Rice irrigation and drainage determination method, device, equipment, medium and product

Abstract

The invention discloses a rice irrigation and drainage determination method and device, equipment, a medium and a product, and relates to the field of agricultural production management.The method comprises the steps that prediction is conducted through a greenhouse gas emission daily flux prediction network and a total yield prediction network according to actual meteorological parameters, actual soil parameters and actual agricultural management parameters; the actual greenhouse gas emission daily flux and the actual predicted total yield are obtained; the greenhouse gas emission daily flux prediction network and the total yield prediction network are both obtained by training a deep neural network by using a deep learning training data set; the deep learning training data set is generated through a denitrification-decomposition model; and by using a multi-objective genetic algorithm, optimizing the irrigation and drainage scheme by taking the actual greenhouse gas emission daily flux and the actual predicted total yield as optimization objectives to obtain an optimal irrigation and drainage scheme. The greenhouse gas emission is reduced to the maximum extent while the yield is guaranteed.

Description

Technical Field

[0001] The present application relates to the field of agricultural production management, and particularly to a method, device, equipment, medium and product for determining rice irrigation and drainage. Background Art

[0002] Rice cultivation is an important part of food production, and its irrigation method is crucial for yield and environmental impact. Traditional irrigation strategies, such as continuous flooding, although ensuring the growth requirements of rice to a certain extent, also lead to a large amount of greenhouse gas (such as methane and nitrous oxide) emissions. Most existing irrigation methods rely on farmers' experience or simple mathematical models, and it is difficult to comprehensively consider the complex interactions of multi-dimensional factors such as soil, meteorology, and agronomy. For example, empirical irrigation strategies often perform poorly in the face of extreme weather events (such as droughts and floods), and are unable to adjust irrigation and drainage volumes in a timely manner, resulting in reduced yields or increased greenhouse gas emissions. Therefore, there is an urgent need for an irrigation and drainage method that can minimize greenhouse gas emissions while ensuring yields. Summary of the Invention

[0003] The purpose of the present application is to provide a method, device, equipment, medium and product for determining rice irrigation and drainage, which can minimize greenhouse gas emissions while ensuring yields.

[0004] To achieve the above object, the present application provides the following solutions: In a first aspect, the present application provides a method for determining rice irrigation and drainage, including: Obtaining actual meteorological parameters, actual soil parameters of the paddy field, and actual agronomic management parameters of the paddy field; Predicting according to the actual meteorological parameters, the actual soil parameters, and the actual agronomic management parameters by using a greenhouse gas daily emission flux prediction network and a total yield prediction network to obtain an actual greenhouse gas daily emission flux and an actual predicted total yield; both the greenhouse gas daily emission flux prediction network and the total yield prediction network are obtained by training a deep neural network using a deep learning training dataset; the deep learning training dataset is generated by an anaerobic digestion-decomposition model; Using a multi-objective genetic algorithm, optimizing the irrigation and drainage plan with the actual greenhouse gas daily emission flux and the actual predicted total yield as optimization objectives to obtain an optimal irrigation and drainage plan.

[0005] In an embodiment, the generation process of the deep learning training dataset specifically includes: Obtaining historical meteorological parameters, historical soil parameters, historical agronomic management parameters, historical greenhouse gas daily emission flux, and historical total yield of the paddy field; Adjust the parameters of the denitrification-decomposition model according to historical meteorological parameters, historical soil parameters, historical agronomic management parameters, historical daily fluxes of greenhouse gas emissions, and historical total yields to obtain a denitrification-decomposition model for simulating daily fluxes of greenhouse gas emissions and total yields. Use different training meteorological parameters, training soil parameters, and training agronomic management parameters to simulate the training daily fluxes of greenhouse gas emissions and training total yields using the denitrification-decomposition model for simulating daily fluxes of greenhouse gas emissions and total yields. Determine a deep learning training set based on the training meteorological parameters, training soil parameters, training agronomic management parameters, training daily fluxes of greenhouse gas emissions, and training total yields.

[0006] In one embodiment, the training process of the daily flux prediction network for greenhouse gas emissions specifically includes: Normalize the training meteorological parameters, training soil parameters, training agronomic management parameters, and training daily fluxes of greenhouse gas emissions in the deep learning training set; the training agronomic management parameters include fertilizer management, tillage management, residue disposal methods, and irrigation and drainage schemes. Use the normalized training soil parameters, fertilizer management, tillage management, and residue disposal methods as the static inputs of the deep neural network, use the training meteorological parameters and irrigation and drainage schemes as the dynamic inputs of the deep neural network, use the normalized training daily fluxes of greenhouse gas emissions as the output of the deep neural network, use the mean squared error as the loss function, and optimize the hyperparameters of the deep neural network using the Bayesian optimization framework to obtain a daily flux prediction network for greenhouse gas emissions.

