Training method and device of irrigation water estimation model and irrigation water estimation method

Abstract

The application provides a training method and device of an irrigation water estimation model and an irrigation water estimation method, and belongs to the technical field of remote sensing and agriculture. The training method comprises the following steps: determining a data set based on historical irrigation water data, soil humidity sample data, evapotranspiration sample data and rainfall sample data; training multiple machine learning models based on the data set to obtain trained multiple machine learning models; and integrating the multiple machine learning models by using a Bayesian triangular cap method to determine an irrigation water estimation model. The multiple machine learning models are trained by using the data set, and the multiple models are integrated by using the Bayesian triangular cap method, so that the irrigation water estimation model can be efficiently determined. The irrigation water estimation model can be used to accurately estimate irrigation water, the complex mechanism relationship between the irrigation water and related variables can be avoided, and the model has strong adaptability and migration ability.

Description

Technical Field

[0001] This invention relates to the fields of remote sensing and agricultural technology, and in particular to a training method, apparatus, and method for estimating irrigation water use as a model. Background Technology

[0002] Accurate estimation of irrigation water use can promote sustainable water resource management, safeguard food security, and assess the environmental impact of human irrigation activities. Traditional methods of obtaining irrigation water mainly rely on surveys and statistics from local water resource management departments. However, due to significant differences in irrigation patterns, infrastructure, and crop types across regions, this traditional method is extremely resource-intensive.

[0003] In recent decades, with the development of remote sensing technology, obtaining irrigation-related parameters such as evapotranspiration and rainfall through remote sensing has become a possible method for estimating irrigation water consumption. Some methods in related technologies typically define irrigation as the residual of changes in precipitation, evapotranspiration, and soil water storage, and use water balance formulas or energy balance formulas to calculate irrigation water consumption.

[0004] However, current methods for estimating irrigation water use using remote sensing are primarily based on hydrological mechanisms such as the water balance formula. These methods require human experience to describe the complex and difficult-to-determine physical relationships between irrigation water use and related variables, and are limited by the accuracy and coverage of parameter products. Furthermore, their applicability on a large scale is difficult to guarantee. How to efficiently and accurately estimate irrigation water use using remote sensing is a crucial problem that urgently needs to be solved in the industry. Summary of the Invention

[0005] To address the problems existing in the prior art, embodiments of the present invention provide a training method, apparatus, and method for estimating irrigation water use for an irrigation water estimation model.

[0006] In a first aspect, the present invention provides a training method for an irrigation water estimation model, comprising:

[0007] The dataset was determined based on historical irrigation water data, soil moisture sample data, evapotranspiration sample data, and rainfall sample data.

[0008] Based on the dataset, train multiple machine learning models and obtain the trained machine learning models.

[0009] By integrating the various machine learning models using the Bayesian triangular hat method, an irrigation water estimation model is determined.

[0010] The soil moisture sample data, the evapotranspiration sample data, and the rainfall sample data were obtained through satellite remote sensing.

[0011] Optionally, according to the training method of the irrigation water estimation model provided by the present invention, the historical irrigation water data is used to characterize the total irrigation amount of each region in each year, and the soil moisture sample data includes satellite soil moisture product samples and reanalysis soil moisture product samples.

[0012] The dataset, determined based on historical irrigation water data, soil moisture sample data, evapotranspiration sample data, and rainfall sample data, includes:

[0013] By analyzing the different responses of the satellite soil moisture product samples and the reanalysis soil moisture product samples to irrigation activities, residual terms corresponding to each region in each year are extracted.

[0014] Based on the soil moisture sample data, the evapotranspiration sample data, and the rainfall sample data, the soil moisture explanatory variables, evapotranspiration explanatory variables, and rainfall explanatory variables for each region in each year are determined through cumulative calculation.

[0015] The dataset is determined based on the historical irrigation water data and the soil moisture explanatory variables, residual terms, evapotranspiration explanatory variables, and rainfall explanatory variables for each region in each year.

