A winter wheat multi-objective irrigation dynamic optimization method based on crop model and artificial intelligence

Abstract

The application discloses a winter wheat multi-target irrigation dynamic optimization method based on a crop model and artificial intelligence, and belongs to the field of agricultural intelligent irrigation optimization and decision-making. The method comprises the following steps: preprocessing meteorological data, soil data, crop parameter data and irrigation management data to obtain standardized input data; performing coupling processing on an AquaCrop model and a PyFAO56 model based on the standardized input data to obtain final output of the coupling model; based on the final output of the coupling model, adopting an NSGA-2 multi-target optimization algorithm to optimize decision variables to obtain a Pareto optimal solution set; and combining a fusion strategy of an entropy weight method and a random forest algorithm to sort candidate schemes in the Pareto optimal solution set to obtain a final irrigation strategy. The application provides a new type of precise irrigation scheme by giving consideration to yield and water efficiency through multi-target optimization, objective weighting and machine learning intelligent evaluation.

Description

Technical Field

[0001] This invention belongs to the field of agricultural intelligent irrigation optimization and decision-making technology, and in particular relates to a multi-objective dynamic optimization method for winter wheat irrigation based on crop models and artificial intelligence. Background Technology

[0002] Traditional irrigation strategies are mostly based on field trials, combined with optimization algorithms such as linear, nonlinear, dynamic programming, and genetic algorithms. However, field trials are often time-consuming and labor-intensive, and it is difficult to account for the impact of different climate types, soil types, and agronomic practices on crop yields in different years and regions. With the development of computer technology, crop growth models and machine learning algorithms are playing an increasingly important role in agricultural production and crop irrigation decisions.

[0003] The AquaCrop model is a widely used tool for simulating crop water response (Steduto et al., 2009). However, the current AquaCrop model has limited flexibility, especially in setting lower limits for staged irrigation control, making it difficult to meet the needs of refined deficit irrigation. Furthermore, the model's irrigation strategy is singular, unable to precisely adjust the percentage of irrigation evapotranspiration (ET), and can only perform 100% full irrigation. While the open-source Python version of AquaCrop, AquaCrop-OSPy, facilitates secondary development, it does not substantially improve these shortcomings. The existing irrigation decision-making technology suffers from the following technical problems:

[0004] 1. Traditional irrigation strategies based on field trials are time-consuming and labor-intensive, and cannot fully consider the impact of climate, soil and agronomic conditions in different years and regions on crop yield and water use efficiency.

[0005] 2. Although the existing AquaCrop model is widely used in crop water response simulation, it lacks flexibility in setting the lower limit of phased irrigation control and adjusting the percentage of irrigation replenishment, and can only achieve a fixed 100% irrigation strategy.

[0006] 3. Although the PyFAO56 model can implement various irrigation strategies (such as fixed cycle, ET-based supplemental irrigation, and soil moisture measurement irrigation), it still cannot achieve combined optimization of the two key decision variables: irrigation evapotranspiration percentage and soil moisture content control limit.

[0007] 4. Single optimization methods (such as traditional genetic algorithms or dynamic programming) are easily affected by the subjectivity of weight setting in the process of multi-objective trade-offs (such as maximizing output, improving irrigation water use efficiency and minimizing irrigation water volume), and the generated Pareto front solution set lacks a mechanism for further objective evaluation and screening.

[0008] 5. Existing simulation methods have low computational efficiency when evaluating multiple winter wheat irrigation schemes, making it difficult to meet the requirements for rapid response and real-time adjustment in actual decision-making processes.

[0009] Therefore, this invention provides a multi-objective dynamic optimization method for winter wheat irrigation based on crop models and artificial intelligence. Summary of the Invention

[0010] To address the aforementioned technical problems, this invention proposes a multi-objective dynamic optimization method for winter wheat irrigation based on crop models and artificial intelligence, thereby resolving the issues present in the existing technologies.

[0011] To achieve the above objectives, this invention provides a multi-objective dynamic optimization method for winter wheat irrigation based on crop models and artificial intelligence, comprising:

[0012] Meteorological data, soil data, crop parameter data, and irrigation management data are preprocessed to obtain standardized input data;

[0013] The AquaCrop model and the PyFAO56 model are coupled based on the standardized input data to obtain the final output of the coupled model.

[0014] Based on the final output of the coupled model, the Pareto optimal solution set is obtained by optimizing the decision variables using the NSGA-2 multi-objective optimization algorithm.

[0015] The final irrigation strategy is obtained by ranking the candidate solutions in the Pareto optimal solution set using a fusion strategy combining entropy weight method and random forest algorithm.

[0016] Optionally, the meteorological data includes: daily maximum temperature, daily minimum temperature, rainfall, radiation, wind speed, and humidity;

[0017] The soil data includes: soil profile information, field water holding capacity, wilting point, saturated water content, soil permeability, and soil stratification.

