An intelligent irrigation and fertilization decision system based on target yield and soil moisture

Abstract

The application discloses an intelligent irrigation and fertilization decision system based on target yield and soil moisture, and belongs to the field of intelligent agriculture, comprising: a regional data acquisition module, which acquires soil types and real-time meteorological data; a characteristic database establishment module, which is used for establishing a soil characteristic database; a prediction data output module, which outputs first and second prediction parameters reaching a target yield; a missing probability output module, which outputs first and second missing probabilities; a parameter balance optimization module, which outputs optimized first and second prediction parameters; and a decision result output module, which outputs a decision result. The application solves the technical problems in the prior art that agricultural irrigation and fertilization rely on experience, the degree of refinement is not high, and the utilization rate of water and fertilizer, resource waste and yield instability are caused, and achieves the technical effects that through a mixed prediction model and a balanced decision model, irrigation and fertilization parameters are finely optimized, resource efficient utilization is realized, and the utilization rate of water and fertilizer and yield are improved.

Description

Technical Field

[0001] The present invention relates to the field of smart agriculture, and in particular to an intelligent irrigation and fertilization decision-making system based on target yield and soil moisture conditions. Background Art

[0002] With the continuous development of modern agriculture, water-saving irrigation and precision fertilization are gaining increasing attention. Reasonable irrigation and fertilization management is not only crucial for crop yield and quality, but also for the efficient use of water and fertilizer resources and the sustainable development of agriculture. Currently, agricultural irrigation and fertilization generally rely on traditional empirical judgment methods. Agricultural managers manually determine irrigation water and fertilizer amounts based on their years of planting experience, soil moisture conditions, and crop growth. This method is simple to operate and easy to promote, but it also has some shortcomings. First, empirical judgments are often subjective and uninformed, making it difficult to adapt to the precise needs of crops under different environmental conditions, resulting in low water and fertilizer utilization rates. Second, manual management lacks quantitative and standardized operating standards, making it difficult to strictly control the timing, frequency, and quotas of irrigation and fertilization, resulting in wasted resources and unstable yields. Summary of the Invention

[0003] This application provides an intelligent irrigation and fertilization decision-making system based on target yield and soil moisture conditions, aiming to solve the technical problems in the existing technology that agricultural irrigation and fertilization rely on experience judgment, have a low degree of refinement, and lead to water and fertilizer utilization rate, resource waste and unstable yield.

[0004] In view of the above problems, the present application provides an intelligent irrigation and fertilization decision-making system based on target yield and soil moisture conditions. The first aspect disclosed in the present application provides an intelligent irrigation and fertilization decision-making system based on target yield and soil moisture conditions, the system comprising: a regional data acquisition module for acquiring the soil type and real-time meteorological data of the target area; a feature database establishment module for establishing a soil feature database, storing the corresponding soil moisture ratio index and soil fertility level according to the soil type; a prediction data output module for establishing a hybrid prediction model, inputting the soil moisture ratio index, soil fertility level and real-time meteorological data into the hybrid prediction model, and outputting a first prediction parameter and a second prediction parameter based on meteorological influences, wherein the first prediction parameter is an irrigation parameter and the second prediction parameter is an irrigation parameter. The parameters are fertilization parameters; a missing probability output module is used to calculate based on the first prediction parameter and the first target parameter, and output the first missing probability, and calculate based on the second prediction parameter and the second target parameter, and output the second missing probability; a parameter balanced optimization module is used to input the first prediction parameter and the second prediction parameter into the balanced decision model, and the balanced decision model performs balanced optimization on the first prediction parameter and the second prediction parameter according to the first missing probability and the second missing probability, and outputs the optimized first prediction parameter and the optimized second prediction parameter; a decision result output module is used to output the decision result based on the optimized first prediction parameter and the optimized second prediction parameter.

[0005] Another aspect disclosed in the present application provides an intelligent irrigation and fertilization decision-making method based on target yield and soil moisture conditions, the method comprising: obtaining soil type and real-time meteorological data of the target area; establishing a soil feature database to store corresponding soil moisture ratio indicators and soil fertility grades according to soil types; establishing a hybrid prediction model, inputting the soil moisture ratio indicator, soil fertility grade and real-time meteorological data into the hybrid prediction model, and outputting a first prediction parameter and a second prediction parameter for achieving the target yield under meteorological influence, wherein the first prediction parameter is an irrigation parameter and the second prediction parameter is a fertilization parameter; calculating based on the first prediction parameter and the first target parameter, outputting a first missing probability, calculating based on the second prediction parameter and the second target parameter, outputting a second missing probability; inputting the first prediction parameter and the second prediction parameter into a balanced decision model, the balanced decision model performing balanced optimization on the first prediction parameter and the second prediction parameter according to the first missing probability and the second missing probability, and outputting an optimized first prediction parameter and an optimized second prediction parameter; outputting a decision result based on the optimized first prediction parameter and the optimized second prediction parameter.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] By acquiring the soil type and real-time meteorological data of the target area, the soil and environmental background in which the crops grow can be fully understood; a soil characteristic database is established, and the corresponding soil moisture ratio index and soil fertility level are stored according to the soil type to provide a basis for subsequent decision-making; a hybrid prediction model is established, and the soil moisture ratio index, soil fertility level and real-time meteorological data are input into it, and the irrigation parameters and fertilization parameters based on the meteorological influence to achieve the target yield are output, and the water and fertilizer requirements of the crops to achieve the target yield are predicted; the first missing probability is calculated based on the irrigation parameters and the target irrigation parameters, and the second missing probability is calculated based on the fertilization parameters and the target fertilization parameters, and quantitative Evaluate the pros and cons of irrigation and fertilization plans; input irrigation parameters and fertilization parameters into a balanced decision model, perform balanced optimization according to the first missing probability and the second missing probability, output the optimized irrigation parameters and fertilization parameters, and obtain the final decision result. Through multi-objective optimization, a technical solution for balanced optimal irrigation and fertilization is achieved, which solves the technical problem that agricultural irrigation and fertilization in the existing technology relies on experience judgment and has a low degree of refinement, resulting in water and fertilizer utilization rate, resource waste and unstable yield. It achieves the technical effect of finely optimizing irrigation and fertilization parameters through a hybrid prediction model and a balanced decision model, realizing efficient resource utilization, and improving water and fertilizer utilization rate and yield.

