A dynamic decision-making method and system for regulating water, fertilizer and salt in saline-alkali soil soybeans

Abstract

The present invention relates to the field of water-fertilizer regulation in saline-alkali land, and specifically to a dynamic decision-making method and system for water-fertilizer-salt regulation of soybeans in saline-alkali land. A dynamic decision-making system for water-fertilizer-salt regulation of soybeans in saline-alkali land comprises: a regional division module, a data monitoring module, an irrigation initial decision module, a fertilization initial decision module, an irrigation decision acquisition module and a fertilization decision acquisition module. The present invention integrates advanced sensor technology and intelligent decision-making models to accurately monitor key parameters, and combines the crop growth stage and historical water and fertilizer records to calculate the total amount of irrigation and fertilization required within a specific time window in the future; at the same time, it can dynamically adjust the irrigation and fertilization frequency according to the real-time monitored soil and environmental data, avoid the waste of water resources and fertilizers, and possible soil salinization problems, help maintain soil health, and promote crop growth.

Description

Technical Field

[0001] The invention relates to the field of water-fertilizer regulation in saline-alkali land, and in particular to a dynamic decision-making method and system for water-fertilizer-salt regulation of soybeans in saline-alkali land. Background Art

[0002] In China and even around the world, the utilization of saline-alkali land is of great significance for increasing agricultural production and ensuring food security. Saline-alkali land limits crop growth, especially soybean cultivation, due to its high salt concentration, poor soil structure and low fertility. Traditional water, fertilizer and salt management methods often rely on experience and lack the ability to respond to changes in soil and environmental conditions in real time, resulting in the inability to accurately meet the crops' needs for water and nutrients at different growth stages. In addition, these methods usually do not take into account the soil's infiltration capacity and water holding capacity, which is crucial for water, fertilizer and salt management in saline-alkali land, because these soil characteristics directly affect the distribution of water and nutrients and the absorption efficiency of crops.

[0003] With the advancement of science and technology, especially the development of the Internet of Things, big data and artificial intelligence technologies, modern agricultural management has gradually developed towards intelligence and precision; deep learning technologies, such as convolutional neural networks (CNN) and long short-term memory networks (LSTM), have provided new possibilities for processing and analyzing complex agricultural data; however, although these technologies have shown great potential in other fields, their application in water, fertilizer and salt regulation of soybeans in saline-alkali land is still limited; existing decision-making models often fail to comprehensively consider multiple factors of soil, environment and crop growth, resulting in the inability to achieve optimal water, fertilizer and salt management.

[0004] Therefore, developing a method and system that can dynamically adjust water, fertilizer and salt management measures to adapt to the complex soil environment and crop growth requirements of saline-alkali land is of great practical significance for improving the planting efficiency and yield of soybeans in saline-alkali land; this can not only increase agricultural output, but also promote the sustainable use of saline-alkali land, and have a far-reaching impact on global food security and sustainable agricultural development. Summary of the invention

[0005] The present invention integrates advanced sensor technology and intelligent decision-making models to accurately monitor key parameters such as soil moisture, temperature, salt concentration, ambient temperature and humidity, and combines the crop growth stage and historical water and fertilizer records to calculate the total amount of irrigation and fertilization required within a specific time window in the future. At the same time, taking into account the real-time changes in soil properties and environmental conditions, the present invention can dynamically adjust the frequency of irrigation and fertilization according to the real-time monitored soil and environmental data to avoid waste of water resources and fertilizers, as well as possible soil salinization problems. This dynamic adjustment capability makes water, fertilizer and salt management more accurate, which not only improves the utilization efficiency of water and fertilizer resources, but also helps maintain soil health and promote crop growth, especially in saline-alkali land, an environment with higher requirements for water, fertilizer and salt management.

[0006] A dynamic decision-making method for water, fertilizer and salt regulation of soybeans in saline-alkali land, comprising:

[0007] Set the region division cycle. At the beginning of any region division cycle, the soybean planting area is evenly divided into several local control areas of the same size in a grid shape. The grid spacing is selected by the particle swarm optimization algorithm.

[0008] For any local control area, first soil data and environmental data are collected at the current first monitoring time point, the first soil data includes soil moisture, soil temperature and soil salt concentration; the environmental data includes ambient temperature and ambient humidity; it is determined whether the currently collected soil moisture is higher than the preset moisture threshold, if so, no operation is performed; if not, the irrigation initial decision model is used to obtain the total irrigation volume data within the first time window in the future;

[0009] The most recent irrigation data is obtained, and the obtained irrigation data is analyzed to obtain the soil infiltration capacity and soil water holding capacity; the irrigation frequency adjustment value is calculated using the soil infiltration capacity, soil water holding capacity and soil salt concentration; the first component value of n is calculated using the obtained irrigation frequency adjustment value, where n is a positive integer; the total irrigation amount data is evenly divided into n preparatory irrigation amount data using the first component value; for any preparatory irrigation amount data, the rainfall data corresponding to the current preparatory irrigation amount data is obtained, and the actual irrigation amount data corresponding to the current preparatory irrigation amount is obtained by subtracting the rainfall data from the preparatory irrigation amount data; finally, the obtained n actual irrigation amount data are combined into an irrigation decision data set;

[0010] For any local control area, the second soil data and environmental data are collected at the current second monitoring time point, and the second soil data includes soil nutrients, soil salt concentration and soil pH value; it is determined whether the soil nutrients are higher than the preset nutrient threshold, if so, no operation is performed; if not, the total amount of fertilizer data and fertilizer type in the future second time window are obtained using the fertilization initial decision model;

[0011] A fertilization frequency adjustment value is calculated based on the obtained fertilization type and soil salt concentration; a second component value of size m is calculated using the obtained fertilization frequency adjustment value, where m is a positive integer; the total fertilization amount data is evenly divided into m preliminary fertilization amount data using the second component value; and the m preliminary fertilization amount data obtained are combined into a fertilization decision data set;

[0012] The obtained irrigation decision data set and fertilization decision data set are used for regulation and management.

