupling. BACKGROUND

[0002] High yield and stable yield of greenhouse crops depend on precise coordinated regulation of environment, crop physiology and inputs (water, fertilizer and pesticide). Traditional greenhouse control systems mostly use "threshold-feedback" or "expert rule" mode:

[0003] Environment side: collect data through temperature and humidity, light, CO2 sensors, and trigger ventilation, shading or light supplementation according to experience threshold;

[0004] Crop side: manual visual inspection or offline detection of leaf area, nitrogen content and other indicators as a rough basis for fertilization and pesticide application;

[0005] Input side: irrigation, fertilization and pesticide application are executed according to fixed formula or periodic plan, lacking dynamic coupling with real-time environment-crop state.

[0006] The traditional greenhouse control method has the following defects:

[0007] a) Data discontinuity: heterogeneous sensors and different sampling frequencies result in difficulty in aligning environment, crop and input data in time and space dimensions, forming "information islands";

[0008] b) Cause and effect missing: threshold rules only reflect the linear relationship between a single environmental factor and crop response, and cannot quantify the nonlinear causal chain of multi-element coupling of temperature and humidity, light, CO2, soil water-fertilizer-pesticide, etc;

[0009] c) Root zone blind area: existing systems lack real-time simulation of the three-dimensional diffusion process of rhizosphere water-nutrient-pesticide, and the control instructions do not match the actual needs of the root zone, resulting in water, fertilizer and pesticide redundancy, root salt damage or pesticide residue exceeding standards;

[0010] d) Rule rigidity: expert rule base is based on historical experience, which is difficult to adapt to different growth stages of crops and sudden climate changes, resulting in control lag or overkill.

[0011] Therefore, there is an urgent need for a greenhouse water-fertilizer-pesticide integrated control method that can integrate multi-modal perception. SUMMARY

[0012] In view of the deficiencies of the prior art, the present application provides a greenhouse water-fertilizer-pesticide optimization control system and method based on environment-crop-input multi-element coupling to solve the problems raised in the background art.

[0013] To achieve the above object, the present application is implemented by the following technical solutions: A greenhouse water, fertilizer and pesticide optimization control method based on environment-crop-input multi-element coupling comprises the following steps:

[0014] S1: Deploy a sensor network to collect greenhouse data, and process the greenhouse data to generate a time-space continuous data set;

[0015] The greenhouse data includes environment data, crop state data and input data.

[0016] S2: Based on the generated time-space continuous data set, quantize the dynamic causal chain of environment, crop and input, and generate a decision rule library;

[0017] S3: Fuse the decision rule library with the real-time simulation results of the root zone three-dimensional digital twin model to generate water, fertilizer and pesticide control instructions;

[0018] S4: Send the water, fertilizer and pesticide control instructions generated by fusion to the execution equipment through the Internet of Things controller.

[0019] As a preferred embodiment of the present application, the specific steps of processing the greenhouse data to generate a time-space continuous data set include:

[0020] Processing the greenhouse data specifically includes outlier rejection, time synchronization, spatial mapping, data standardization for data cleaning and time-space alignment to obtain a time-stamped multi-modal data stream;

[0021] A time series database is constructed to store the time-stamped multi-modal data stream, and the time-stamped multi-modal data stream generates a time-space continuous data set by fusing historical-real-time data.

[0022] As a preferred embodiment of the present application, the specific process of step S2 includes: based on the generated time-space continuous data set, real-time extraction of time series data of environment, crop and input, quantization of dynamic causal relationship among the three, and obtaining of dynamic causal chain;

[0023] Based on the dynamic causal chain, a causal graph of environment, crop and input is constructed;

[0024] The causal graph is modeled by a graph neural network using an attention mechanism, and a dynamic causal chain and a readable decision rule library are outputted;

[0025] The decision rule library contains water, fertilizer and pesticide control rules based on causal relationship.

[0026] As a preferred embodiment of the present application, the specific process of real-time extraction of time series data of environment, crop and input based on the generated time-space continuous data set, quantization of dynamic causal relationship among the three includes:

[0027] Screening environmental elements, crop elements, input elements and real-time data from the data pool;

[0028] The dynamic causal strength among the three elements is quantified by combining the time-delay Granger causality test and transfer entropy.

