Intelligent decision-making method and device for agricultural water use based on big data

Abstract

The present invention relates to the field of artificial intelligence and data processing technology, and specifically to a method and device for intelligent agricultural water use decision-making based on big data. Specifically, the method comprises the following steps: collecting soil data and meteorological monitoring data, recording crop growth cycle data, and then forming a data set, which is then manually labeled based on field plots; normalizing the data collected in the data set; constructing an intelligent agricultural water use decision-making classification model, and training the model multiple times through multiple iterations to obtain a trained model by constructing a cost-sensitive feature cross-mapping network, calculating the loss function of an extreme learning machine, optimizing the parameters of the extreme learning machine based on evapotranspiration constraints, and dynamically adjusting the threshold. The newly collected data is pre-processed and then input into the trained intelligent agricultural water use decision-making classification model for processing, ultimately outputting a decision result. The present invention can address the accuracy, robustness, and interpretability challenges in agricultural water use management.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence and data processing technology, and in particular to a method and device for intelligent agricultural water use decision-making based on big data. Background Art

[0002] Focusing on the ecological protection and high-quality development of the Yellow River Basin is a major national strategy. To achieve this major national strategy, it can be specifically implemented in the development of water resources through science and technology, digital water management, smart water management and scientific water use. Water conservancy development and information technology can be deeply integrated to carry out comprehensive management of water conservancy. Among them, water source management in agricultural production is a complex and critical issue in smart water management and scientific water supply. Reasonable irrigation decisions can not only increase crop yields, but also effectively save water resources. However, agricultural water use decisions face many challenges, especially the significant differences in water demand at different stages of crop growth. The spatiotemporal heterogeneity and multimodal characteristics of soil and meteorological data make irrigation demand prediction more complicated.

[0003] Traditional agricultural irrigation decision-making methods mostly rely on fixed rules or simple statistical models, ignoring the nonlinear coupling relationship between crop growth stages, soil properties and meteorological conditions; in addition, traditional data processing methods such as global normalization often cannot effectively retain the water demand patterns of different growth stages, and are difficult to adapt to problems such as uneven sensor distribution and data noise; existing technologies have significant deficiencies in accuracy and adaptability when processing agricultural data, affecting the scientific nature and effectiveness of irrigation decisions.

[0004] The problems with existing technologies are as follows: Most existing technologies use global normalization or simple time series alignment methods, which cannot effectively handle the spatiotemporal heterogeneity of agricultural data and the differences between different crop growth stages. As a result, traditional methods cannot retain key farmland moisture characteristics, affecting the accuracy and adaptability of decision-making. Traditional agricultural water use decision-making models lack sensitivity to the misclassification costs of different categories, especially the high misclassification costs of the "water shortage" category. Existing methods cannot accurately balance the cost differences between different categories. Many existing agricultural water use decision-making models rely on data-driven methods and do not consider the physical constraints of the agricultural field. As a result, the models may output results that are inconsistent with agricultural common sense, making the models lack interpretability and scientific basis in practical applications. Existing agricultural decision-making models have poor adaptability to the time-varying characteristics of the agricultural environment and cannot flexibly respond to the influence of factors such as seasonal changes and meteorological mutations.

[0005] Therefore, the present invention proposes an intelligent decision-making method and device for agricultural water use based on big data to solve the above problems. Summary of the Invention

[0006] In response to the shortcomings of the existing technology, the present invention develops an intelligent decision-making method and device for agricultural water use based on big data. The present invention can solve the problems of accuracy, robustness and explainability in agricultural water management by constructing an intelligent decision-making classification model for agricultural water use.

[0007] On the one hand, the technical solution to the technical problem of the present invention is a method for intelligent agricultural water use decision-making based on big data, comprising the following steps:

[0008] S1. Data Collection: Soil data is collected through soil moisture sensors, meteorological monitoring data is collected through meteorological detection equipment, and crop growth cycle data is recorded. The collected data constitutes a dataset for training, and irrigation decisions are manually labeled for the corresponding fields based on the collected data;

[0009] S2. Data preprocessing: The data collected in the dataset are normalized using a segmented normalization strategy based on dynamic space-time windows and spatial classes;

[0010] S3. Training the constructed intelligent decision-making classification model for agricultural water use: Build an intelligent decision-making classification model for agricultural water use by constructing a cost-sensitive feature cross-mapping network, calculating the loss function of the extreme learning machine, optimizing the parameters of the extreme learning machine based on evapotranspiration constraints, and dynamically adjusting the threshold;

[0011] The preprocessed data is input into the agricultural water use intelligent decision-making classification model for training until the preset stop iteration condition is met, that is, the training of the agricultural water use intelligent decision-making classification model is completed;

[0012] S4. Intelligent decision-making on agricultural water use: The newly collected data is pre-processed and input into the trained intelligent decision-making classification model for agricultural water use, and the decision result is finally output.