[0007] In one embodiment, the training process of the total yield prediction network specifically includes: Normalize the training meteorological parameters, training soil parameters, training agronomic management parameters, and training total yields in the deep learning training set; the training agronomic management parameters include fertilizer management, tillage management, residue disposal methods, and irrigation and drainage schemes. Use the normalized training soil parameters, fertilizer management, tillage management, and residue disposal methods as the static inputs of the deep neural network, use the training meteorological parameters and irrigation and drainage schemes as the dynamic inputs of the deep neural network, use the normalized training total yields as the output of the deep neural network, use the mean squared error as the loss function, and optimize the hyperparameters of the deep neural network using the Bayesian optimization framework to obtain a total yield prediction network.

[0008] In one embodiment, the deep neural network adopted by the greenhouse gas daily emission flux prediction network is a long short-term memory neural network; the deep neural network adopted by the total output prediction network is an improved long short-term memory neural network; the improved long short-term memory neural network includes a long short-term memory neural network and a fully connected neural network layer connected to the output layer of the long short-term memory neural network.

[0009] In one embodiment, using the multi-objective genetic algorithm, the actual greenhouse gas daily emission flux and the actual predicted total output are used as optimization objectives to optimize the irrigation and drainage plan, and the optimal irrigation and drainage plan is obtained, which specifically includes: Using the multi-objective genetic algorithm R-NSGA-II to optimize the Pareto front curve, normalizing and weighting the actual greenhouse gas daily emission flux and the actual predicted total output to obtain the irrigation and drainage plan score; the Pareto front curve includes the actual greenhouse gas daily emission flux and the actual predicted total output; Select the irrigation and drainage plan with the highest irrigation and drainage plan score as the optimal irrigation and drainage plan.

[0010] In a second aspect, the present application provides a rice irrigation and drainage determination device, including: An acquisition module, configured to acquire actual meteorological parameters, actual soil parameters of the paddy field, and actual agronomic management parameters of the paddy field; A prediction module, configured to perform predictions using the greenhouse gas daily emission flux prediction network and the total output prediction network according to the actual meteorological parameters, the actual soil parameters, and the actual agronomic management parameters to obtain the actual greenhouse gas daily emission flux and the actual predicted total output; both the greenhouse gas daily emission flux prediction network and the total output prediction network are obtained by training a deep neural network using a deep learning training dataset; the deep learning training dataset is generated by a denitrification-decomposition model; An optimization module, configured to use the multi-objective genetic algorithm to optimize the irrigation and drainage plan with the actual greenhouse gas daily emission flux and the actual predicted total output as optimization objectives to obtain the optimal irrigation and drainage plan.

[0011] In a third aspect, the present application provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the computer program to implement the rice irrigation and drainage determination method described above.

[0012] In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the rice irrigation and drainage determination method described above is implemented.

[0013] Fifth aspect, the present application provides a computer program product, including a computer program which, when executed by a processor, implements the rice irrigation and drainage determination method described above.