[0008] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 A flowchart of an intelligent fertilization decision-making method based on target yield and soil moisture conditions is provided for an embodiment of the present application;

[0010] Figure 2 A flowchart of establishing a hybrid prediction model in an intelligent fertilization decision-making method based on target yield and soil moisture conditions is provided for an embodiment of the present application;

[0011] Figure 3 A structural schematic diagram of an intelligent irrigation and fertilization decision-making system based on target yield and soil moisture conditions is provided for an embodiment of the present application.

[0012] Explanation of the accompanying symbols: regional data acquisition module 11, feature database establishment module 12, prediction data output module 13, missing probability output module 14, parameter balance optimization module 15, decision result output module 16. DETAILED DESCRIPTION

[0013] The overall idea of ​​the technical solution provided by this application is as follows:

[0014] The embodiment of the present application provides an intelligent irrigation and fertilization decision-making system based on target yield and soil moisture conditions, aiming to solve the problems of extensive management of irrigation and fertilization, low resource utilization efficiency, and unstable crop yield in existing agricultural production. First, the soil type and real-time meteorological data of the target area are obtained to fully grasp the dynamic information of the crop growth environment, providing a basis for subsequent analysis and decision-making. Then, a soil characteristic database is established to store the corresponding soil moisture ratio index and soil fertility level according to the soil type. Next, a hybrid prediction model and a balanced decision-making model are constructed to respectively realize the preliminary generation and dynamic optimization of irrigation and fertilization parameters, taking into account the temporal and spatial differences in soil moisture conditions and meteorological conditions, and taking into account the diversified needs such as yield targets and resource constraints, so as to obtain the optimal irrigation and fertilization plan.

[0015] Compared with existing technologies, this method comprehensively utilizes multi-source heterogeneous data to depict the dynamic changes in the soil environment. It also integrates various model algorithms to achieve intelligent optimization of irrigation and fertilization decisions. It balances multiple objectives, including yield increase, cost reduction, and ecological conservation, achieving balanced agricultural development. This effectively addresses the pain points and difficulties of agricultural irrigation and fertilization management, supporting the refinement, intelligence, and sustainable development of agricultural production.

[0016] After introducing the basic principles of the present application, various non-limiting implementation methods of the present application will be specifically introduced in conjunction with the drawings in the specification.

[0017] Example 1

[0018] like Figure 1 As shown, the embodiment of the present application provides an intelligent irrigation and fertilization decision-making method based on target yield and soil moisture conditions, the method comprising:

[0019] S1: Obtain soil type and real-time meteorological data of the target area;

[0020] In the embodiment of the present application, the target area is the farmland area where irrigation and fertilization decisions are to be made. In order to achieve accurate irrigation and fertilization management, it is necessary to obtain the soil type and real-time meteorological data of the area. Among them, the soil type is determined by soil parameters. The soil parameters are obtained by deploying a certain number of soil sensors in the target area to monitor and collect soil moisture, temperature, conductivity, pH and other indicator parameters in real time, thereby compensating for the static and hysteresis of traditional soil surveys and achieving dynamic monitoring of soil types; real-time meteorological data includes parameters such as light, temperature, humidity, wind speed, and precipitation, which are obtained by setting up automatic weather stations or small meteorological sensor nodes in the fields. The weather stations or sensor nodes are deployed in areas away from obstacles and with flat and open terrain, and the collected data is uploaded to the data center at regular intervals.

[0021] The soil type and real-time meteorological data are transmitted back to a data center server in real time through a wireless communication network, which can be a low-power wide-area network such as NB-IoT, 4G, LoRa, ZigBee, or a short-range wireless communication technology, to provide input for subsequent irrigation and fertilization decisions.

[0022] By obtaining the soil type and real-time meteorological data of the target area, the soil type and real-time climate conditions of the farmland are mastered, providing a basis for scientific irrigation and fertilization decisions. Compared with conventional manual sampling and experience-based judgment, more timely, accurate, and comprehensive data support can be provided to lay the foundation for intelligent decision-making.

[0023] S2: Establish a soil characteristic database for storing corresponding soil moisture ratio indicators and soil fertility levels according to soil types;

[0024] In the embodiments of the present application, after obtaining the soil type of the target area, the soil moisture ratio indicator and the soil fertility level of the soil type are obtained.

[0025] The soil moisture ratio indicator is the ratio of the actual soil water content to the soil water holding capacity, reflecting the moisture level of the target area. The actual soil water content is directly measured by a soil moisture sensor or calculated according to the soil moisture characteristic curve; the soil water holding capacity depends on factors such as soil texture, bulk density, and organic matter content, and is obtained by interpolation from the soil moisture characteristic curve.

[0026] The soil fertility level refers to the abundance or deficiency level of soil nutrients, taking into account soil organic matter content, total nitrogen, available nitrogen, available phosphorus, available potassium, and other nutrient indicators in the target area, and referring to the pre-set soil nutrient classification standard by experts, the soil fertility is divided into six levels: I, II, III, IV, V, and VI, corresponding to the fertility level.

[0027] Subsequently, the obtained soil moisture ratio indicator and soil fertility level are stored to establish a soil characteristic database, providing an important reference and decision variable for intelligent irrigation and fertilization decisions.

[0028] S3: Establish a hybrid prediction model, input the soil moisture ratio indicator, the soil fertility level, and the real-time meteorological data into the hybrid prediction model, and output a first prediction parameter and a second prediction parameter based on the target yield under the influence of meteorology, wherein the first prediction parameter is an irrigation parameter, and the second prediction parameter is a fertilization parameter;

[0029] In this embodiment, to achieve intelligent fertigation decisions, a hybrid prediction model is established that comprehensively considers soil moisture, soil fertility, and meteorological conditions to predict crop water and fertilizer requirements. This model takes a soil moisture ratio indicator, soil fertility level, and real-time meteorological data as input and outputs a first prediction parameter and a second prediction parameter required for the target area. The first prediction parameter is the irrigation parameter, and the second prediction parameter is the fertilization parameter.