[0013] S1 is as follows:

[0014] Soil data collected include, but are not limited to, volumetric moisture content, electrical conductivity, and temperature;

[0015] Meteorological monitoring data include but are not limited to precipitation, evaporation, temperature and wind speed;

[0016] Crop growth cycle data includes monitoring data at different growth stages, including germination, heading, and maturity;

[0017] During agricultural production, several soil moisture sensors and meteorological monitoring equipment are installed in the fields. These sensors and meteorological monitoring equipment collect data on different fields and record soil and meteorological data in real time. The collection time granularity can be set to hours or days according to specific needs.

[0018] The collected data is transmitted to the central database in real time via a wireless network for storage. The data stored in the central database constitutes the data set for training. The data in the central database is then read, and experts manually label irrigation decisions for each field based on the data collected from each field. The classification labels for irrigation decisions include water shortage, suitable, and oversaturated.

[0019] S2 is as follows:

[0020] The segmented normalization strategy specifically divides the crop growth stages, retains the water demand characteristics of each stage, and uses the spatial correction factor to introduce spatial correlation correction. The calculation formula of the segmented normalization strategy is as follows:

[0021] ,

[0022] in, Indicates the The first The data collected in this category are Normalized value of the growth stage, Indicates the The first The data collected in this category are The original observations collected during the growth stage, Indicates the The plots are in the first The spatial correction factor for each growth stage, Indicates that the time window is divided into A growing period, Indicates that the crop within the growth stage Time point The first Class collected data.

[0023] S3 is as follows:

[0024] The pre-processed data is input into the agricultural water use intelligent decision-making classification model to train the model. The model includes an extreme learning machine model. The input data is first passed through the extreme learning machine model. Through the calculation of the extreme learning machine hidden layer, the pre-processed data is mapped to the decision space to generate a preliminary decision.

[0025] Improve the accuracy of your decision by:

[0026] Construct a cost-sensitive feature cross-mapping network and dynamically adjust the parameter weights through a cost-sensitive learning mechanism;

[0027] A dual weight adjustment mechanism based on cost-sensitive matrix and sample density is used to calculate the loss function of the extreme learning machine;

[0028] Optimize the model parameters of the extreme learning machine based on evapotranspiration constraints;

[0029] Use a dynamic threshold adjustment mechanism to automatically adjust the decision threshold based on historical data fluctuations.

[0030] The cost-sensitive learning mechanism is as follows:

[0031] A dual-channel cross-mapping mechanism is used to calculate the activation values ​​of soil and meteorological data at different hidden layer nodes. Specifically, the cross-term design enables the agricultural water use intelligent decision-making classification model to explicitly learn the synergistic effect of soil and meteorological data, and uses the dynamic coupling coefficient through a gating mechanism to adjust the contribution of different data combinations.

[0032] The activation value is calculated as follows:

[0033] ,

[0034] ,

[0035] in, represents the set of preprocessed soil data, represents the set of preprocessed meteorological data, express A specific parameter index in for A specific parameter index in Indicates soil data channel The activation value of the hidden layer nodes, Indicates the meteorological data channel The activation value of the hidden layer nodes, Represents the rectified linear unit activation function ReLU, Represents soil data To The connection weights of hidden layer nodes, Represents soil data To The connection weights of hidden layer nodes, Represents the preprocessed soil data The input value of Indicates weather data Dynamic coupling weights in soil channels, Indicates weather data Dynamic coupling weights in the meteorological channel, Represents preprocessed meteorological data The input value of represents the Hadamard product used to capture the local interaction of soil-meteorological data, represents the Kronecker product used to capture the global coupling effect, Represents the dynamic coupling coefficient for learning the nonlinear synergistic effect of soil-meteorological data.

[0036] The loss function of the extreme learning machine is as follows:

[0037] A dual-weight adjustment mechanism based on a cost-sensitive matrix and sample density is used to calculate the loss function of the extreme learning machine. Specifically, the sample density weight is used to alleviate the imbalance in the number of categories, and the cost matrix is ​​used to introduce cost differences defined by business knowledge.

[0038] The calculation formula of the extreme learning machine loss function is as follows:

[0039] ,

[0040] in, represents the loss function of the extreme learning machine, Indicates the The true category labels of samples, express The density weight of , Indicates that the label is The number of samples, represents the total number of categories of irrigation decisions, is the total number of training samples, represents the preset cost matrix, Indicates that the The true category label of the sample is misclassified as The penalty coefficient for each category, Indicates the The output vector of the sample in the hidden layer of the extreme learning machine is transposed. Indicates that the output layer of the extreme learning machine corresponds to The weight vector of each category, Indicates that the output layer corresponds to The weight vector of each category, represents the focusing factor, Set to 2.

[0041] The model parameter optimization process of the extreme learning machine based on evapotranspiration constraint is as follows:

[0042] The FAO Penman-Monteith equation for calculating potential evapotranspiration in high latitudes is incorporated into the optimization process as a soft constraint, ensuring that the model predictions conform to the basic laws of water transport through physical constraints.

