Irrigation decision optimization method and platform based on multi-source information fusion

Abstract

The present invention discloses an irrigation decision optimization method and platform for multi-source information fusion, which involves data processing related fields. The method includes: determining data acquisition nodes, which are based on preset time intervals and interval irrigation factor characteristics as compensation amounts; triggering source acquisition equipment, synchronously collecting data from the irrigation area, and determining multi-source information; performing fuzzy set analysis on irrigation factors, mining confidence matrices, and supervising the training of irrigation decision models; transmitting multi-source information to the irrigation decision model, performing de-heterogeneity and matrix operation decisions, and determining irrigation strategies; carrying an intelligent irrigation decision system, responding to irrigation strategies, and managing and controlling irrigation equipment. The method solves the technical problems of lack of intelligence and low decision accuracy in existing irrigation decisions, and achieves the technical effect of realizing intelligent and precise irrigation management.

Description

Technical Field

[0001] The present application relates to data processing related fields, and in particular to an irrigation decision optimization method and platform based on multi-source information fusion. Background Art

[0002] As an indispensable part of agricultural production, irrigation management and decision-making quality are directly related to the growth of crops, yield and water resource utilization efficiency. With the increasing shortage of global water resources, how to improve the utilization efficiency of irrigation water and reduce water resource waste while ensuring the normal growth of crops has become an important challenge facing the sustainable development of agriculture. Traditional irrigation methods are often based on experience or fixed irrigation plans, without fully considering the differences in crop types, growth stages, and environmental conditions (such as soil moisture, rainfall, temperature, etc.). For example, in arid areas, different types of crops have very different water requirements, and traditional irrigation often adopts a "one-size-fits-all" strategy, resulting in some crops being damaged by excessive water, while others are hindered from growing due to lack of water, low irrigation efficiency, and serious waste of water resources. In addition, the traditional irrigation decision-making process lacks the support of intelligent technology, and cannot obtain and analyze the irrigation needs of farmland in real time and accurately, so that irrigation decisions often lag behind the actual needs of crops. When crops are in urgent need of water, irrigation decisions may not be made in time, resulting in hindered crop growth; and when the soil moisture is sufficient, irrigation may still continue, resulting in unnecessary waste of water resources.

[0003] Among the current related technologies, irrigation decisions have technical problems such as lack of intelligence and low decision-making accuracy. Summary of the invention

[0004] This application provides an irrigation decision optimization method and platform for multi-source information fusion, integrates multiple data sources, and uses methods such as fuzzy set analysis and confidence matrix to explore the degree of support of each irrigation factor for irrigation decision-making, and constructs an irrigation decision model including a multi-source fusion layer, an irrigation support layer and a conflict balance layer. The model is placed on the cloud computing end and can process and analyze multi-source information in real time, determine the optimal irrigation strategy, and use technical means such as precise control of irrigation equipment through intelligent irrigation decision-making systems, thereby achieving the technical effect of realizing intelligent and precise irrigation management.

[0005] This application provides an irrigation decision optimization method based on multi-source information fusion, including:

[0006] Determine a data acquisition node, wherein the data acquisition node is determined based on a preset time interval and the interval irrigation factor characteristics as the compensation amount; based on the data acquisition node, trigger the source end acquisition device to synchronously collect data in the irrigation area to determine multi-source information; perform fuzzy set analysis on the irrigation factors, mine the confidence matrix, and supervise the training of the irrigation decision model, wherein the confidence matrix is ​​the support degree of each irrigation factor for the irrigation decision, the irrigation decision model includes a multi-source fusion layer, an irrigation support layer and a conflict balance layer, and the irrigation decision model is placed on the cloud computing end; transmit the multi-source information to the irrigation decision model, perform de-heterogeneity and matrix operation decision, and determine the irrigation strategy; carry out an intelligent irrigation decision system, respond to the irrigation strategy, and control the irrigation equipment.

[0007] In a possible implementation, a data collection node is determined and the following processing is performed:

[0008] Determine the growth stage of the target crop in the irrigation area, initialize and set a preset time interval, wherein the preset interval settings of each growth stage are different; determine the lower-level acquisition nodes based on the upper-level acquisition nodes and the preset time interval, and determine the variable trend of the interval irrigation factor characteristics; perform expected bias measurement on the variable trend, compensate the lower-level acquisition nodes, and determine the data acquisition nodes.

[0009] In a possible implementation, the irrigation decision model includes a multi-source fusion layer, which performs the following processing:

[0010] Based on the irrigation factors, a probability allocation matrix is ​​determined, wherein the probability allocation matrix represents the basic probability of each irrigation factor for irrigation decision; based on the probability allocation matrix, a self-attention mechanism is set; based on the self-attention mechanism, a multi-source fusion layer is supervised for training.