[0029] As a preferred embodiment of the present application, the calculation formula of the time-delay Granger causality test is:

[0030] ;

[0031] wherein, Yt represents the value of the dependent variable at time t, and p represents the lag order, βt represents the regression coefficient of the lagged term of the dependent variable itself, Yt-i represents the i-th lagged term of the dependent variable, βt represents the regression coefficient of the lagged term of the independent variable, Xt-i represents the j-th lagged term of the independent variable, εt represents the residual term;

[0032] The calculation formula for quantifying the nonlinear contribution of the causal relationship by combining the transfer entropy calculation is:

[0033] ;

[0034] wherein, p(·) represents the probability density function, Yt+1 represents the value of the variable Y at time t+1, Yt represents the value of the variable Y at time t, Xt represents the value of the variable X at time t, P(Yt+1|Yt) represents the probability of Y at time t+1 given only the value of Y at time t, P(Yt+1|Xt,Yt) represents the probability of Y at time t+1 given the values of X and Y at time t.

[0035] As a preferred embodiment of the present application, the calculation formula for modeling the causal graph by the attention mechanism of the graph neural network is:

[0036] ;

[0037] ;

[0038] wherein, αi,j represents the attention weight of the neighbor node j of the node i, W represents the transpose of the learnable attention parameter vector, σ represents the activation function, W represents the linear transformation matrix, xi represents the original feature vector of the node i, represents a concatenation operation, represents the result of linearly transforming the feature of neighbor node j, represents the neighbor set of node i, represents the updated feature of node i.

[0039] As a preferred embodiment of the present embodiment, the rule base generation process comprises:

[0040] extracting a key causal path from the dynamic causal chain;

[0041] generating an “If-Then” rule based on the associated causal path.

[0042] As a preferred embodiment of the present embodiment, the step S3 specifically comprises:

[0043] receiving environmental data, crop growth data and input data from a multi-modal sensor network in real time, and completing missing data through a spatio-temporal interpolation algorithm to construct a dynamically updated time series database;

[0044] calling a pre-generated decision rule base, calculating the causal weight between nodes based on the current time series database state through a graph neural network, matching historical decision rules with a similarity greater than a preset threshold to the current greenhouse state, and outputting a candidate set of control strategies;

[0045] constructing a root zone three-dimensional digital twin model, inputting real-time soil moisture, nutrient concentration and drip irrigation belt partition state data, dynamically simulating the diffusion process of water and nutrients in the rhizosphere, and generating a three-dimensional permeation rate field and a nutrient distribution thermal map of the root zone;

[0046] inputting the candidate set of control strategies and the digital twin simulation results into a Bayesian causal reasoning module to calculate the expected effect probability of each strategy under the current root zone environment, and selecting strategies with a probability value greater than a preset threshold as preliminary instructions;

[0047] applying the preliminary instructions to the greenhouse execution equipment, synchronously deploying soil sensors to monitor soil data changes in real time, feeding back to the digital twin model for parameter calibration, and iteratively optimizing the control instructions until the measured value and the simulation value error is less than a preset threshold, generating a control instruction containing specific parameters.

[0048] As a preferred embodiment of the present embodiment, a greenhouse water, fertilizer and pesticide optimization control system based on the coupling of environment, crop and input factors is used to implement the above-mentioned greenhouse water, fertilizer and pesticide optimization control method based on the coupling of environment, crop and input factors, comprising:

[0049] a data acquisition module: acquiring greenhouse data through a multi-modal sensor network, and processing the greenhouse data to generate a spatio-temporal continuous data set;

[0050] decision rule generation module: for generating a decision rule library based on the generated spatiotemporal continuous data set, quantifying the dynamic causal chain of the environment, crops and input products;

[0051] regulation instruction generation module: for fusing the decision rule library with the real-time simulation results of the root zone three-dimensional digital twin model, generating water, fertilizer and pesticide regulation instructions;

[0052] instruction execution module: sending the fused water, fertilizer and pesticide regulation instructions to the execution equipment through the Internet of Things controller.