[0043] The calculation formula is as follows:

[0044] ,

[0045] ,

[0046] in, The model predicts the The evapotranspiration of the samples, Indicates the first The theoretical evapotranspiration of a sample, Indicates that the parameter is The loss function of the extreme learning machine when represents the parameters of the extreme learning machine, represents the weight coefficient of the physical constraint term, Set to 0.3, represents the total number of training samples, represents the total number of time steps in the crop growth cycle, It represents the vapor pressure difference influence coefficient, which is obtained by optimizing the historical evapotranspiration data fitting through the gradient descent method after random initialization. Represents the net radiation influence coefficient, which is obtained by optimizing the historical evapotranspiration data fitting through the gradient descent method after random initialization. Indicates the The sample in The hidden layer output vector of time period; Indicates the The sample in Vapor pressure difference over time periods; Indicates the The sample in Net radiation amount in a time period;

[0047] Then, based on the loss function of the extreme learning machine, the improved Nesterov accelerated gradient method is used to update the parameters of the extreme learning machine.

[0048] The dynamic threshold adjustment is as follows:

[0049] The sliding time window threshold calibration method is used to calibrate the output of the extreme learning machine, and the decision function is defined as:

[0050] ,

[0051] in, Indicates the The irrigation decision prediction results for each time period are: 1 means irrigation is needed, corresponding to water shortage, and 0 means no irrigation is needed, corresponding to suitable and oversaturated conditions. Indicates that the extreme learning machine model is The confidence level of the forecast for the time period, Indicates the Multi-source input feature vectors for time periods, Indicates the The decision threshold is dynamically adjusted for each time period.

[0052] S4 is as follows:

[0053] Adjust the irrigation amount of the field in each time period according to the decision results, which include water shortage, suitable and oversaturation;

[0054] If there is water shortage, increase the amount of irrigation.

[0055] If the output is suitable, it remains unchanged;

[0056] If the output is oversaturated, drain the water in time.

[0057] On the other hand, the present invention also provides a computer storage medium storing instructions, which, when executed on a computer, enables the computer to execute an intelligent agricultural water use decision-making method driven by big data.

[0058] The effects provided in the summary of the invention are only the effects of the embodiments, rather than all the effects of the invention. The above technical solution has the following advantages or beneficial effects:

[0059] The present invention adopts a segmented normalization method based on dynamic time windows and spatial clustering. Through segmented normalization based on dynamic time windows and spatial clustering, the present invention can dynamically divide time periods according to the crop growth cycle, retain the water demand characteristics of different growth periods, and use spatial correction factors to resolve local deviations caused by uneven sensor density. This allows the water demand patterns of each stage to be accurately preserved, avoids global normalization blurring key features, and improves the accuracy of farmland water management.

[0060] The present invention adopts a dual-channel cross-mapping mechanism to explicitly learn the synergistic effects of soil and meteorological parameters through cross-terms, and uses a cost-sensitive method to adjust the model weights, thereby enhancing the sensitivity to important decision categories and improving the accuracy and convergence speed of the model.

[0061] A dual-weight adjustment mechanism based on a cost-sensitive matrix and sample density is used to adjust the weights of different categories. This improves the class imbalance problem and improves the recall rate of the water-scarce category, especially when dealing with categories with large cost differences such as "water shortage" and "oversaturation".

[0062] This invention incorporates physical constraints into the model training process to ensure that the model's predictions conform to the basic laws of water transport, avoiding the possibility that purely data-driven models will produce predictions that violate agricultural common sense, and improving the model's interpretability and physical rationality.

[0063] This invention uses a dynamic threshold adjustment mechanism to address the time-varying nature of agricultural water use decisions. It uses a sliding time window to adjust the decision threshold, automatically senses environmental changes, and adjusts the conservatism of decisions based on historical data fluctuations. This enhances the model's adaptability to meteorological changes and seasonal fluctuations, and reduces decision-making fluctuations caused by sudden meteorological changes.

[0064] In summary, the problems of accuracy, robustness and interpretability in agricultural water management have been systematically solved, providing a set of innovative and practical technical solutions for precision agricultural water use. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0066] Figure 1 Schematic diagram of the method of the present invention.

[0067] Figure 2 Comparison diagram of feature distinction between dynamic segmentation normalization and global normalization.

[0068] Figure 3 The figure is a line graph showing the classification accuracy of the method in the present invention and the existing method under different sample sizes.

[0069] Figure 4 This is a performance comparison chart of the confusion matrix of the present invention and the traditional confusion matrix.

[0070] Figure 5 It is a line graph of the evapotranspiration constraint effect in the present invention.

[0071] Figure 6 Schematic diagram of evapotranspiration prediction for the unconstrained model.

[0072] Figure 7 Schematic diagram of the prediction of physical constraints in the present invention. DETAILED DESCRIPTION

[0073] To clearly illustrate the technical features of this solution, the present invention is described in detail below through specific embodiments and in conjunction with the accompanying drawings. The following disclosure provides many different embodiments or examples for implementing different structures of the present invention. To simplify the disclosure of the present invention, the components and configurations of specific examples are described below.