[0013] Preferably, the irrigation initial decision model is used to obtain the total irrigation volume data in the first future time window, and the specific operations are as follows:

[0014] Set the duration to The first time window of the future is used to obtain the predicted environmental data within the first time window, and the first crop growth data and historical irrigation records are obtained. The first crop growth data includes the crop growth stage and the crop water requirement. The first soil data, environmental data, predicted environmental data, first crop growth data and historical irrigation records currently obtained are combined into a first non-time series data set. At the same time, the most recent All the first soil data and environmental data collected within the time are sorted in time to form a first time series data set; the first non-time series data set and the first time series data set are used as inputs of the initial irrigation decision model, and the total irrigation volume data within the first time window in the future is output;

[0015] The initial fertilization decision model is used to obtain the total fertilization data and fertilization type in the second time window in the future. The specific operations are as follows:

[0016] Set the duration to The second time window, > , obtain the predicted environmental data within the second time window in the future, and obtain the second crop growth data and historical fertilization records; the second crop growth data includes the crop growth stage, crop nutrient deficiency symptoms and crop nutrient requirements, and the currently acquired second soil data, environmental data, predicted environmental data, second crop growth data and historical fertilization records are combined into a second non-time series data set; at the same time, the most recent All the second soil data and environmental data collected within the time are sorted in time to form a second time series data set; the second non-time series data set and the second time series data set are used as inputs of the initial fertilization decision model, and the total amount of fertilizer data and fertilization type in the future second time window are output.

[0017] Preferably, the most recent irrigation data is obtained, and the obtained irrigation data is analyzed to obtain the soil permeability and soil water holding capacity; the irrigation frequency adjustment value is calculated using the soil permeability, soil water holding capacity and soil salt concentration. The specific operations are as follows:

[0018] Get the most recent irrigation data, including irrigation duration t and irrigation amount x;

[0019] Using the formula Calculate the irrigation water flow Q; then use the formula Calculate and obtain the soil permeability K of the current area; where L represents the soil thickness of the current area, s represents the area of ​​the current area, is the water head difference in the current area;

[0020] Using the formula Calculate and obtain the moisture change rate B; where, It represents the time interval from the start of irrigation until the soil moisture in the current area reaches the field capacity; Indicated in The soil moisture changes over time; then use the formula Calculate and obtain the soil water holding capacity P of the current area;

[0021] Set the value range of the first component to to ; Also set the maximum soil penetration capacity and minimum soil infiltration capacity ; Maximum soil water holding capacity and minimum soil water holding capacity ; Maximum soil salt concentration and minimum soil salt concentration ; Among them, the maximum soil penetration capacity , maximum soil water holding capacity and minimum soil salt concentration The value corresponding to the first component is ;Minimum soil penetration capacity , minimum soil water holding capacity and maximum soil salt concentration The value corresponding to the first component is ;

[0022] For the soil permeability K, use the formula Calculate the penetration adjustment value ; For soil water holding capacity P, use the formula Calculate the water holding capacity adjustment value ; For soil salt concentration S, use the formula Calculate the salt concentration adjustment value ;

[0023] Finally, using the formula Calculate the irrigation frequency adjustment value .

[0024] Preferably, the fertilization frequency adjustment value is calculated based on the obtained fertilization type and soil salt concentration, and the specific operation is as follows:

[0025] For any fertilization type, set the value range of the second component corresponding to the current fertilization type to to , i = 1, 2, ..., u; u represents the number of fertilization types; soil salt concentration is When , the corresponding second component value is ; Soil salt concentration is When , the corresponding second component value is ; Apply the obtained salt concentration adjustment value , as the fertilization frequency adjustment value .

[0026] Preferably, the first component value of size n is obtained by calculating the obtained irrigation frequency adjustment value, and the specific operation is as follows:

[0027] Using the formula Calculation acquisition ;judge Is it an integer? If so, apply as the first component value; if not, the closest An integer as the first component value;

[0028] The obtained fertilization frequency adjustment value is used to calculate the second component value of size m. The specific operation is as follows:

[0029] Based on the output of the initial fertilization decision model, the type of fertilization required in the second future time window is calculated using the formula Calculate the corresponding fertilization type ;judge Is it an integer? If so, apply as the second component value; if not, the closest An integer as the second component value.

[0030] Preferably, the irrigation initial decision model is established based on CNN and LSTM, including a first input layer, a first CNN layer, a first LSTM layer, a first fully connected layer and a first output layer; the first input layer is used to receive the first non-time series data set and the first time series data set; the first CNN layer is used to extract the non-time series features of the first non-time series data set; the first LSTM layer is used to extract the time series features of the first time series data set; the first fully connected layer is used to fuse the non-time series features and time series features extracted by the first CNN layer and the first LSTM layer, and generate an output result; the first output layer is used to output the total irrigation volume data within the first time window in the future;

[0031] The specific operations for training the initial irrigation decision model are as follows:

[0032] A number of irrigation decision training samples are obtained, each of which contains a historical first non-time series data set, a historical first time series data set and corresponding historical total irrigation data; all irrigation decision training samples are divided into an irrigation decision training set and an irrigation decision verification set, the historical first non-time series data set and the historical first time series data set are used as inputs of the irrigation initial decision model, the historical total irrigation data is used as output of the irrigation initial decision model, the irrigation decision training set is used to train the irrigation initial decision model, the irrigation decision verification set is input into the irrigation initial decision model, and a first verification result is obtained; a first training condition is set, and it is determined whether the first verification result meets the first training condition. If so, the trained irrigation initial decision model is output; if not, the irrigation decision training set is continued to be used to train the irrigation initial decision model.

