A brackish water drip irrigation water and fertilizer control method, device and system

Abstract

The application relates to the technical field of intelligent drip irrigation, and discloses a brackish water drip irrigation water and fertilizer control method, equipment and system, the control method comprising the following steps: acquiring soil data, crop growth state data and meteorological data of an irrigation area; adopting a geostatistical method to process the collected soil data of the irrigation area, and drawing a spatial distribution graph of soil salinity of the irrigation area; according to the spatial distribution graph of soil salinity of the irrigation area, the irrigation area is divided into a plurality of blocks; by inputting soil characteristics, crop characteristics and meteorological characteristics into a drip irrigation control strategy network, the drip irrigation control strategy network outputs a drip irrigation control strategy; through obtaining an optimal drip irrigation strategy, the drip irrigation flow and the salt concentration are accurately controlled, it is ensured that every drop of water can be efficiently absorbed by crops, and the water evaporation and seepage loss are reduced, and especially in arid and semiarid areas, the utilization rate of limited fresh water resources is greatly improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent drip irrigation, and more specifically, to a method, equipment, and system for controlling water and fertilizer in brackish water drip irrigation. Background Technology

[0002] The greenhouse brackish water drip irrigation technology refers to the use of brackish water for irrigation in greenhouses. The drip irrigation system evenly drips water with low salt content onto the roots of plants, controlling the supply of soil moisture and salt, thereby achieving the purpose of regulating the distribution of soil water and salt.

[0003] Due to the enclosed environment of greenhouses and the limitations of existing brackish water drip irrigation methods, soil salinity accumulation is severe, especially near the roots. Secondly, uneven soil moisture distribution in greenhouses leads to some areas being overly moist while others are waterlogged. This uneven water and salt distribution causes salt accumulation near the roots, damaging the grape roots, reducing photosynthesis, and resulting in lower yields.

[0004] Therefore, a new method for controlling water and fertilizer in brackish water drip irrigation needs to be designed to precisely control the salt concentration and flow rate of the water from each drip head during multiple drip irrigations in the irrigated area throughout the grape production cycle, in order to control the water and salt distribution in the soil within the irrigated area. Summary of the Invention

[0005] This invention provides a method, equipment, and system for controlling water and fertilizer in slightly saline drip irrigation, which solves the technical problems mentioned in the background art.

[0006] This invention provides a method for controlling water and fertilizer use in drip irrigation of slightly saline water, comprising the following steps:

[0007] Step 101: Obtain soil data, crop growth status data and meteorological data for the irrigated area. Soil data includes soil moisture, salinity concentration and temperature. Crop growth status data includes bud germination rate, new shoot length and number and fruit set rate of grape crops. Meteorological data includes average temperature and extreme temperature, precipitation and sunshine duration and intensity.

[0008] Step 102: Soil data collected from the irrigation area are processed using geostatistical methods, and a spatial distribution map of soil salinity in the irrigation area is drawn. Based on the spatial distribution map of soil salinity in the irrigation area, the irrigation area is divided into several blocks. The division rule is that the similarity index of soil salinity and moisture conditions between any two sampling points in the same block meets the threshold requirement.

[0009] Step 103: Generate soil characteristics for each divided block using soil data, crop characteristics for each divided block using crop growth status data, and meteorological characteristics using meteorological data. Input the soil characteristics, crop characteristics, and meteorological characteristics into the drip irrigation control strategy network, and the drip irrigation control strategy network outputs the drip irrigation control strategy. The drip irrigation control strategy includes the flow rate and salinity concentration corresponding to each drip head.

[0010] In a preferred embodiment, in step 102, the rule for dividing the blocks is that the similarity index of soil salinity and moisture conditions of any two sampling points within the same block meets the threshold requirement. The similarity index is calculated using the moisture similarity index method, and the threshold is set empirically. When the similarity index between two points is greater than the threshold, the soil salinity and moisture conditions of the two sampling points are considered to be similar and belong to the same block; otherwise, they do not belong to the same block.

[0011] In a preferred embodiment, the drip irrigation control strategy network includes a graph data encoding layer, a feature combination layer, a hidden layer, and an output layer. The graph data encoding layer takes soil features and crop features as input and then outputs graph-encoded features. The feature combination layer takes graph-encoded features and meteorological features as input, splices them together, and outputs combined features to the hidden layer. The hidden layer outputs hidden features to the output layer. The output layer outputs the Q-value of each selectable drip irrigation control strategy and selects the drip irrigation control strategy with the largest Q-value as the output.

