Irrigation optimization method based on groundwater level threshold control

Abstract

The present invention relates to the field of ecohydrological technology, and specifically discloses an irrigation optimization method based on groundwater level threshold control, the method including data collection and preprocessing, crop growth model construction, feature analysis, groundwater level prediction, irrigation demand prediction, irrigation strategy optimization and continuous optimization. This scheme extracts the long-term trend, periodic components and random components of groundwater level data, integrates multi-source features to provide a more refined, clear and stable feature set; uses stationary differential processing to improve the adaptability of the model to non-stationary time series, combines the AquaCrop model and the groundwater level prediction model, simulates the impact of different water levels on crop growth, and provides a dynamic irrigation scheme; takes crop yield and water use efficiency as the core, balances the relationship between irrigation efficiency and yield by setting the weight coefficients of yield and water consumption, and ensures the robustness and reliability of the model results through dynamic adjustment and iterative improvement of model parameters.

Description

Technical Field

[0001] The present invention relates to the field of ecohydrology technology, and in particular to an irrigation optimization method based on groundwater level threshold regulation. Background Art

[0002] As the main water supply source, groundwater resources are responsible for agricultural irrigation and domestic water use. Groundwater sources are widely distributed and the water volume is relatively stable. Rational regulation of groundwater levels is crucial for sustainable agricultural development. Traditional groundwater level threshold regulation methods often only use groundwater level data from a single source, ignoring the potential impact of meteorology, soil moisture, etc. on groundwater level changes. It is difficult to capture complex time series features during feature processing and lacks the ability to fuse multi-source information. General time series prediction methods often ignore the assessment of fluctuation factors and uncertainty factors, and the credibility of prediction accuracy is low in actual decision-making. The coupling between groundwater level prediction and crop growth is insufficient, increasing the risk of misjudgment of the model. General irrigation strategy optimization methods often rely on empirical settings and lack feedback adjustment processes, making it difficult to ensure the rationality and scientificity of irrigation plans. Summary of the invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an irrigation optimization method based on groundwater level threshold control. Traditional groundwater level threshold control methods often only use groundwater level data from a single source, ignoring the potential impact of meteorology, soil moisture, etc. on groundwater level changes, and are difficult to capture complex time series features during feature processing, and lack the ability to fuse multi-source information. This solution extracts the long-term trend, periodic components and random components of groundwater level data, eliminates short-term random fluctuations that are not conducive to prediction, integrates multi-source features to provide a more refined, clear and stable feature set, and improves the accuracy of subsequent model predictions; general time series prediction methods often ignore the evaluation of fluctuation factors and uncertainty factors, and the credibility of prediction accuracy is low in actual decision-making, and groundwater level prediction and crop In order to solve the problem of insufficient coupling between growth and increased risk of misjudgment of the model, this solution improves the adaptability of the model to non-stationary time series. By combining the groundwater level prediction model with the AquaCrop model, it simulates the impact of different groundwater levels on crop growth, provides dynamic irrigation plans for crop water demand, and ensures that the prediction results meet actual needs. In view of the general irrigation strategy optimization methods that often rely on empirical settings and lack feedback adjustment processes, and the rationality and scientificity of irrigation plans are difficult to guarantee, this solution takes crop yield and water use efficiency as the core, balances the relationship between irrigation efficiency and yield by setting weight coefficients for yield and water consumption, and achieves the optimal groundwater level threshold setting and water use efficiency through dynamic adjustment and iterative improvement of model parameters, ensuring the robustness and reliability of model results.

[0004] The technical solution adopted by the present invention is as follows: The irrigation optimization method based on groundwater level threshold control provided by the present invention comprises the following steps:

[0005] Step S1: Data collection and preprocessing, collecting historical environmental data and crop growth data. The historical environmental data includes meteorological data, topography, irrigation history, soil moisture, and groundwater level, and processing missing values ​​and outliers in the data;

[0006] Step S2: Construction of crop growth model: Based on historical environmental data and crop growth data, an AquaCrop model of crop response to water productivity is preliminarily constructed to simulate crop water demand under different groundwater levels and preliminarily determine the irrigation demand of crops at various stages;

[0007] Step S3: feature analysis, analyzing the time series features from the historical environmental data, identifying the changing patterns of the groundwater level, and generating a feature set;

[0008] Step S4: Groundwater level prediction: construct a SARIMA model based on the feature set as a groundwater level prediction model, output the future groundwater level prediction value, and combine it with the AquaCrop model to simulate the crop growth status;

[0009] Step S5: Irrigation demand prediction: using the prediction results of the groundwater level prediction model, the crop growth stages are divided according to the crop life cycle, the water demand of crops at different stages is predicted, and corresponding irrigation strategies are formulated;

[0010] Step S6: Optimizing irrigation strategy: using optimization algorithm to adjust the model parameters of the groundwater level prediction model, finding the optimal groundwater level threshold, and finding the optimal irrigation strategy based on the simulation results of the AquaCrop model;

[0011] Step S7: Continuously optimize, regularly observe and provide feedback on irrigation effects, and adjust and optimize irrigation strategies based on the feedback on irrigation effects.