[0011] In a possible implementation, the irrigation decision model includes an irrigation support layer, which performs the following processing:

[0012] Determine irrigation indicators, which are control factors of irrigation equipment; calculate similarities between the irrigation indicators based on a distance function method, and determine indicator weight distribution; combine the probability allocation matrix with the indicator weight distribution to construct the confidence matrix, which represents the degree of support of each irrigation factor for irrigation decision-making; and construct a decision support layer based on the confidence matrix and the basis for irrigation decision-making.

[0013] In a possible implementation, the irrigation decision basis is determined and the following processing is performed:

[0014] For the target crops in the irrigation area, based on the correlation between plant phenotypic characteristics and tuber development, first correlation information is determined; the fertilizer and water demand rules in the growth stage are determined to determine second correlation information; and the first correlation information and the second correlation information are fitted to determine the basis for irrigation decision.

[0015] In a possible implementation, the irrigation strategy is determined by performing the following processing:

[0016] The multi-source information is transmitted to the multi-source fusion layer to perform data de-heterogeneity and determine homogeneous multi-source data; the homogeneous multi-source data is traversed and the initialization irrigation factor characteristics are determined based on the self-attention mechanism; the initialization irrigation factor characteristics are homogenously fused and decision probability is allocated, and the characteristics are transferred to the irrigation support layer to perform probability calculation, determine the strategy points, and optimize the strategy points based on the conflict balance layer to determine the irrigation strategy.

[0017] In a possible implementation, the probability calculation is performed, the strategy points are determined, and the strategy points are optimized based on the conflict balance layer, the irrigation strategy is determined, and the following processing is performed:

[0018] Based on the decision probability, by matching the irrigation indicators, indicator decisions are made to determine the strategy points, wherein the strategy points correspond to the irrigation indicators one by one; the strategy points are traversed, and based on the indicator weight distribution, conflict judgment and iterative optimization of the strategy point combination are performed to determine the irrigation strategy; wherein the conflict iterative optimization is based on the normalization coefficient, and the priority is adjusted according to the indicator weight distribution constraint, and the normalization coefficient is a measure of the conflict degree of the strategy point.

[0019] This application also provides an irrigation decision optimization platform based on multi-source information fusion, including:

[0020] A data acquisition node determination module, the data acquisition node determination module is used to determine the data acquisition node, the data acquisition node is determined based on the preset time interval and the interval irrigation factor characteristics as the compensation amount; a data acquisition module, the data acquisition module is used to trigger the source end acquisition device based on the data acquisition node, synchronously collect data for the irrigation area, and determine multi-source information; an irrigation decision model training module, the irrigation decision model training module is used to perform fuzzy set analysis on the irrigation factors, mine the confidence matrix, and supervise the training of the irrigation decision model, wherein the confidence matrix is ​​the support degree of each irrigation factor for the irrigation decision, the irrigation decision model includes a multi-source fusion layer, an irrigation support layer and a conflict balance layer, and the irrigation decision model is placed on the cloud computing end; an irrigation strategy determination module, the irrigation strategy determination module is used to transmit the multi-source information to the irrigation decision model, perform de-heterogeneity and matrix operation decision, and determine the irrigation strategy; an irrigation control module, the irrigation control module is used to carry an intelligent irrigation decision system, respond to the irrigation strategy, and control the irrigation equipment.

[0021] The irrigation decision optimization method and platform of multi-source information fusion proposed in this application first determine the data acquisition node, which is based on the preset time interval and the interval irrigation factor characteristics as the compensation amount. Then, based on the data acquisition node, the source end acquisition equipment is triggered to synchronously collect data in the irrigation area to determine the multi-source information. Then, fuzzy set analysis is performed on the irrigation factors, and the confidence matrix is ​​mined to supervise the training of the irrigation decision model, wherein the confidence matrix is ​​the support degree of each irrigation factor for the irrigation decision. The irrigation decision model includes a multi-source fusion layer, an irrigation support layer and a conflict balance layer. The irrigation decision model is placed on the cloud computing end, and then the multi-source information is transmitted to the irrigation decision model, de-heterogeneity and matrix operation decisions are performed, and the irrigation strategy is determined. Finally, an intelligent irrigation decision system is installed to respond to the irrigation strategy and control the irrigation equipment, thereby achieving the technical effect of realizing intelligent and precise irrigation management. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the accompanying drawings of the embodiment of the present invention are briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the platform according to the embodiment of the present application. It should be understood that the previous or following operations are not necessarily performed precisely in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0023] Figure 1 A flowchart of an irrigation decision optimization method based on multi-source information fusion provided in an embodiment of the present application.