[0016] Optionally, according to the training method of the irrigation water estimation model provided by the present invention, the satellite soil moisture product samples include satellite soil moisture product samples during non-irrigation periods and satellite soil moisture product samples during irrigation periods;

[0017] The method involves analyzing the different responses of the satellite soil moisture product samples and the reanalysis soil moisture product samples to irrigation activities, and extracting the residual terms corresponding to each region in each year, including:

[0018] Based on the satellite soil moisture product samples during the non-irrigation period and the original reanalysis soil moisture product samples during the non-irrigation period, the systematic error between the satellite soil moisture product and the reanalysis soil moisture product is extracted by the least squares method.

[0019] Based on the system error and the satellite soil moisture product samples during the irrigation period, the reanalysis soil moisture product samples during the irrigation period are corrected, and the corrected reanalysis soil moisture product samples are determined.

[0020] Based on the satellite soil moisture product samples during the irrigation period and the corrected reanalysis soil moisture product samples, the residual terms corresponding to each region in each year are extracted.

[0021] Optionally, according to the training method of the irrigation water estimation model provided by the present invention, the step of determining the soil moisture explanatory variable, evapotranspiration explanatory variable, and rainfall explanatory variable for each region in each year through cumulative calculation based on the soil moisture sample data, the evapotranspiration sample data, and the rainfall sample data includes:

[0022] For each region, based on the corresponding phenological period of each year, the soil moisture sample data, the evapotranspiration sample data, and the rainfall sample data are cumulatively calculated to obtain the explanatory variables of soil moisture, evapotranspiration, and rainfall for the region.

[0023] Optionally, according to the training method of the irrigation water estimation model provided by the present invention, the step of integrating the multiple machine learning models through the Bayesian triangular hat method to determine the irrigation water estimation model includes:

[0024] The irrigation water estimation results output by the various machine learning models are analyzed using the Bayesian triangular hat method to determine the estimation weights corresponding to each machine learning model.

[0025] Based on the various machine learning models and their corresponding estimation weights, model integration is performed to determine the irrigation water estimation model.

[0026] Optionally, in the training method of the irrigation water estimation model provided by the present invention, the various machine learning models include: random forest model, extreme gradient boosting model, support vector machine model, multiple linear regression model and artificial neural network model.

[0027] Secondly, the present invention also provides a method for estimating irrigation water use, comprising:

[0028] Soil moisture, evapotranspiration, and rainfall data are acquired through satellite remote sensing.

[0029] Input the soil moisture data, the evapotranspiration data, and the rainfall data into the irrigation water estimation model, and obtain the irrigation water estimation results output by the irrigation water estimation model;

[0030] The irrigation water estimation model is obtained through the training method of the irrigation water estimation model as described in any of the above.

[0031] Thirdly, the present invention also provides a training apparatus for an irrigation water estimation model, comprising:

[0032] The dataset construction module is used to determine the dataset based on historical irrigation water data, soil moisture sample data, evapotranspiration sample data, and rainfall sample data.

[0033] The model training module is used to train multiple machine learning models based on the dataset and obtain the trained machine learning models.

[0034] The model integration module is used to integrate the various machine learning models using the Bayesian triangular hat method to determine the irrigation water estimation model.

[0035] The soil moisture sample data, the evapotranspiration sample data, and the rainfall sample data were obtained through satellite remote sensing.

[0036] Fourthly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement a training method for an irrigation water estimation model as described above, or the processor executes the program to implement an irrigation water estimation method as described above.

[0037] Fifthly, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a training method for an irrigation water estimation model as described above, or, when executed by a processor, implements an irrigation water estimation method as described above.

[0038] The irrigation water estimation model training method, apparatus, and method provided by this invention determine a dataset based on historical irrigation water data, soil moisture sample data, evapotranspiration sample data, and rainfall sample data. This dataset can be used to train multiple machine learning models, resulting in a variety of trained machine learning models. Furthermore, the Bayesian triangular hat method can be used to integrate these multiple machine learning models, efficiently determining the irrigation water estimation model. This model accurately estimates irrigation water usage, avoiding the need to describe complex mechanisms between irrigation water and related variables. The model, built using machine learning, exhibits high accuracy in its predictions and strong adaptability and transferability. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0040] Figure 1 This is one of the flowcharts illustrating the training method for the irrigation water estimation model provided by the present invention;

[0041] Figure 2 This is a schematic diagram of the process for extracting residual terms provided by the present invention;

[0042] Figure 3 This is the second flowchart illustrating the training method for the irrigation water estimation model provided by this invention.