[0014] According to the specific embodiments provided by the present application, the following technical effects are disclosed: The present application provides a rice irrigation and drainage determination method, device, equipment, medium and product. When predicting the actual daily greenhouse gas emission flux and the actual predicted total output, a deep neural network trained by a deep learning training dataset is used, where the deep learning training dataset is generated by a denitrification-decomposition model. Combining the mechanism model and the deep neural network not only retains the physical interpretability of the mechanism model but also improves the prediction accuracy of the deep neural network. In addition, a genetic algorithm is used to optimize the irrigation and drainage plan, which can ensure that the irrigation and drainage plan adapts to environmental changes, so as to reduce greenhouse gas emissions while ensuring rice yield. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0016] Figure 1 It is an application environment diagram of a rice irrigation and drainage determination method in an embodiment of the present application; Figure 2 It is a schematic flowchart of a rice irrigation and drainage determination method provided by an embodiment of the present application; Figure 3 It is a schematic diagram of the rice irrigation and drainage determination method; Figure 4 It is a schematic diagram of the daily greenhouse gas emission flux prediction network; Figure 5 It is a schematic diagram of the total output prediction network; Figure 6 It is a schematic diagram of optimizing the irrigation and drainage plan; Figure 7 It is a schematic diagram of the functional modules of a rice irrigation and drainage determination device provided by another embodiment of the present application; Figure 8 It is a schematic diagram of the structure of a computer device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0018] With the development of artificial intelligence technology and the improvement of farmland environment datasets (such as meteorology, soil, crop growth, etc.), data-driven methods such as machine learning and deep learning have become effective methods for accurately predicting rice greenhouse gas emissions and yields. It can more accurately simulate the complex behaviors of farmland ecosystems, thus providing a scientific basis for optimizing irrigation strategies. At the same time, combining mechanism models (such as DNDC) with data-driven models can generate a large amount of interpretable virtual training data for data-driven models, thereby improving their simulation accuracy. In addition, the application of multi-objective optimization algorithms (such as genetic algorithms, particle swarm algorithms, etc.) makes it possible to coordinate the conflict between yield and greenhouse gas emissions. By generating Pareto optimal solution sets, multiple optional solutions are provided for decision-makers, enabling them to screen solutions according to actual needs. The combination of real-time data collection and online learning technology further promotes the development of dynamic adjustment of irrigation strategies, enabling them to adapt to environmental changes (such as sudden rainfall, drought), so as to achieve precise irrigation in complex and changeable farmland environments.

[0019] The present application aims to optimize the irrigation strategy in the rice planting process through a data-driven approach, reduce greenhouse gas emissions, and ensure rice yields at the same time. The core lies in hybrid modeling, dynamic optimization, and multi-objective trade-off. First, by combining the denitrification-decomposition model (DNDC) in the mechanism model with a data-driven model (an LSTM-based neural network), the physical interpretability of the mechanism model is retained, and the prediction accuracy of the model is improved. Second, the genetic algorithm (R-NSGA-II) is used to optimize the irrigation strategy in real time, and the drainage and irrigation plan is dynamically adjusted according to the daily updated meteorological and soil data to ensure that the irrigation strategy can adapt to environmental changes. Finally, the conflict between yield and greenhouse gas emissions is coordinated through a multi-objective optimization algorithm to generate a Pareto optimal solution set and provide the optimal daily drainage and irrigation strategy under trade-off. Through the above innovations, the present application provides a scientific and intelligent decision-making method for low-carbon drainage and irrigation of rice, which has certain theoretical significance and important practical application value.

[0020] To make the above objects, features, and advantages of the present application more obvious and understandable, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0021] The rice irrigation and drainage determination method provided by the embodiments of this application can be applied to, for example, Figure 1 the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, integrated on the server 104, or placed on the cloud or other servers. The terminal 102 can send the actual meteorological parameters to be processed, the actual soil parameters of the paddy field, and the actual agronomic management parameters of the paddy field to the server 104. After receiving the actual meteorological parameters to be processed, the actual soil parameters of the paddy field, and the actual agronomic management parameters of the paddy field, for the actual meteorological parameters to be processed, the actual soil parameters of the paddy field, and the actual agronomic management parameters of the paddy field, the server 104 uses the greenhouse gas emission daily flux prediction network and the total output prediction network to predict according to the actual meteorological parameters, the actual soil parameters, and the actual agronomic management parameters, and obtains the actual greenhouse gas emission daily flux and the actual predicted total output; uses the multi-objective genetic algorithm, and uses the actual greenhouse gas emission daily flux and the actual predicted total output as the optimization objectives to optimize the irrigation and drainage plan, and obtains the optimal irrigation and drainage plan. The server 104 can feedback the obtained optimal irrigation and drainage plan to the terminal 102. In addition, in some embodiments, the rice irrigation and drainage determination method can also be implemented by the server 104 or the terminal 102 alone. For example, the terminal 102 can directly determine the irrigation and drainage for the actual meteorological parameters to be processed, the actual soil parameters of the paddy field, and the actual agronomic management parameters of the paddy field, or the server 104 can obtain the actual meteorological parameters to be processed, the actual soil parameters of the paddy field, and the actual agronomic management parameters of the paddy field from the data storage system, and determine the irrigation and drainage for the actual meteorological parameters to be processed, the actual soil parameters of the paddy field, and the actual agronomic management parameters of the paddy field.

[0022] Among them, the terminal 102 can be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, Internet of Things devices, and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. The portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented by an independent server or a server cluster composed of multiple servers, and can also be a cloud server.