[0018] The crop parameter data includes: winter wheat planting date, maturity period, root distribution, canopy growth parameters, biomass conversion coefficient, harvest index, and crop water sensitivity parameters;

[0019] The irrigation management data includes: the initial irrigation plan, the irrigation time window for each stage, and the irrigation volume.

[0020] Optionally, the process of coupling the AquaCrop model and the PyFAO56 model based on the standardized input data to obtain the final output of the coupled model includes:

[0021] The standardized input data, including meteorological data, soil data, and crop parameter data, are input into the AquaCrop model to obtain crop growth indicators and water use data.

[0022] The meteorological and soil data from the standardized input data are input into the PyFAO56 model to generate irrigation events;

[0023] The irrigation event is fed back as input to the AquaCrop model to obtain the final output of the coupled model.

[0024] Optionally, the final output of the coupled model includes several evaluation indicators, including: final yield, total irrigation volume, irrigation water use efficiency, and number of irrigations.

[0025] Optionally, the decision variables include: the lower limit of soil moisture content control and the percentage of irrigation evapotranspiration recharge.

[0026] Optionally, the process of ranking the candidate solutions in the Pareto optimal solution set using a fusion strategy combining the entropy weight method and the random forest algorithm to obtain the final irrigation strategy includes:

[0027] The entropy weight method is used to calculate the weights of each evaluation index of the candidate solutions in the Pareto optimal solution set, and the index weights based on information entropy are obtained.

[0028] The contribution of decision variables to each evaluation index is determined based on the random forest algorithm, and the index weights based on the model contribution are obtained.

[0029] The weights of the indicators based on information entropy and the weights of the indicators based on model contribution are fused to obtain a comprehensive weight.

[0030] Based on the comprehensive weights, the candidate solutions in the Pareto optimal solution set are comprehensively evaluated and ranked to obtain the final irrigation strategy.

[0031] Optionally, the calculation process of the comprehensive weight includes:

[0032] The weights of the information entropy-based indicators are normalized.

[0033] The weights of the indicators based on model contribution are normalized.

[0034] Based on a preset fusion factor, the normalized index weights based on information entropy and the normalized index weights based on model contribution are weighted and averaged to obtain the comprehensive weight.

[0035] Compared with the prior art, the present invention has the following advantages and technical effects:

[0036] This invention presents a multi-objective dynamic optimization method for winter wheat irrigation based on crop models and artificial intelligence. By coupling the AquaCrop and PyFAO56 models, and combining the NSGA-2 multi-objective optimization algorithm with a fusion strategy of entropy weighting and random forest algorithms, it achieves refined management and scientific decision-making for irrigation strategies. This method can dynamically adjust irrigation strategies for different growth stages of winter wheat, ensuring adequate water supply during critical growth periods while avoiding unnecessary irrigation, significantly improving irrigation water use efficiency. Through the multi-objective optimization algorithm, this method achieves an optimal balance between maximizing yield, minimizing irrigation water consumption, and maximizing irrigation water use efficiency. Furthermore, the combination of entropy weighting and random forest algorithms further enhances the objectivity and accuracy of scheme evaluation, providing a scientific basis for irrigation decisions. This technology saves water resources, reduces irrigation costs, and ensures high and stable yields of winter wheat, which is of great significance for sustainable agricultural development, especially suitable for water-scarce areas. Attached Figure Description

[0037] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0038] Figure 1 This is the Pareto front for NSGA multi-objective optimization in an embodiment of the present invention;

[0039] Figure 2 This invention uses a random forest algorithm to analyze the contribution of eight decision variables to decision indicators in an embodiment of the invention.

[0040] Figure 3 This is a dependency graph of seedling irrigation threshold generated by random forest analysis in an embodiment of the present invention;

[0041] Figure 4 This is a dependency graph of irrigation threshold during the slow canopy growth period in random forest analysis according to an embodiment of the present invention;

[0042] Figure 5 This is a graph showing the dependence of maximum cover on the irrigation threshold at the onset of canopy senescence in random forest analysis according to an embodiment of the present invention.

[0043] Figure 6 This is a graph showing the dependence of the irrigation threshold at the onset of canopy senescence to maturity in a random forest analysis according to an embodiment of the present invention.

[0044] Figure 7 This is a dependency graph of the percentage of irrigation evapotranspiration during the seedling stage generated by random forest analysis in an embodiment of the present invention;

[0045] Figure 8 This is a graph showing the dependence of the percentage of irrigation evapotranspiration during the slow canopy growth period in a random forest analysis according to an embodiment of the present invention.

[0046] Figure 9 This is a graph showing the dependence of maximum cover on the percentage of irrigation evapotranspiration at the onset of canopy senescence in random forest analysis according to an embodiment of the present invention.

[0047] Figure 10 This is a graph showing the dependence of the percentage of irrigation evapotranspiration at the onset of canopy senescence to maturity in a random forest analysis according to an embodiment of the present invention.

[0048] Figure 11 This is a heatmap showing the correlation between eight decision variables in a random forest analysis according to an embodiment of the present invention.