[0043] Figure 4 This is a schematic diagram of the model prediction results provided by the present invention;

[0044] Figure 5 This is a flowchart illustrating the irrigation water estimation method provided by the present invention;

[0045] Figure 6 This is a schematic diagram of the structure of the training device for the irrigation water estimation model provided by the present invention;

[0046] Figure 7 This is a schematic diagram of the physical structure of the electronic device provided by the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0048] Figure 1 This is one of the flowcharts illustrating the training method for the irrigation water estimation model provided by the present invention, such as... Figure 1 As shown, the training method for the irrigation water estimation model can be performed by electronic devices, such as servers. The training method includes:

[0049] Step 101: Determine the dataset based on historical irrigation water data, soil moisture sample data, evapotranspiration sample data, and rainfall sample data;

[0050] The soil moisture sample data, the evapotranspiration sample data, and the rainfall sample data were obtained through satellite remote sensing.

[0051] Specifically, in order to train various machine learning models, soil moisture sample data, evapotranspiration sample data, and rainfall sample data can be obtained through satellite remote sensing. Then, a dataset can be constructed based on historical irrigation water data, soil moisture sample data, evapotranspiration sample data, and rainfall sample data, with historical irrigation water data serving as labels.

[0052] Step 102: Based on the dataset, train multiple machine learning models to obtain the trained machine learning models.

[0053] Specifically, after constructing the dataset, supervised training can be performed on various machine learning models based on the dataset to obtain various trained machine learning models.

[0054] Understandably, the model's input includes soil moisture sample data, evapotranspiration sample data, and rainfall sample data, while the model's output is irrigation water estimation data, with historical irrigation water data used as training labels.

[0055] For example, supervised training can be performed on random forest models, extreme gradient boosting models, support vector machine models, multiple linear regression models, and artificial neural network models based on datasets to obtain trained random forest models, extreme gradient boosting models, support vector machine models, multiple linear regression models, and artificial neural network models.

[0056] Step 103: Integrate the various machine learning models using the Bayesian triangular hat method to determine the irrigation water estimation model.

[0057] Specifically, integrating multiple machine learning models using the Bayesian triangular hat method can improve the accuracy and reliability of irrigation water estimation models. This method combines the prediction results of different models to obtain more accurate predictions. Simultaneously, the Bayesian triangular hat method can optimize the weights of different models, ensuring that the contribution of each model is fully utilized. Therefore, the irrigation water estimation model derived through this method can better meet practical needs and improve prediction accuracy.

[0058] The training method for the irrigation water estimation model provided by this invention determines a dataset based on historical irrigation water data, soil moisture sample data, evapotranspiration sample data, and rainfall sample data. This dataset can be used to train multiple machine learning models, resulting in a variety of trained machine learning models. Furthermore, the Bayesian triangular hat method can be used to integrate these multiple machine learning models, efficiently determining the irrigation water estimation model. This model accurately estimates irrigation water usage, avoiding the need to describe complex mechanisms between irrigation water and related variables. The model, built using machine learning, exhibits high accuracy in its predictions and strong adaptability and transferability.

[0059] Optionally, according to the training method of the irrigation water estimation model provided by the present invention, the historical irrigation water data is used to characterize the total irrigation amount of each region in each year, and the soil moisture sample data includes satellite soil moisture product samples and reanalysis soil moisture product samples.

[0060] The dataset, determined based on historical irrigation water data, soil moisture sample data, evapotranspiration sample data, and rainfall sample data, includes:

[0061] By analyzing the different responses of the satellite soil moisture product samples and the reanalysis soil moisture product samples to irrigation activities, residual terms corresponding to each region in each year are extracted.

[0062] Based on the soil moisture sample data, the evapotranspiration sample data, and the rainfall sample data, the soil moisture explanatory variables, evapotranspiration explanatory variables, and rainfall explanatory variables for each region in each year are determined through cumulative calculation.

[0063] The dataset is determined based on the historical irrigation water data and the soil moisture explanatory variables, residual terms, evapotranspiration explanatory variables, and rainfall explanatory variables for each region in each year.