[0053] The application provides a greenhouse water, fertilizer and pesticide optimization control system and method based on environment-crop-input product multi-element coupling, which has the following beneficial effects: through deployment of a sensor network to collect greenhouse data and process the greenhouse data to generate a spatiotemporal continuous data set, based on the generated spatiotemporal continuous data set, quantifying the dynamic causal chain of the environment, crops and input products, generating a decision rule library, fusing the decision rule library with the real-time simulation results of the root zone three-dimensional digital twin model, generating water, fertilizer and pesticide regulation instructions, and sending the fused water, fertilizer and pesticide regulation instructions to the execution equipment through the Internet of Things controller; through a multi-modal sensor network, spatiotemporal continuous collection and fusion processing of greenhouse environment-crop-input product data are realized, a decision rule library containing a dynamic causal chain is constructed, real-time simulation of a root zone three-dimensional digital twin model and Bayesian causal reasoning are combined to generate precise water, fertilizer and pesticide regulation instructions, and a closed-loop control mechanism is formed by using the Internet of Things for execution and feedback optimization, effectively quantifying the nonlinear causal relationship among multiple elements and improving the scientific nature of decision-making; through dynamic simulation of the root zone water-nutrient diffusion process by the digital twin model, precise matching and real-time calibration of the regulation instructions are realized; based on reinforcement learning, the rule library is dynamically optimized to adapt to environmental changes and crop growth needs, reduce water, fertilizer and pesticide redundant input, improve resource utilization efficiency, reduce pesticide residue risk, and promote intelligentization and sustainability of greenhouse production. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 A greenhouse water, fertilizer and pesticide optimization control method flowchart based on environment-crop-input product multi-element coupling of the application;

[0055] Figure 2 A greenhouse water, fertilizer and pesticide optimization control system block diagram based on environment-crop-input product multi-element coupling of the application. DETAILED DESCRIPTION

[0056] The embodiments of the application are described in detail below, and examples of the embodiments are shown in the drawings, wherein the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by reference to the drawings are exemplary and are only used to explain the application and cannot be understood as a limitation of the application.

[0057] The disclosure below provides many different embodiments or examples for implementing different structures of the application. For the purpose of simplicity, the description below for a particular embodiment refers only to certain examples and specific configurations. Purposely, this is only in the interest of clarity and managing a reasonable specific number of issues. It is therefore not intended to limit the application to the particular examples described. Furthermore, the application can refer to certain features throughout different examples. Such repetition is for the purpose of simplicity and clarity and does not itself dictate the relationship between the various embodiments and / or configurations discussed.

[0058] As shown in Figure 1 The embodiment of the application provides a greenhouse water, fertilizer and pesticide optimization control method based on environment-crop-input multi-element coupling, which comprises the following steps:

[0059] S1: deploying a sensor network to collect greenhouse data, and processing the greenhouse data to generate a time-space continuous data set;

[0060] The greenhouse data comprises environment data, crop state data and input data.

[0061] Specifically, by deploying environment sensors (temperature and humidity, illumination, CO2), soil sensors (moisture, EC value, pH), crop growth sensors (leaf area index, stem flow rate) and pesticide and fertilizer residue detection devices in the greenhouse, a spatially distributed multi-modal sensing network is formed, and the greenhouse data is collected through the multi-modal sensing network.

[0062] In the embodiment, processing the greenhouse data specifically comprises data cleaning and time-space alignment of outlier rejection, time synchronization, space mapping and data standardization, to obtain a multi-modal data stream with a time stamp.

[0063] A time series database is constructed to store the multi-modal data stream with a time stamp, and the multi-modal data stream with a time stamp generates a time-space continuous data set by fusing historical-real-time data.

[0064] Specifically, the sliding Forest algorithm is used for outlier rejection to filter sensor noise; the DTW algorithm is used for time synchronization to align environment data and crop image data; the ICP point cloud registration is used for space mapping to unify root fiber data and soil moisture sensor coordinates to a three-dimensional root zone model; and the Z-Score normalization is used for data standardization to process heterogeneous indexes such as soil EC value (μS / cm) and plant nitrogen content (g / kg). The past 7 days of historical data and the current data are spliced through a sliding window to generate a time-space continuous data set.

[0065] It should be noted that the time series database stores a spatiotemporal continuous data set.

[0066] Specifically, by sliding the Forest algorithm to eliminate sensor noise, the DTW algorithm to align heterogeneous time series data, and the ICP point cloud registration to unify the spatial coordinate system, combined with Z-Score standardization processing of heterogeneous indicators, a timestamped spatiotemporal continuous data set is constructed, solving the data fault problem caused by the heterogeneity of the greenhouse environment and providing a high-quality data basis for accurate decision-making.