[0012] In a preferred embodiment, the drip irrigation control strategy network includes:

[0013] Constructing a graph structure:

[0014] The irrigation area is divided into blocks as nodes in a graph structure. When there is a direct adjacency between blocks, it means that there is an edge between the two nodes.

[0015] Graph data encoding layer:

[0016] Input: The initial node feature of each node is a concatenation of the soil and crop features of the block, denoted as . ;

[0017] The calculation formula for the graph data encoding layer is as follows:

[0018]

[0019] in, This represents the block coding characteristics of the i-th block. This represents the initial block characteristics of the i-th block. This represents the attention coefficients of the i-th block and the j-th block. Represents a non-linear activation function. This represents the first weight matrix. This represents the set of neighboring blocks of the i-th block;

[0020] Attention coefficient The calculation formula is:

[0021]

[0022] in, This represents the attention coefficients of the i-th block and the j-th block. This represents the weight parameter, and T represents the transpose. LeakyRelu represents the natural exponential function, and LeakyRelu represents the LeakyRelu activation function. , This represents the first weight matrix. , Let i and j represent the initial block features of the i-th and j-th blocks, respectively. express and Perform vector concatenation;

[0023]

[0024] This represents the block encoding feature of the i-th block, and CONCAT represents vector concatenation. Represents graph encoding features;

[0025] The calculation formula for the feature combination layer is as follows:

[0026]

[0027] Where A represents meteorological characteristics, CONCAT represents vector concatenation, indicating combined features. Represents graph encoding features;

[0028] The formula for calculating the hidden layer is as follows:

[0029]

[0030] in, This indicates the output characteristics of the irrigated area. Represents combined features, This represents the second weight matrix. This represents the bias coefficient of the hidden layer. Represents a non-linear activation function;

[0031] The formula for calculating the output layer is as follows:

[0032]

[0033] in, This represents the Q value of the drip irrigation strategy for the irrigated area. This indicates the output characteristics of the irrigated area. Represents a non-linear activation function. This represents the third weight matrix. This represents the bias parameters of the output layer.

[0034] In a preferred embodiment, the status includes soil data, crop growth status data, and meteorological data;

[0035] Action space: The action space is generated by combining the controllable flow rate of each drip head with the salinity of the effluent.

[0036] One action includes the flow rate and salt concentration for each drip head;

[0037] The flow rate corresponding to each drip head represents the total amount of water output during one drip irrigation process.

[0038] In a preferred embodiment, training the drip control strategy network includes the following steps:

[0039] Step 1: Initialize the experience pool, randomly initialize the drip control strategy network, and initialize... The network parameters are the same as those of the drip control strategy network.

[0040] Step two, reset the environment to obtain the first state;

[0041] Step 3: Generate an action using the ϵ-greedy strategy: There is a probability ϵ that an action will be randomly selected, which is the exploration mode; otherwise, ... , choose in In this state, The most significant action is the experience mode. Calculations are performed via a drip irrigation control strategy network.

[0042] Feedback is obtained based on the interaction between actions and the environment. Next state And whether the termination condition "done" has been triggered, the termination condition being the end of the grape production cycle;

[0043] Experience Storing experience into the experience pool, one experience is a quaternion. or ;

[0044] Step 4: Randomly obtain one experience point from the experience pool;

[0045]

[0046] Representing state Down The network's maximum output;

[0047] according to and Calculate the loss and update the dropper control strategy network using gradient descent.

[0048]

[0049] Representing state and actions The output Q value of the drip irrigation control strategy network;

[0050] It is a discount factor, a value between 0 and 1, used to balance the weight of immediate rewards and future rewards, with a default value of 0.6;

[0051] Step 5: Update every fixed number of network updates. Configure the network so that its parameters are the same as those of the current drip control strategy network;

[0052] Step six, until Once the number of convergences or updates reaches the set value, the process terminates.

[0053] In a preferred embodiment, the feedback The calculation is performed one week after a drip irrigation cycle is completed, using the following formula:

[0054]

[0055] in, This indicates the actual grape yield per acre in the irrigated area. This indicates the standard grape yield per acre in the irrigated area. This represents the soil salinity content of the i-th block. Indicates the number of all blocks. This represents the average soil salinity content across all plots. The soil salinity data for all plots was collected one week after drip irrigation was completed. The standard yield per mu is the average yield per mu of the county or city where the irrigated area is located in the previous year.