[0030] Hybrid prediction models can employ a variety of machine learning algorithms, such as support vector machines, random forests, and neural networks. The appropriate algorithm should be selected based on actual needs and data characteristics. The model's training data consists of historical crop growth records, soil moisture monitoring data, soil nutrient testing data, and meteorological observation data. By analyzing and mining this data, quantitative relationships between crop growth and soil moisture and nutrient status, as well as meteorological conditions, are discovered, guiding model parameter optimization and structural design.

[0031] During model training, crop yield and quality indicators are optimized, and soil moisture ratio, soil fertility level, and meteorological factors are used as input features. Model parameters are repeatedly iterated to ensure that the predicted results are as close as possible to the actual observed values. At the same time, the model undergoes cross-validation and independent testing to evaluate its predictive performance and generalization ability, ensuring that the model can adapt to the task of predicting crop water and fertilizer requirements across different years, varieties, and growth stages.

[0032] Once the model training is complete, it can be put into practical use. Simply inputting the target area's current soil moisture ratio, soil fertility level, and real-time meteorological data, the model predicts the crop's water and fertilizer requirements at each growth stage. From these, it extracts the water and fertilizer requirements necessary to achieve the crop's target yield, generating the irrigation and fertilization parameters (i.e., the primary and secondary prediction parameters).

[0033] By establishing a hybrid prediction model, we fully utilize multi-source data such as soil moisture, soil fertility and meteorological conditions, explore the intrinsic connection between them and crop growth, and dynamically predict the changing trends of water and fertilizer requirements in the target area, providing a basis for precise regulation of irrigation and fertilization.

[0034] S4: Calculate based on the first prediction parameter and the first target parameter, and output a first missing probability; calculate based on the second prediction parameter and the second target parameter, and output a second missing probability;

[0035] In an embodiment of the present application, after obtaining the first prediction parameter and the second prediction parameter output by the hybrid prediction model, they are compared with the first target parameter (target irrigation amount) and the second target parameter (target fertilization amount) of the target area, and the first missing probability (irrigation missing probability) and the second missing probability (fertilization missing probability) are calculated to quantify the degree of insufficiency of the first prediction parameter and the second prediction parameter.

[0036] The first missing probability is calculated by subtracting the first prediction parameter from the first target parameter, dividing the result by the first target parameter, and then multiplying the result by 100%. The first target parameter represents the ideal irrigation amount required for crops in the target area at the current growth stage and is pre-determined based on factors such as crop type, growth period, soil type, and evapotranspiration. The first prediction parameter represents the irrigation amount predicted by the hybrid prediction model based on current soil moisture and meteorological conditions. A greater first missing probability indicates a greater insufficiency in the predicted irrigation amount, and a greater need for supplemental irrigation.

[0037] The second missing probability is calculated by subtracting the second prediction parameter from the second target parameter, dividing the result by the second target parameter, and then multiplying the result by 100%. The second target parameter represents the ideal fertilizer amount required for crops in the target area at the current growth stage and is pre-determined based on factors such as crop type, growth period, soil fertility, and yield target. The second prediction parameter represents the fertilizer amount predicted by the hybrid prediction model based on current soil fertility and meteorological conditions. A greater second missing probability indicates a less-than-ideal predicted fertilizer amount, and a greater amount of additional fertilizer is needed.

[0038] When calculating the missing probability, when the first prediction parameter is greater than the first target parameter or the second prediction parameter is greater than the second target parameter, the missing probability should be zero, indicating that no supplementation is required.

[0039] By calculating the first missing probability and the second missing probability, the gap between the current predicted amount of irrigation and fertilization and the crop growth requirements of the target area is intuitively reflected, revealing the degree of restriction of soil moisture and fertility on crop growth, providing a quantitative basis for subsequent irrigation and fertilization decisions, and helping to achieve precise irrigation and fertilization.

[0040] S5: Inputting the first prediction parameter and the second prediction parameter into a balanced decision model, wherein the balanced decision model performs balanced optimization on the first prediction parameter and the second prediction parameter according to the first missing probability and the second missing probability, and outputs an optimized first prediction parameter and an optimized second prediction parameter;

[0041] In this embodiment of the present application, after obtaining the first and second missing probabilities, further first and second prediction parameters are required to balance the water and nutrient requirements of crop growth. A balanced decision model is introduced to comprehensively consider the degree of irrigation and fertilization deficiencies, collaboratively optimize the irrigation and fertilization parameters, and generate more balanced and reasonable optimized first and second prediction parameters.

[0042] The balanced decision model takes as input the first prediction parameter, the second prediction parameter, the first missing probability, and the second missing probability. Its output is the optimized first prediction parameter (optimized irrigation parameter) and the optimized second prediction parameter (optimized fertilization parameter). The model balances the importance of irrigation and fertilization and adjusts the irrigation and fertilization parameters to meet crop growth needs while minimizing irrigation and fertilization costs and resource consumption.

[0043] The equilibrium decision model first determines the optimization priority of irrigation and fertilization based on the magnitude of the first and second missing probabilities. If the first missing probability is higher, irrigation parameters are prioritized; if the second missing probability is higher, fertilization parameters are prioritized. Parameter optimization is then performed on the higher-priority side, and an iterative search is performed to find an irrigation or fertilization parameter that meets crop growth requirements. Finally, while maintaining the higher-priority side, the other side is optimized to achieve an overall balanced irrigation and fertilization plan.

[0044] The equilibrium optimization process can employ a variety of optimization techniques, such as gradient descent, genetic algorithms, and particle swarm optimization, to approximate the optimal solution through repeated iterations. During the optimization process, various constraints, such as upper limits on irrigation water volume, upper limits on fertilization costs, and soil nutrient carrying capacity, are considered to ensure that the optimization results meet actual production requirements. Furthermore, domain knowledge graphs and empirical rules, such as those based on differences in water and fertilizer requirements during different growth stages and nutrient supply capacity across different soil types, can be incorporated to provide a priori guidance for optimization decisions.