[0064] Specifically, considering the inherent systematic errors between the two sets of products (satellite soil moisture products and reanalysis soil moisture product samples), the residual terms corresponding to each region in each year can be extracted by analyzing the different response levels of the satellite soil moisture product samples and reanalysis soil moisture product samples to irrigation activities, so as to correct the systematic errors. This allows the constructed dataset to more accurately represent the soil moisture status of each region in each year, thereby improving the training effect of the model.

[0065] Historical irrigation water data reflects the total annual irrigation volume of a region, while parametric products (such as soil moisture sample data, evapotranspiration sample data, and rainfall sample data) reflect the value of a single pixel. Therefore, it is necessary to perform cumulative calculations on parametric products to determine the explanatory variables for soil moisture, evapotranspiration, and rainfall for each region in each year.

[0066] Optionally, according to the training method of the irrigation water estimation model provided by the present invention, the satellite soil moisture product samples include satellite soil moisture product samples during non-irrigation periods and satellite soil moisture product samples during irrigation periods;

[0067] The method involves analyzing the different responses of the satellite soil moisture product samples and the reanalysis soil moisture product samples to irrigation activities, and extracting the residual terms corresponding to each region in each year, including:

[0068] Based on the satellite soil moisture product samples during the non-irrigation period and the original reanalysis soil moisture product samples during the non-irrigation period, the systematic error between the satellite soil moisture product and the reanalysis soil moisture product is extracted by the least squares method.

[0069] Based on the system error and the satellite soil moisture product samples during the irrigation period, the reanalysis soil moisture product samples during the irrigation period are corrected, and the corrected reanalysis soil moisture product samples are determined.

[0070] Based on the satellite soil moisture product samples during the irrigation period and the corrected reanalysis soil moisture product samples, the residual terms corresponding to each region in each year are extracted.

[0071] Specifically, Figure 2 This is a schematic diagram of the process for extracting residual terms provided by the present invention, as shown below. Figure 2 As shown, the process for extracting residual terms may include steps 201 to 204.

[0072] Step 201: The original soil moisture value is converted into its logarithmic value by utilizing the nonlinear relationship between satellite soil moisture data and irrigation water.

[0073] Optionally, satellite soil moisture product samples can be converted to their logarithms, and reanalyzed soil moisture product samples can be converted to their logarithms.

[0074] Step 202: Based on the satellite soil moisture product samples during the non-irrigation period and the original reanalysis soil moisture product samples during the non-irrigation period, the systematic error between the satellite soil moisture product and the reanalysis soil moisture product is extracted using the least squares method.

[0075] Specifically, the systematic error between satellite soil moisture products and reanalysis soil moisture products can be extracted using the following systematic error extraction formula:

[0076]

[0077] Where Bias and y represent the slopes of the deviation, which can be obtained using the least squares method. This indicates satellite soil moisture products for non-irrigation periods. This indicates the original reanalysis soil moisture product during the non-irrigation period.

[0078] Understandably, systematic errors can be characterized by Bias and y.

[0079] Step 203: Based on system errors and satellite soil moisture product samples during the irrigation period, correct the reanalysis soil moisture product samples during the irrigation period, and determine the corrected reanalysis soil moisture product samples.

[0080] Specifically, the corrected soil moisture product sample after reanalysis can be determined using the following systematic error correction formula:

[0081]

[0082] Where Bias and y represent the slopes of the deviation, respectively. Indicates satellite soil moisture product during the irrigation period. This indicates the corrected reanalysis soil moisture product.

[0083] Step 204: Based on the satellite soil moisture product samples during the irrigation period and the corrected reanalysis soil moisture product samples, extract the residual terms corresponding to each region in each year.

[0084] Specifically, the residual terms for each region in each year can be extracted using the following residual term calculation formula:

[0085]

[0086] in, Indicates satellite soil moisture product during the irrigation period. Represents the corrected reanalysis soil moisture product, where N represents the number of days in a year during the irrigation period (i.e., the dry season), and ResSM represents the residual term.

[0087] It can be understood that the correction is due to the inherent systematic errors between the two sets of products (satellite soil moisture products and reanalysis soil moisture product samples). By correcting the two during the non-irrigation period (i.e., the rainy day sequence), and then using the fitting coefficients (i.e., Bias and y) to correct the reanalysis product during the irrigation period, the inherent systematic errors between the two sets of products can be corrected. By correcting the systematic errors, the constructed dataset can more accurately represent the soil moisture status of each region in each year, thereby improving the training effect of the model and enabling the trained model to more accurately estimate irrigation water.