[0012] Furthermore, in step S3, the feature analysis specifically includes the following steps:

[0013] Step S31: Time series segmentation, generating periodic time tags according to year, month, season, and solar term;

[0014] Step S32: extracting environmental periodic features, specifically comprising: extracting periodic components in meteorological data according to periodic time tags, analyzing periodic fluctuations in soil moisture, and extracting seasonal and annual periodic variations in groundwater levels; the environmental periodic features extracted include periodic components in meteorological data, periodic fluctuations in soil moisture, and seasonal and annual periodic variations in groundwater levels;

[0015] Step S33: Trend decomposition, extracting the long-term change trend, periodic fluctuation component and random fluctuation component of the groundwater level, including the following steps:

[0016] Step S331: performing trend decomposition on groundwater level data and extracting long-term change trends using a moving average method;

[0017] Step S332: removing the long-term variation trend from the original data of the groundwater level to obtain residual data containing periodic fluctuation components and random fluctuation components;

[0018] Step S333: further calculating the moving average of the residual data to eliminate the low-frequency components of the periodic fluctuation components;

[0019] Step S334: using high-frequency filtering to remove high-frequency components of periodic fluctuation components from the residual data, and finally obtaining random fluctuation components;

[0020] Step S34: introducing hysteresis characteristics, adding hysteresis characteristics of groundwater level, meteorological data, and soil moisture of the previous day, week, and month;

[0021] Step S35: Cumulative and differential feature extraction, calculating the cumulative value of irrigation history and soil moisture, identifying the phased water resource replenishment or consumption, calculating the first-order difference of the groundwater level, and capturing the incremental change of the groundwater level. The formula used is as follows:

[0022] ;

[0023] In the formula, represents the incremental change in groundwater level, Indicates Groundwater level data at the moment, Indicates Groundwater level data at all times;

[0024] Step S36: feature selection and dimensionality reduction, integrating all the features extracted in steps S32 to S35 into time series features, using feature correlation analysis method to remove redundant features in the time series features, evaluating the importance of each feature to groundwater level prediction, setting weight parameters, and using principal component analysis method to reduce the dimension of the time series features;

[0025] Step S37: normalize the selected and dimension-reduced time series features to generate a feature set containing multi-source information.

[0026] Furthermore, in step S4, the groundwater level prediction specifically includes the following steps:

[0027] Step S41: feature set division, dividing the feature set into a training set and a validation set;

[0028] Step S42: Check the stability of the groundwater level data. If it is not stable, perform differential processing on the data. For time series data with obvious seasonal fluctuations, perform seasonal differential processing;

[0029] Step S43: Establish a SARIMA model as a groundwater level prediction model, select the non-seasonal difference number and seasonal difference number according to the stability check result, select the seasonal cycle according to the data type, and determine the model order. The model form is as follows:

[0030] ;

[0031] In the formula, represents the SARIMA model, represents the order of the non-seasonal autoregressive term, represents the order of the non-seasonal moving average term, represents the number of non-seasonal differences, represents the order of seasonal autoregressive term, represents the number of seasonal differences, represents the order of seasonal moving average term, Represents seasonal cycles;

[0032] Step S44: model fitting, inputting the selected parameters into the groundwater level prediction model, inputting the training set fitting model, and using grid search to optimize the model parameters;

[0033] Step S45: Calculate the mean square error on the validation set to evaluate the model effect, and adjust the model parameters to improve the prediction accuracy of the model;

[0034] Step S46: Groundwater level prediction, setting the prediction time range and step size, outputting the future groundwater level prediction value, adding a confidence interval to the prediction result, and conducting uncertainty assessment.