[0074] Example 1

[0075] A big data-driven intelligent decision-making method for agricultural water use includes the following steps:

[0076] S1. Data Collection: Soil data is collected through soil moisture sensors, meteorological monitoring data is collected through meteorological detection equipment, and crop growth cycle data is recorded. The collected data constitutes a dataset for training, and irrigation decisions are manually labeled for the corresponding fields based on the collected data;

[0077] S2. Data preprocessing: The data collected in the dataset are normalized using a segmented normalization strategy based on dynamic space-time windows and spatial classes;

[0078] S3. Training the constructed intelligent decision-making classification model for agricultural water use: Build an intelligent decision-making classification model for agricultural water use by constructing a cost-sensitive feature cross-mapping network, calculating the loss function of the extreme learning machine, optimizing the parameters of the extreme learning machine based on evapotranspiration constraints, and dynamically adjusting the threshold;

[0079] The preprocessed data is input into the agricultural water use intelligent decision-making classification model for training until the preset stop iteration condition is met, that is, the training of the agricultural water use intelligent decision-making classification model is completed;

[0080] S4. Intelligent decision-making on agricultural water use: The newly collected data is pre-processed and input into the trained intelligent decision-making classification model for agricultural water use, and the decision result is finally output.

[0081] In a specific implementation manner, S1 is specifically as follows:

[0082] The collected soil data include but are not limited to volumetric moisture content, electrical conductivity and temperature, which reflect the soil's moisture status, electrical conductivity and temperature changes;

[0083] Meteorological monitoring data include but are not limited to precipitation, evaporation, temperature and wind speed, which reflect the impact of the external environment on soil moisture;

[0084] Crop growth cycle data includes monitoring data at different growth stages, including germination, heading, and maturity. The monitoring data at different growth stages reflect the changes in crop water requirements as the growth stages change.

[0085] During agricultural production, several soil moisture sensors and meteorological monitoring equipment are installed in the fields. These sensors and meteorological monitoring equipment collect data on different fields and record soil and meteorological data in real time. The collection time granularity can be set to hours or days according to specific needs.

[0086] The collected data is transmitted to the central database in real time via a wireless network for storage. The data stored in the central database constitutes the data set for training. The data in the central database is then read, and experts manually label irrigation decisions for each field based on the data collected from each field. The classification labels for irrigation decisions include water shortage, suitable, and oversaturated.

[0087] In a specific implementation manner, S2 is specifically as follows:

[0088] The segmented normalization strategy specifically divides the crop growth stages, retains the water demand characteristics of each stage, and uses the spatial correction factor to introduce spatial correlation correction. The calculation formula of the segmented normalization strategy is as follows:

[0089] ,

[0090] in, Indicates the The first The data collected in this category are Normalized value of the growth stage, Indicates the The first The data collected in this category are The original observations collected during the growth stage, Indicates the The plots are in the first The spatial correction factor for each growth stage, Indicates that the time window is divided into The problem of different water demand patterns in different growth stages is solved by dynamically dividing the time periods. For example, the water demand in the heading stage is significantly higher than that in the budding stage. Segmented normalization can retain the characteristics of each stage and avoid global normalization from blurring the key water demand characteristics. Indicates that the crop within the growth stage Time point The first Data collected by the class;

[0091] Furthermore, the calculation formula of the spatial correction factor is as follows:

[0092] ,

[0093] in, Represents the spatial correlation coefficient. By adjusting the spatial correlation coefficient to balance local characteristics and global distribution, it can adapt to the heterogeneity of different farmland areas. Set to 0.2, Indicates the The multidimensional feature mean vector of the plot, Indicates the The characteristic mean vector of the spatial cluster center to which each field belongs, It represents the weighted Mahalanobis distance, which can measure the spatial difference between a field and its cluster center, thereby solving the local deviation caused by uneven sensor deployment density.

[0094] In a specific implementation, S3 is as follows:

[0095] The pre-processed data is input into the agricultural water use intelligent decision-making classification model to train the model. The model includes an extreme learning machine model. The input data is first passed through the extreme learning machine model. Through the calculation of the extreme learning machine hidden layer, the pre-processed data is mapped to the decision space to generate a preliminary decision.

[0096] Improve the accuracy of your decision by:

[0097] Construct a cost-sensitive feature cross-mapping network and dynamically adjust the parameter weights through a cost-sensitive learning mechanism;

[0098] A dual weight adjustment mechanism based on cost-sensitive matrix and sample density is used to calculate the loss function of the extreme learning machine;

[0099] Optimize the model parameters of the extreme learning machine based on evapotranspiration constraints;

[0100] Use a dynamic threshold adjustment mechanism to automatically adjust the decision threshold based on historical data fluctuations.

[0101] Furthermore, the cost-sensitive learning mechanism is as follows:

[0102] A dual-channel cross-mapping mechanism is used to calculate the activation values ​​of soil and meteorological data at different hidden layer nodes. Specifically, the cross-term design enables the agricultural water use intelligent decision-making classification model to explicitly learn the synergistic effect of soil and meteorological data, and uses the dynamic coupling coefficient through a gating mechanism to adjust the contribution of different data combinations.