[0088] Optionally, according to the training method of the irrigation water estimation model provided by the present invention, the step of determining the soil moisture explanatory variable, evapotranspiration explanatory variable, and rainfall explanatory variable for each region in each year through cumulative calculation based on the soil moisture sample data, the evapotranspiration sample data, and the rainfall sample data includes:

[0089] For each region, based on the corresponding phenological period of each year, the soil moisture sample data, the evapotranspiration sample data, and the rainfall sample data are cumulatively calculated to obtain the explanatory variables of soil moisture, evapotranspiration, and rainfall for the region.

[0090] Understandably, historical irrigation water data reflects the total annual irrigation volume of a region, while parametric products (such as soil moisture sample data, evapotranspiration sample data, and rainfall sample data) reflect the value of a specific pixel. Therefore, it is necessary to accumulate parametric products spatially. Furthermore, it is generally believed that parameter changes within phenological periods have an impact on irrigation; therefore, parameters must also be accumulated temporally (within phenological periods). By accumulating parametric products based on the corresponding phenological periods for each year, the resulting explanatory variables can better correspond to historical irrigation water data. Furthermore, based on historical irrigation water data and the corresponding soil moisture explanatory variables, residuals, evapotranspiration explanatory variables, and rainfall explanatory variables for each region in each year, the model can be trained more effectively, improving its training performance.

[0091] Optionally, according to the training method of the irrigation water estimation model provided by the present invention, the step of integrating the multiple machine learning models through the Bayesian triangular hat method to determine the irrigation water estimation model includes:

[0092] The irrigation water estimation results output by the various machine learning models are analyzed using the Bayesian triangular hat method to determine the estimation weights corresponding to each machine learning model.

[0093] Based on the various machine learning models and their corresponding estimation weights, model integration is performed to determine the irrigation water estimation model.

[0094] Specifically, the principle behind the above-mentioned integration of multiple machine learning models using the Bayesian triangular hat method is as follows:

[0095] For the prediction result of the i-th machine learning model, its probability density function (PDF) can be expressed as:

[0096]

[0097] ε i =IWU i -IWU t ;

[0098] Among them, IWU t This represents the actual value of irrigation water use (IWU) in year t (this actual value can be obtained from historical irrigation water use data). i ε represents the prediction result of the i-th machine learning model. i It is white noise with a mean of 0, σ i It's IWU i The error variance is given by L, which is the likelihood function.

[0099] For the predicted values ​​given by N machine learning models, IWU t The likelihood function can be expressed as:

[0100]

[0101] To calculate IWU t The maximum likelihood value, the loss function can be defined as:

[0102]

[0103] By making J(IWU) t ) is 0, IWU t It can be represented as:

[0104]

[0105]

[0106] Among them, w i Representing IWU i The weights of each IWU predicted value. i corresponding It can be obtained using the classic triangular hat method.

[0107] Optionally, in the training method of the irrigation water estimation model provided by the present invention, the various machine learning models include: random forest model, extreme gradient boosting model, support vector machine model, multiple linear regression model and artificial neural network model.

[0108] Specifically, supervised training can be performed on random forest models, extreme gradient boosting models, support vector machine models, multiple linear regression models, and artificial neural network models based on datasets to obtain trained random forest models, extreme gradient boosting models, support vector machine models, multiple linear regression models, and artificial neural network models. Then, the Bayesian triangular hat method can be used to integrate the random forest models, extreme gradient boosting models, support vector machine models, multiple linear regression models, and artificial neural network models to determine the irrigation water estimation model.

[0109] Optionally, Figure 3 This is the second flowchart illustrating the training method for the irrigation water estimation model provided by this invention. Figure 3 As shown, the training method for the irrigation water estimation model includes steps 301 to 305.

[0110] Step 301: Obtain historical irrigation water data, soil moisture sample data, evapotranspiration sample data, and rainfall sample data.