[0033] Preferably, the initial fertilization decision model is established based on CNN and LSTM, including a second input layer, a second CNN layer, a second LSTM layer, a second fully connected layer and a second output layer; the second input layer is used to receive the second non-time series data set and the second time series data set; the second CNN layer is used to extract the non-time series features of the second non-time series data set; the second LSTM layer is used to extract the time series features of the second time series data set; the second fully connected layer is used to fuse the non-time series features and time series features extracted by the second CNN layer and the second LSTM layer, and generate an output result; the second output layer is used to output the total amount of fertilization data and the type of fertilization required in the second time window in the future;

[0034] The specific operations for training the initial fertilization decision model are as follows:

[0035] A number of fertilization decision training samples are obtained, each of which contains a historical second non-time series data set, a historical second time series data set, and corresponding historical total fertilization data and fertilization type; all fertilization decision training samples are divided into a fertilization decision training set and a fertilization decision verification set, the historical second non-time series data set and the historical second time series data set are used as inputs of the initial fertilization decision model, the historical total fertilization data and the fertilization type are used as outputs of the initial fertilization decision model, the initial fertilization decision model is trained using the fertilization decision training set, the fertilization decision verification set is input into the initial fertilization decision model, and a second verification result is obtained; a second training condition is set, and it is determined whether the second verification result meets the second training condition. If so, the trained initial fertilization decision model is output; if not, the initial fertilization decision model is continuously trained using the fertilization decision training set.

[0036] Preferably, a dynamic decision-making system for regulating water, fertilizer and salt in saline-alkali soil soybeans comprises:

[0037] The area division module is used to divide the soybean planting area into several local control areas of equal size in a grid-like manner through a particle swarm optimization algorithm;

[0038] The data monitoring module includes a first data monitoring unit and a second data monitoring unit; the first data monitoring unit is used to collect first soil data and environmental data for any local control area at a first monitoring time point, the first soil data including soil moisture, soil temperature and soil salt concentration; the environmental data including ambient temperature and ambient humidity; the second data monitoring unit is used to collect second soil data and environmental data for any local control area at a second monitoring time point, the second soil data including soil nutrients, soil salt concentration and soil pH value;

[0039] The irrigation initial decision module is used to determine whether the currently collected soil moisture is higher than the preset moisture threshold. If so, no operation is performed; if not, the irrigation initial decision model is used to obtain the total irrigation volume data in the first time window in the future;

[0040] The initial fertilization decision module is used to determine whether the soil nutrients are higher than the preset nutrient threshold. If so, no operation is performed; if not, the initial fertilization decision model is used to obtain the total amount of fertilization data and fertilization type in the second time window in the future;

[0041] The irrigation decision acquisition module includes a soil permeability calculation unit, a soil water holding capacity calculation unit and an irrigation decision data set acquisition unit; the soil permeability calculation unit is used to obtain the most recent irrigation data, analyze the obtained irrigation data, and obtain the soil permeability; the soil water holding capacity calculation unit is used to analyze the obtained irrigation data and obtain the soil water holding capacity; the irrigation decision data set acquisition unit is used to calculate and obtain the irrigation frequency adjustment value using the soil permeability, soil water holding capacity and soil salt concentration; the obtained irrigation frequency adjustment value is used to calculate and obtain the first component value of size n; the first component value is used to evenly divide the total irrigation amount data into n preparatory irrigation amount data; for any preparatory irrigation amount data, the rainfall data corresponding to the current preparatory irrigation amount data is obtained, and the rainfall data is subtracted from the preparatory irrigation amount data to obtain the actual irrigation amount data corresponding to the current preparatory irrigation amount; and finally the obtained n actual irrigation amount data are formed into an irrigation decision data set;

[0042] The fertilization decision acquisition module is used to calculate the fertilization frequency adjustment value using the obtained fertilization type and soil salt concentration; use the obtained fertilization frequency adjustment value to calculate the second component value of size m; use the second component value to evenly divide the total fertilization amount data into m preliminary fertilization amount data; and form a fertilization decision data set with the obtained m preliminary fertilization amount data.

[0043] The present invention has the following advantages:

[0044] 1. The present invention integrates advanced sensor technology and intelligent decision-making models to accurately monitor key parameters such as soil moisture, temperature, salt concentration, ambient temperature and humidity, and combines the crop growth stage and historical water and fertilizer records to calculate the total amount of irrigation and fertilization required within a specific time window in the future; at the same time, taking into account the real-time changes in soil properties and environmental conditions, the present invention can dynamically adjust the frequency of irrigation and fertilization according to the real-time monitored soil and environmental data to avoid waste of water resources and fertilizers, as well as possible soil salinization problems. This dynamic adjustment capability makes water, fertilizer and salt management more accurate, which not only improves the utilization efficiency of water and fertilizer resources, but also helps maintain soil health and promote crop growth, especially in saline-alkali land, which has higher requirements for water, fertilizer and salt management.

[0045] 2. The present invention adopts a particle swarm optimization algorithm to grid the soybean planting area, thereby achieving precise management of each local control area; this grid management strategy enables each area to obtain a customized water, fertilizer and salt management plan based on its unique soil and environmental conditions, significantly improving the soil management efficiency and crop response speed of saline-alkali land; in addition, the present invention uses a real-time data-driven method to enable management measures to respond quickly to changes in soil conditions, thereby adjusting irrigation and fertilization strategies in a timely manner and reducing adverse effects on crop growth; through this rapid response mechanism, the present invention not only improves the flexibility and adaptability of saline-alkali land soil management, but also enhances the adaptability of crops to environmental changes, providing a strong guarantee for the efficient utilization of saline-alkali land and the stable growth of crop yields. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a schematic diagram of the structure of the dynamic decision-making system for water, fertilizer and salt regulation of soybeans in saline-alkali land adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0048] Embodiment 1, a dynamic decision-making method for regulating water, fertilizer and salt in saline-alkali soil soybeans, comprising:

[0049] Set the area division cycle. At the beginning of any area division cycle, the soybean planting area is evenly divided into several local control areas of the same size in a grid shape. The grid spacing is selected by the particle swarm optimization algorithm. This division allows for refined management of each small area, making water, fertilizer and salt management more accurate and able to be adjusted according to specific conditions in different areas. The particle swarm optimization algorithm is an optimization technology that simulates the foraging behavior of bird flocks and is used to determine the optimal grid spacing to ensure the representativeness of data collection and the uniformity of irrigation and fertilization. For any local control area, the first soil data and environmental data are collected at the current first monitoring time point. The first soil data includes soil moisture, soil temperature and soil salt concentration. These parameters directly affect the growth conditions and irrigation needs of crops. Soil moisture The humidity and temperature affect the water absorption and metabolic activities of crops, while the soil salt concentration is related to the salt stress of crops. Environmental data include ambient temperature and humidity. Environmental conditions have an important impact on the transpiration and water demand of crops, and also affect the evaporation of soil moisture and the decomposition of fertilizers. It is judged whether the currently collected soil moisture is higher than the preset moisture threshold. If so, no operation is performed. If not, the irrigation initial decision model is used to obtain the total irrigation volume data in the first time window in the future. This decision-making process ensures the timeliness and appropriateness of irrigation. By comparing with the preset threshold, over-irrigation can be avoided, water resource waste and soil salinization risks can be reduced. If irrigation is required, the irrigation initial decision model will calculate the required irrigation volume in the first time window in the future based on the collected data, providing a scientific basis for precision irrigation.

[0050] Obtain the most recent irrigation data, analyze the obtained irrigation data, and obtain soil infiltration capacity and soil water holding capacity. Soil infiltration capacity and soil water holding capacity are key soil physical properties that affect irrigation efficiency and effectiveness. By analyzing the most recent irrigation data, the soil's ability to retain and infiltrate water can be evaluated, which is crucial for adjusting subsequent irrigation strategies; use soil infiltration capacity, soil water holding capacity and soil salt concentration to calculate the irrigation frequency adjustment value; use the obtained irrigation frequency adjustment value to calculate the first component value of size n, where n is a positive integer; use the first component value to evenly divide the total irrigation volume data into n preparatory irrigation volume data; for any preparatory irrigation volume data , obtain the rainfall data corresponding to the current prepared irrigation data, use the prepared irrigation data to subtract the rainfall data, and obtain the actual irrigation data corresponding to the current prepared irrigation data; finally, the obtained n actual irrigation data will form an irrigation decision data set; this step will subdivide the total irrigation amount into multiple small irrigations, making the irrigation more uniform and controllable; by adjusting the value of n, the irrigation frequency can be flexibly adjusted according to soil characteristics and environmental conditions to meet the actual soil demand for water; at the same time, rainfall is an important factor affecting the soil moisture state. By considering rainfall, unnecessary irrigation can be reduced, water resources can be saved, and the excessive soil moisture caused by rainfall can be avoided to affect crop growth;

[0051] The practice of conducting fractional irrigation after obtaining the total irrigation amount is to manage soil moisture more accurately. The advantage of doing so is that it can reduce water waste and soil nutrient loss caused by excessive single irrigation, and avoid uneven crop growth caused by uneven irrigation; fractional irrigation helps maintain soil structure, reduce soil compaction and erosion, and reduce water stress on crops by reducing drastic fluctuations in soil moisture; in addition, this method can also avoid the risk of soil salinization caused by excessive irrigation, especially under saline-alkali conditions, fractional irrigation helps reduce soil salinity through leaching, thereby creating a more favorable growth environment for crops;

[0052] For any local control area, the second soil data and environmental data are collected at the current second monitoring time point. The second soil data includes soil nutrients, soil salt concentration and soil pH value. These data are crucial for evaluating the nutrient status of crops and soil environmental conditions. The soil nutrient level directly affects the growth and yield of crops, while the soil salt concentration and pH value affect the fertility of the soil and the efficiency of crop absorption of nutrients. It is determined whether the soil nutrients are higher than the preset nutrient threshold. If so, no operation is performed. If not, the initial fertilization decision model is used to obtain the total amount of fertilizer data and fertilization type in the second time window in the future. A customized fertilization plan is provided for each local control area, which helps to optimize the use of fertilizers and ensure that crops obtain appropriate nutrient supply, while reducing excessive use of fertilizers and reducing the impact on the environment.

[0053] The fertilization frequency adjustment value is calculated based on the obtained fertilization type and soil salt concentration; different types of fertilizers have different effects on crop growth, and the release and absorption rates in soils with different salt concentrations will also be different; by calculating the fertilization frequency adjustment value, the fertilization frequency can be optimized to ensure that the effectiveness of the fertilizer and the needs of the crop are met, while avoiding the problem of soil salinization caused by excessive fertilization; the obtained fertilization frequency adjustment value is used to calculate the second component value of size m, where m is a positive integer; the second component value is used to evenly divide the total fertilization amount data into m preliminary fertilization amount data; the obtained m preliminary fertilization amount data are combined into a fertilization decision data set; this step subdivides the total fertilization amount into multiple small portions to achieve more uniform and controllable fertilization; by adjusting the value of m, that is, the number of fertilization times, the fertilization plan can be flexibly adjusted according to soil characteristics and crop needs; this method helps to improve the utilization efficiency of fertilizers, reduce nutrient loss, and ensure that crops obtain the nutrients they need at critical growth stages;

[0054] The obtained irrigation decision data set and fertilization decision data set are used for regulation and management, that is, the n actual irrigation amount data in the irrigation decision data set are used to perform n irrigations at the same time interval in the first time window in the future; at the same time, the m prepared fertilization amount data in the fertilization decision data set are used to perform m fertilizations at the same time interval in the second time window in the future.