[0103] The activation value is calculated as follows:

[0104] ,

[0105] ,

[0106] in, represents the set of preprocessed soil data, represents the set of preprocessed meteorological data, express A specific parameter index in for A specific parameter index in Indicates soil data channel The activation value of the hidden layer nodes, Indicates the meteorological data channel The activation value of the hidden layer nodes, Represents the rectified linear unit activation function ReLU, Represents soil data To The connection weights of hidden layer nodes, Represents soil data To The connection weights of hidden layer nodes, Represents the preprocessed soil data The input value of Indicates weather data Dynamic coupling weights in soil channels, Indicates weather data Dynamic coupling weights in the meteorological channel, Represents preprocessed meteorological data The input value of represents the Hadamard product used to capture the local interaction of soil-meteorological data, represents the Kronecker product used to capture the global coupling effect, Represents the dynamic coupling coefficient for learning the nonlinear synergistic effect of soil-meteorological data.

[0107] Calculated by cost-sensitive method and , the calculation formula is as follows:

[0108] ,

[0109] ,

[0110] in, represents the first basic coupling coefficient, The default value is 0.5, represents the second basic coupling coefficient, The default value is 0.3, Represents the true category The misclassification cost, Represents the predicted category The price, It means that the weight is dynamically adjusted according to the misclassification cost. By adjusting the feature weight by the cost ratio, the sensitivity to high-cost categories can be enhanced.

[0111] For example, the weight of misjudging water shortage as oversaturation (cost 5) is higher than that of misjudging the opposite (cost 1), thus increasing the sensitivity to the high-cost category;

[0112] About the dynamic coupling coefficient The calculation formula is as follows:

[0113] ,

[0114] ,

[0115] in, represents the Sigmoid activation function, and represent the original soil data and meteorological data without preprocessing, represents parameter interaction function; represents the first embedding matrix, which is obtained by random initialization and updated as the model is trained. The update method can be updated using the gradient descent method; represents the second embedding matrix, which is randomly initialized and updated as the model is trained. The update method can be updated using the gradient descent method; Represents the bias term, which is updated as the model is trained. The update method can be updated using the gradient descent method.

[0116] Furthermore, the loss function of the extreme learning machine is as follows:

[0117] A dual-weight adjustment mechanism based on a cost-sensitive matrix and sample density is used to calculate the loss function of the extreme learning machine. Specifically, the sample density weight is used to alleviate the imbalance in the number of categories, and the cost matrix is ​​used to introduce cost differences defined by business knowledge.

[0118] The calculation formula of the extreme learning machine loss function is as follows:

[0119] ,

[0120] in, represents the loss function of the extreme learning machine, Indicates the The true category labels of samples, express The density weight of , Indicates that the label is The number of samples, represents the total number of categories of irrigation decisions, is the total number of training samples, Indicates the The output vector of the sample in the hidden layer of the extreme learning machine is transposed. Indicates that the output layer of the extreme learning machine corresponds to The weight vector of each category, Dynamic sparsification is achieved through the L1 regularization term, which suppresses redundant features and solves the problem of feature redundancy in high-dimensional agricultural data. Indicates that the output layer corresponds to The weight vector of each category, represents the focusing factor, Set to 2;

[0121] represents the preset cost matrix, Indicates that the The true category label of the sample is misclassified as The penalty coefficients for each category are predefined by domain experts based on the actual irrigation cost differences. For example, for the categories of water shortage (0), suitable (1), and oversaturated (2), the constructed cost matrix is ​​as follows:

[0122] ,

[0123] Among them, the rows of the cost matrix represent the true categories, the columns represent the predicted categories, the cost of the diagonal (correct classification) is 0, the cost of misclassifying water shortage (0) as oversaturation (2) is 5 (the highest penalty), the cost of misclassifying water shortage (0) as suitable (1) is 3, the cost of misclassifying oversaturation (2) as water shortage (0) is 1 (the lowest penalty), and the cost of misclassifying suitable (1) as oversaturation (2) is 2;

[0124] The update rules are as follows:

[0125] ,

[0126] in, Indicates the updated class weight vectors, represents the learning rate of the weight vector, Set to 0.01, represents the L1 regularization strength coefficient, Set to 0.001, Indicates the sign function, the sign of the output parameter (+1 / -1), is the L1 norm.

[0127] Furthermore, the model parameter optimization process of the extreme learning machine based on evapotranspiration constraint is as follows:

[0128] The FAO Penman-Monteith equation for calculating potential evapotranspiration in high latitudes is incorporated into the optimization process as a soft constraint, ensuring that the model predictions conform to the basic laws of water transport through physical constraints.