[0049] Figure 12 This is a flowchart illustrating the integrated technology of AquaCrop, pyfao56, ​​NSGA-II, random forest, and entropy weight method in an embodiment of the present invention.

[0050] Figure 13 This is a comparison diagram of the effects of embodiments of the present invention. Detailed Implementation

[0051] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0052] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0053] Example 1

[0054] The objective of this invention is to combine the AquaCrop model with the pyfao56 model, utilizing the NSGA-2 multi-objective optimization algorithm to optimize two key decision variables—the lower limit of soil moisture control and the percentage of irrigation evapotranspiration recharge—at each growth stage. Simultaneously, a random forest algorithm is introduced to evaluate the contribution of each indicator, and the entropy weight method is used to objectively assign weights to the schemes, forming a comprehensive intelligent optimization decision-making method that considers yield, irrigation water use efficiency, and irrigation water volume. This method not only enables precise control of staged deficit irrigation strategies but also improves the efficiency and accuracy of irrigation decisions through intelligent, multi-indicator comprehensive evaluation, thus providing reliable technical support for the sustainable production of winter wheat in water-scarce areas such as the North China Plain.

[0055] This embodiment provides a multi-objective dynamic optimization method for winter wheat irrigation based on crop models and artificial intelligence, including the following steps:

[0056] I. Data Input. The data required for this technical solution mainly includes the following categories: (1) Meteorological data: including daily temperature (highest and lowest temperatures), rainfall, radiation, wind speed, humidity, etc. These data are usually monitored and recorded by meteorological stations and stored in text files or tables. This data is the basis for AquaCrop and PyFAO56 models to simulate crop growth and soil moisture balance. (2) Soil data: including soil profile information (thickness of each soil layer), field capacity (FC), wilting point (WP), saturated water content (S), soil permeability (Ksat), soil stratification structure, etc. These data are obtained through field measurements or existing soil databases, used to construct soil models, and jointly called by PyFAO56 and AquaCrop. (3) Crop parameter data: including winter wheat planting date, maturity period, root distribution, canopy growth parameters (such as canopy expansion, critical days of the decline stage, maximum canopy coverage, etc.), biomass conversion coefficient, harvest index (HI), and crop sensitivity parameters to water, etc. These parameters play a decisive role in the AquaCrop model and are usually calibrated using field trial data. (4) Irrigation management data: including initial irrigation plans (such as irrigation event records during full or partial irrigation), irrigation time windows for each stage, irrigation volume, etc., to provide initial scheduling basis for subsequent models when simulating irrigation regulation.

[0057] II. Data Preprocessing: Meteorological data, soil data, crop parameter data, and irrigation management data are preprocessed to obtain standardized input data. All input data undergoes standardized formatting and preprocessing steps before computation, including:

[0058] Meteorological data undergoes format conversion and missing value handling using specialized utility functions (such as prepare_weather);

[0059] Soil and crop parameter files are read using the get_filepath and corresponding loadfile methods and saved as objects for internal use within the model;

[0060] Data preprocessing ensures that the data called by different models (AquaCrop and PyFAO56) is consistent, so that the two models can run in the same environment and achieve coupled simulation.

[0061] III. Model Calling and Coupling Process: Based on standardized input data, the AquaCrop model and the PyFAO56 model are coupled to obtain the final output of the coupled model.

[0062] Furthermore, the process of coupling the AquaCrop model and the PyFAO56 model based on standardized input data to obtain the final output of the coupled model includes: inputting meteorological data, soil data, and crop parameter data from the standardized input data into the AquaCrop model to obtain crop growth indicators and water use; inputting the meteorological data and soil data from the standardized input data into the PyFAO56 model to generate irrigation events; and feeding the irrigation events back to the AquaCrop model as input to obtain the final output of the coupled model.

[0063] The AquaCrop model is used to simulate the growth process of winter wheat under specific meteorological, soil, and irrigation conditions, including canopy cover, evapotranspiration, yield generation, and biomass accumulation.

[0064] Input preprocessed meteorological data, soil parameters, and crop parameters to construct an AquaCrop model object. Set the crop planting date, maturity date, and parameters for each growth stage (e.g., development index, canopy growth, root expansion, etc.), and load the initial water content. Run the model according to the simulation time range (e.g., from 2023 / 10 / 25 to 2024 / 06 / 15), and output crop growth indicators (e.g., Fresh Yield) and water use (e.g., croptranspiration).

[0065] The PyFAO56 model is primarily used for irrigation scheduling control through soil moisture balance simulation. The process involves building the PyFAO56 soil moisture model using the same meteorological and soil data as AquaCrop. The model calculates reference evapotranspiration (ET0) based on daily meteorological conditions and combines this with soil parameters to calculate soil moisture changes.

[0066] The model sets an irrigation trigger condition: irrigation is initiated when the root zone water content falls below a preset threshold (decision variable 1). The water supplied during irrigation is supplemented according to the percentage of the crop's actual evapotranspiration demand at that stage (decision variable 2), thereby generating the specific irrigation amount and irrigation time series for each irrigation event.