[0056] A system for controlling water and fertilizer application in slightly saline drip irrigation includes the following modules:

[0057] The data acquisition module is used to collect soil data, crop growth status data, and meteorological data from multiple irrigation sources.

[0058] The data processing module is used to generate a soil salinity distribution map of the irrigated area based on soil data, crop growth status data and meteorological data, and to divide the irrigated area into multiple blocks based on the soil salinity distribution map;

[0059] The feature generation module is used to generate corresponding feature vectors based on the divided blocks;

[0060] A drip irrigation control strategy network is used to generate drip irrigation control strategies;

[0061] The drip irrigation execution module is used to set and implement precise flow rates and salt concentrations for each drip head according to the drip irrigation control strategy.

[0062] A brackish water drip irrigation fertilization control device includes a memory that stores non-transitory computer-readable instructions. When the non-transitory computer-readable instructions are executed by a computer, the steps in the brackish water drip irrigation fertilization control method described above can be performed.

[0063] The beneficial effects of this invention are as follows:

[0064] 1. This invention obtains the optimal drip irrigation strategy, precisely controls the drip irrigation flow rate and salt concentration, ensures that every drop of water can be efficiently absorbed by crops, reduces water evaporation and seepage losses, and greatly improves the utilization rate of limited freshwater resources, especially in arid and semi-arid regions.

[0065] 2. Through geostatistical analysis and zoning management, this invention can make more scientific use of saline-slightly brackish water resources, rationally arrange irrigation plans, prevent salt accumulation in the soil surface, help improve and control the trend of soil salinization, and protect arable land resources. Attached Figure Description

[0066] Figure 1 This is a flowchart of the water and fertilizer control method for drip irrigation with slightly saline water according to the present invention.

[0067] Figure 2 This is a block diagram of the brackish water drip irrigation and fertilization control system of the present invention.

[0068] In the diagram: 100, data acquisition module; 200, data processing module; 300, feature generation module; 400, drip irrigation control strategy network; 500, drip irrigation execution module. Detailed Implementation

[0069] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0070] like Figure 1 As shown, a method for controlling water and fertilizer in slightly saline drip irrigation includes the following steps:

[0071] Step 101: Obtain soil data, crop growth status data and meteorological data for the irrigated area. Soil data includes soil moisture, salinity concentration and temperature. Crop growth status data includes bud germination rate, new shoot length and number and fruit set rate of grape crops. Meteorological data includes average temperature and extreme temperature, precipitation and sunshine duration and intensity.

[0072] It should be noted that soil data is acquired through an established monitoring network. Soil moisture, salinity, and temperature sensors are evenly deployed within the irrigation area to ensure comprehensive coverage of different management blocks. To collect time-series data, the same time schedule is used for data collection. The simplest method is to sample at fixed time intervals, such as 6 hours. Crop growth status data is acquired by deploying various sensor networks in the vineyard to monitor environmental parameters and crop status in real time. The data is transmitted wirelessly to the central management system. Meteorological data is acquired by connecting to weather stations covering the vineyard irrigation area. Weather forecast information from the weather stations is directly imported into the central management system to obtain meteorological data.

[0073] Step 102: Soil data collected from the irrigation area are processed using geostatistical methods, and a spatial distribution map of soil salinity in the irrigation area is drawn. Based on the spatial distribution map of soil salinity in the irrigation area, the irrigation area is divided into several blocks. The division rule is that the similarity index of soil salinity and moisture conditions between any two sampling points in the same block meets the threshold requirement.

[0074] Geostatistical methods were used to process the collected soil data from historical irrigation areas, including the use of geostatistical methods such as Kriging interpolation and semivariogram analysis, along with GIS (Geographic Information System) technology, to analyze the spatial variability of soil salinity and to draw spatial distribution maps of soil salinity.

[0075] The division rule is that the similarity index of soil salinity and moisture conditions of any two sampling points within the same block meets the threshold requirement. The similarity index is calculated using the moisture similarity index method. The threshold is set empirically, for example, the threshold is 0.8. When the similarity index between two points is greater than 0.8, they are considered to have similar soil salinity and moisture conditions and belong to the same block. Otherwise, they do not belong to the same block.

[0076] Step 103: Generate soil characteristics for each divided block using soil data, crop characteristics for each divided block using crop growth status data, and meteorological characteristics using meteorological data. Input the soil characteristics, crop characteristics, and meteorological characteristics into the drip irrigation control strategy network, and the drip irrigation control strategy network outputs the drip irrigation control strategy. The drip irrigation control strategy includes the flow rate and salinity concentration corresponding to each drip head.