[0055] The irrigation initial decision model is used to obtain the total irrigation volume data required in the first time window in the future. The specific operations are as follows:

[0056] Set the duration to The first time window of the future is used to obtain the predicted environmental data within the first time window, and the first crop growth data and historical irrigation records are obtained. The first crop growth data includes the crop growth stage and the crop water requirement. The first soil data, environmental data, predicted environmental data, first crop growth data and historical irrigation records currently obtained are combined into a first non-time series data set. At the same time, the most recent All the first soil data and environmental data collected within the time are sorted in time to form a first time series data set; the first non-time series data set and the first time series data set are used as inputs of the initial irrigation decision model, and the total irrigation volume data within the first time window in the future is output;

[0057] The initial fertilization decision model is used to obtain the total amount of fertilization data and fertilization type required in the second time window in the future. The specific operations are as follows:

[0058] Set the duration to The second time window, > , obtain the predicted environmental data within the second time window in the future, and obtain the second crop growth data and historical fertilization records; the second crop growth data includes the crop growth stage, crop nutrient deficiency symptoms and crop nutrient requirements, and the currently acquired second soil data, environmental data, predicted environmental data, second crop growth data and historical fertilization records are combined into a second non-time series data set; at the same time, the most recent All the second soil data and environmental data collected within the time are sorted in time to form a second time series data set; the second non-time series data set and the second time series data set are used as inputs of the initial fertilization decision model, and the total amount of fertilizer data and fertilization type in the future second time window are output.

[0059] Get the most recent irrigation data, analyze the obtained irrigation data, and obtain the soil infiltration capacity and soil water holding capacity; use the soil infiltration capacity, soil water holding capacity and soil salt concentration to calculate the irrigation frequency adjustment value. The specific operations are as follows:

[0060] Get the most recent irrigation data, including irrigation duration t and irrigation amount x;

[0061] Using the formula Calculate the irrigation water flow Q; then use the formula Calculate and obtain the soil permeability K of the current area; where L represents the soil thickness of the current area, s represents the area of ​​the current area, is the water head difference in the current area;

[0062] Using the formula Calculate and obtain the moisture change rate B; where, It represents the time interval from the start of irrigation until the soil moisture in the current area reaches the field capacity; Indicated in The soil moisture changes over time; then use the formula Calculate and obtain the soil water holding capacity P of the current area;

[0063] Set the value range of the first component to to ; Also set the maximum soil penetration capacity and minimum soil infiltration capacity ; Maximum soil water holding capacity and minimum soil water holding capacity ; Maximum soil salt concentration and minimum soil salt concentration ; Among them, the maximum soil penetration capacity , maximum soil water holding capacity and minimum soil salt concentration The value corresponding to the first component is ;Minimum soil penetration capacity , minimum soil water holding capacity and maximum soil salt concentration The value corresponding to the first component is ;

[0064] For the soil permeability K, use the formula Calculate the penetration adjustment value ; For soil water holding capacity P, use the formula Calculate the water holding capacity adjustment value ; For soil salt concentration S, use the formula Calculate the salt concentration adjustment value ;

[0065] Finally, using the formula Calculate the irrigation frequency adjustment value .

[0066] Soil infiltration capacity, soil water holding capacity and soil salt concentration are key factors in adjusting irrigation frequency because they directly affect the movement, retention and crop absorption efficiency of water in the soil; soil infiltration capacity determines the speed at which water penetrates into the soil, soil water holding capacity reflects the soil's ability to retain water, and soil salt concentration affects the absorption of water and nutrients and the health of crop roots; when soil infiltration capacity is high, moisture drops quickly and more frequent irrigation may be required to keep the soil moist; conversely, soil with low infiltration capacity retains moisture for a long time and the irrigation frequency can be reduced; soil with strong soil water holding capacity can retain more water without increasing the risk of salinization, thereby reducing the number of irrigations; and when soil salt concentration is high, crop water absorption is hindered and more frequent light irrigation may be required to reduce salt stress on crops; therefore, adjusting irrigation frequency by combining these soil characteristics can optimize water management, improve irrigation efficiency, reduce resource waste, and maintain a good soil environment required for crop growth.

[0067] The fertilization frequency adjustment value is calculated based on the obtained fertilization type and soil salt concentration. The specific operations are as follows:

[0068] For any fertilization type, set the value range of the second component corresponding to the current fertilization type to to , i = 1, 2, ..., u; u represents the number of fertilization types; soil salt concentration is When , the corresponding second component value is ; Soil salt concentration is When , the corresponding second component value is ; Apply the obtained salt concentration adjustment value , as the fertilization frequency adjustment value ; The reason why soil salt concentration is used to determine the fertilization frequency adjustment value is that when the soil salt concentration is high, it may have an adverse effect on crop growth, so more frequent fertilization may be required to avoid excessive single fertilization and increase the salt concentration;

[0069] The obtained irrigation frequency adjustment value is used to calculate the first component value of size n. The specific operation is as follows:

[0070] Using the formula Calculation acquisition ;judge Is it an integer? If so, apply as the first component value; if not, the closest An integer as the first component value;

[0071] The obtained fertilization frequency adjustment value is used to calculate the second component value of size m. The specific operation is as follows:

[0072] Based on the output of the initial fertilization decision model, the type of fertilization required in the second future time window is calculated using the formula Calculate the corresponding fertilization type ;judge Is it an integer? If so, apply as the second component value; if not, the closest An integer as the second component value.

[0073] The irrigation initial decision model is established based on CNN and LSTM, including the first input layer, the first CNN layer, the first LSTM layer, the first fully connected layer and the first output layer; the first input layer is used to receive the first non-time series data set and the first time series data set; the first CNN layer is used to extract the non-time series features of the first non-time series data set; the first LSTM layer is used to extract the time series features of the first time series data set; the first fully connected layer is used to fuse the non-time series features and time series features extracted by the first CNN layer and the first LSTM layer, and generate output results; the first output layer is used to output the total irrigation volume data in the first time window in the future;

[0074] The specific operations for training the initial irrigation decision model are as follows:

[0075] A number of irrigation decision training samples are obtained, each of which contains a historical first non-time series data set, a historical first time series data set and corresponding historical total irrigation data; all irrigation decision training samples are divided into an irrigation decision training set and an irrigation decision verification set, the historical first non-time series data set and the historical first time series data set are used as inputs of the irrigation initial decision model, the historical total irrigation data is used as output of the irrigation initial decision model, the irrigation decision training set is used to train the irrigation initial decision model, the irrigation decision verification set is input into the irrigation initial decision model, and a first verification result is obtained; a first training condition is set, and it is determined whether the first verification result meets the first training condition. If so, the trained irrigation initial decision model is output; if not, the irrigation decision training set is continued to be used to train the irrigation initial decision model.