[0067] Coupling and data interaction between the two models: The purpose of coupling is to ensure that changes in soil moisture (calculated by PyFAO56) can be dynamically fed back to AquaCrop, so that crop growth simulation can be updated according to actual irrigation events, thereby truly reflecting changes in yield and growth status caused by changes in irrigation strategies.

[0068] The coupling process is as follows: When running PyFAO56, irrigation events (including irrigation date and water volume) are automatically generated based on preset irrigation decision variables. These irrigation scheduling results are then passed as input parameters to the AquaCrop model, allowing AquaCrop to update soil moisture status using these irrigation events during growth simulation. The model output includes key indicators such as final yield, total irrigation volume, water use efficiency (IWUE), and irrigation frequency, which are used for subsequent optimization and evaluation.

[0069] IV. NSGA-2 Multi-Objective Optimization Process: Based on the final output of the coupled model, the NSGA-2 multi-objective optimization algorithm is used to optimize the decision variables to obtain the Pareto optimal solution set.

[0070] The decision variables used in this invention include:

[0071] Soil moisture content control limit: The percentage of soil moisture content when irrigation is triggered at each growth stage.

[0072] Irrigation evapotranspiration replenishment percentage: The percentage of actual water replenishment after irrigation is triggered relative to the crop reference evapotranspiration at that stage.

[0073] These two types of decision variables constitute an 8-dimensional parameter space (for the four irrigation stages), with the ranges set within reasonable physical intervals (moisture content of 0.25–0.55 and water replenishment percentage of 0.5–1).

[0074] (1) Population initialization and evaluation

[0075] Population generation: FloatRandomSampling is used to randomly generate an initial population (e.g., 50 individuals) within a defined parameter space.

[0076] Objective function calculation: For each individual, the above coupled model is called to perform simulation to obtain three main objective values: total irrigation water volume (the smaller the better); crop yield (the larger the better, usually taken as a negative value to achieve minimization when converted into an objective function); irrigation water use efficiency (IWUE, the larger the better, also converted into a minimization problem).

[0077] During the simulation, the parameters of each individual are passed to PyFAO56 to generate irrigation events, and then AquaCrop simulates crop growth and outputs various indicators.

[0078] (2) Non-dominated sorting and genetic operations

[0079] Non-dominated sorting: Sort all individuals according to the objective function value and select a set of non-dominated solutions (Pareto front), that is, no other individual is better than this solution in all objectives.

[0080] Crowding distance calculation: To maintain the diversity of the solution set, the crowding distance of each object in the target space is calculated.

[0081] Selection, crossover, and mutation: Select a subset of superior individuals for genetic manipulation based on non-dominated ordination and crowding distance.

[0082] Crossover: Simulated Binary Crossover is used to mix the decision variables of two parent individuals to generate offspring.

[0083] Mutation: Polynomial Mutation is used to slightly perturb the decision variables of offspring individuals to ensure a comprehensive exploration of the search space.

[0084] Iterative Update: After a certain number of generations (e.g., n_gen generations), the NSGA-2 algorithm converges, outputting a Pareto optimal solution set that covers the trade-offs between different objectives. Each candidate solution consists of 8 decision variables:

[0085] Phase 1 (seedling stage): Decision variable 1 is the lower limit of soil moisture content (the critical value for starting irrigation); decision variable 2 is the percentage of actual irrigation amount to reference evapotranspiration during irrigation.

[0086] Phase 2 (slow canopy growth period): Decision variable 3 is the lower limit of soil moisture content; decision variable 4 is the percentage of irrigation evapotranspiration.

[0087] Phase 3 (maximum cover - canopy aging period): Decision variable 5 is the lower limit of soil moisture content; decision variable 6 is the percentage of irrigation evapotranspiration.

[0088] Stage 4 (canopy senescence begins - maturity): Decision variable 7 is the lower limit of soil moisture content; decision variable 8 is the percentage of irrigation evapotranspiration.

[0089] For each candidate scheme, the output indicators obtained through simulation mainly include: yield, total irrigation amount, irrigation water use efficiency (WUE), and number of irrigations.

[0090] These outputs constitute the evaluation matrix, which forms the basis for subsequent entropy weight calculations and random forest sensitivity analysis.

[0091] V. A unified computational framework for the entropy weight method and the random forest algorithm: The computational process combines the entropy weight method and the random forest algorithm to rank candidate solutions in the Pareto optimal solution set, resulting in the final irrigation strategy. This system integrates random forest and the entropy weight method into a unified decision support system for winter wheat irrigation schemes, rather than using them independently.

[0092] Furthermore, the process of ranking candidate solutions in the Pareto optimal solution set to obtain the final irrigation strategy by combining the entropy weight method and the random forest algorithm includes: calculating the weights of each evaluation index of the candidate solutions in the Pareto optimal solution set using the entropy weight method to obtain the index weights based on information entropy; determining the contribution of the decision variable to each evaluation index based on the random forest algorithm to obtain the index weights based on model contribution; fusing the index weights based on information entropy and the index weights based on model contribution to obtain the comprehensive weight; and comprehensively evaluating and ranking the candidate solutions in the Pareto optimal solution set based on the comprehensive weight to obtain the final irrigation strategy.