[0129] The calculation formula is as follows:

[0130] ,

[0131] ,

[0132] in, The model predicts the The evapotranspiration of the samples, Indicates the first The theoretical evapotranspiration of a sample, Indicates that the parameter is The loss function of the extreme learning machine when represents the parameters of the extreme learning machine, represents the weight coefficient of the physical constraint term, Set to 0.3, represents the total number of training samples, represents the total number of time steps in the crop growth cycle, It represents the vapor pressure difference influence coefficient, which is obtained by optimizing the historical evapotranspiration data fitting through the gradient descent method after random initialization. Represents the net radiation influence coefficient, which is obtained by optimizing the historical evapotranspiration data fitting through the gradient descent method after random initialization. Indicates the The sample in The hidden layer output vector of time period; Indicates the The sample in Vapor pressure difference over time periods; Indicates the The sample in Net radiation amount in a time period;

[0133] Then, based on the loss function of the extreme learning machine, the improved Nesterov accelerated gradient method is used to update the parameters of the extreme learning machine. The calculation formula is as follows:

[0134] ,

[0135] ,

[0136] in, It means that the Nesterov momentum is The update amount of the iteration, It means that the Nesterov momentum is The update amount of the iteration, represents the momentum decay coefficient, Set to 0.9, represents the basic learning rate of the extreme learning machine parameters, Set to 0.005, represents the joint gradient of the extreme learning machine loss function and the physical constraint term with respect to the extreme learning machine parameters; represents the mean square loss of evapotranspiration prediction error, ; represents the extreme learning machine Model parameters for the iteration; represents the extreme learning machine Model parameters for the iteration; represents the momentum correction factor, ; represents the L2 norm.

[0137] Furthermore, the dynamic threshold adjustment is as follows:

[0138] The sliding time window threshold calibration method is used to calibrate the output of the extreme learning machine, and the decision function is defined as:

[0139] ,

[0140] in, Indicates the The irrigation decision prediction results for each time period are: 1 means irrigation is needed, corresponding to water shortage, and 0 means no irrigation is needed, corresponding to suitable and oversaturated conditions. Indicates that the extreme learning machine model is The confidence level of the forecast for the time period, Indicates the Multi-source input feature vectors for time periods, Indicates the Dynamically adjusted decision thresholds for each time period;

[0141] The update rules are as follows:

[0142] ,

[0143] in, Indicates the first Input feature vector of historical moments; represents the threshold smoothing coefficient, Set to 0.7; Indicates the Dynamically adjusted decision thresholds for each time period; Indicates the The set of historical moment indexes contained in the sliding time window of a time period; represents the sensitivity coefficient, Set to 0.1; Indicates the The standard deviation of the model output within the time window of the time period is obtained by Automatically sense the severity of environmental changes and when data fluctuates greatly Raising means increasing the conservatism of the threshold, thereby resolving the decision-making shocks caused by sudden changes in weather conditions;

[0144] The calculation formula is as follows:

[0145] ,

[0146] in, Indicates the first The input feature vector of each historical moment.

[0147] Furthermore, S4 is as follows:

[0148] Adjust the irrigation amount of the field in each time period according to the decision results, which include water shortage, suitable and oversaturation;

[0149] If there is water shortage, increase the amount of irrigation.

[0150] If the output is suitable, it remains unchanged;

[0151] If the output is oversaturated, drain the water in time.

[0152] Example 2

[0153] A computer storage medium stores instructions, which, when executed on a computer, enable the computer to execute an intelligent decision-making method for agricultural water use driven by big data.

[0154] Example 3

[0155] like Figure 2 As shown in the figure, it is a feature comparison diagram of dynamic segment normalization and global normalization. Figure 2 It can be seen that by comparing the effects of dynamic segmented normalization, the effectiveness of segmented normalization in feature retention is verified. In terms of key features such as volumetric moisture content and electrical conductivity, the method of the present invention improves the discrimination by about 15% to 20% compared with global normalization. The experimental results show that dynamic time window division effectively retains the characteristic differences of crops in different growth stages.

[0156] Example 4

[0157] like Figure 3As shown in the figure, the classification accuracy line graph of the method of the present invention and the existing method under different sample sizes is shown in the figure. Figure 3 It can be seen that by comparing the performance of feature cross-networks, the effectiveness of dual-channel cross-mapping is verified. When the sample size is greater than 2000, the accuracy of the present invention is improved by 8% to 12% compared with the traditional extreme learning machine, and the convergence speed is faster. The experimental results show that the feature cross-network mechanism significantly improves the model's ability to model parameter coupling relationships.

[0158] Example 5

[0159] like Figure 4 As shown in the figure, the performance comparison diagram of the confusion matrix of the present invention and the traditional confusion matrix is ​​shown in the figure. Figure 4 It can be seen that by analyzing the cost-sensitive loss effect, the effectiveness of the misclassification cost adjustment mechanism is verified. In terms of the recall rate of the water shortage category, the present invention improves it by 9% and reduces the oversaturation misjudgment rate by 30%. The experimental results show that the cost-sensitive mechanism effectively balances the business needs of different misjudgment costs.