[0076] The initial fertilization decision model is established based on CNN and LSTM, including a second input layer, a second CNN layer, a second LSTM layer, a second fully connected layer and a second output layer; the second input layer is used to receive the second non-time series data set and the second time series data set; the second CNN layer is used to extract the non-time series features of the second non-time series data set; the second LSTM layer is used to extract the time series features of the second time series data set; the second fully connected layer is used to fuse the non-time series features and time series features extracted by the second CNN layer and the second LSTM layer, and generate an output result; the second output layer is used to output the total amount of fertilization data and fertilization type required in the second time window in the future;

[0077] The specific operations for training the initial fertilization decision model are as follows:

[0078] A number of fertilization decision training samples are obtained, each of which contains a historical second non-time series data set, a historical second time series data set, and corresponding historical total fertilization data and fertilization type; all fertilization decision training samples are divided into a fertilization decision training set and a fertilization decision verification set, the historical second non-time series data set and the historical second time series data set are used as inputs of the initial fertilization decision model, the historical total fertilization data and the fertilization type are used as outputs of the initial fertilization decision model, the initial fertilization decision model is trained using the fertilization decision training set, the fertilization decision verification set is input into the initial fertilization decision model, and a second verification result is obtained; a second training condition is set, and it is determined whether the second verification result meets the second training condition. If so, the trained initial fertilization decision model is output; if not, the initial fertilization decision model is continuously trained using the fertilization decision training set.

[0079] Example 2, a dynamic decision-making system for regulating water, fertilizer and salt in saline-alkali land soybeans, such as Figure 1 As shown, including:

[0080] The area division module is used to divide the soybean planting area into several local control areas of equal size in a grid-like manner through a particle swarm optimization algorithm;

[0081] The data monitoring module includes a first data monitoring unit and a second data monitoring unit; the first data monitoring unit is used to collect first soil data and environmental data for any local control area at a first monitoring time point, the first soil data including soil moisture, soil temperature and soil salt concentration; the environmental data including ambient temperature and ambient humidity; the second data monitoring unit is used to collect second soil data and environmental data for any local control area at a second monitoring time point, the second soil data including soil nutrients, soil salt concentration and soil pH value;

[0082] The irrigation initial decision module is used to determine whether the currently collected soil moisture is higher than the preset moisture threshold. If so, no operation is performed; if not, the irrigation initial decision model is used to obtain the total irrigation volume data in the first time window in the future;

[0083] The initial fertilization decision module is used to determine whether the soil nutrients are higher than the preset nutrient threshold. If so, no operation is performed; if not, the initial fertilization decision model is used to obtain the total amount of fertilization data and fertilization type in the second time window in the future;

[0084] The irrigation decision acquisition module includes a soil permeability calculation unit, a soil water holding capacity calculation unit and an irrigation decision data set acquisition unit; the soil permeability calculation unit is used to obtain the most recent irrigation data, analyze the obtained irrigation data, and obtain the soil permeability; the soil water holding capacity calculation unit is used to analyze the obtained irrigation data and obtain the soil water holding capacity; the irrigation decision data set acquisition unit is used to calculate and obtain the irrigation frequency adjustment value using the soil permeability, soil water holding capacity and soil salt concentration; the obtained irrigation frequency adjustment value is used to calculate and obtain the first component value of size n; the first component value is used to evenly divide the total irrigation amount data into n preparatory irrigation amount data; for any preparatory irrigation amount data, the rainfall data corresponding to the current preparatory irrigation amount data is obtained, and the rainfall data is subtracted from the preparatory irrigation amount data to obtain the actual irrigation amount data corresponding to the current preparatory irrigation amount; and finally the obtained n actual irrigation amount data are formed into an irrigation decision data set;

[0085] The fertilization decision acquisition module is used to calculate the fertilization frequency adjustment value using the obtained fertilization type and soil salt concentration; use the obtained fertilization frequency adjustment value to calculate the second component value of size m; use the second component value to evenly divide the total fertilization amount data into m preliminary fertilization amount data; and form a fertilization decision data set with the obtained m preliminary fertilization amount data.