[0077] The drip irrigation control strategy network consists of a graph data encoding layer, a feature combination layer, a hidden layer, and an output layer. The graph data encoding layer takes soil and crop features as input and outputs graph-encoded features. The feature combination layer takes graph-encoded features and meteorological features as input, concatenates them, and outputs combined features to the hidden layer. The hidden layer outputs hidden features to the output layer. The output layer outputs the Q-value of each optional drip irrigation control strategy and selects the drip irrigation control strategy with the largest Q-value as the output.

[0078] Constructing a graph structure:

[0079] The irrigation area is divided into blocks, which are treated as nodes in a graph structure. If there is a direct adjacency between blocks, it means that there is an edge between the two nodes, thus forming a graph. ;

[0080] Graph data encoding layer:

[0081] Input: Each node The initial node features are a combination of soil and crop features of the block, denoted as . ;

[0082] The calculation formula for the graph data encoding layer is as follows:

[0083]

[0084] in, This represents the block coding characteristics of the i-th block. This represents the initial block characteristics of the i-th block. This represents the attention coefficients of the i-th block and the j-th block. This represents a non-linear activation function such as the ReLU function. This represents the first weight matrix. This represents the set of neighboring blocks of the i-th block;

[0085] Attention coefficient The calculation formula is:

[0086]

[0087] in, This represents the attention coefficients of the i-th block and the j-th block. This represents the weight parameter, and T represents the transpose. LeakyRelu represents the natural exponential function, and LeakyRelu represents the LeakyRelu activation function. , This represents the first weight matrix. , Let i and j represent the initial block features of the i-th and j-th blocks, respectively. express and Perform vector concatenation;

[0088]

[0089] This represents the block encoding feature of the i-th block, and CONCAT represents vector concatenation. Represents graph encoding features;

[0090] The calculation formula for the feature combination layer is as follows:

[0091]

[0092] Where A represents meteorological characteristics, CONCAT represents vector concatenation, indicating combined features. Represents graph encoding features;

[0093] The formula for calculating the hidden layer is as follows:

[0094]

[0095] in, This indicates the output characteristics of the irrigated area. Represents combined features, This represents the second weight matrix. This represents the bias coefficient of the hidden layer. This represents a non-linear activation function such as the ReLU function;

[0096] The formula for calculating the output layer is as follows:

[0097]

[0098] in, This represents the Q value of the drip irrigation strategy for the irrigated area. This indicates the output characteristics of the irrigated area. This represents a non-linear activation function such as the ReLU function. This represents the third weight matrix. Indicates the bias parameters of the output layer;

[0099] It should be noted that soil data, growth status data, and meteorological data need to be updated before each drip irrigation operation to regenerate the drip irrigation control strategy;

[0100] Training the drip irrigation control strategy network includes the following steps:

[0101] Step 1: Initialize the experience pool, randomly initialize the drip control strategy network, and initialize... The network parameters are the same as those of the drip control strategy network.

[0102] Step two, reset the environment to obtain the first state;

[0103] Step 3: Generate an action using the ϵ-greedy strategy: There is a probability ϵ that an action will be randomly selected, which is the exploration mode; otherwise, ... , choose in In this state, The most significant action is the experience mode. Calculations are performed via a drip irrigation control strategy network.

[0104] Feedback is obtained based on the interaction between actions and the environment. Next state And whether the termination condition "done" has been triggered, the termination condition being the end of the grape production cycle;

[0105] Experience Storing experience into the experience pool, one experience is a quaternion. or ;

[0106] Step 4: Randomly obtain one experience point from the experience pool;

[0107]

[0108] Representing state Down The network's maximum output;

[0109] according to and Calculate the loss and update the dropper control strategy network using gradient descent.

[0110]

[0111] Representing state and actions The output Q value of the drip irrigation control strategy network;

[0112] It is a discount factor, a value between 0 and 1, used to balance the weight of immediate rewards and future rewards, with a default value of 0.6;

[0113] Step 5: Update every fixed number of network updates. Configure the network so that its parameters are the same as those of the current drip control strategy network;

[0114] Step six, until Once the network convergence or the number of updates reaches the set value, the process terminates.

[0115] Among them, feedback The calculation is performed one week after a drip irrigation cycle is completed, using the following formula:

[0116]

[0117] in, This indicates the actual grape yield per acre in the irrigated area. This indicates the standard grape yield per acre in the irrigated area. This represents the soil salinity content of the i-th block. Indicates the number of all blocks. This represents the average soil salinity across all blocks.