[0067] S2: Based on the generated spatiotemporal continuous data set, quantifying the dynamic causal chain of the environment, crops, and inputs, generating a decision rule library;

[0068] In this embodiment, the specific process of step S2 includes: based on the generated spatiotemporal continuous data set, real-time extraction of time series data of the environment, crops, and inputs, quantification of the dynamic causal relationship among the three, and obtaining a dynamic causal chain;

[0069] Based on the dynamic causal chain, a causal graph of the environment, crops, and inputs is constructed;

[0070] Through a graph neural network with an attention mechanism, the causal graph is modeled, and a dynamic causal chain and a readable decision rule library are outputted;

[0071] Among them, the decision rule library includes water, pesticide, and fertilizer regulation rules based on causal relationships.

[0072] Specifically, based on the generated spatiotemporal continuous data set, real-time extraction of time series data of the environment, crops, and inputs, and quantification of the dynamic causal relationship among the three, the specific process includes:

[0073] Filtering environmental elements from the data pool: for example, temperature, humidity, light, CO2 concentration, crop elements: for example, leaf area index, stem flow rate, root activity, nutrient content, input elements: for example, irrigation amount, fertilizer concentration, pesticide dosage, historical and real-time data;

[0074] Using a method combining time-delay Granger causality test and transfer entropy, the dynamic causal strength among the three elements is quantified;

[0075] Among them, the calculation formula of time-delay Granger causality test is:

[0076] ;

[0077] Among them, Yt represents the value of the dependent variable at time t, p represents the lag order, βt represents the regression coefficient of the lagged term of the dependent variable itself, Yt-i represents the i-th lagged term of the dependent variable, βt-i represents the regression coefficient of the lagged term of the independent variable, This represents the j-th lag term of the independent variable. This represents the residual term.

[0078] The nonlinear contribution of causal relationships is calculated by combining transfer entropy with computation, and the formula is as follows:

[0079] ;

[0080] Where p(·) represents the probability density function, This represents the value of variable Y at time t+1. This represents the value of variable Y at time t. This represents the value of variable X at time t. This represents the probability of Y at time t+1, given only the value of variable Y at time t. This represents the probability of Y at time t+1, given the values ​​of variables X and Y at time t.

[0081] It should be noted that the cause-effect graph G=(V,E) is a graph where the set of nodes V represents the state variables of the environment, crops, and inputs, and the set of edges E represents the causal relationships between the variables.

[0082] In this embodiment, the dynamic causal chain representation formula is:

[0083] ;

[0084] in, This represents the environment state vector at time t. Let represent the crop state vector at time t. Denotes the input vector at time t. The nonlinear function representing the parameterization of GNN, Indicates the causal delay time. This indicates the noise term.

[0085] Specifically, the nonlinear causal quantification method based on time-delay Granger causality test and transfer entropy breaks through the limitations of traditional single environmental factor regulation. It realizes multi-dimensional dynamic correlation analysis between environmental factors such as temperature, humidity, light, and CO2 concentration and crop leaf area index, stem flow rate, nutrient content, and water, fertilizer, and pesticide input, significantly improving the scientific nature of regulation strategies.

[0086] Specifically, graph neural networks use an attention mechanism to model causal graphs, and their calculation formula is as follows:

[0087] ;

[0088] ;

[0089] wherein, represents the attention weight of node i to neighbor node j, represents the transpose of the learnable attention parameter vector, represents an activation function, represents a linear transformation matrix, represents the original feature vector of node i, represents a concatenation operation, represents the result of the linear transformation of the feature of neighbor node j, represents the neighbor set of node i, represents the updated node i feature.

[0090] Specifically, the generation process of the rule base includes:

[0091] extracting a key causal path from a dynamic causal chain;

[0092] generating an "If-Then" rule based on the associated causal path.

[0093] Specifically, the causal graph is modeled using a graph neural network, and the key causal path is extracted in real time to generate an "If-Then" rule base. Compared with a static expert rule base, the water, fertilizer, and pesticide control parameters can be dynamically adjusted to adapt to changes in crop growth stages and the environment.