[0086] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A dynamic decision-making method for regulating water, fertilizer and salt in saline-alkali soil soybeans, characterized in that: include: Set the area division cycle. At the beginning of any area division cycle, the soybean planting area is evenly divided into several local control areas of the same size in a grid shape. The grid spacing is selected by the particle swarm optimization algorithm. For any local control area, the first soil data and environmental data are collected at the current first monitoring time point. The first soil data includes soil moisture, soil temperature and soil salt concentration; the environmental data includes ambient temperature and ambient humidity. It is determined whether the currently collected soil moisture is higher than the preset moisture threshold. If so, no operation is performed; if not, the irrigation initial decision model is used to obtain the total irrigation volume data in the first time window in the future. The most recent irrigation data is obtained, and the obtained irrigation data is analyzed to obtain the soil infiltration capacity and soil water holding capacity; the irrigation frequency adjustment value is calculated using the soil infiltration capacity, soil water holding capacity and soil salt concentration; the first component value of n is calculated using the obtained irrigation frequency adjustment value, where n is a positive integer; the total irrigation amount data is evenly divided into n preparatory irrigation amount data using the first component value; for any preparatory irrigation amount data, the rainfall data corresponding to the current preparatory irrigation amount data is obtained, and the actual irrigation amount data corresponding to the current preparatory irrigation amount is obtained by subtracting the rainfall data from the preparatory irrigation amount data; finally, the obtained n actual irrigation amount data are combined into an irrigation decision data set; For any local control area, the second soil data and environmental data are collected at the current second monitoring time point, and the second soil data includes soil nutrients, soil salt concentration and soil pH value; it is determined whether the soil nutrients are higher than the preset nutrient threshold, if so, no operation is performed; if not, the total amount of fertilizer data and fertilizer type in the future second time window are obtained using the fertilization initial decision model; Calculate and obtain a fertilization frequency adjustment value based on the obtained fertilization type and soil salt concentration; calculate and obtain a second component value of size m using the obtained fertilization frequency adjustment value, where m is a positive integer; The second component value is used to evenly divide the total amount of fertilizer application data into m pieces of preliminary fertilizer application data; the m pieces of preliminary fertilizer application data are used to form a fertilization decision data set; The obtained irrigation decision data set and fertilization decision data set are used for regulation and management, that is, the n actual irrigation amount data in the irrigation decision data set are used to perform n irrigations at the same time interval in the first time window in the future; At the same time, m preliminary fertilizer amount data in the fertilization decision data set are applied to perform m fertilizations at the same time interval in the second time window in the future; The irrigation initial decision model is used to obtain the total irrigation volume data in the first future time window. The specific operations are as follows: A first time window with a duration of t1 is set, and predicted environmental data within the first time window in the future is obtained, and first crop growth data and historical irrigation records are obtained. The first crop growth data includes the crop growth stage and the crop water requirement. The first soil data, environmental data, predicted environmental data, first crop growth data and historical irrigation records currently obtained are combined into a first non-time series data set; at the same time, all first soil data and environmental data collected within the most recent t1 time are sorted in time to form a first time series data set; the first non-time series data set and the first time series data set are used as inputs of the initial irrigation decision model, and the total irrigation volume data within the first time window in the future is output; The initial fertilization decision model is used to obtain the total fertilization data and fertilization type in the second time window in the future. The specific operations are as follows: A second time window with a duration of t2 is set, t2>t1, and the predicted environmental data within the future second time window is obtained, and the second crop growth data and historical fertilization records are obtained; the second crop growth data includes the crop growth stage, crop nutrient deficiency symptoms and crop nutrient requirements, and the currently acquired second soil data, environmental data, predicted environmental data, second crop growth data and historical fertilization records are combined into a second non-time series data set; at the same time, all the second soil data and environmental data collected within the recent t2 time are sorted by time to form a second time series data set; the second non-time series data set and the second time series data set are used as inputs of the initial fertilization decision model, and the total amount of fertilizer data and fertilization type within the future second time window are output.

2. The method for dynamic decision-making on water, fertilizer and salt regulation of soybeans in saline-alkali land according to claim 1, characterized in that: Get the most recent irrigation data, analyze the obtained irrigation data, and obtain the soil infiltration capacity and soil water holding capacity; use the soil infiltration capacity, soil water holding capacity and soil salt concentration to calculate the irrigation frequency adjustment value. The specific operations are as follows: Get the most recent irrigation data, including irrigation duration t and irrigation amount x; Using the formula Calculate the irrigation water flow Q; then use the formula Calculate and obtain the soil permeability K of the current area; where L represents the soil thickness of the current area, s represents the area of ​​the current area, and Δh is the water head difference of the current area; Using the formula Calculate the moisture change rate B; where Δt represents the time interval from the start of irrigation until the soil moisture in the current area reaches the field water holding capacity; Δθ represents the change in soil moisture within Δt; then use the formula Calculate and obtain the soil water holding capacity P of the current area; Set the value range of the first component to n min to n max ; At the same time, set the maximum soil penetration capacity K max and minimum soil permeability K min ; Maximum soil water holding capacity P max and minimum soil water holding capacity P min ; Maximum soil salt concentration S max and minimum soil salt concentration S mim ; Among them, the maximum soil penetration capacity K max , maximum soil water holding capacity P max and minimum soil salt concentration S min The value corresponding to the first component is n min ; Minimum soil permeability K min , minimum soil water holding capacity P min and the maximum soil salt concentration S max The value corresponding to the first component is n max ; For the soil permeability K, use the formula Calculate the penetration adjustment value R K ; For soil water holding capacity P, use the formula Calculate the water holding capacity adjustment value R P ; For soil salt concentration S, use the formula Calculate the salt concentration adjustment value R S ; Finally, using the formula Calculate and obtain the irrigation frequency adjustment value R1.

3. The method for dynamic decision-making on water, fertilizer and salt regulation of soybeans in saline-alkali land according to claim 2, characterized in that: The fertilization frequency adjustment value is calculated based on the obtained fertilization type and soil salt concentration. The specific operations are as follows: For any fertilization type, set the value range of the second component value corresponding to the current fertilization type to m(i) min to m(i) max , i = 1, 2, ..., u; u represents the number of fertilization types; soil salt concentration is S min When , the corresponding second component value is m(i) min ; Soil salt concentration is S max When , the corresponding second component value is m(i) max ; Apply the obtained salt concentration adjustment value R S , as the fertilization frequency adjustment value R2.

4. The method for dynamic decision-making on water, fertilizer and salt regulation of soybeans in saline-alkali land according to claim 3, characterized in that: The obtained irrigation frequency adjustment value is used to calculate the first component value of size n. The specific operation is as follows: Using the formula n adjusted =n min +R1(n max -n min ) Calculate and obtain n adjusted ; Judge n adjusted Is it an integer? If so, apply n adjusted as the first component value; if not, the value closest to n adjusted An integer as the first component value; The obtained fertilization frequency adjustment value is used to calculate the second component value of size m. The specific operation is as follows: Based on the output of the initial fertilization decision model, the type of fertilization required in the second future time window is calculated using the formula m(i) adjusted =m(i) min +R2(m(i) max -m(i) min ) Calculate and obtain the corresponding fertilization type m(i) adjusted ; Judge m(i) adjusted Is it an integer? If so, apply m(i) adjusted as the second component value; if not, the closest to m(i) adjusted An integer as the second component value.