[0035] Furthermore, in step S6, the irrigation strategy optimization specifically includes the following steps:

[0036] Step S61: define the optimization goal as improving water utilization efficiency and design the objective function. The formula used is as follows:

[0037] ;

[0038] In the formula, represents the objective function, and is the weight coefficient used to balance the impact of production and water consumption, Indicates the growth status of crops. represents water use efficiency, To prevent the random constant from having a zero denominator;

[0039] Step S62: preliminarily setting the groundwater level threshold according to actual conditions;

[0040] Step S63: combining the crop growth status and the groundwater level prediction value, using a particle swarm optimization algorithm to optimize the parameters of the groundwater level prediction model;

[0041] Step S64: iteratively run the model parameters, calculate the objective function value, and when the objective function value reaches the maximum, output the optimal solution of the optimal model parameters and the groundwater level threshold;

[0042] Step S65: Simulate using the AquaCrop model under the optimal model parameters to evaluate the actual impact of the optimal model parameters on crop growth.

[0043] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0044] (1) Traditional groundwater level threshold control methods often only use groundwater level data from a single source, ignoring the potential impact of meteorology, soil moisture, etc. on groundwater level changes. It is difficult to capture complex time series features during feature processing and lacks the ability to fuse multi-source information. This scheme extracts the long-term trend, periodic components, and random components of groundwater level data, eliminates short-term random fluctuations that are not conducive to prediction, and integrates multi-source features to provide a more refined, clear, and stable feature set, thereby improving the accuracy of subsequent model predictions.

[0045] (2) In view of the fact that the evaluation of fluctuation factors and uncertainty factors is often ignored in general time series prediction methods, the credibility of prediction accuracy is low in actual decision-making, the coupling between groundwater level prediction and crop growth is insufficient, and the risk of misjudgment of the model is increased. This solution improves the adaptability of the model to non-stationary time series. By combining the groundwater level prediction model with the AquaCrop model, it simulates the impact of different groundwater levels on crop growth and provides a dynamic irrigation plan for crop water demand, ensuring that the prediction results meet actual needs.

[0046] (3) In view of the fact that general irrigation strategy optimization methods often rely on empirical settings, lack feedback adjustment processes, and are difficult to ensure the rationality and scientificity of irrigation plans, this plan focuses on crop yield and water use efficiency. By setting weight coefficients for yield and water consumption, it balances the relationship between irrigation efficiency and yield. Through dynamic adjustment and iterative improvement of model parameters, it achieves the optimal groundwater level threshold setting and water use efficiency, ensuring the robustness and reliability of the model results. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A schematic diagram of the flow of the irrigation optimization method based on groundwater level threshold control proposed by the present invention;

[0048] Figure 2 is a schematic flow chart of step S3;

[0049] Figure 3 It is a schematic diagram of the process of step S4.

[0050] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0051] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0052] Example 1, see Figure 1 The irrigation optimization method based on groundwater level threshold control provided by the present invention comprises the following steps:

[0053] Step S1: Data collection and preprocessing, collecting historical environmental data and crop growth data. The historical environmental data includes meteorological data, topography, irrigation history, soil moisture, and groundwater level, and processing missing values ​​and outliers in the data;

[0054] Step S2: Construction of crop growth model: Based on historical environmental data and crop growth data, an AquaCrop model of crop response to water productivity is preliminarily constructed to simulate crop water demand under different groundwater levels and preliminarily determine the irrigation demand of crops at various stages;

[0055] Step S3: feature analysis, analyzing the time series features from the historical environmental data, identifying the changing patterns of the groundwater level, and generating a feature set;

[0056] Step S4: Groundwater level prediction: construct a SARIMA model based on the feature set as a groundwater level prediction model, output the future groundwater level prediction value, and combine it with the AquaCrop model to simulate the crop growth status;

[0057] Step S5: Irrigation demand prediction: using the prediction results of the groundwater level prediction model, the crop growth stages are divided according to the crop life cycle, the water demand of crops at different stages is predicted, and corresponding irrigation strategies are formulated;

[0058] Step S6: Irrigation strategy optimization, using the optimization algorithm to adjust the model parameters of the groundwater level prediction model, calibrate the initial conditions of soil, water, and crop growth, find the optimal groundwater level threshold, and find the optimal irrigation strategy based on the simulation results of the AquaCrop model;

[0059] Step S7: Continuously optimize, regularly observe and provide feedback on irrigation effects, and adjust and optimize irrigation strategies based on the feedback on irrigation effects.

[0060] Example 2, see Figure 1 This embodiment is based on the above embodiment. In step S1, the missing value processing adopts the interpolation method, and the outlier value processing adopts the mean filling method.