[0093] (1) Data preparation and preprocessing: Collect historical irrigation scheme data of Pareto optimal solution set, including parameter settings of different schemes (emergence irrigation threshold, slow canopy growth irrigation threshold, maximum coverage - canopy senescence stage irrigation threshold, canopy senescence stage - maturity stage irrigation threshold, percentage of evapotranspiration during emergence irrigation, percentage of evapotranspiration during slow canopy growth irrigation, percentage of evapotranspiration during maximum coverage - canopy senescence stage irrigation, percentage of evapotranspiration during canopy senescence stage - maturity stage irrigation) and corresponding result indicators (yield, total irrigation volume, irrigation water use efficiency, irrigation frequency).

[0094] (2) Calculating the weights of multi-objective evaluation indicators using the entropy weight method. The entropy weight method determines the weights of each indicator by utilizing the differences in the distribution of each indicator's data among candidate solutions. The steps are as follows:

[0095] Data normalization: For each evaluation indicator (e.g., yield, irrigation volume, WUE, irrigation frequency), normalization is performed first to eliminate the influence of dimensions.

[0096] Calculate information entropy: Based on the normalized data, calculate the information entropy of each indicator. If an indicator is distributed relatively discretely across different schemes, its information entropy is low, indicating that the indicator contributes significantly to distinguishing the schemes.

[0097] Determine the entropy weights: Calculate the entropy weights based on the entropy values ​​of each indicator. The result is a set of weight vectors:

[0098] Wentropy=[Wyield,Wirrigation,WWUE,Wfrequency]

[0099] This vector reflects the objective importance of each evaluation metric from the perspective of candidate solutions. In the code, this part has been implemented by the function get_entropy_weight(data).

[0100] (3) Random forest extraction of the contribution of decision variables to the output. Using a random forest model, the influence of each decision variable (seedling stage irrigation threshold, slow canopy growth stage irrigation threshold, maximum cover - canopy senescence stage irrigation threshold, canopy senescence stage - maturity stage irrigation threshold, seedling stage irrigation evapotranspiration percentage, slow canopy growth stage irrigation evapotranspiration percentage, maximum cover - canopy senescence stage irrigation evapotranspiration percentage, canopy senescence stage - maturity stage irrigation evapotranspiration percentage) on the simulated output can be evaluated. The basic steps are as follows:

[0101] First, construct training data. Input: candidate schemes obtained from NSGA-2 (8 decision variables). Output: corresponding simulation results (e.g., yield, irrigation amount, WUE, irrigation number) to construct a multi-output regression model.

[0102] Secondly, the random forest model is trained, mainly through cross-validation and parameter tuning (such as adjusting n_estimators, max_features, etc.) to ensure that the model has a good fitting effect.

[0103] Then, feature importance is extracted. After training, the importance of each decision variable is extracted using `model.feature_importances_`. This set of importance values ​​can be denoted as:

[0104] RFimp = [rf1, rf2, ..., rf8]

[0105] This indicates the degree of impact on output at each stage, such as "lower limit of soil moisture content" or "percentage of irrigation evapotranspiration".

[0106] Finally, mapping to the output metrics, since the candidate solutions output four evaluation metrics, and decision variables usually appear in pairs, further analysis is possible:

[0107] (4) Detailed steps of the fusion strategy: ① Entropy weight vector normalization: After obtaining the weights, normalize them to ensure that the sum of each term is 1. ② Random forest contribution vector normalization: Normalize the contribution values ​​RFoutput of each evaluation index obtained by the random forest through sensitivity mapping, and obtain:

[0108] wRF=[wRF, Yield, wRF, Irrigation, wRF, WUE, wRF, Frequency].

[0109] ③Weight fusion: The two weights are fused using a weighted average method. The weight of the fused index is calculated as follows: wfsion=α×wentrop+(1-α)×wRF.

[0110] Here, α is the fusion factor, which can be determined based on practical experience or through cross-validation (e.g., when α = 0.5, the two parts are considered equally). The resulting wfusion reflects both the scheme's ability to distinguish among candidate data (entropy weight) and the sensitivity of each decision variable to the final output (random forest contribution).

[0111] ④ Overall Score Calculation: The output metrics of each candidate solution are weighted and summed using the fused weights (wfusion) to obtain the final overall score. The calculation formula can be written as:

[0112] Wherein: S i Let be the overall score of the i-th candidate solution; The j-th output index of candidate scheme i after normalization (e.g., normalized yield, irrigation amount, WUE, irrigation frequency).

[0113] In the code, this step can modify the original get_score function to use the fused weights instead of the simple entropy weights.