[0111] Optionally, the example area can be China, and the historical irrigation water data includes approximately 341 administrative cities and three crop types: rice, wheat, and maize. The historical irrigation water data, soil moisture sample data, evapotranspiration sample data, and rainfall sample data span from 2003 to 2013.

[0112] Step 302: By analyzing the different responses of satellite soil moisture product samples and reanalyzing soil moisture product samples to irrigation activities, the residual terms corresponding to each region in each year are extracted.

[0113] Step 303: Use phenological data to extract and accumulate relevant variables to determine the explanatory variables for soil moisture, evapotranspiration, and rainfall in each region for each year.

[0114] Since historical irrigation water data reflects the total annual irrigation volume within an administrative city, while parametric products (such as soil moisture sample data, evapotranspiration sample data, and rainfall sample data) reflect the value of a single pixel, it is necessary to accumulate parametric products spatially. Furthermore, it is generally believed that parameter changes within phenological periods have an impact on irrigation; therefore, it is also necessary to accumulate parameters temporally (within phenological periods).

[0115] Step 304: Based on historical irrigation water data and the soil moisture explanatory variables, residual terms, evapotranspiration explanatory variables, and rainfall explanatory variables corresponding to each region in each year, train multiple machine learning models.

[0116] Alternatively, various machine learning models can be used, including random forest models, extreme gradient boosting models, support vector machine models, multiple linear regression models, and artificial neural network models.

[0117] Alternatively, data from 2003 to 2012 can be used as the training set, and data from 2013 can be used as the test set.

[0118] Step 305: Integrate the predictions from various machine learning models using the Bayesian triangular hat method.

[0119] Specifically, the prediction results based on multiple machine learning models are processed using the Bayesian triangular hat method. This method assigns different weights to each model according to the sequential relationship between the multiple machine learning models, thereby performing a weighted average of the prediction results.

[0120] Optionally, Figure 4 This is a schematic diagram of the model prediction results provided by the present invention. Figure 4The results of irrigation water prediction for rice are shown, including predictions from random forest, extreme gradient boosting, support vector machine, multiple linear regression, artificial neural network, and the ensemble result obtained using the Bayesian triangular hat method. Here, RMSE represents the root mean square error, MAE represents the mean absolute error, and R0... 2 This represents the coefficient of determination.

[0121] Figure 4 The results also show the predictions for irrigation water for wheat, including predictions from random forest models, extreme gradient boosting models, support vector machine models, multiple linear regression models, artificial neural network models, and predictions after integration using the Bayesian triangular hat method.

[0122] Figure 4 The results also show the predictions for irrigation water for maize, including predictions from random forest models, extreme gradient boosting models, support vector machine models, multiple linear regression models, artificial neural network models, and predictions after integration using the Bayesian triangular hat method.

[0123] pass Figure 4 The prediction results show that, compared with using a single model to predict irrigation water, this application integrates multiple machine learning models through the Bayesian triangular hat method, which can improve the estimation performance of the model.

[0124] Figure 5 This is a flowchart illustrating the irrigation water estimation method provided by the present invention, as shown below. Figure 5 As shown, the entity executing the irrigation water estimation method can be an electronic device, such as a server. The irrigation water estimation method includes:

[0125] Step 501: Obtain soil moisture data, evapotranspiration data, and rainfall data through satellite remote sensing;

[0126] Step 502: Input the soil moisture data, the evapotranspiration data and the rainfall data into the irrigation water estimation model, and obtain the irrigation water estimation result output by the irrigation water estimation model;

[0127] The irrigation water estimation model is obtained through any of the above-mentioned training methods for the irrigation water estimation model.

[0128] Understandably, irrigation water estimation models can accurately estimate irrigation water usage, avoiding the need to describe the complex mechanisms between irrigation water usage and related variables. By using machine learning models for modeling, the prediction results have high accuracy and strong adaptability and transferability.

[0129] The training apparatus for the irrigation water estimation model provided by the present invention will be described below. The training apparatus for the irrigation water estimation model described below and the training method for the irrigation water estimation model described above can be referred to in correspondence.