[0094] S3: Fuse the decision rule base with the real-time simulation results of the root zone three-dimensional digital twin model to generate water, fertilizer, and pesticide control instructions;

[0095] Specifically, environmental data, crop growth data, and input data from a multi-modal sensor network are received in real time, and missing data is completed through a spatiotemporal interpolation algorithm to build a dynamically updated time series database;

[0096] Call the pre-generated decision rule base, based on the current time series database state, calculate the causal weight between nodes through a graph neural network, match the historical decision rules with a similarity greater than a preset threshold to the current greenhouse state, and output a candidate control strategy set;

[0097] The constructed root zone three-dimensional digital twin model inputs real-time soil moisture, nutrient concentration, and drip irrigation belt partition state data, dynamically simulates the diffusion process of water and nutrients in the rhizosphere, and generates a root zone three-dimensional permeation rate field and a nutrient distribution thermal map;

[0098] Input the candidate control strategy set and the digital twin simulation results into the Bayesian causal reasoning module, calculate the expected effect probability of each strategy under the current root zone environment, and select the strategies with a probability value greater than a preset threshold as the preliminary instructions;

[0099] The initial selection command is applied to the greenhouse execution equipment, and the soil sensors deployed simultaneously monitor changes in soil data in real time. The data is fed back to the digital twin model for parameter calibration, and the control command is iteratively optimized until the error between the measured value and the simulated value is less than the preset threshold.

[0100] Generate control instructions containing specific parameters.

[0101] It should be noted that the control instructions include: drip irrigation zone switch status, fertilizer injection rate and pesticide spraying dosage, and expected root zone humidity distribution and nutrient concentration curve.

[0102] S4: The integrated water, fertilizer and pesticide control commands are sent to the execution device via the IoT controller.

[0103] In this embodiment, after S4, the method further includes: continuously collecting environmental parameters, crop physiological indicators, root zone status, and input residue data after execution through a deployed sensor network to form a dynamic feedback dataset;

[0104] Real-time feedback data is input into the root zone 3D digital twin model, the model parameters are dynamically corrected, the virtual space and physical space are synchronously mapped, and the corrected environment-crop response prediction is output.

[0105] Based on feedback data and model correction results, a reinforcement learning algorithm is used to dynamically optimize the decision rule base generated by S2.

[0106] Optimizing the decision rule base generated by S2 includes: adjusting the weights of the environment-crop causal chain, updating the input effect function, and introducing constraints.

[0107] This embodiment provides a greenhouse water, fertilizer, and pesticide optimization control method based on the coupling of multiple factors including environment, crop, and inputs. It utilizes a multimodal sensor network to achieve continuous spatiotemporal acquisition and fusion processing of greenhouse environment, crop, and input data. This constructs a decision rule base containing dynamic causal chains. Combined with real-time simulation of a three-dimensional digital twin model of the root zone and Bayesian causal inference, precise water, fertilizer, and pesticide control commands are generated. The method leverages the Internet of Things (IoT) for execution and feedback optimization, forming a closed-loop control mechanism. This effectively quantifies the nonlinear causal relationships among multiple factors, enhancing the scientific nature of decision-making. The method dynamically simulates the water-nutrient diffusion process in the root zone using a digital twin model, achieving precise matching and real-time calibration of control commands. Based on reinforcement learning, the method dynamically optimizes the rule base, adapting to environmental changes and crop growth needs, reducing redundant water, fertilizer, and pesticide inputs, improving resource utilization efficiency, and simultaneously reducing pesticide residue risks, thus promoting intelligent and sustainable greenhouse production.

[0108] like Figure 2As shown, the embodiment also provides a greenhouse water, fertilizer and pesticide optimization control system based on environment-crop-input coupling, for realizing the greenhouse water, fertilizer and pesticide optimization control method based on environment-crop-input coupling, comprising:

[0109] The data acquisition module acquires greenhouse data through a multi-modal sensor network, processes the greenhouse data, and generates a spatio-temporal continuous data set;

[0110] The decision rule generation module is used to quantify the dynamic causal chain of the environment, the crop and the input based on the generated spatio-temporal continuous data set, and generate a decision rule library;

[0111] The control instruction generation module is used to fuse the decision rule library with the real-time simulation results of the root zone three-dimensional digital twin model, and generate water, fertilizer and pesticide control instructions;

[0112] The instruction execution module sends the water, fertilizer and pesticide control instructions generated by fusion to the execution equipment through the Internet of Things controller.

[0113] In the embodiment, processing the greenhouse data specifically includes outlier rejection, time synchronization, spatial mapping, data standardization for data cleaning and spatio-temporal alignment, to obtain a multi-modal data stream with a timestamp;

[0114] The time-series database is constructed to store the multi-modal data stream with a timestamp, and the multi-modal data stream with a timestamp generates a spatio-temporal continuous data set by fusing historical-real-time data.