[0114] ⑤ Scheme Ranking and Decision Output: Rank all candidate schemes according to their comprehensive score Si, and the scheme with the highest score is the recommended scheme; at the same time, the top few candidate schemes can be output for actual decision-making reference. You can further illustrate the trade-offs between the schemes through visualization methods (such as Pareto front plots, box plots, etc.).

[0115] This fusion framework ensures information sharing and mutual complementarity between the two algorithms: Random Forest provides predictive data and objective importance, while Entropy Weight provides multi-indicator trade-offs and objective weighting. Their combination automates and scientizes the decision-making process.

[0116] Based on candidate irrigation schemes generated by AquaCrop and PyFAO56 simulations, a series of Pareto optimal solutions were formed, such as... Figure 1 As shown, each scheme exhibits different performance indicators in terms of yield, irrigation volume, irrigation water use efficiency, and irrigation frequency, demonstrating a clear trade-off.

[0117] Table 1

[0118]

[0119] On the one hand, the highest-yielding schemes often involve greater irrigation inputs (higher volume and frequency of irrigation), resulting in lower water use efficiency. On the other hand, extremely water-saving schemes significantly reduce irrigation volume and frequency, but crop yields also decrease. Most compromise schemes fall somewhere in between, reducing water use while moderately sacrificing yield to improve IWUE.

[0120] The entropy weight method assigns the following weights to each indicator: irrigation frequency 0.4722, irrigation volume 0.3009, yield 0.1207, and water use efficiency (WUE) 0.1060. It can be seen that the comprehensive evaluation places great emphasis on reducing irrigation frequency and water consumption (the sum of these two weights is close to 0.77), while the weights for yield and water use efficiency are relatively low. This weight allocation directly determines the final ranking: top-ranked schemes have a significant advantage in water-saving indicators. First, extremely few irrigations are a common characteristic of high-ranking schemes. Many top-ranking schemes reduce irrigation frequency to only 1-2 times, far fewer than other schemes. This allows them to obtain the highest evaluation score on the irrigation frequency indicator. A reduction in irrigation frequency often also means a reduction in total water consumption—and indeed, this is the case: leading schemes typically reduce total irrigation volume significantly; some schemes reduce total irrigation water consumption by tens of percentage points compared to high-yield, high-water-consumption schemes, and also score highly on the irrigation volume indicator. Despite significant water savings, these winning schemes still maintain high yield levels. The top-ranked strategies generally achieved yields exceeding 80% of those under fully irrigated conditions, with some even approaching 90%. Since yield has a weight of only 0.12, this slight difference in yield has a relatively limited impact on the overall score. Simultaneously, water use efficiency (WUE) significantly improved due to water conservation: achieving near-maximum yields with less irrigation water resulted in a substantial increase in output per unit of irrigation water. This allowed the top-ranked strategies to also achieve good scores on the WUE indicator (although WUE has a small weight, high WUE further solidified their overall advantage). In summary, the high scores of the top-ranked strategies stemmed from "trading significant water-saving benefits for smaller yield losses." Since the weight of irrigation frequency is approximately four times that of yield, the score improvement from reducing irrigation frequency far outweighs the gains from a small increase in yield. Therefore, without exception, these strategies chose to minimize irrigation volume and frequency within acceptable yield loss limits, achieving an optimal balance between output and input. This is the main reason they excelled in the overall ranking. Those schemes with slightly higher yields but more frequent irrigation suffer losses due to excessive irrigation frequency and water consumption, resulting in significant point deductions. Their increased yields cannot compensate for the high cost of irrigation, ultimately leading to a lower overall score compared to water-saving schemes.

[0121] Table 2

[0122]

[0123]

[0124] After comprehensively evaluating the above multiple indicators using the entropy weight method, the final ranking of each scheme was obtained. The results show that the top-ranked schemes mostly maintained a stable yield of around 9 t·hm⁻², and their irrigation water use efficiency also remained at a high level. This indicates that a reasonable irrigation strategy (appropriate irrigation quota and suitable irrigation frequency) can maintain a stable and high yield while conserving water resources. Many of the top-ranked schemes involved 8-10 irrigations, resulting in higher costs, but these were not excessively increased to an unacceptable level. This demonstrates that on-demand irrigation and multiple irrigations have significant advantages in improving yield and water use efficiency. For water-scarce areas or areas with limited funds, a moderate to fewer irrigation frequency can be chosen, but precise scheduling is crucial to avoid a significant reduction in yield.

[0125] Analysis of irrigation schemes derived from the Pareto front, examining the contributions of eight variables across four irrigation periods to four evaluation indicators, reveals that the soil water threshold during the critical growth stage (peak soil water demand) is the most important variable. This threshold determines the level of soil water content at which irrigation should begin. During the peak water demand phase for crops (typically corresponding to heading and flowering to grain filling), the threshold setting directly impacts yield and irrigation frequency: a high threshold ("conservative irrigation," irrigating when the soil is slightly dry) increases irrigation frequency and water consumption to ensure crop survival from drought, guaranteeing yield but decreasing IWUE; a low threshold ("relaxed irrigation," allowing for irrigating only when the soil is drier) reduces irrigation frequency and water consumption, increasing IWUE, but excessively low thresholds can exacerbate crop water stress during this stage, leading to a significant yield decrease. In the random forest model, this threshold contributes most to the overall score fluctuation, demonstrating its crucial role. Figure 2 As shown.