[0045] Through the balanced decision-making model, the irrigation and fertilization parameters are adaptively adjusted according to the lack of irrigation and fertilization. While meeting the needs of crop growth, the cost-effectiveness of irrigation and fertilization is taken into account, the optimal allocation of water and fertilizer resources is achieved, and intelligent and refined irrigation and fertilization parameters are provided for agricultural production.

[0046] S6: Output a decision result based on the optimized first prediction parameter and the optimized second prediction parameter.

[0047] In the embodiments of the present application, after the equalization decision model completes the optimization of the irrigation and fertilization parameters, a set of optimal irrigation parameters (optimized first prediction parameters) and optimal fertilization parameters (optimized second prediction parameters) that meet the crop growth demand and take into account resource conservation can be obtained. Then, the optimized first prediction parameters and the optimized second prediction parameters are summarized as a decision result. The decision result at least includes irrigation decision, fertilization decision, execution instructions, and effect estimation. The irrigation decision explicitly gives the optimized irrigation amount, as well as the time, frequency, and duration of irrigation, and also gives suggestions on irrigation water quality and irrigation methods. The fertilization decision explicitly gives the optimized fertilization amount, as well as the time, frequency, and method of fertilization, and also gives suggestions on fertilizer types, fertilizer ratios, and fertilization sites. The execution instructions are necessary instructions on the execution process and matters needing attention of irrigation and fertilization, to ensure that the decision result can be accurately conveyed to the irrigation and fertilization equipment and effectively executed. The effect estimation is a prediction of the changes in crop growth and yield in the target area after irrigation and fertilization, as well as the saving of water and fertilizer resources, to provide a reference for decision effect evaluation and continuous improvement.

[0048] The output mode of the decision result can adopt various forms, such as digital signals, text instructions, voice broadcast, etc., to adapt to different control devices and operation habits. At the same time, it also has human-computer interaction function, allowing users to query, modify, confirm, etc. operation on the decision result, to improve the flexibility and usability of the decision result.

[0049] By outputting the decision result, the agricultural production practice in the target area can be effectively guided, the decision deviation and execution error can be reduced, the fine irrigation and fertilization can be realized, and the utilization rate of water and fertilizer can be improved, thereby improving the crop yield and quality.

[0050] Further, the embodiments of the present application also include:

[0051] A first strategy combination is generated for the first prediction parameters and the first target parameters.

[0052] A second strategy combination is generated for the second prediction parameters and the second target parameters.

[0053] The first strategy combination and the second strategy combination are used for segmented optimization by using an equalization algorithm, and a decision result is output.

[0054] In a feasible implementation manner, the strategy combination and segmented optimization are introduced to improve the accuracy and operability of the decision result.

[0055] First, a set of irrigation control schemes is generated for the first prediction parameter and the first target parameter, as the first strategy combination. Each scheme in this combination is a possible irrigation decision, including specific parameters such as irrigation time, irrigation quota, and irrigation duration. These schemes are generated based on the deviation between the first prediction parameter and the first target parameter, as well as constraints such as irrigation threshold and irrigation intensity. Simultaneously, a set of fertilization control schemes is generated for the second prediction parameter and the second target parameter, as the second strategy combination. Each scheme in this combination is a possible fertilization decision, including specific parameters such as fertilization time, fertilizer amount, and fertilizer type. These schemes are generated based on the deviation between the first prediction parameter and the first target parameter, as well as constraints such as fertilization threshold and fertilizer utilization rate.

[0056] Then, using a balancing algorithm, the first and second strategy combinations are used as inputs. A piecewise optimization approach is used to find the optimal fertigation decision and output it as the decision result. This process involves first fixing one strategy combination while searching for the optimal solution in another. The fixed and optimized combinations are then swapped, and multiple iterations are repeated until neither strategy combination can be further optimized. This piecewise optimization approach effectively balances the mutual influence and constraints between the two strategy combinations, avoiding the local optimum and decision imbalance that can result from single-pronged optimization.

[0057] By generating the first strategy combination and the second strategy combination and using the equilibrium algorithm for segmented optimization, more refined and balanced optimization decisions can be made in the two dimensions of irrigation and fertilization, which can effectively improve the effects and benefits of intelligent irrigation and fertilization.

[0058] Furthermore, the embodiment of the present application also includes:

[0059] The first strategy combination and the second strategy combination are segmented and iteratively optimized with the required soil moisture index as the optimization target until they reach a convergence state, and the control parameters in the convergence state are output as the optimized first prediction parameters and the optimized second prediction parameters;

[0060] The process of completing a segmented optimization includes selecting a first strategy from the first strategy combination, optimizing the second strategy combination while keeping the first strategy unchanged and outputting the second strategy, and optimizing the first strategy while keeping the second strategy unchanged.

[0061] In a feasible implementation, a soil moisture index is introduced as an indicator to measure the effect of irrigation and fertilization, and a segmented iterative optimization method is used to continuously approach the convergent optimal solution.

[0062] First, the desired soil moisture index is used as the optimization target for the equilibrium algorithm. This index reflects the equilibrium state of multiple factors, including soil moisture, nutrients, temperature, and aeration. It is designed based on the actual needs and data conditions of the target area. For example, a weighted average or comprehensive score of multiple indicators, such as soil moisture, nutrient content, temperature, and pH, can be used. Then, a stepwise iterative optimization approach is used, alternating between optimizing the first and second strategy combinations to gradually improve the soil moisture index until convergence is achieved.

[0063] To complete a segmented optimization process, the optimal irrigation strategy is first selected from the first strategy combination. This strategy, as the first strategy, maximizes the soil moisture index under the current fertilization strategy. Then, while the first strategy is fixed, the optimal fertilization strategy is selected from the second strategy combination. This strategy, as the second strategy, maximizes the soil moisture index under the current irrigation strategy. Next, while the second strategy is fixed, the optimal irrigation strategy is again selected from the first strategy combination. This strategy, as the second strategy, maximizes the soil moisture index under the current fertilization strategy. This iterative optimization process is repeated, continuously updating the first and second strategies to continuously improve the soil moisture index. The balancing algorithm is considered to have converged to the optimal solution after the improvement in several consecutive iterations is less than a preset threshold, or the maximum number of iterations is reached. At this point, the irrigation parameters corresponding to the current first strategy and the fertilization parameters corresponding to the current second strategy are output as the optimized first and second prediction parameters.