[0115] Specifically, the sliding Forest algorithm is used to filter sensor noise for outlier rejection; the DTW algorithm is used for time synchronization to align the environment data and the crop image data; the ICP point cloud registration is used for spatial mapping to unify the root fiber data and the soil moisture sensor coordinates to the three-dimensional root zone model; the Z-Score normalization is used for data standardization to process heterogeneous indexes such as soil EC value (μS / cm) and plant nitrogen content (g / kg). The past 7 days of historical data and the current data are spliced through a sliding window to generate a spatio-temporal continuous data set.

[0116] It should be noted that the time-series database stores the spatio-temporal continuous data set.

[0117] In the embodiment, based on the generated spatio-temporal continuous data set, the dynamic causal chain of the environment, the crop and the input is quantified, and the specific process of generating the decision rule library includes: based on the generated spatio-temporal continuous data set, the time series data of the environment, the crop and the input are extracted in real time, the dynamic causal relationship among the three is quantified, and a dynamic causal chain is obtained;

[0118] Based on the dynamic causal chain, a causal graph of the environment, the crop and the input is constructed;

[0119] The causal graph is modeled by a graph neural network with an attention mechanism, and a dynamic causal chain and a readable decision rule library are output.

[0120] The decision rule library includes water, fertilizer, and pesticide control rules based on causal relationships.

[0121] Specifically, based on the generated spatiotemporal continuous data set, the temporal data of the environment, crops, and inputs is extracted in real time, and the specific process of quantifying the dynamic causal relationship among the three is as follows:

[0122] Filtering environmental factors from the data pool, such as temperature, humidity, light, CO2 concentration, crop factors such as leaf area index, stem flow rate, root activity, nutrient content, and input factors such as irrigation amount, fertilizer concentration, and pesticide dosage, historical and real-time data;

[0123] Using a combination of time-delay Granger causality test and transfer entropy, the dynamic causal strength among the three elements is quantified.

[0124] The calculation formula of time-delay Granger causality test is:

[0125] ;

[0126] Where, Yt represents the value of the dependent variable at time t, p represents the lag order, βt represents the regression coefficient of the lagged term of the dependent variable itself, Yt-i represents the i-th lagged term of the dependent variable, βt represents the regression coefficient of the lagged term of the independent variable, Xt-j represents the j-th lagged term of the independent variable, εt represents the residual term.

[0127] The nonlinear contribution of the causal relationship is quantified by transfer entropy, and the calculation formula is:

[0128] ;

[0129] Where p(·) represents the probability density function, Yt+1 represents the value of variable Y at time t+1, Yt represents the value of variable Y at time t, Xt represents the value of variable X at time t, P(Yt+1|Yt) represents the probability of Y at time t+1 given only the value of Y at time t, P(Yt+1|Xt,Yt) represents the probability of Y at time t+1 given the values of variables X and Y at time t.

[0130] It should be noted that the causal graph G = (V, E), wherein the node set V represents the state variables of the environment, crop and input, and the edge set E represents the causal relationship between variables;

[0131] In this embodiment, the dynamic causal chain representation formula is:

[0132] ;

[0133] wherein, represents the environment state vector at time t, represents the crop state vector at time t, represents the input vector at time t, represents a nonlinear function parameterized by GNN, represents the causal delay time, represents a noise term.

[0134] Specifically, the graph neural network adopts an attention mechanism to model the causal graph, and the calculation formula is:

[0135] ;

[0136] ;

[0137] wherein, represents the attention weight of node i to neighbor node j, represents the transpose of the learnable attention parameter vector, represents an activation function, represents a linear transformation matrix, represents the original feature vector of node i, represents a concatenation operation, represents the result of the linear transformation of the feature of neighbor node j, represents the neighbor set of node i, represents the updated node i feature.

[0138] Specifically, the generation process of the rule base includes:

[0139] extracting key causal paths from the dynamic causal chain;

[0140] generating "If-Then" rules based on the associated causal paths.