[0126] The amount of water applied per irrigation during the same critical stage (irrigation intensity) is a highly important factor, closely following the critical stage. The amount of water applied each time during critical irrigation has a profound impact: insufficient irrigation intensity fails to adequately alleviate soil drought, potentially requiring more frequent irrigation (increasing the number of applications) or subjecting crops to persistent mild water stress (affecting yield); while excessive irrigation intensity, exceeding the soil's effective capacity per application, leads to water leakage and waste, reducing irrigation frequency but increasing total water consumption and decreasing water use efficiency (WUE). Random forest analysis shows that, within the scope of the model, this variable is second only to the critical stage threshold in importance, indicating that appropriate irrigation dosage is crucial for balancing yield and water conservation.

[0127] Furthermore, the soil moisture threshold during the later stages of growth (maturity) also demonstrates significant importance. Many of the top-ranked schemes employed a lower soil moisture threshold during the later stages of crop growth (allowing for drier soil before irrigation or even no irrigation at all) to reduce irrigation inputs near maturity. Since crops are less sensitive to water shortages in the later stages, this approach results in minimal yield loss while significantly conserving water. Therefore, the positive impact of the later-stage threshold on overall performance is also reflected in the models.

[0128] like Figures 3-10 As shown, the partial dependency analysis of random forest further reveals the relationship between various important variables and performance indicators. For the most important variable, the medium-term soil water threshold, there exists an optimal range: the partial dependency plot shows that when the threshold is too high, the overall score decreases due to over-irrigation; when the threshold is too low, the score decreases due to yield loss. This indicates that a moderate threshold balances yield and water conservation, maximizing overall performance. Similarly, the dependency relationship of the irrigation intensity variable also has an inflection point: moderate irrigation volume per application is beneficial to improving the overall score, while too low or too high intensity will cause the score to decrease (either due to increased frequency or decreased marginal benefits of water). These findings provide a basis for irrigation decisions: managers should focus on these highly performance-sensitive variables and optimize threshold and irrigation volume settings within the above reasonable range to achieve optimal balance between yield and water efficiency.

[0129] like Figure 11 As shown, the correlations among the eight variables reveal a strong negative correlation between the slow growth period and the maximum coverage period, indicating that the water control strategies employed during these stages are contradictory. Optimizing irrigation strategies for these stages requires balancing the water demands of both periods.

[0130] The weak correlation between the emergence period and other stages means that water management during the emergence period has little impact on subsequent stages, especially the fundamental role of early water control in later growth.

[0131] The positive correlation between the maximum coverage period and the aging period suggests that these two stages may have some commonalities in water demand, and irrigation strategies may be relatively consistent between the two stages.

[0132] Correlation analysis between stages provides clues for optimizing irrigation programs. For example, the strong negative correlation between the slow growth period and the maximum coverage period suggests that different irrigation strategies may be needed in these two stages, as shown in Tables 3-4.

[0133] Table 3

[0134]

[0135] Table 4

[0136]

[0137]

[0138] From the perspective of the decision-making schemes, all schemes showed relatively consistent irrigation control and evapotranspiration during the seedling stage, with generally higher irrigation amounts during the maximum coverage stage, while evapotranspiration decreased during the aging stage. The irrigation evapotranspiration varied significantly among different schemes during the slow growth stage, which may be related to differences in soil moisture control. Particularly during the slow growth and aging stages, this suggests that the schemes may compensate for insufficient soil moisture by increasing evapotranspiration.

[0139] Secondly, in order to improve the efficiency of water resource utilization, a balance can be considered between the slow growth period and the aging period to avoid over-irrigation and waste of water resources, while maintaining appropriate soil moisture and evapotranspiration.

[0140] Based on the above detailed analysis, the unified irrigation decision calculation framework of the present invention has the following technical effects:

[0141] Multi-stage water management: By setting irrigation control limits and irrigation evapotranspiration percentages in stages, refined irrigation management is carried out for different key stages of crop growth to ensure sufficient water during the critical period of yield formation (especially from the period of maximum canopy coverage to senescence), while avoiding over-irrigation in the early and mature stages.

[0142] Efficient water use: Image analysis shows that adjusting the irrigation evapotranspiration percentage in the later stages can significantly improve IWUE. This framework can maximize water use efficiency while ensuring yield.

[0143] Reduced irrigation frequency and cost savings: By setting reasonable irrigation thresholds, the framework effectively controls the number of irrigations, avoiding overly frequent irrigation and reducing energy consumption and operating costs.

[0144] The integration of objective and model prediction results: the entropy weight method assigns objective weights to each output indicator (yield, irrigation amount, IWUE, irrigation frequency), while random forest analysis reveals the actual contribution of decision variables at each stage to these indicators. The integration of the two constructs a scientific and reasonable comprehensive scoring system, providing solid data support for decision-making.