[0023] In an exemplary embodiment, as Figure 2 shown, a rice irrigation and drainage determination method is provided. This method is executed by a computer device, and can specifically be executed alone by a computer device such as a terminal or a server, or jointly executed by a terminal and a server. In the embodiments of this application, taking this method applied to Figure 1 the server 104 therein as an example for illustration, it includes the following steps 201 to step 203. Among them: Step 201: Obtain the actual meteorological parameters, the actual soil parameters of the paddy field, and the actual agronomic management parameters of the paddy field.

[0024] Step 202: According to the actual meteorological parameters, the actual soil parameters, and the actual agronomic management parameters, use the greenhouse gas daily emission flux prediction network and the total yield prediction network to make predictions, and obtain the actual greenhouse gas daily emission flux and the actual predicted total yield; both the greenhouse gas daily emission flux prediction network and the total yield prediction network are obtained by training a deep neural network using a deep learning training dataset; the deep learning training dataset is generated by a denitrification-decomposition (DNDC) model.

[0025] Step 203: Use a multi-objective genetic algorithm to optimize the irrigation and drainage scheme with the actual greenhouse gas daily emission flux and the actual predicted total yield as the optimization objectives, and obtain the optimal irrigation and drainage scheme.

[0026] This application optimizes the irrigation strategy in the rice planting process in a data-driven manner, reduces greenhouse gas emissions, and at the same time ensures the rice yield. Specifically, when predicting the actual greenhouse gas daily emission flux and the actual predicted total yield, a deep neural network trained using a deep learning training dataset is used, where the deep learning training dataset is generated by a denitrification-decomposition model. By combining the mechanism model and the deep neural network, the physical interpretability of the mechanism model is retained, and the prediction accuracy of the deep neural network is improved. In addition, using a genetic algorithm to optimize the irrigation and drainage scheme can ensure that the irrigation and drainage scheme adapts to environmental changes, so as to achieve the reduction of greenhouse gas emissions while ensuring the rice yield.

[0027] In an exemplary embodiment, the generation process of the deep learning training dataset specifically includes: Obtain the historical meteorological parameters, historical soil parameters, historical agronomic management parameters, historical greenhouse gas daily emission flux, and historical total yield of the paddy field.

[0028] Adjust the parameters of the denitrification-decomposition model according to the historical meteorological parameters, historical soil parameters, historical agronomic management parameters, historical greenhouse gas daily emission flux, and historical total yield to obtain a denitrification-decomposition model that simulates the greenhouse gas daily emission flux and the total yield.

[0029] Use different training meteorological parameters, training soil parameters, and training agronomic management parameters to simulate and train the greenhouse gas daily emission flux and the training total yield using the denitrification-decomposition model that simulates the greenhouse gas daily emission flux and the total yield.

[0030] Determine the deep learning training set based on the training meteorological parameters, training soil parameters, training agronomic management parameters, training greenhouse gas daily emission flux, and training total yield.

[0031] As Figure 3 shown, specifically, establish a dataset of greenhouse gases and yields for rice, and adjust the parameters of the DNDC model based on the established dataset to enable it to accurately simulate the daily emission flux of greenhouse gases and the total yield under different meteorological, soil, and agronomic management parameters.

[0032] Specifically, it includes: extracting historical meteorological parameters, historical soil parameters, historical agronomic management parameters, and their corresponding daily emission fluxes of greenhouse gases and final total yields of paddy fields from the published literature on various literature websites (such as Web of Science, CNKI, etc.). Further, the historical meteorological parameters include the daily maximum temperature, minimum temperature, precipitation, average wind speed, evaporation, and relative humidity during the growth period of the paddy field; the historical soil parameters include the soil texture (proportion of sand particles, proportion of silt particles, proportion of clay particles), soil organic matter content, total soil nitrogen content, soil pH value, and total soil nitrogen content at the location of the paddy field; the historical agronomic management parameters include: fertilizer management (usage amount, usage time, usage frequency, and usage method of inorganic nitrogen, phosphorus, potassium fertilizers, and organic fertilizers), tillage management (tillage method, tillage frequency), residue disposal method (amount of straw returned to the field, method of returning to the field), and irrigation and drainage plan (irrigation amount and irrigation time during the growth period of rice). Substitute the above input parameters into the denitrification-decomposition model and fine-tune the meteorological parameters and soil parameters; through the coefficient of determination R 2 , mean square error MSE, and mean absolute error MAE indicators to evaluate the simulation accuracy of the total yield and daily emission flux of greenhouse gases simulated by the DNDC model. In this application, the adjusted denitrification-decomposition model is used as the data generator Coefficient of determination R 2 , mean square error MSE and mean absolute error MAE The indicators are evaluated as follows: .