[0141] Specifically, the detailed process of generating water, pesticide and fertilizer regulation instructions by fusing the decision rule base with the real-time simulation results of the root zone three-dimensional digital twin model includes:

[0142] Real-time receive environment data, crop growth data and input data from multi-modal sensor network, and complete the missing data through spatio-temporal interpolation algorithm to construct a dynamically updated time series database;

[0143] Call the pre-generated decision rule library, based on the current time series database state, calculate the causal weight between nodes through the graph neural network, match the historical decision rules with similarity greater than the preset threshold to the current greenhouse state, and output the candidate regulation strategy set;

[0144] The constructed root zone three-dimensional digital twin model inputs the current soil moisture, nutrient concentration, and drip irrigation belt partition state data in real time, dynamically simulates the diffusion process of water and nutrients in the rhizosphere, and generates a three-dimensional permeation rate field and a nutrient distribution thermal map of the root zone;

[0145] The candidate regulation strategy set and the digital twin simulation results are input into the Bayesian causal reasoning module to calculate the expected effect probability of each strategy under the current root zone environment, and the strategies with probability values greater than the preset threshold are selected as the preliminary instructions;

[0146] The preliminary instructions are applied to the greenhouse execution equipment, and the soil sensors deployed in synchronization monitor the soil data changes in real time, which are fed back to the digital twin model for parameter calibration, and the regulation instructions are iteratively optimized until the measured value and the simulation value error is less than the preset threshold;

[0147] Generate a regulation instruction containing specific parameters.

[0148] It should be noted that the regulation instruction includes: drip irrigation partition switch state, fertilizer injection rate, pesticide spraying dose, and expected root zone humidity distribution and nutrient concentration curve.

[0149] Furthermore, the greenhouse water, fertilizer and pesticide optimization control system based on the environment-crop-input multi-element coupling of the present application also includes a closed-loop optimization module for optimizing the decision rule library.

[0150] Specifically, the optimization process includes: continuously collecting the environment parameters, crop physiological indicators, root zone state and input residue data after execution through the deployed sensor network to form a dynamic feedback data set;

[0151] Input the real-time feedback data into the root zone three-dimensional digital twin model to dynamically correct the model parameters, realize the synchronous mapping of virtual space and physical space, and output the corrected environment-crop response prediction;

[0152] Based on the feedback data and the model correction results, the decision rule library generated is dynamically optimized using a reinforcement learning algorithm.

[0153] The optimized decision rule library includes: adjusting the environment-crop causal chain weight, updating the input effect function, and introducing constraint conditions

[0154] While embodiments of the application have been shown and described, it is to be understood that the application is not limited to the details of the embodiments described, since numerous changes, modifications, substitutions and variations can be made thereto without departing from the spirit and scope of the application as defined by the appended claims and their equivalents.

Claims

1. A greenhouse water, fertilizer, and pesticide optimization control method based on multi-factor coupling of environment, crop, and inputs, characterized in that, Includes the following steps: S1: Deploy a sensor network to collect greenhouse data and process the data to generate a spatiotemporally continuous dataset; Greenhouse data includes environmental data, crop status data, and input data; S2: Based on the generated spatiotemporal continuous dataset, quantify the dynamic causal chain of environment, crops and inputs, and generate a decision rule base; S3: Integrate the decision rule base with the real-time simulation results of the root zone three-dimensional digital twin model to generate water, pesticide and fertilizer regulation instructions; S4: The integrated water, fertilizer and pesticide control commands are sent to the execution device via the IoT controller.

2. The greenhouse water, fertilizer, and pesticide optimization control method based on multi-factor coupling of environment, crop, and inputs as described in claim 1, is characterized in that... The specific steps for processing greenhouse data to generate a spatiotemporally continuous dataset include: The processing of greenhouse data specifically includes outlier removal, time synchronization, spatial mapping, data standardization, data cleaning and spatiotemporal alignment to obtain a multimodal data stream with timestamps; A time-series database is constructed to store timestamped multimodal data streams. The timestamped multimodal data streams are used to generate spatiotemporally continuous datasets by fusing historical and real-time data.

3. The greenhouse water, fertilizer, and pesticide optimization control method based on multi-factor coupling of environment, crop, and inputs as described in claim 1, is characterized in that... The specific process of step S2 includes: based on the generation of a spatiotemporal continuous dataset, extracting time-series data of the environment, crops and inputs in real time, quantifying the dynamic causal relationship among the three, and obtaining a dynamic causal chain; Based on dynamic causal chains, a causal graph is constructed to extract the environment, crops, and inputs. By using a graph neural network and an attention mechanism to model causal graphs, dynamic causal chains and a readable decision rule base are output. The decision rule base includes water, pesticide, and fertilizer regulation rules based on causal relationships.