[0130] Figure 6 This is a schematic diagram of the training device for the irrigation water estimation model provided by the present invention, as shown below. Figure 6 As shown, the training device includes: a dataset construction module 601, a model training module 602, and a model ensemble module 603, wherein:

[0131] Dataset construction module 601 is used to determine the dataset based on historical irrigation water data, soil moisture sample data, evapotranspiration sample data, and rainfall sample data;

[0132] The model training module 602 is used to train multiple machine learning models based on the dataset and obtain the trained multiple machine learning models.

[0133] The model integration module 603 is used to integrate the multiple machine learning models using the Bayesian triangular hat method to determine the irrigation water estimation model.

[0134] The soil moisture sample data, the evapotranspiration sample data, and the rainfall sample data were obtained through satellite remote sensing.

[0135] Figure 7 This is a schematic diagram of the physical structure of the electronic device provided by the present invention, such as... Figure 7 As shown, the electronic device may include: a processor 710, a communication interface 720, a memory 730, and a communication bus 740, wherein the processor 710, the communication interface 720, and the memory 730 communicate with each other via the communication bus 740. The processor 710 can call logical instructions in the memory 730 to execute a training method for an irrigation water estimation model, the method including:

[0136] The dataset was determined based on historical irrigation water data, soil moisture sample data, evapotranspiration sample data, and rainfall sample data.

[0137] Based on the dataset, train multiple machine learning models and obtain the trained machine learning models.

[0138] By integrating the various machine learning models using the Bayesian triangular hat method, an irrigation water estimation model is determined.

[0139] The soil moisture sample data, the evapotranspiration sample data, and the rainfall sample data were obtained through satellite remote sensing.

[0140] Furthermore, the logical instructions in the aforementioned memory 730 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0141] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a training method for the irrigation water estimation model provided by the methods described above, the method comprising:

[0142] The dataset was determined based on historical irrigation water data, soil moisture sample data, evapotranspiration sample data, and rainfall sample data.

[0143] Based on the dataset, train multiple machine learning models and obtain the trained machine learning models.

[0144] By integrating the various machine learning models using the Bayesian triangular hat method, an irrigation water estimation model is determined.

[0145] The soil moisture sample data, the evapotranspiration sample data, and the rainfall sample data were obtained through satellite remote sensing.

[0146] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0147] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0148] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A training method for an irrigation water estimation model, characterized in that, include: The dataset was determined based on historical irrigation water data, soil moisture sample data, evapotranspiration sample data, and rainfall sample data. Based on the dataset, train multiple machine learning models and obtain the trained machine learning models. By integrating the various machine learning models using the Bayesian triangular hat method, an irrigation water estimation model is determined. The soil moisture sample data, the evapotranspiration sample data, and the rainfall sample data were obtained through satellite remote sensing. The historical irrigation water data is used to characterize the total irrigation amount in each region in each year, and the soil moisture sample data includes satellite soil moisture product samples and reanalysis soil moisture product samples. The dataset, determined based on historical irrigation water data, soil moisture sample data, evapotranspiration sample data, and rainfall sample data, includes: By analyzing the different responses of the satellite soil moisture product samples and the reanalysis soil moisture product samples to irrigation activities, residual terms corresponding to each region in each year are extracted. Based on the soil moisture sample data, the evapotranspiration sample data, and the rainfall sample data, the soil moisture explanatory variables, evapotranspiration explanatory variables, and rainfall explanatory variables for each region in each year are determined through cumulative calculation. The dataset is determined based on the historical irrigation water data and the soil moisture explanatory variables, residual terms, evapotranspiration explanatory variables, and rainfall explanatory variables for each region in each year; The satellite soil moisture product samples include satellite soil moisture product samples during non-irrigation periods and satellite soil moisture product samples during irrigation periods; The method involves analyzing the different responses of the satellite soil moisture product samples and the reanalysis soil moisture product samples to irrigation activities, and extracting the residual terms corresponding to each region in each year, including: Based on the satellite soil moisture product samples during the non-irrigation period and the original reanalysis soil moisture product samples during the non-irrigation period, the systematic error between the satellite soil moisture product and the reanalysis soil moisture product is extracted by the least squares method. Based on the system error and the satellite soil moisture product samples during the irrigation period, the reanalysis soil moisture product samples during the irrigation period are corrected, and the corrected reanalysis soil moisture product samples are determined. Based on the satellite soil moisture product samples during the irrigation period and the corrected reanalysis soil moisture product samples, the residual terms corresponding to each region in each year are extracted.