[0033] .

[0034] .

[0035] Among them, is the regression sum of squares; is the total sum of squares; represents the true value of the daily emission flux of greenhouse gases or the total yield; represents the simulated value of the DNDC for the daily emission flux of greenhouse gases or the total yield; nis the number of data entries in the dataset; i is the numbering of the data starting from 1. The above R 2 The closer it is to 1, the mean square error MSE and the mean absolute error MAE The closer it is to 0, the better the simulation effect of the DNDC model.

[0036] In an exemplary embodiment, the training process of the greenhouse gas daily emission flux prediction network specifically includes: Normalize the training meteorological parameters, training soil parameters, training agronomic management parameters, and training greenhouse gas daily emission flux in the deep learning training set; the training agronomic management parameters include fertilizer management, tillage management, residue disposal method, and irrigation and drainage scheme.

[0037] Using the normalized training soil parameters, fertilizer management, tillage management, and residue disposal method as the static input of the deep neural network, the training meteorological parameters and irrigation and drainage scheme as the dynamic input of the deep neural network, the normalized training greenhouse gas daily emission flux as the output of the deep neural network, and the mean square error as the loss function, optimize the hyperparameters of the deep neural network using the Bayesian optimization framework to obtain the greenhouse gas daily emission flux prediction network.

[0038] In an exemplary embodiment, the training process of the total yield prediction network specifically includes: Normalize the training meteorological parameters, training soil parameters, training agronomic management parameters, and training total yield in the deep learning training set; the training agronomic management parameters include fertilizer management, tillage management, residue disposal method, and irrigation and drainage scheme.

[0039] Using the normalized training soil parameters, fertilizer management, tillage management, and residue disposal method as the static input of the deep neural network, the training meteorological parameters and irrigation and drainage scheme as the dynamic input of the deep neural network, the normalized training total yield as the output of the deep neural network, and the mean square error as the loss function, optimize the hyperparameters of the deep neural network using the Bayesian optimization framework to obtain the total yield prediction network.

[0040] In an exemplary embodiment, the deep neural network adopted by the greenhouse gas daily emission flux prediction network is a long short-term memory neural network; the deep neural network adopted by the total yield prediction network is an improved long short-term memory neural network; the improved long short-term memory neural network includes a long short-term memory neural network and a fully connected neural network layer connected to the output layer of the long short-term memory neural network.

[0041] Such as Figure 4 AndFigure 5 As shown, the training processes of the greenhouse gas daily emission flux prediction network and the total output prediction network are processed as follows in practical applications.

[0042] Using the calibrated DNDC model, a large number of different meteorological, soil, and agronomic management parameters are input to obtain the corresponding greenhouse gas daily emission flux and total output data.

[0043] Based on this data, a deep learning model is trained to accurately predict various greenhouse gas daily emission fluxes and total output data.

[0044] Specifically, through the calibrated DNDC model, a large number of different meteorological parameters, soil parameters, and agronomic management parameters are input using random numbers to obtain the deep learning training dataset.

[0045] The greenhouse gas daily emission flux prediction network uses a long short-term memory neural network (LSTM) to train on this deep learning dataset, enabling it to accurately predict the denitrification-decomposition model. Specifically, for the LSTM model of each specific farmland, each day is taken as a time step of the LSTM, and the total number of time steps should be equal to the total number of days in the rice growing period. The input of each time step consists of two parts. One is the static input, including soil parameters and the parameters of agronomic management except for the irrigation scheme. The other is the dynamic input, including daily meteorological parameters and the irrigation scheme. The output of each time step is the predicted greenhouse gas daily emission flux.

[0046] The total output prediction network uses an LSTM network integrated with a fully connected neural network layer to predict the final output. The input of each time step of the LSTM is exactly the same as that of the greenhouse gas prediction LSTM network, and the output of each time step is aggregated through the fully connected neural network layer into the predicted total output value.

[0047] All input and output values of the above deep learning networks need to be normalized before training to eliminate the influence of different dimensions on model prediction. The formula is as follows: .

[0048] Among them, is the normalization result; is the original value of the input or output quantity in the training set; is the maximum value of the input or output quantity in the training set; is the minimum value of the input or output quantity in the training set.