4. The greenhouse water, fertilizer, and pesticide optimization control method based on multi-factor coupling of environment, crop, and inputs as described in claim 3, is characterized in that... The specific process for extracting time-series data of the environment, crops, and inputs in real time based on the generated spatiotemporal continuous dataset, and quantifying the dynamic causal relationship among the three, includes: Filter environmental factors, crop factors, input factors, and real-time data from the data pool; A method combining time-delay Granger causality test and transit entropy is used to quantify the dynamic causal strength among the three factors.

5. The greenhouse water, fertilizer, and pesticide optimization control method based on multi-factor coupling of environment, crop, and inputs as described in claim 4, is characterized in that... The formula for calculating the Granger causality test with time delay is as follows: ； in, This represents the value of the dependent variable at time t, and p represents the lag order. The regression coefficient represents the lagged term of the dependent variable itself. This represents the i-th lag term of the dependent variable. The regression coefficient represents the lagged term of the independent variable. This represents the j-th lag term of the independent variable. Represents the residual term; The formula for calculating the nonlinear contribution of causal relationships by combining transfer entropy with computation is as follows: ； Where p(·) represents the probability density function, This represents the value of variable Y at time t+1. This represents the value of variable Y at time t. This represents the value of variable X at time t. This represents the probability of Y at time t+1, given only the value of variable Y at time t. This represents the probability of Y at time t+1, given the values ​​of variables X and Y at time t.

6. The greenhouse water, fertilizer, and pesticide optimization control method based on multi-factor coupling of environment, crop, and inputs as described in claim 3, is characterized in that... The graph neural network uses an attention mechanism to model the causal graph. The calculation formula is as follows: ； ； in, This represents the attention weight of node i to its neighbor node j. This represents the transpose of the learnable attention parameter vector. This represents the activation function. Represents a linear transformation matrix. This represents the original feature vector of node i. This indicates a splicing operation. This represents the result of a linear transformation of the features of neighbor node j. Let i represent the set of neighbors of node i. This represents the updated feature of node i.

7. The greenhouse water, fertilizer, and pesticide optimization control method based on multi-factor coupling of environment, crop, and inputs as described in claim 6, is characterized in that... The process of generating the rule base includes: Extracting key causal paths from dynamic causal chains; "If-Then" rules are generated based on the aforementioned causal path.

8. The greenhouse water, fertilizer, and pesticide optimization control method based on multi-factor coupling of environment, crop, and inputs as described in claim 7, is characterized in that... The specific steps of S3 are as follows: It receives environmental data, crop growth data, and input data from a multimodal sensor network in real time, and completes the missing data through a spatiotemporal interpolation algorithm to build a dynamically updated time-series database. The system calls a pre-generated decision rule library, calculates the causal weights between nodes based on the current time series database state using a graph neural network, matches historical decision rules with a similarity greater than a preset threshold to the current greenhouse state, and outputs a set of candidate control strategies. The constructed three-dimensional digital twin model of the root zone is dynamically simulated by inputting real-time data on current soil moisture, nutrient concentration, and drip irrigation tape zoning status, generating a three-dimensional infiltration rate field and nutrient distribution heat map of the root zone. The candidate control strategy set and the results of the digital twin simulation are input into the Bayesian causal inference module to calculate the expected effect probability of each strategy in the current root region environment, and the strategies with probability values ​​greater than the preset threshold are selected as the initial selection instructions. The initial selection command is applied to the greenhouse execution equipment, and the synchronously deployed soil sensors monitor changes in soil data in real time. The data is fed back to the digital twin model for parameter calibration, and the control command is iteratively optimized until the error between the measured value and the simulated value is less than the preset threshold, thus generating a control command containing specific parameters.

9. A greenhouse water, fertilizer, and pesticide optimization control system based on multi-factor coupling of environment, crop, and inputs, used to implement the greenhouse water, fertilizer, and pesticide optimization control method based on multi-factor coupling of environment, crop, and inputs as described in any one of claims 1-8, characterized in that, include: Data acquisition module: Collects greenhouse data through a multimodal sensor network, processes the greenhouse data, and generates a spatiotemporal continuous dataset; Decision rule generation module: used to quantify the dynamic causal chain of environment, crops and inputs based on the generated spatiotemporal continuous dataset, and generate a decision rule base; Regulation command generation module: used to integrate the decision rule base with the real-time simulation results of the root zone three-dimensional digital twin model to generate water, pesticide and fertilizer regulation commands; Command execution module: Sends the integrated water, fertilizer and pesticide regulation commands to the execution device through the IoT controller.

Images