2. The training method for the irrigation water estimation model according to claim 1, characterized in that, The process of determining the explanatory variables for soil moisture, evapotranspiration, and rainfall for each region in each year through cumulative calculation, based on the soil moisture sample data, the evapotranspiration sample data, and the rainfall sample data, includes: For each region, based on the corresponding phenological period of each year, the soil moisture sample data, the evapotranspiration sample data, and the rainfall sample data are cumulatively calculated to obtain the explanatory variables of soil moisture, evapotranspiration, and rainfall for the region.

3. The training method for the irrigation water estimation model according to claim 1, characterized in that, The method of integrating multiple machine learning models using the Bayesian triangular hat approach to determine the irrigation water estimation model includes: The irrigation water estimation results output by the various machine learning models are analyzed using the Bayesian triangular hat method to determine the estimation weights corresponding to each machine learning model. Based on the various machine learning models and their corresponding estimation weights, model integration is performed to determine the irrigation water estimation model.

4. The training method for the irrigation water estimation model according to any one of claims 1-3, characterized in that, The various machine learning models include: random forest model, extreme gradient boosting model, support vector machine model, multiple linear regression model, and artificial neural network model.

5. A method for estimating irrigation water use, characterized in that, include: Soil moisture, evapotranspiration, and rainfall data are acquired through satellite remote sensing. Input the soil moisture data, the evapotranspiration data, and the rainfall data into the irrigation water estimation model, and obtain the irrigation water estimation results output by the irrigation water estimation model; The irrigation water estimation model is obtained by the training method of the irrigation water estimation model as described in any one of claims 1-4.

6. A training device for an irrigation water estimation model, characterized in that, include: The dataset construction module is used to determine the dataset based on historical irrigation water data, soil moisture sample data, evapotranspiration sample data, and rainfall sample data. The model training module is used to train multiple machine learning models based on the dataset and obtain the trained machine learning models. The model integration module is used to integrate the various machine learning models using the Bayesian triangular hat method to determine the irrigation water estimation model. The soil moisture sample data, the evapotranspiration sample data, and the rainfall sample data were obtained through satellite remote sensing. The historical irrigation water data is used to characterize the total irrigation amount in each region in each year, and the soil moisture sample data includes satellite soil moisture product samples and reanalysis soil moisture product samples. The dataset, determined based on historical irrigation water data, soil moisture sample data, evapotranspiration sample data, and rainfall sample data, includes: By analyzing the different responses of the satellite soil moisture product samples and the reanalysis soil moisture product samples to irrigation activities, residual terms corresponding to each region in each year are extracted. Based on the soil moisture sample data, the evapotranspiration sample data, and the rainfall sample data, the soil moisture explanatory variables, evapotranspiration explanatory variables, and rainfall explanatory variables for each region in each year are determined through cumulative calculation. The dataset is determined based on the historical irrigation water data and the soil moisture explanatory variables, residual terms, evapotranspiration explanatory variables, and rainfall explanatory variables for each region in each year; The satellite soil moisture product samples include satellite soil moisture product samples during non-irrigation periods and satellite soil moisture product samples during irrigation periods; The method involves analyzing the different responses of the satellite soil moisture product samples and the reanalysis soil moisture product samples to irrigation activities, and extracting the residual terms corresponding to each region in each year, including: Based on the satellite soil moisture product samples during the non-irrigation period and the original reanalysis soil moisture product samples during the non-irrigation period, the systematic error between the satellite soil moisture product and the reanalysis soil moisture product is extracted by the least squares method. Based on the system error and the satellite soil moisture product samples during the irrigation period, the reanalysis soil moisture product samples during the irrigation period are corrected, and the corrected reanalysis soil moisture product samples are determined. Based on the satellite soil moisture product samples during the irrigation period and the corrected reanalysis soil moisture product samples, the residual terms corresponding to each region in each year are extracted.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the training method of the irrigation water estimation model as described in any one of claims 1 to 4, or when the processor executes the program, it implements the irrigation water estimation method as described in claim 5.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the training method of the irrigation water estimation model as described in any one of claims 1 to 4, or when the computer program is executed by the processor, it implements the irrigation water estimation method as described in claim 5.

Images