[0049] For the two deep learning networks mentioned above, the training dataset and the test dataset are divided in a ratio of 8:2, and the mean squared error function ( MSE), and the Adam optimizer is adopted. Bayesian optimization frameworks such as Optuna and Hyperopt are used to optimize the hyperparameters of the deep neural network. For the greenhouse gas daily emission flux prediction network, the hyperparameters include the number of epochs and the batch size; for the total output prediction network, in addition to the number of epochs and the batch size, the hyperparameters also include the number of hidden layers \(n_{hiddenLayer}\) of the fully connected neural network layer and the number of neurons in each hidden layer \(n_{hiddenLayer\_num}\). i . For the prediction of the daily greenhouse gas emission flux, the multi-objective mean square error is used to measure the prediction accuracy of each hyperparameter for the daily greenhouse gas emissions. The formula is: .

[0050] where is the multi-objective mean square error for the prediction of the daily greenhouse gas emission flux; is the total number of planting days; is the total number of data in the test set; is the th day's greenhouse gas emission value of the th test data; is the measured greenhouse gas emission value of the

[0051] th .

[0052] where is the mean square error value for the output prediction; represents the true value of the daily greenhouse gas emission flux or the total output, represents the simulated value of the daily greenhouse gas emission flux or the total output by DNDC. The closer it is to 0, the better the model performance.

[0053] For these two neural networks, the Bayesian optimization framework is used to select the hyperparameters with the best above indicators as the final hyperparameters, and all training data is used as the training set to train the final greenhouse gas daily emission flux prediction model and the total output prediction model.

[0054] In an exemplary embodiment, a multi-objective genetic algorithm is used to optimize the irrigation and drainage scheme with the actual greenhouse gas daily emission flux and the actual predicted total output as the optimization objectives, and the optimal irrigation and drainage scheme is obtained, which specifically includes: The Pareto front curve is optimized using the multi-objective genetic algorithm R-NSGA-II, and the actual daily greenhouse gas emission flux and the actual predicted total output are normalized and weighted to obtain the irrigation and drainage plan score; the Pareto front curve includes the actual daily greenhouse gas emission flux and the actual predicted total output; the irrigation and drainage plan with the highest score is selected as the optimal irrigation and drainage plan.

[0055] As Figure 6 shown, in practical applications, after rice sowing, the genetic algorithm is used daily to perform dual-objective optimization on the trained deep learning model to obtain drainage and irrigation strategies.

[0056] Specifically, after rice sowing, the current meteorological data input into the daily greenhouse gas emission flux prediction network and the total output prediction network every day, where the current meteorological data includes the current day's meteorology and future meteorological data, are all updated in real time according to the local meteorological station and weather forecast results to realize the dynamic adjustment of the optimization plan of the irrigation and drainage system. In addition, the meteorological data updated on the current day and the optimal irrigation and drainage strategy executed on the current day need to be stored as historical input data.

[0057] If the current day is the sowing day, there are no historical parameters, and only the meteorological parameters of the current day and the predicted future meteorological parameters need to be input into the daily greenhouse gas emission flux prediction network and the total output prediction network for the first irrigation and drainage decision. If the current day is after the second day after sowing, the daily greenhouse gas emission flux prediction network and the total output prediction network also need to input the stored historical meteorological data, and input all agronomic management parameters except irrigation and the optimal irrigation and drainage strategy that has been executed historically.

[0058] Using the multi-objective genetic algorithm R-NSGA-II, according to the daily greenhouse gas emission flux prediction network and the total output prediction network, search for the Pareto front curve of the optimal management mode for the current day and every future day, and perform optimal plan screening, as shown in the following formula: .

[0059] Among them, is the total output predicted by deep learning; is the daily greenhouse gas emission flux predicted by deep learning; is the total number of planting days; is the number of days from the current day to the sowing day; is the yield under the local traditional irrigation mode; , are respectively the day's irrigation volume and drainage volume. The current day's meteorological parameters input into the deep learning are input through the data of the nearest meteorological station on the current day, and the future meteorological parameters are input through the weather forecast results.

[0060] For the irrigation and drainage Pareto front obtained by genetic algorithm optimization, the optimal irrigation and drainage scheme is selected by using the normalized weighted value of yield and greenhouse gas emissions, as shown in the following formula: .