[0064] By taking soil moisture indicators as the optimization target and adopting a piecewise iterative optimization equilibrium algorithm, the optimal balance between irrigation and fertilization is achieved, which not only meets the crop's needs for water and nutrients, but also avoids resource waste and environmental pollution caused by excessive irrigation and fertilization; at the same time, the convergence characteristics of the piecewise iterative optimization ensure the efficiency and controllability of the optimization process, and can quickly approach the optimal solution within a limited number of iterations, adapting to the dynamically changing agricultural production environment.

[0065] Furthermore, the embodiment of the present application also includes:

[0066] Get the associated impact feedback network layer;

[0067] The correlation impact feedback network layer is obtained by training based on the mutual correlation impact between irrigation parameters and fertilization parameters, and the training data includes irrigation parameters, fertilization parameters, and identification information identifying irrigation-fertilization impact indicators and identification information identifying fertilization-irrigation impact indicators;

[0068] A second strategy under the association influence of the first strategy is acquired based on the association influence feedback network layer.

[0069] In a feasible implementation, on the basis of segmented optimization, a policy optimization method based on correlation influence feedback network layer is introduced, which is used to consider the mutual influence and correlation constraints between irrigation strategy and fertilization strategy in the segmented optimization process, so as to realize the collaborative optimization between the two strategies.

[0070] Firstly, a pre-trained correlation influence feedback network layer is obtained, which can model the mutual influence relationship between irrigation parameters and fertilization parameters. The network layer can adopt various forms such as feedforward neural network, recurrent neural network and graph neural network, the input of the network layer is irrigation parameters and fertilization parameters, and the output is irrigation-fertilization influence indicators and fertilization-irrigation influence indicators. When training the correlation influence feedback network layer, training data is collected and labeled in advance, including historical irrigation parameters, fertilization parameters and corresponding irrigation-fertilization influence indicators and fertilization-irrigation influence indicators. These influence indicators can be qualitative grade labels (such as high, medium and low), or quantitative numerical indicators (such as correlation coefficient, causal strength, etc.), which are used to measure the influence degree and direction of the change of one strategy parameter on another strategy parameter. Through supervised learning of the correlation influence feedback network layer on the training data, the correlation influence law between irrigation parameters and fertilization parameters is mastered to predict the possible change or influence of another strategy parameter under a given strategy parameter.

[0071] Then, in the process of segmented optimization, the obtained correlation influence feedback network layer is used to optimize the correlation between the first strategy and the second strategy. In each segmented optimization, when the first strategy is fixed to optimize the second strategy combination, the current first strategy is input into the correlation influence feedback network layer to obtain the irrigation-fertilization influence indicators under the influence of the irrigation strategy, which is used to guide the search and selection of the second strategy combination, so that the selected optimal fertilization strategy not only performs well under the current irrigation condition, but also matches and coordinates with the current irrigation strategy. Similarly, when the second strategy is fixed to optimize the first strategy combination, the current second strategy is also input into the correlation influence feedback network layer to obtain the fertilization-irrigation influence indicators under the influence of the fertilization strategy, which is used to guide the search and selection of the first strategy combination, so that the selected optimal irrigation strategy is adapted and supplemented with the current fertilization strategy.

[0072] By introducing correlation influence feedback in segmented optimization, irrigation strategy and fertilization strategy can influence and dynamically adjust each other in the optimization process, avoiding the strategy mismatch and negative effects that may be caused by simple independent optimization, realizing the collaborative optimization of irrigation strategy and fertilization strategy, and enhancing the collaboration and robustness of irrigation and fertilization decision-making.

[0073] Further, the embodiments of the application also include:

[0074] The first missing probability and the second missing probability are judged. If the first missing probability is greater than or equal to the second missing probability, the priority of the first strategy combination is set to be higher than the priority of the second strategy combination during segmented optimization.

[0075] If the first missing probability is smaller than the second missing probability, during segmented optimization, the priority of the first strategy combination is set to be smaller than the priority of the second strategy combination.

[0076] In a feasible implementation, a priority adjustment mechanism based on missing probability is introduced to dynamically adjust the optimization priorities of the first strategy combination and the second strategy combination during the segmented optimization process to adapt to real-time changes in soil moisture and fertility conditions.

[0077] First, the calculated first and second missing probabilities are extracted. These two missing probabilities reflect the degree of deviation between the current irrigation and fertilization status and the crop growth requirements, and serve as the basis for determining the optimization priority of the first and second strategy combinations. Then, by comparing the first and second missing probabilities, the optimization priorities of the first and second strategy combinations are determined. If the first missing probability is greater than or equal to the second missing probability, it means that the deviation of the current irrigation status is greater than or equal to the fertilization status, and the irrigation strategy needs to be optimized first. Therefore, the priority of the first strategy combination is higher than that of the second strategy combination. If the first missing probability is less than the second missing probability, it means that the deviation of the current fertilization status is greater than that of the irrigation status, and the fertilization strategy needs to be optimized first. Therefore, the priority of the first strategy combination is lower than that of the second strategy combination.

[0078] Based on the priority judgment results of the first and second missing probabilities, the optimization order of the first and second strategy combinations is dynamically adjusted during the segmented optimization process. When the priority of the first strategy combination is higher than that of the second strategy combination, in each segmented optimization process, the optimal irrigation strategy is first selected from the first strategy combination, and then the second strategy combination is optimized with this fixed irrigation strategy, and then the first strategy combination is optimized. This prioritizes the optimization effect of the irrigation strategy and promotes the improvement of soil moisture conditions. When the priority of the first strategy combination is lower than that of the second strategy combination, in each segmented optimization process, the optimal fertilization strategy is first selected from the second strategy combination, and then the first strategy combination is optimized with this fixed fertilization strategy, and then the second strategy combination is optimized. This prioritizes the optimization effect of the fertilization strategy and promotes the improvement of soil nutrient conditions.