[0118] The soil salinity data for all blocks was collected one week after drip irrigation was completed; the standard yield per mu was the average yield per mu of the county or city where the irrigated area was located in the previous year.

[0119] The purpose of this incentive design is to encourage a more even distribution of soil salinity across different zones, thereby avoiding the adverse effects of excessively high or low salinity concentrations on the plants and maximizing yield.

[0120] The status includes soil data, crop growth status data, and meteorological data;

[0121] Action space: The action space is generated by combining the controllable flow rate of each drip head with the salinity of the effluent.

[0122] One action includes the flow rate and salt concentration for each drip head;

[0123] Here, flow rate is defined as the total amount of water discharged during a single drip irrigation process;

[0124] Since both flow rate and salinity concentration are continuous values, the controllable flow rate and effluent salinity concentration range are discretized into point values ​​and then combined to generate actions.

[0125] like Figure 2 As shown, a slightly saline water drip irrigation and fertilization control system includes the following modules:

[0126] Data acquisition module 100 is used to collect soil data, crop growth status data and meteorological data from multiple irrigation sources.

[0127] The data processing module 200 is used to generate a soil salinity distribution map of the irrigation area based on soil data, crop growth status data and meteorological data of the irrigation area, and to divide the irrigation area into multiple blocks based on the soil salinity distribution map;

[0128] The feature generation module 300 is used to generate corresponding feature vectors based on the divided blocks;

[0129] Drip irrigation control strategy network 400 is used to generate drip irrigation control strategies;

[0130] The drip irrigation execution module 500 is used to set and implement precise flow rates and salt concentrations for each drip head according to the drip irrigation control strategy.

[0131] A brackish water drip irrigation fertilization control device includes a memory that stores non-transitory computer-readable instructions. When the non-transitory computer-readable instructions are executed by a computer, the steps in the brackish water drip irrigation fertilization control method described above can be performed.

[0132] The embodiments of this example have been described above. However, this example is not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of this example, and all of them are within the protection scope of this example.

Claims

1. A method for controlling water and fertilizer application in slightly saline drip irrigation, characterized in that, Includes the following steps: Step 101: Obtain soil data, crop growth status data and meteorological data for the irrigated area. Soil data includes soil moisture, salinity concentration and temperature. Crop growth status data includes bud germination rate, new shoot length and number and fruit set rate of grape crops. Meteorological data includes average temperature and extreme temperature, precipitation and sunshine duration and intensity. Step 102: Soil data collected from the irrigation area are processed using geostatistical methods, and a spatial distribution map of soil salinity in the irrigation area is drawn. Based on the spatial distribution map of soil salinity in the irrigation area, the irrigation area is divided into several blocks. The division rule is that the similarity index of soil salinity and moisture conditions between any two sampling points in the same block meets the threshold requirement. Step 103: Generate soil characteristics for each divided block using soil data, generate crop characteristics for each divided block using crop growth status data, and generate meteorological characteristics using meteorological data. Input the soil characteristics, crop characteristics, and meteorological characteristics into the drip irrigation control strategy network, and output the drip irrigation control strategy network. The drip irrigation control strategy includes the flow rate and salt concentration corresponding to each drip head. In step 102, the rule for dividing the blocks is that the similarity index of soil salinity and moisture conditions of any two sampling points within the same block meets the threshold requirement. The similarity index is calculated using the moisture similarity index method. The threshold is set empirically. When the similarity index between two points is greater than the threshold, the soil salinity and moisture conditions of the two sampling points are considered to be similar and belong to the same block. Otherwise, they do not belong to the same block. The drip irrigation control strategy network includes a graph data encoding layer, a feature combination layer, a hidden layer, and an output layer. The graph data encoding layer takes soil features and crop features as input and then outputs graph-encoded features. The feature combination layer takes input image encoding features and meteorological features, splices them together, outputs the combined features to the hidden layer, the hidden layer outputs the hidden features to the output layer, the output layer outputs the Q value of each optional drip irrigation control strategy, and selects the drip irrigation control strategy with the largest Q value as the output; The drip irrigation control strategy network includes: Constructing a graph structure: The irrigation area is divided into blocks as nodes in a graph structure. When there is a direct adjacency between blocks, it means that there is an edge between the two nodes. Graph data encoding layer: Input: The initial node feature of each node is a concatenation of the soil and crop features of the block, denoted as . ; The calculation formula for the graph data encoding layer is as follows: in, This represents the block coding characteristics of the i-th block. This represents the initial block characteristics of the i-th block. This represents the attention coefficients of the i-th block and the j-th block. Represents a non-linear activation function. This represents the first weight matrix. This represents the set of neighboring blocks of the i-th block; Attention coefficient The calculation formula is: in, This represents the attention coefficients of the i-th block and the j-th block. This represents the weight parameter, and T represents the transpose. LeakyRelu represents the natural exponential function, and LeakyRelu represents the LeakyRelu activation function. , This represents the first weight matrix. , Let i and j represent the initial block features of the i-th and j-th blocks, respectively. express and Perform vector concatenation; This represents the block encoding feature of the i-th block, and CONCAT represents vector concatenation. Represents graph encoding features; The calculation formula for the feature combination layer is as follows: Where A represents meteorological characteristics, CONCAT represents vector concatenation, indicating combined features. Represents graph encoding features; The formula for calculating the hidden layer is as follows: in, This indicates the output characteristics of the irrigated area. Represents combined features, This represents the second weight matrix. This represents the bias coefficient of the hidden layer. Represents a non-linear activation function; The calculation formula for the output layer is as follows: in, This represents the Q value of the drip irrigation strategy for the irrigated area. This indicates the output characteristics of the irrigated area. Represents a non-linear activation function. This represents the third weight matrix. Indicates the bias parameters of the output layer; The status includes soil data, crop growth status data, and meteorological data; Action space: The action space is generated by combining the controllable flow rate of each drip head with the salinity of the effluent. One action includes the flow rate and salt concentration for each drip head; The flow rate corresponding to each drip head represents the total amount of water output during one drip irrigation process.