[0061] Wherein, is the score of the th irrigation and drainage scheme; is the predicted global warming potential (GWP) value of the th irrigation and drainage scheme; , are respectively the maximum and minimum predicted GWP values among all irrigation and drainage schemes; is the predicted total yield of the current irrigation and drainage scheme; , are respectively the maximum and minimum predicted yield values among all irrigation and drainage schemes.

[0062] Select the scheme with the highest score as the irrigation and drainage operation for each day from the current day to the harvest day, and execute the irrigation and drainage operation of the current day. After the daily irrigation and drainage strategy is executed, the current day's meteorological parameters and irrigation and drainage strategy are stored for subsequent input of model historical data.

[0063] This application can make real-time decisions on paddy field irrigation and drainage and execute them according to information such as the soil, meteorology, and agronomic management parameters of the paddy field, so as to reduce the greenhouse gas emissions caused by rice production while maintaining the yield.

[0064] Based on the same inventive concept, an embodiment of this application also provides a rice irrigation and drainage determination device for implementing the above-mentioned rice irrigation and drainage determination method. The solution provided by this device to solve the problem is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the rice irrigation and drainage determination device provided below can refer to the limitations on the rice irrigation and drainage determination method in the above text, and will not be repeated here.

[0065] As Figure 7 shown, in an exemplary embodiment, a rice irrigation and drainage determination device is provided, including: An acquisition module 701, configured to acquire actual meteorological parameters, actual soil parameters of the paddy field, and actual agronomic management parameters of the paddy field.

[0066] A prediction module 702, configured to perform predictions by using a greenhouse gas daily emission flux prediction network and a total output prediction network based on the actual meteorological parameters, the actual soil parameters, and the actual agronomic management parameters, so as to obtain an actual greenhouse gas daily emission flux and an actual predicted total output; both the greenhouse gas daily emission flux prediction network and the total output prediction network are obtained by training a deep neural network by using a deep learning training dataset; the deep learning training dataset is generated by a denitrification-decomposition model.

[0067] An optimization module 703, configured to optimize the irrigation and drainage scheme by using a multi-objective genetic algorithm, taking the actual greenhouse gas daily emission flux and the actual predicted total output as optimization objectives, so as to obtain an optimal irrigation and drainage scheme.

[0068] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal, and its internal structure diagram may be as Figure 8 shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store rice irrigation and drainage determination data. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a rice irrigation and drainage determination method is implemented.

[0069] Those skilled in the art can understand that Figure 8 the structure shown in

[0070] is only a block diagram of some structures related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0071] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, which when executed by a processor implements the above method embodiments.

[0072] In an exemplary embodiment, a computer program product is provided, including a computer program, which when executed by a processor implements the above method embodiments.

[0073] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data that have been authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.

[0074] In this application, all actions of obtaining signals, information, or data are carried out on the premise of complying with the corresponding data protection regulations and policies of the country where it is located and obtaining authorization from the owner of the corresponding device.

[0075] Those of ordinary skill in the art can understand that all or part of the processes in the above method embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above method embodiments. Among them, any reference to a memory, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0076] In each of the embodiments provided in the present application, the database involved may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on a blockchain, etc., without limitation. In each of the embodiments provided in the present application, the processor involved may be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without limitation.

[0077] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0078] Specific examples are used in this article to elaborate on the principles and implementation manners of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A method for determining rice irrigation and drainage, characterized in that, The rice irrigation and drainage determination method includes: Obtaining actual meteorological parameters, actual soil parameters of the paddy field, and actual agronomic management parameters of the paddy field; Predicting according to the actual meteorological parameters, the actual soil parameters, and the actual agronomic management parameters by using a greenhouse gas daily emission flux prediction network and a total yield prediction network to obtain the actual greenhouse gas daily emission flux and the actual predicted total yield; both the greenhouse gas daily emission flux prediction network and the total yield prediction network are obtained by training a deep neural network using a deep learning training dataset; the deep learning training dataset is generated by a denitrification-decomposition model; Using a multi-objective genetic algorithm, taking the actual greenhouse gas daily emission flux and the actual predicted total yield as optimization objectives to optimize the irrigation and drainage scheme, and obtaining the optimal irrigation and drainage scheme.