5. The method for dynamic decision-making on water, fertilizer and salt regulation of soybeans in saline-alkali land according to claim 4, characterized in that: The irrigation initial decision model is established based on CNN and LSTM, including the first input layer, the first CNN layer, the first LSTM layer, the first fully connected layer and the first output layer; the first input layer is used to receive the first non-time series data set and the first time series data set; the first CNN layer is used to extract the non-time series features of the first non-time series data set; The first LSTM layer is used to extract the temporal features of the first time series data set; The first fully connected layer is used to fuse the non-temporal features and temporal features extracted by the first CNN layer and the first LSTM layer, and generate output results; the first output layer is used to output the total irrigation volume data within the first time window in the future; The specific operations for training the initial irrigation decision model are as follows: Acquire a number of irrigation decision training samples, each of which contains a historical first non-time series data set, a historical first time series data set, and corresponding historical total irrigation data; divide all irrigation decision training samples into an irrigation decision training set and an irrigation decision verification set, use the historical first non-time series data set and the historical first time series data set as inputs of an initial irrigation decision model, use the historical total irrigation data as outputs of the initial irrigation decision model, use the irrigation decision training set to train the initial irrigation decision model, input the irrigation decision verification set into the initial irrigation decision model, and obtain a first verification result; A first training condition is set, and it is determined whether the first verification result satisfies the first training condition. If so, the trained irrigation initial decision model is output; if not, the irrigation decision training set is continued to be used to train the irrigation initial decision model.

6. The method for dynamic decision-making on water, fertilizer and salt regulation of soybeans in saline-alkali land according to claim 5, characterized in that: The initial fertilization decision model is established based on CNN and LSTM, including a second input layer, a second CNN layer, a second LSTM layer, a second fully connected layer, and a second output layer; the second input layer is used to receive the second non-time series data set and the second time series data set; the second CNN layer is used to extract the non-time series features of the second non-time series data set; The second LSTM layer is used to extract the temporal features of the second time series data set; The second fully connected layer is used to fuse the non-temporal features and temporal features extracted by the second CNN layer and the second LSTM layer, and generate output results; the second output layer is used to output the total amount of fertilizer data and fertilizer type required in the second time window in the future; The specific operations for training the initial fertilization decision model are as follows: Acquire several fertilization decision training samples, each of which contains a historical second non-time series data set, a historical second time series data set, and corresponding historical total fertilization data and fertilization type; divide all fertilization decision training samples into a fertilization decision training set and a fertilization decision verification set, use the historical second non-time series data set and the historical second time series data set as inputs of an initial fertilization decision model, use the historical total fertilization data and the fertilization type as outputs of the initial fertilization decision model, use the fertilization decision training set to train the initial fertilization decision model, input the fertilization decision verification set into the initial fertilization decision model, and obtain a second verification result; A second training condition is set to determine whether the second verification result satisfies the second training condition. If so, the trained initial fertilization decision model is output; if not, the initial fertilization decision model is continuously trained using the fertilization decision training set.

7. A dynamic decision-making system for regulating water, fertilizer and salt in saline-alkali land soybeans, characterized in that: The system is applied to a dynamic decision-making method for water, fertilizer and salt regulation of soybeans in saline-alkali land as described in any one of claims 1 to 6, comprising: The area division module is used to divide the soybean planting area into several local control areas of equal size in a grid-like manner through a particle swarm optimization algorithm; The data monitoring module includes a first data monitoring unit and a second data monitoring unit; the first data monitoring unit is used to collect first soil data and environmental data for any local control area at a first monitoring time point, the first soil data including soil moisture, soil temperature and soil salt concentration; the environmental data including ambient temperature and ambient humidity; the second data monitoring unit is used to collect second soil data and environmental data for any local control area at a second monitoring time point, the second soil data including soil nutrients, soil salt concentration and soil pH value; The irrigation initial decision module is used to determine whether the currently collected soil moisture is higher than the preset moisture threshold. If so, no operation is performed; if not, the irrigation initial decision model is used to obtain the total irrigation volume data in the first time window in the future; The initial fertilization decision module is used to determine whether the soil nutrients are higher than the preset nutrient threshold. If so, no operation is performed; if not, the initial fertilization decision model is used to obtain the total amount of fertilization data and fertilization type in the second time window in the future; The irrigation decision acquisition module includes a soil permeability calculation unit, a soil water holding capacity calculation unit and an irrigation decision data set acquisition unit; the soil permeability calculation unit is used to obtain the most recent irrigation data, analyze the obtained irrigation data, and obtain the soil permeability; the soil water holding capacity calculation unit is used to analyze the obtained irrigation data and obtain the soil water holding capacity; the irrigation decision data set acquisition unit is used to calculate and obtain the irrigation frequency adjustment value using the soil permeability, soil water holding capacity and soil salt concentration; the obtained irrigation frequency adjustment value is used to calculate and obtain the first component value of size n; the first component value is used to evenly divide the total irrigation amount data into n preparatory irrigation amount data; for any preparatory irrigation amount data, the rainfall data corresponding to the current preparatory irrigation amount data is obtained, and the rainfall data is subtracted from the preparatory irrigation amount data to obtain the actual irrigation amount data corresponding to the current preparatory irrigation amount; and finally the obtained n actual irrigation amount data are formed into an irrigation decision data set; The fertilization decision acquisition module is used to calculate the fertilization frequency adjustment value using the obtained fertilization type and soil salt concentration; use the obtained fertilization frequency adjustment value to calculate the second component value of size m; use the second component value to evenly divide the total fertilization amount data into m preliminary fertilization amount data; and form a fertilization decision data set with the obtained m preliminary fertilization amount data.

Images