[0079] Through the priority adjustment mechanism based on missing probability, the optimization order in the segmented optimization is dynamically adjusted. According to the real-time status of soil moisture and fertility, the priority of irrigation and fertilization optimization is flexibly determined, so as to achieve dynamic matching between strategy optimization and demand changes, and improve the pertinence and timeliness of farmland management.

[0080] Further, such as Figure 2 As shown, the embodiment of the present application also includes:

[0081] Obtain growth status samples of target crops;

[0082] Obtaining soil moisture samples, soil fertility samples, and real-time meteorological samples of the target crop, wherein the soil moisture samples include irrigation belt laying method, soil water infiltration rate, and real-time wetting ratio; and the soil fertility samples include fertilization type, soil nutrient value, and fertilizer utilization rate;

[0083] Build a fully connected neural network;

[0084] The fully connected neural network is trained using the soil moisture sample, the soil fertility sample, the real-time meteorological sample, and identification information identifying the growth status of the target crop to establish a hybrid prediction model, wherein the hybrid prediction model stores the required growth index of the target crop.

[0085] In one feasible implementation, to establish a hybrid prediction model, first, growth status samples of the target crop are obtained. These crop growth status samples refer to a series of growth indicator data obtained by measuring and recording crops at different growth stages, such as plant height, stem diameter, leaf area index, chlorophyll content, and biomass, reflecting the growth and development status of the crop under specific environmental conditions. When obtaining growth status samples, representative plots and plants are selected for tracking and monitoring based on factors such as crop type, planting area, and climatic characteristics. At the same time, attention is paid to the temporal frequency and spatial distribution of sampling to fully capture the dynamic changes and differences in the crop growth process. Then, while collecting crop growth status samples, soil moisture, soil fertility, and real-time meteorological data are collected at the same plots and time points to form soil moisture, soil fertility, and real-time meteorological samples of the target crop. Among them, soil moisture samples include parameters such as irrigation tape laying method, soil water infiltration rate, real-time moisture ratio, etc., which reflect the soil's water supply capacity and moisture status; soil fertility samples include parameters such as fertilization type, soil nutrient value, fertilizer utilization rate, etc., which reflect the soil's fertilizer supply capacity and nutrient level; real-time meteorological samples include temperature, humidity, light, wind speed and other factors, which reflect the atmospheric environment conditions for the growth of target crops.

[0086] Then, a fully connected neural network is established, including an input layer, a hidden layer and an output layer, and information transmission and feature transformation are realized between layers through dense connection weights, so as to establish a high-dimensional mapping relationship between soil moisture, soil fertility and real-time weather and crop growth state. At the same time, according to the characteristics and magnitude of the sample data, the activation function, regularization method and other hyperparameters of the fully connected neural network are set. Then, the soil moisture sample, soil fertility sample, real-time weather sample and corresponding crop growth state label in the growth state sample are used to supervise the training of the fully connected neural network, and a nonlinear prediction model between environmental factors and crop growth, i.e. a hybrid prediction model, is established. At the same time, the demand growth index of the target crop is set in the hybrid prediction model, wherein the demand growth index is an ideal growth state parameter preset according to the crop type, yield target and other factors, such as the optimal leaf area index and suitable plant height. When the hybrid prediction model is used to predict the growth of the target crop, the demand growth index is used as a reference standard to analyze the difference between the predicted value and the demand value, diagnose the limiting factors of crop growth, and optimize the irrigation and fertilization decision according to the demand growth index, dynamically control the growth process of the crop, and realize precision management.

[0087] Further, the embodiments of the present application also include:

[0088] The soil wetness ratio index, the soil fertility level and the real-time weather data are input into the hybrid prediction model to obtain a predicted growth index.

[0089] The demand growth index stored in the hybrid prediction model is fed back according to the predicted growth index, and irrigation parameters and fertilization parameters under the influence of weather are output.

[0090] The irrigation parameters at least include irrigation water quantity and irrigation time, and the fertilization parameters at least include fertilization type and fertilization content.

[0091] In a feasible implementation, after obtaining the soil wetness ratio index, the soil fertility level and the real-time weather data, the soil wetness ratio index, the soil fertility level and the real-time weather data are input into the hybrid prediction model, so that the soil moisture, soil fertility and weather conditions of the target area are comprehensively considered, and the growth index of the crop in the target area, such as the leaf area index, the plant height and the biomass, is predicted as the predicted growth index, which represents the growth state and development potential of the crop in the target area under the current environment.

[0092] After obtaining the predicted growth index, it is compared with the demand growth index of the target crop stored in the hybrid prediction model, and the irrigation parameters and the fertilization parameters are dynamically optimized through a feedback mechanism to obtain the optimal irrigation parameters and the fertilization parameters under the real-time weather data.

[0093] By comparing the predicted growth index with the target growth index, the hybrid prediction model determines the current limiting factors for crop growth and adjusts irrigation and fertilization parameters accordingly to better meet crop growth requirements. When the predicted growth index falls below the target growth index, the hybrid prediction model adjusts the irrigation and fertilization parameters accordingly. When the predicted growth index reaches the target growth index, the hybrid prediction model maintains the irrigation and fertilization parameters. Irrigation parameters include at least the amount and duration of irrigation, which affect the water supply of the target crop. Fertilization parameters include at least the type and amount of fertilizer applied, which influence the nutrient absorption and metabolism of the target crop.

[0094] Through the feedback optimization mechanism of the hybrid prediction model, the irrigation parameters and fertilization parameters of the target crops are adjusted dynamically and timely according to local conditions, so that the irrigation parameters and fertilization parameters always match the growth requirements of the target crops, realizing real-time synchronization and dynamic adaptation of farmland management, thereby promoting high-quality and high-yield crops and improving resource utilization efficiency.