[0061] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, the AquaCrop model is used to simulate the growth process of crops under different water supply conditions, including the simulation of the water supply of shallow groundwater to crops, the simulation of the water absorption of crop roots, and the simulation of the impact of groundwater level changes on crop yields.

[0062] Example 4, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S3, feature analysis specifically includes the following steps:

[0063] Step S31: Time series segmentation, generating periodic time tags according to year, month, season, and solar term;

[0064] Step S32: Extraction of environmental periodic features, specifically comprising: extracting periodic components in meteorological data according to periodic time tags, analyzing periodic fluctuations of soil moisture, and extracting seasonal and annual periodic variation patterns of groundwater levels; the extracted environmental periodic features include periodic components in meteorological data, periodic fluctuations of soil moisture, and seasonal and annual periodic variation patterns of groundwater levels; wherein the periodic components in meteorological data include seasonal variations of precipitation, temperature, and evaporation; periodic fluctuations of soil moisture include alternating dry and wet changes of soil; seasonal and annual periodic variation patterns of groundwater levels include changes in flood season, normal water season, and dry season;

[0065] Step S33: Trend decomposition, extracting the long-term change trend, periodic fluctuation component and random fluctuation component of the groundwater level, including the following steps:

[0066] Step S331: performing trend decomposition on groundwater level data and extracting long-term change trends using a moving average method, including the following steps:

[0067] Step S3311: Select the window size, and select the time window size according to the change law of groundwater level ;

[0068] Step S3312: Calculate the moving average and calculate the moving average before each time point. The average value of the daily data is used to replace the groundwater level data at each time point with the average value within the window. The formula used is as follows:

[0069] ;

[0070] In the formula, Indicates The moving average of time, is the window size, For the Groundwater level at the moment;

[0071] Step S3313: Generate a moving average sequence, arrange the moving averages at each moment in chronological order, and the obtained moving average sequence is the long-term trend of the groundwater level data;

[0072] Step S332: Remove the long-term variation trend from the original data of the groundwater level to obtain residual data containing periodic fluctuation components and random fluctuation components. The formula used is as follows:

[0073] ;

[0074] In the formula, represents the residual data, Raw data representing groundwater levels;

[0075] Step S333: further calculating the moving average of the residual data to eliminate the low-frequency components of the periodic fluctuation components;

[0076] Step S334: Use high-frequency filtering to remove the high-frequency components of the periodic fluctuation components from the residual data, and finally obtain the random fluctuation components. The formula used is as follows:

[0077] ;

[0078] In the formula, represents the periodic fluctuation component, represents the random fluctuation component;

[0079] Step S34: introducing hysteresis characteristics, adding hysteresis characteristics of groundwater level, meteorological data, and soil moisture of the previous day, week, and month;

[0080] Step S35: Cumulative and differential feature extraction, calculating the cumulative value of irrigation history and soil moisture, identifying the phased water resource replenishment or consumption, calculating the first-order difference of the groundwater level, and capturing the incremental change of the groundwater level. The formula used is as follows:

[0081] ;

[0082] In the formula, represents the incremental change in groundwater level, Indicates Groundwater level data at the moment, Indicates Groundwater level data at all times;

[0083] Step S36: feature selection and dimensionality reduction, integrating all features extracted from steps S32 to S35 into time series features, including the following steps:

[0084] Step S361: feature correlation analysis, calculating the Pearson correlation coefficient between each feature in the time series feature and the groundwater level, generating a correlation matrix to analyze the correlation between features, setting the correlation coefficient, retaining features whose correlation calculation results exceed the correlation coefficient, and removing redundant features;

[0085] Step S362: feature importance evaluation, evaluating the importance score of each feature to groundwater level prediction, sorting the features according to the feature importance score and setting the weight parameters of the features;

[0086] Step S363: using principal component analysis to reduce feature dimensions and retain the main information in the features;

[0087] Step S37: normalize the selected and dimensionally reduced time series features to generate a feature set containing multi-source information.