2. The method for controlling water and fertilizer application in slightly saline drip irrigation according to claim 1, characterized in that, Training the drip irrigation control strategy network includes the following steps: Step 1: Initialize the experience pool, randomly initialize the drip control strategy network, and initialize... The network parameters are the same as those of the drip control strategy network. Step two, reset the environment to obtain the first state; Step 3: Generate an action using the ϵ-greedy strategy: There is a probability ϵ that an action will be randomly selected, which is the exploration mode; otherwise, ... , choose in In this state, The most significant action is the experience mode. Calculations are performed via a drip irrigation control strategy network. Feedback is obtained based on the interaction between actions and the environment. Next state And whether the termination condition "done" has been triggered, the termination condition being the end of the grape production cycle; Experience Storing experience into the experience pool, one experience is a quaternion. or ; Step 4: Randomly obtain one experience point from the experience pool; Representing state Down The network's maximum output; according to and Calculate the loss and update the dropper control strategy network using gradient descent. Representing state and actions The output Q value of the drip irrigation control strategy network; It is a discount factor, a value between 0 and 1, used to balance the weight of immediate rewards and future rewards, with a default value of 0.6; Step 5: Update every fixed number of network updates. Configure the network so that its parameters are the same as those of the current drip control strategy network; Step six, until Once the number of convergences or updates reaches the set value, the process terminates.

3. The method for controlling water and fertilizer application in slightly saline drip irrigation according to claim 2, characterized in that, in, Feedback The calculation is performed one week after a drip irrigation cycle is completed, using the following formula: in, This indicates the actual grape yield per acre in the irrigated area. This indicates the standard grape yield per acre in the irrigated area. This represents the soil salinity content of the i-th block. Indicates the number of all blocks. This represents the average soil salinity content across all plots. The soil salinity data for all plots was collected one week after drip irrigation was completed. The standard yield per mu is the average yield per mu of the county or city where the irrigated area is located in the previous year.

4. A system for controlling water and fertilizer application in a slightly saline drip irrigation method, used to perform the steps in the slightly saline drip irrigation water and fertilizer application method as described in claim 1, characterized in that, Includes the following modules: The data acquisition module is used to collect soil data, crop growth status data, and meteorological data from multiple irrigation sources. The data processing module is used to generate a soil salinity distribution map of the irrigated area based on soil data, crop growth status data and meteorological data, and to divide the irrigated area into multiple blocks based on the soil salinity distribution map; The feature generation module is used to generate corresponding feature vectors based on the divided blocks; A drip irrigation control strategy network is used to generate drip irrigation control strategies; The drip irrigation execution module is used to set and implement precise flow rates and salt concentrations for each drip head according to the drip irrigation control strategy.

5. A brackish water drip irrigation fertilization control device, comprising a memory storing non-transitory computer-readable instructions, which, when executed by a computer, are capable of performing the steps in the brackish water drip irrigation fertilization control method as described in claim 1.

Images