[0095] In summary, the intelligent fertilization decision-making method based on target yield and soil moisture provided by the embodiments of the present application has the following technical effects:

[0096] Obtain soil type and real-time meteorological data for the target area to understand the soil moisture status and environmental meteorological conditions in which crops grow, providing basic input for subsequent analysis and decision-making. Establish a soil characteristic database to store soil moisture ratio indicators and soil fertility levels corresponding to soil type, providing a basis for fertigation parameter prediction. Develop a hybrid prediction model, inputting soil moisture ratio indicators, soil fertility levels, and real-time meteorological data into the hybrid prediction model. The model outputs first and second prediction parameters based on meteorological influences to achieve the target yield, predicting the water and fertilizer requirements required to achieve the target yield during the crop growth period. Calculations are performed based on the first prediction parameter and the first target parameter, outputting the first missing probability. Calculations are performed based on the second prediction parameter and the second target parameter, outputting the second missing probability. This quantitatively assesses the applicability and feasibility of fertigation, providing quantitative indicators for subsequent optimization and adjustment. The first prediction parameter and the second prediction parameter are input into the balanced decision model, and the balanced decision model performs balanced optimization on the first prediction parameter and the second prediction parameter according to the first missing probability and the second missing probability, outputs the optimized first prediction parameter and the optimized second prediction parameter, and obtains the optimized irrigation parameter and fertilization parameter through multi-objective optimization; according to the optimized first prediction parameter and the optimized second prediction parameter, the decision result is output, and based on the balanced optimized irrigation and fertilization parameters, the preferred irrigation and fertilization plan is formed, so as to realize the fine optimization of irrigation and fertilization parameters and efficient utilization of resources, thereby improving the water and fertilizer utilization rate and yield.

[0097] Example 2

[0098] Based on the same inventive concept as the intelligent fertilization decision-making method based on target yield and soil moisture in the aforementioned embodiment, Figure 3 As shown, the embodiment of the present application provides an intelligent irrigation and fertilization decision-making system based on target yield and soil moisture conditions, the system comprising:

[0099] The regional data acquisition module 11 is used to obtain the soil type and real-time meteorological data of the target area;

[0100] A feature database establishment module 12 is used to establish a soil feature database, wherein the soil feature database is used to store corresponding soil moisture ratio indicators and soil fertility levels according to soil types;

[0101] a prediction data output module 13 for establishing a hybrid prediction model, inputting the soil moisture ratio index, the soil fertility level, and the real-time meteorological data into the hybrid prediction model, and outputting a first prediction parameter and a second prediction parameter based on achieving a target yield under meteorological influences, wherein the first prediction parameter is an irrigation parameter and the second prediction parameter is a fertilization parameter;

[0102] a missing probability output module 14, configured to calculate based on the first prediction parameter and the first target parameter and output a first missing probability, and to calculate based on the second prediction parameter and the second target parameter and output a second missing probability;

[0103] a parameter balancing optimization module 15, configured to input the first prediction parameter and the second prediction parameter into a balancing decision model, wherein the balancing decision model performs balancing optimization on the first prediction parameter and the second prediction parameter according to the first missing probability and the second missing probability, and outputs an optimized first prediction parameter and an optimized second prediction parameter;

[0104] The decision result output module 16 is used to output a decision result based on the optimized first prediction parameter and the optimized second prediction parameter.

[0105] Furthermore, the parameter balancing optimization module 15 includes the following execution steps:

[0106] generating a first strategy combination for the first prediction parameter and the first target parameter;

[0107] generating a second strategy combination for the second prediction parameter and the second target parameter;

[0108] Using the equilibrium algorithm, segmented optimization is performed using the first strategy combination and the second strategy combination, and a decision result is output.

[0109] Furthermore, the parameter equalization optimization module 15 further includes the following execution steps:

[0110] segmented iteration optimization is completed, the control parameters at the time of convergence are output as the optimized first prediction parameter and the optimized second prediction parameter;

[0111] The process of completing one segmented optimization includes screening a first strategy from the first strategy combination, optimizing the second strategy combination while fixing the first strategy, and optimizing the first strategy while fixing the second strategy.

[0112] Further, the parameter balance optimization module 15 further includes the following execution steps:

[0113] Obtain an associated influence feedback network layer;

[0114] The associated influence feedback network layer is trained based on the mutual influence between irrigation parameters and fertilization parameters, and the training data includes irrigation parameters, fertilization parameters, and identification information of irrigation-fertilization influence indicators and identification information of fertilization-irrigation influence indicators.

[0115] Obtain the second strategy under the influence of the first strategy based on the associated influence feedback network layer.

[0116] Further, the parameter balance optimization module 15 further includes the following execution steps:

[0117] Judge the first missing probability and the second missing probability, if the first missing probability is greater than or equal to the second missing probability, the priority of the first strategy combination is greater than the priority of the second strategy combination during segmented optimization;

[0118] If the first missing probability is less than the second missing probability, the priority of the first strategy combination is less than the priority of the second strategy combination during segmented optimization.

[0119] Further, the prediction data output module 13 includes the following execution steps:

[0120] Obtain the growth state sample of the target crop;

[0121] Obtain the soil moisture condition sample, soil fertility sample, and real-time weather sample of the target crop, wherein the soil moisture condition sample includes irrigation belt laying method, soil water infiltration rate, and real-time wetness ratio, and the soil fertility sample includes fertilization type, soil nutrient value, and fertilizer utilization rate.

[0122] Establish a fully connected neural network;

[0123] The soil moisture sample, the soil fertility sample, the real-time weather sample, and identification information identifying the growth state of the target crop are used to train the full connection neural network, and a hybrid prediction model is established, wherein the hybrid prediction model stores the demand growth index of the target crop.

[0124] Further, the prediction data output module 13 further includes the following execution steps:

[0125] The soil wetness ratio index, the soil fertility level, and the real-time weather data are input into the hybrid prediction model to obtain a predicted growth index.

[0126] According to the predicted growth index and the demand growth index stored in the hybrid prediction model, feedback is performed, and irrigation parameters and fertilization parameters under the influence of weather are output.