[0145] like Figure 12As shown, the main objective of this technical solution is to improve the yield and water use efficiency of winter wheat by simulating and optimizing irrigation strategies. Specifically, this technology combines multiple models (AquaCrop, FAO56) and algorithms (NSGA-2, multi-objective optimization, entropy weight method, random forest). By comprehensively considering factors such as irrigation strategies, soil characteristics, and crop growth, it simulates and optimizes the irrigation strategy for winter wheat. Then, through the combination of multi-objective optimization and intelligent algorithms, it can find the optimal irrigation scheme under various decision variables and constraints to improve crop yield and water use efficiency. Simultaneously, the entropy weight method and random forest are combined for the optimization and decision-making of the irrigation scheme.

[0146] like Figure 13 As shown, the integrated technical solution based on AquaCrop, pyfao56, ​​NSGA-II, random forest, and entropy weight method can effectively reduce water consumption, improve irrigation water use efficiency, and ensure high yields with fewer irrigation cycles. Compared with traditional agricultural solutions and literature-based solutions, this technology demonstrates significant advantages in various key indicators, particularly in water conservation and efficient water resource utilization, showing high application value in the context of water scarcity.

[0147] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A multi-objective dynamic optimization method for irrigation of winter wheat based on crop models and artificial intelligence, characterized in that, Includes the following steps: Meteorological data, soil data, crop parameter data, and irrigation management data are preprocessed to obtain standardized input data; The AquaCrop model and the PyFAO56 model are coupled based on the standardized input data to obtain the final output of the coupled model. The specific process includes: inputting meteorological data, soil data, and crop parameter data from the standardized input data into the AquaCrop model to obtain crop growth indicators and water use; inputting the meteorological data and soil data from the standardized input data into the PyFAO56 model to generate irrigation events; and feeding the irrigation events back to the AquaCrop model as input to obtain the final output of the coupled model. The final output of the coupled model includes several evaluation indicators, including: final yield, total irrigation volume, irrigation water use efficiency, and number of irrigations. Decision variables include: lower limit of soil moisture content control and percentage of irrigation evapotranspiration recharge; Based on the final output of the coupled model, the NSGA-2 multi-objective optimization algorithm is used to optimize the decision variables to obtain the Pareto optimal solution set. The specific process includes: randomly generating an initial population within the decision variable parameter space; simulating the decision variables of each individual in the initial population based on the coupled model to obtain the total irrigation water volume, crop yield, and irrigation water use efficiency corresponding to each individual; stratifying all individuals in the initial population using a non-dominated sorting method to obtain non-dominated levels and calculating the crowding distance of each individual in the objective space; performing selection, crossover, and mutation operations on the population using genetic operators based on the non-dominated levels and crowding distance to generate offspring populations; and outputting the Pareto optimal solution set composed of non-dominated solutions after a preset number of iterations. The final irrigation strategy is obtained by ranking the candidate solutions in the Pareto optimal solution set using a fusion strategy combining entropy weight method and random forest algorithm.

2. The method for dynamic optimization of multi-objective irrigation for winter wheat based on crop models and artificial intelligence according to claim 1, characterized in that, The meteorological data includes: daily maximum temperature, daily minimum temperature, rainfall, radiation, wind speed, and humidity; The soil data includes: soil profile information, field water holding capacity, wilting point, saturated water content, soil permeability, and soil stratification. The crop parameter data includes: winter wheat planting date, maturity period, root distribution, canopy growth parameters, biomass conversion coefficient, harvest index, and crop water sensitivity parameters; The irrigation management data includes: the initial irrigation plan, the irrigation time window for each stage, and the irrigation volume.

3. The method for dynamic optimization of multi-objective irrigation for winter wheat based on crop models and artificial intelligence according to claim 1, characterized in that, The process of ranking candidate solutions in the Pareto optimal solution set using a fusion strategy combining entropy weighting and random forest algorithms to obtain the final irrigation strategy includes: The entropy weight method is used to calculate the weights of each evaluation index of the candidate solutions in the Pareto optimal solution set, and the index weights based on information entropy are obtained. The contribution of decision variables to each evaluation index is determined based on the random forest algorithm, and the index weights based on the model contribution are obtained. The weights of the indicators based on information entropy and the weights of the indicators based on model contribution are combined to obtain a comprehensive weight. Based on the comprehensive weights, the candidate solutions in the Pareto optimal solution set are comprehensively evaluated and ranked to obtain the final irrigation strategy.

4. The method for dynamic optimization of multi-objective irrigation for winter wheat based on crop models and artificial intelligence according to claim 3, characterized in that, The calculation process for the comprehensive weight includes: The weights of the information entropy-based indicators are normalized. The weights of the indicators based on model contribution are normalized. Based on a preset fusion factor, the normalized index weights based on information entropy and the normalized index weights based on model contribution are weighted and averaged to obtain the comprehensive weight.

Images