[0088] Example 5, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. In step S4, groundwater level prediction specifically includes the following steps:

[0089] Step S41: feature set division, dividing the feature set into a training set and a validation set;

[0090] Step S42: Use unit root test to check the stability of groundwater level data. If it is not stable, perform differential processing on the data. For time series data with obvious seasonal fluctuations, perform seasonal differential processing;

[0091] Step S43: Establish a SARIMA model as a groundwater level prediction model. According to the stationarity check results, select the number of non-seasonal differences as 1 and the number of seasonal differences as 1, select the seasonal period as 12, and use the autocorrelation function and partial autocorrelation function to determine the model orders as 2, 1, 1, and 1 respectively. The formula used is as follows:

[0092] ;

[0093] ;

[0094] In the formula, represents the SARIMA model, express The groundwater level at the time, , represents the non-seasonal autoregressive coefficient, represents the non-seasonal moving average coefficient, represents the seasonal autoregressive coefficient, represents the seasonal moving average coefficient, represents the white noise error term;

[0095] Step S44: model fitting, inputting the selected parameters into the groundwater level prediction model, fitting the model using the training set, and optimizing the model parameters using grid search;

[0096] Step S45: Calculate the mean square error on the validation set to evaluate the model effect, and adjust the model parameters to improve the prediction accuracy of the model;

[0097] Step S46: Groundwater level prediction, setting the prediction time range and step size, outputting the future groundwater level prediction value, adding a confidence interval to the prediction result, and conducting uncertainty assessment.

[0098] Example 5, see Figure 1 This embodiment is based on the above embodiment. In step S5, irrigation demand prediction specifically includes the following steps:

[0099] Step S51: dividing the crop growth stages, including sowing stage, seedling stage, jointing stage, flowering stage, and maturity stage;

[0100] Step S52: Obtain the predicted value of the future groundwater level, combine it with the crop water demand output by the AquaCrop model simulation, and analyze the irrigation demand of crops at different stages. The formula used is as follows:

[0101] ;

[0102] In the formula, Indicates the irrigation demand at a specific stage, Indicates the water requirement of crops. represents the moisture sensitivity coefficient of crops, A time interval representing a specific phase;

[0103] Step S53: According to the calculated irrigation demand, formulate corresponding irrigation strategies, including irrigation frequency and irrigation amount: when the groundwater level is high, reduce the irrigation frequency and irrigation amount; when the groundwater level is low, increase the irrigation frequency and irrigation amount.

[0104] Example 6, see Figure 1 This embodiment is based on the above embodiment. In step S6, the irrigation strategy is optimized, specifically including the following steps:

[0105] Step S61: define the optimization goal as improving water utilization efficiency and design the objective function. The formula used is as follows:

[0106] ;

[0107] In the formula, represents the objective function, and is the weight coefficient used to balance the impact of production and water consumption, Indicates the growth status of crops. represents water use efficiency, To prevent the random constant from having a zero denominator;

[0108] Step S62: preliminarily setting the groundwater level threshold according to actual conditions;

[0109] Step S63: combining the crop growth status and the groundwater level prediction value, using a particle swarm optimization algorithm to optimize the parameters of the groundwater level prediction model;

[0110] Step S64: iteratively run the model parameters, calculate the objective function value, and when the objective function value reaches the maximum, output the optimal solution of the optimal model parameters and the groundwater level threshold;

[0111] Step S65: Simulate using the AquaCrop model under the optimal model parameters to evaluate the actual impact of the optimal model parameters on crop growth.

[0112] Embodiment 7, see Figure 1 This embodiment is based on the above embodiment. In step S7, continuous optimization is performed, specifically, regularly monitoring environmental data and crop growth data, extracting features from updated data, and adding them to the feature set to train a groundwater level prediction model, establish a feedback mechanism between irrigation strategies and crop growth effects, and verify and adjust irrigation effects.

[0113] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0114] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