[0127] The irrigation parameters at least include irrigation water quantity and irrigation time, and the fertilization parameters at least include fertilization type and fertilization content.

[0128] Any step of the system described above can be stored in a computer memory without limitation as computer instructions or programs, and can be called and recognized by a computer processor without limitation to realize any system in the embodiments of the present application, and no redundant limitation is made herein.

[0129] Further, the above-mentioned first or second may not only represent a sequence relationship, but also may represent a certain specific concept, and / or refer to the selection of multiple elements individually or collectively. Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technologies, the present application is intended to include these modifications and variations.

Claims

1. An intelligent irrigation and fertilization decision-making system based on target yield and soil moisture, characterized by: The system comprises: A regional data acquisition module, wherein the regional data acquisition module is used to obtain soil type and real-time meteorological data of the target area; A feature database establishment module, wherein the feature database establishment module is used to establish a soil feature database, wherein the soil feature database is used to store corresponding soil moisture ratio indicators and soil fertility levels according to soil types; a prediction data output module, the prediction data output module being used to establish a hybrid prediction model, inputting the soil moisture ratio index, the soil fertility level, and the real-time meteorological data into the hybrid prediction model, and outputting a first prediction parameter and a second prediction parameter based on achieving a target yield under meteorological influences, wherein the first prediction parameter is an irrigation parameter and the second prediction parameter is a fertilization parameter; a missing probability output module, configured to calculate based on the first prediction parameter and the first target parameter and output a first missing probability, and to calculate based on the second prediction parameter and the second target parameter and output a second missing probability; a parameter balancing optimization module, the parameter balancing optimization module being configured to input the first prediction parameter and the second prediction parameter into a balancing decision model, the balancing decision model performing balancing optimization on the first prediction parameter and the second prediction parameter based on the first missing probability and the second missing probability, and outputting an optimized first prediction parameter and an optimized second prediction parameter; a decision result output module, the decision result output module being used to output a decision result based on the optimized first prediction parameter and the optimized second prediction parameter; The balanced decision model performs balanced optimization on the first prediction parameter and the second prediction parameter according to the first missing probability and the second missing probability, and the system includes: generating a first strategy combination for the first prediction parameter and the first target parameter; generating a second strategy combination for the second prediction parameter and the second target parameter; Using a balancing algorithm, performing segmented optimization with the first strategy combination and the second strategy combination, and outputting a decision result; The first strategy combination and the second strategy combination are segmented and iteratively optimized with the required soil moisture index as the optimization target until they reach a convergence state, and the control parameters in the convergence state are output as the optimized first prediction parameters and the optimized second prediction parameters; The process of completing a segmented optimization includes selecting a first strategy from the first strategy combination, optimizing the second strategy combination while keeping the first strategy unchanged and outputting the second strategy, and optimizing the first strategy while keeping the second strategy unchanged.

2. The system according to claim 1, wherein The first strategy is fixed unchanged, the second strategy is optimized, and the second strategy is output, the system includes: Get the associated impact feedback network layer; The correlation impact feedback network layer is obtained by training based on the mutual correlation impact between irrigation parameters and fertilization parameters, and the training data includes irrigation parameters, fertilization parameters, and identification information identifying irrigation-fertilization impact indicators and identification information identifying fertilization-irrigation impact indicators; A second strategy under the association influence of the first strategy is acquired based on the association influence feedback network layer.

3. The system according to claim 1, wherein: The first missing probability and the second missing probability are judged. If the first missing probability is greater than or equal to the second missing probability, the priority of the first strategy combination is set to be higher than the priority of the second strategy combination during segmented optimization. If the first missing probability is smaller than the second missing probability, during segmented optimization, the priority of the first strategy combination is set to be smaller than the priority of the second strategy combination.

4. The system according to claim 1, wherein: Establish a hybrid prediction model. The system includes: Obtain growth status samples of target crops; Obtaining soil moisture samples, soil fertility samples, and real-time meteorological samples of the target crop, wherein the soil moisture samples include irrigation belt laying method, soil water infiltration rate, and real-time wetting ratio; and the soil fertility samples include fertilization type, soil nutrient value, and fertilizer utilization rate; Build a fully connected neural network; The fully connected neural network is trained using the soil moisture sample, the soil fertility sample, the real-time meteorological sample, and identification information identifying the growth status of the target crop to establish a hybrid prediction model, wherein the hybrid prediction model stores the required growth index of the target crop.

5. The system according to claim 4, wherein: The system comprises: Inputting the soil moisture ratio index, the soil fertility level and the real-time meteorological data into the hybrid prediction model to obtain a predicted growth index; Feedback is performed based on the demand growth index stored in the hybrid prediction model according to the predicted growth index, and irrigation parameters and fertilization parameters based on meteorological influences are output; The irrigation parameters include at least the amount of irrigation water and the irrigation time, and the fertilization parameters include at least the type of fertilization and the content of fertilization.

6. An intelligent fertilization decision-making method based on target yield and soil moisture, characterized in that: The method uses the intelligent fertilization and irrigation decision-making system based on target yield and soil moisture according to any one of claims 1 to 5, and the method comprises: Obtain soil type and real-time meteorological data for the target area; Establishing a soil characteristic database, wherein the soil characteristic database is used to store corresponding soil moisture ratio indicators and soil fertility grades according to soil types; Establishing a hybrid prediction model, inputting the soil moisture ratio index, the soil fertility level, and the real-time meteorological data into the hybrid prediction model, and outputting a first prediction parameter and a second prediction parameter for achieving a target yield under meteorological influences, wherein the first prediction parameter is an irrigation parameter and the second prediction parameter is a fertilization parameter; Calculate based on the first prediction parameter and the first target parameter to output a first missing probability, and calculate based on the second prediction parameter and the second target parameter to output a second missing probability; Inputting the first prediction parameter and the second prediction parameter into a balanced decision model, wherein the balanced decision model performs balanced optimization on the first prediction parameter and the second prediction parameter according to the first missing probability and the second missing probability, and outputs optimized first prediction parameter and optimized second prediction parameter; Output a decision result based on the optimized first prediction parameter and the optimized second prediction parameter.

Images