[0160] Example 6

[0161] like Figure 5 As shown, Figure 5 is a line graph of the evapotranspiration constraint effect in the present invention, such as Figure 5 As shown in the figure, in order to analyze the optimization effect of evapotranspiration constraints and verify the guiding role of physical constraints on model optimization, after the constraints were introduced, the evapotranspiration prediction error was reduced by about 42%, and the training loss converged more stably. The experimental results show that physical constraints effectively improve the interpretability of the model and the rationality of the prediction.

[0162] Further, if Figure 6 and Figure 7 As shown, Figure 6 Schematic diagram of evapotranspiration prediction for the unconstrained model. Figure 7 This is a schematic diagram of the prediction of physical constraints in the present invention, Figure 6 and Figure 7 It can be seen that the role of physical constraints in improving the rationality of model predictions is verified by contour line comparison. The unconstrained model violates the law of water transfer in the key parameter range of vapor pressure difference and net radiation, and its predicted value has a systematic deviation from the theoretical calculated value. However, this technology considers evapotranspiration strategy as an optimization constraint to make the predicted surface highly consistent with the theoretical contour line, especially in extreme parameter combination areas such as high temperature and low humidity. Physical consistency is still maintained. This optimization strategy that integrates data-driven and physical laws not only improves the prediction accuracy, but also ensures that the model output conforms to the basic principles of agricultural hydrology, making irrigation decisions both data intelligent and scientifically interpretable.

[0163] Although the above describes the specific implementation methods of the invention in conjunction with the accompanying drawings, it does not limit the scope of protection of the invention. Based on the technical solution of the present invention, various modifications or variations that can be made by those skilled in the art without creative work are still within the scope of protection of the present invention.

Claims

1. A big data-driven intelligent decision-making method for agricultural water use, characterized by: The following steps are involved: S1. Data Collection: Soil data is collected through soil moisture sensors, meteorological monitoring data is collected through meteorological detection equipment, and crop growth cycle data is recorded. The collected data constitutes a dataset for training, and irrigation decisions are manually labeled for the corresponding fields based on the collected data; S2. Data preprocessing: normalize the data collected in the dataset through a segmented normalization strategy based on dynamic space-time windows and spatial classes; S3. Training the constructed intelligent decision-making classification model for agricultural water use: Build an intelligent decision-making classification model for agricultural water use by constructing a cost-sensitive feature cross-mapping network, calculating the loss function of the extreme learning machine, optimizing the parameters of the extreme learning machine based on evapotranspiration constraints, and dynamically adjusting the threshold; The preprocessed data is input into the agricultural water use intelligent decision-making classification model for training until the preset stop iteration condition is met, that is, the training of the agricultural water use intelligent decision-making classification model is completed; The model includes an extreme learning machine model. The input data first passes through the extreme learning machine model. Through the calculation of the extreme learning machine hidden layer, the pre-processed data is mapped to the decision space to generate a preliminary decision; Improve the accuracy of your decision by: A cost-sensitive feature cross-mapping network is constructed, and parameter weights are dynamically adjusted through a cost-sensitive learning mechanism. The cost-sensitive learning mechanism uses a dual-channel cross-mapping mechanism to calculate the activation values ​​of soil and meteorological data at different hidden layer nodes. Specifically, through the cross-term design, the agricultural water intelligent decision-making classification model can explicitly learn the synergistic effect of soil and meteorological data, and use the dynamic coupling coefficient through a gating mechanism to adjust the contribution of different data combinations. A dual-weight adjustment mechanism based on a cost-sensitive matrix and sample density is used to calculate the loss function of the extreme learning machine. Specifically, the sample density weight is used to alleviate the imbalance in the number of categories, and the cost matrix is ​​used to introduce cost differences defined by business knowledge. Optimize the model parameters of the extreme learning machine based on evapotranspiration constraints; Use a dynamic threshold adjustment mechanism to automatically adjust the decision threshold based on historical data fluctuations; The model parameter optimization process of the extreme learning machine based on evapotranspiration constraint is as follows: The FAO Penman-Monteith equation for calculating potential evapotranspiration in high-latitude regions was incorporated into the optimization process as a soft constraint. Physical constraints ensured that the model predictions adhered to the fundamental laws of water transport. Then, based on the extreme learning machine's loss function, an improved Nesterov accelerated gradient method was used to update the extreme learning machine's parameters. S4. Intelligent decision-making on agricultural water use: The newly collected data is pre-processed and input into the trained intelligent decision-making classification model for agricultural water use, and the decision result is finally output.

2. The intelligent decision-making method for agricultural water use based on big data drive according to claim 1 is characterized in that: S1 is as follows: Soil data collected include, but are not limited to, volumetric moisture content, electrical conductivity, and temperature; Meteorological monitoring data include but are not limited to precipitation, evaporation, temperature and wind speed; Crop growth cycle data includes monitoring data at different growth stages, including germination, heading, and maturity; During agricultural production, several soil moisture sensors and meteorological monitoring equipment are installed in the fields. These sensors and meteorological monitoring equipment collect data on different fields and record soil and meteorological data in real time. The collection time granularity can be set to hours or days according to specific needs. The collected data is transmitted to the central database in real time via a wireless network for storage. The data stored in the central database constitutes the data set for training. The data in the central database is then read, and experts manually label irrigation decisions for each field based on the data collected from each field. The classification labels for irrigation decisions include water shortage, suitable, and oversaturated.