[0115] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. An irrigation optimization method based on groundwater level threshold control, characterized in that: The method comprises the following steps: Step S1: Data collection and preprocessing, collecting historical environmental data and crop growth data. The historical environmental data includes meteorological data, topography, irrigation history, soil moisture, and groundwater level, and processing missing values ​​and outliers in the data; Step S2: Construction of crop growth model: Based on historical environmental data and crop growth data, an AquaCrop model of crop response to water productivity is preliminarily constructed to simulate crop water demand under different groundwater levels and preliminarily determine the irrigation demand of crops at various stages; Step S3: feature analysis, analyzing the time series features from the historical environmental data, identifying the changing patterns of the groundwater level, and generating a feature set; Step S4: Groundwater level prediction: construct a SARIMA model based on the feature set as a groundwater level prediction model, output the future groundwater level prediction value, and combine it with the AquaCrop model to simulate the crop growth status; Step S5: Irrigation demand prediction: using the prediction results of the groundwater level prediction model, the crop growth stages are divided according to the crop life cycle, the water demand of crops at different stages is predicted, and corresponding irrigation strategies are formulated; Step S6: Optimizing irrigation strategy: using optimization algorithm to adjust the model parameters of the groundwater level prediction model, finding the optimal groundwater level threshold, and finding the optimal irrigation strategy based on the simulation results of the AquaCrop model; Step S7: Continuously optimize, regularly observe and feedback the irrigation effect, and adjust and optimize the irrigation strategy according to the feedback irrigation effect; In step S4, the groundwater level prediction includes the following steps: Step S41: feature set division, dividing the feature set into a training set and a validation set; Step S42: Check the stability of the groundwater level data. If it is not stable, perform differential processing on the data. For time series data with obvious seasonal fluctuations, perform seasonal differential processing; Step S43: Establish a SARIMA model as a groundwater level prediction model, select the non-seasonal difference number and seasonal difference number according to the stability check result, select the seasonal cycle according to the data type, and determine the model order. The model form is as follows: ； In the formula, represents the SARIMA model, represents the order of the non-seasonal autoregressive term, represents the order of the non-seasonal moving average term, represents the number of non-seasonal differences, represents the order of seasonal autoregressive term, represents the number of seasonal differences, represents the order of seasonal moving average term, Represents seasonal cycles; Step S44: model fitting, inputting the selected parameters into the groundwater level prediction model, inputting the training set fitting model, and using grid search to optimize the model parameters; Step S45: Calculate the mean square error on the validation set to evaluate the model effect, and adjust the model parameters to improve the prediction accuracy of the model; Step S46: predicting groundwater level, setting the prediction time range and step size, outputting the predicted value of future groundwater level, adding confidence interval to the prediction result, and conducting uncertainty assessment; In step S6, the irrigation strategy optimization includes the following steps: Step S61: define the optimization goal as improving water utilization efficiency and design the objective function. The formula used is as follows: ； In the formula, represents the objective function, and is the weight coefficient used to balance the impact of production and water consumption, Indicates the growth status of crops. represents water use efficiency, To prevent the random constant from having a zero denominator; Step S62: preliminarily setting the groundwater level threshold according to actual conditions; Step S63: combining the crop growth status and the groundwater level prediction value, using a particle swarm optimization algorithm to optimize the parameters of the groundwater level prediction model; Step S64: iteratively run the model parameters, calculate the objective function value, and when the objective function value reaches the maximum, output the optimal solution of the optimal model parameters and the groundwater level threshold; Step S65: Simulate using the AquaCrop model under the optimal model parameters to evaluate the actual impact of the optimal model parameters on crop growth.

2. The irrigation optimization method based on groundwater level threshold control according to claim 1 is characterized in that: In step S3, the feature analysis includes the following steps: Step S31: time segmentation, setting periodic time markers; Step S32: extracting environmental periodic features, specifically comprising: extracting periodic components in meteorological data according to periodic time tags, analyzing periodic fluctuations in soil moisture, and extracting seasonal and annual periodic variations in groundwater levels; the environmental periodic features extracted include periodic components in meteorological data, periodic fluctuations in soil moisture, and seasonal and annual periodic variations in groundwater levels; Step S33: trend decomposition, extracting the long-term change trend, periodic fluctuation component and random fluctuation component of the groundwater level; Step S34: introducing hysteresis characteristics, adding hysteresis characteristics of groundwater level, meteorological data, and soil moisture of the previous day, week, and month; Step S35: Cumulative and differential feature extraction, calculating the cumulative value of irrigation history and soil moisture, identifying the phased water resource replenishment or consumption, calculating the first-order difference of the groundwater level, and capturing the incremental change of the groundwater level; Step S36: feature selection and dimensionality reduction, integrating all the features extracted in steps S32 to S35 into time series features, using feature correlation analysis method to remove redundant features in the time series features, evaluating the importance of each feature to groundwater level prediction, setting weight parameters, and using principal component analysis method to reduce the dimension of the time series features; Step S37: normalize the selected and dimension-reduced time series features to generate a feature set containing multi-source information.

3. The irrigation optimization method based on groundwater level threshold control according to claim 2 is characterized in that: In step S33, the trend decomposition includes the following steps: Step S331: performing trend decomposition on groundwater level data and extracting long-term change trends using a moving average method; Step S332: removing the long-term variation trend from the original data of the groundwater level to obtain residual data containing periodic fluctuation components and random fluctuation components; Step S333: further calculating the moving average of the residual data to eliminate the low-frequency components of the periodic fluctuation components; Step S334: Use high-frequency filtering to remove the high-frequency components of the periodic fluctuation components from the residual data, and finally obtain the random fluctuation components.

Images