2. The rice irrigation and drainage determination method according to claim 1, characterized in that The specific process of generating the deep learning training dataset includes: Obtaining historical meteorological parameters, historical soil parameters, historical agronomic management parameters, historical greenhouse gas daily emission flux, and historical total yield of the paddy field; Adjusting the parameters of the denitrification-decomposition model according to the historical meteorological parameters, historical soil parameters, historical agronomic management parameters, historical greenhouse gas daily emission flux, and historical total yield to obtain a denitrification-decomposition model for simulating the greenhouse gas daily emission flux and the total yield; Using different training meteorological parameters, training soil parameters, and training agronomic management parameters to simulate the training greenhouse gas daily emission flux and the training total yield by using the denitrification-decomposition model for simulating the greenhouse gas daily emission flux and the total yield; Determining a deep learning training set according to the training meteorological parameters, training soil parameters, training agronomic management parameters, training greenhouse gas daily emission flux, and training total yield.

3. The rice irrigation and drainage determination method according to claim 2, characterized in that, The specific process of training the greenhouse gas daily emission flux prediction network includes: Normalizing the training meteorological parameters, training soil parameters, training agronomic management parameters, and training greenhouse gas daily emission flux in the deep learning training set; the training agronomic management parameters include fertilizer management, tillage management, residue disposal method, and irrigation and drainage scheme; Taking the normalized training soil parameters, fertilizer management, tillage management, and residue disposal method as the static input of the deep neural network, taking the training meteorological parameters and the irrigation and drainage scheme as the dynamic input of the deep neural network, taking the normalized training greenhouse gas daily emission flux as the output of the deep neural network, using the mean square error as the loss function, and optimizing the hyperparameters of the deep neural network by using a Bayesian optimization framework to obtain the greenhouse gas daily emission flux prediction network.

4. The method for determining rice irrigation and drainage according to claim 2, wherein The specific process of training the total yield prediction network includes: Normalizing the training meteorological parameters, training soil parameters, training agronomic management parameters, and training total yield in the deep learning training set; the training agronomic management parameters include fertilizer management, tillage management, residue disposal method, and irrigation and drainage scheme; Taking the normalized training soil parameters, fertilizer management, tillage management, and residue disposal methods as the static inputs of the deep neural network, the training meteorological parameters and irrigation and drainage schemes as the dynamic inputs of the deep neural network, the normalized training total output as the output of the deep neural network, and the mean square error as the loss function, the hyperparameters of the deep neural network are optimized using the Bayesian optimization framework to obtain the total output prediction network.

5. The method for determining rice irrigation and drainage according to claim 1, wherein The deep neural network used in the greenhouse gas daily emission flux prediction network is a long short-term memory neural network; the deep neural network used in the total output prediction network is an improved long short-term memory neural network; the improved long short-term memory neural network includes a long short-term memory neural network and a fully connected neural network layer connected to the output layer of the long short-term memory neural network.

6. The rice irrigation and drainage determination method according to claim 1, wherein, Using the multi-objective genetic algorithm, the actual greenhouse gas daily emission flux and the actual predicted total output are used as optimization objectives to optimize the irrigation and drainage scheme, and the optimal irrigation and drainage scheme is obtained, specifically including: Using the multi-objective genetic algorithm R-NSGA-II to optimize the Pareto front curve, normalizing and weighting the actual greenhouse gas daily emission flux and the actual predicted total output to obtain the irrigation and drainage scheme score; the Pareto front curve includes the actual greenhouse gas daily emission flux and the actual predicted total output; Selecting the irrigation and drainage scheme with the highest irrigation and drainage scheme score as the optimal irrigation and drainage scheme.

7. A rice irrigation and drainage determination device, characterized in that, The rice irrigation and drainage determination device includes: An acquisition module for acquiring actual meteorological parameters, actual soil parameters of the paddy field, and actual agronomic management parameters of the paddy field; A prediction module for predicting the actual greenhouse gas daily emission flux and the actual predicted total output according to the actual meteorological parameters, the actual soil parameters, and the actual agronomic management parameters using the greenhouse gas daily emission flux prediction network and the total output prediction network; both the greenhouse gas daily emission flux prediction network and the total output prediction network are obtained by training a deep neural network using a deep learning training dataset; the deep learning training dataset is generated by a denitrification-decomposition model; An optimization module for using the multi-objective genetic algorithm to optimize the irrigation and drainage scheme with the actual greenhouse gas daily emission flux and the actual predicted total output as optimization objectives to obtain the optimal irrigation and drainage scheme.

8. A computer device, comprising: A memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the rice irrigation and drainage determination method according to any one of claims 1-6.

9. A 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 rice irrigation and drainage determination method according to any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the rice irrigation and drainage determination method according to any one of claims 1-6.

Images