3. The intelligent decision-making method for agricultural water use based on big data drive according to claim 2 is characterized in that: S2 is as follows: The segmented normalization strategy specifically divides the crop growth stages, retains the water demand characteristics of each stage, and uses the spatial correction factor to introduce spatial correlation correction. The calculation formula of the segmented normalization strategy is as follows: ， in, Indicates the The first The data collected in this category are Normalized value of the growth stage, Indicates the The first The data collected in this category are The original observations collected during the growth stage, Indicates the The plots are in the first The spatial correction factor for each growth stage, Indicates that the time window is divided into A growing period, Indicates that the crop within the growth stage Time point The first Class collected data.

4. The intelligent decision-making method for agricultural water use based on big data according to claim 3 is characterized in that the cost The calculation formula of the activation value in the sensitive learning mechanism is as follows: ， ， in, represents the set of preprocessed soil data, represents the set of preprocessed meteorological data, express A specific parameter index in for A specific parameter index in Indicates soil data channel The activation value of the hidden layer nodes, Indicates the meteorological data channel The activation value of the hidden layer nodes, Represents the rectified linear unit activation function ReLU, Represents soil data To The connection weights of hidden layer nodes, Represents soil data To The connection weights of hidden layer nodes, Represents the preprocessed soil data The input value of Indicates weather data Dynamic coupling weights in soil channels, Indicates weather data Dynamic coupling weights in the meteorological channel, Represents meteorological data after preprocessing The input value of represents the Hadamard product used to capture the local interaction of soil-meteorological data, represents the Kronecker product used to capture the global coupling effect, Represents the dynamic coupling coefficient for learning the nonlinear synergistic effect of soil-meteorological data.

5. The intelligent decision-making method for agricultural water use based on big data drive according to claim 4 is characterized in that: The calculation formula of the extreme learning machine loss function is as follows: ， in, represents the loss function of the extreme learning machine, Indicates the The true category labels of samples, express The density weight of , Indicates that the label is The number of samples, represents the total number of categories of irrigation decisions, is the total number of training samples, represents the preset cost matrix, Indicates that the The true category label of the sample is misclassified as The penalty coefficient for each category, Indicates the The output vector of the sample in the hidden layer of the extreme learning machine is transposed. Indicates that the output layer of the extreme learning machine corresponds to The weight vector of each category, Indicates that the output layer corresponds to The weight vector of each category, represents the focusing factor, Set to 2.

6. The intelligent decision-making method for agricultural water use based on big data drive according to claim 5 is characterized in that: The calculation formula for the model parameter optimization process of the extreme learning machine based on evapotranspiration constraint is as follows: ， ， in, The model predicts the The evapotranspiration of the samples, Indicates the first The theoretical evapotranspiration of a sample, Indicates that the parameter is The loss function of the extreme learning machine when represents the parameters of the extreme learning machine, represents the weight coefficient of the physical constraint term, Set to 0.3, represents the total number of training samples, represents the total number of time steps in the crop growth cycle, It represents the vapor pressure difference influence coefficient, which is obtained by optimizing the historical evapotranspiration data fitting through the gradient descent method after random initialization. Represents the net radiation influence coefficient, which is obtained by optimizing the historical evapotranspiration data fitting through the gradient descent method after random initialization. Indicates the The sample in The hidden layer output vector of each time period; Indicates the The sample in Vapor pressure difference over time periods; Indicates the The sample in Net radiation amount in a time period; Then, based on the loss function of the extreme learning machine, the improved Nesterov accelerated gradient method is used to update the parameters of the extreme learning machine.

7. The big data-driven intelligent decision-making method for agricultural water use according to claim 6 is characterized in that: The dynamic threshold adjustment is as follows: The sliding time window threshold calibration method is used to calibrate the output of the extreme learning machine, and the decision function is defined as: ， in, Indicates the The irrigation decision prediction results for each time period are: 1 means irrigation is needed, corresponding to water shortage, and 0 means no irrigation is needed, corresponding to suitable and oversaturated conditions. Indicates that the extreme learning machine model is The confidence level of the forecast for the time period, Indicates the Multi-source input feature vectors for time periods, Indicates the The decision threshold is dynamically adjusted for each time period.

8. The big data-driven intelligent decision-making method for agricultural water use according to claim 7 is characterized in that: S4 is as follows: Adjust the irrigation amount of the field in each time period according to the decision results, which include water shortage, suitable and oversaturation; If there is a shortage of water, increase the amount of irrigation. If the output is suitable, it remains unchanged; If the output is oversaturated, drain the water in time.

9. A computer storage medium, characterized in that: Instructions are stored on a computer storage medium. When the instructions are executed on a computer, the computer executes the big data-driven intelligent decision-making method for agricultural water use as described in any one of claims 1 to 8.

Images