[0024] Figure 2 A schematic diagram of the structure of an irrigation decision optimization platform for multi-source information fusion provided in an embodiment of the present application.

[0025] Explanation of the reference numerals: data collection node determination module 10 , data collection module 20 , irrigation decision model training module 30 , irrigation strategy determination module 40 , irrigation management and control module 50 . DETAILED DESCRIPTION

[0026] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0027] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0028] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, platform, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0029] The present application embodiment provides an irrigation decision optimization method based on multi-source information fusion, such as Figure 1 As shown, the method includes:

[0030] Step S100, determining the data collection node, the data collection node is determined based on the preset time interval as the benchmark, and the interval irrigation factor characteristics as the compensation amount. Specifically, first, according to the growth cycle of the crop (such as potatoes), irrigation requirements and other factors, a preset time interval is determined as the basic cycle for data collection, and then, according to the changing law and demand of the irrigation factor characteristics (irrigation factor characteristics that affect the data collection interval, such as soil moisture, temperature, rainfall, crop physiological indicators, etc.), the preset time interval is appropriately adjusted to form the final collection time interval. These irrigation factor characteristics are used as compensation amounts to optimize the accuracy and timeliness of data collection. Based on the adjusted collection time interval, the data collection node, that is, the time node of data collection, is determined to guide the actual data collection work.

[0031] In a possible implementation, the data collection node is determined, and step S100 further includes step S110, determining the growth stage of the target crop in the irrigation area, and initializing the preset time interval, wherein the preset interval settings of each growth stage are different. Specifically, the current growth stage of the target crop in the irrigation area (the type of crop that needs to be irrigated in the irrigation area) is determined by means of historical data or expert system. Exemplarily, the target crop can be a crop type such as potato, corn, cotton, etc. The growth stage of the crop generally includes the seedling stage, the growth stage, the flowering stage, the fruiting stage, etc., and the water demand and irrigation strategy of each stage are different. According to the growth stage of the target crop, the preset time interval is initialized and set, and the preset time interval refers to the preset data collection cycle without considering the change of the irrigation factor characteristics. This cycle is determined according to factors such as the growth cycle of the crop, the irrigation demand, and the performance of the irrigation system. There are differences in the preset time intervals of different growth stages because the crops have different water demands and irrigation strategies at different growth stages. Step S120, according to the upper acquisition node, based on the preset time interval, the lower acquisition node is determined, and the variable trend of the interval irrigation factor characteristics is determined. Specifically, the upper acquisition node refers to the time point of the last data acquisition, which is the basis for determining the next data acquisition time (lower acquisition node). The time point of the next data acquisition is determined according to the time point of the last acquisition and the preset time interval. At the same time, by continuously observing and analyzing the data collected last time, the variable trend of the irrigation factor characteristics, that is, the change trend of the irrigation factor characteristics over time, is determined. These trends reflect the actual demand of crops for water and the performance of the irrigation system. Step S130, the variable trend is expected to be measured, the lower acquisition node is compensated, and the data acquisition node is determined. Specifically, according to the variable trend and the irrigation demand of the crop, the expected change of the irrigation factor characteristics is measured, including considering the influence of factors such as the crop growth stage, weather conditions, and soil type on the change of the irrigation factor characteristics. Based on the result of the expected bias measurement, the data acquisition time of the lower acquisition node is compensated and adjusted. If the change trend of the expected irrigation factor characteristics indicates that more frequent or sparse data collection is required, the data collection time of the lower time node can be adjusted accordingly. After compensation adjustment, the final data collection time node is determined. Data collection work is carried out according to the adjusted time to provide accurate and timely irrigation factor data. This implementation method optimizes the efficiency and accuracy of data collection by setting up upper and lower collection nodes and compensating according to variable trends. Crops at different growth stages have different water requirements and irrigation strategies. Through this implementation method, the data collection cycle can be flexibly adjusted according to the growth stage of the crop and the changing trend of the irrigation factor characteristics, reducing unnecessary data collection work and improving the overall performance of the irrigation system.

[0032] Step S200, based on the data collection node, trigger the source end collection device, synchronously collect data for the irrigation area, and determine the multi-source information. Specifically, at each data collection time node, trigger the source end collection device (equipment or system for collecting irrigation factor data, such as sensors, monitoring stations, etc.) to start working, so that all source end collection devices synchronously collect data at the specified time node, and obtain multi-source information in the irrigation area, for example, meteorological data (rainfall, light radiation, ET0, etc.), soil data (soil moisture, soil nutrients, soil bulk density, field support, pH, etc.), plant data (canopy temperature, water content, fresh dry weight, leaf water potential, etc.).

[0033] Step S300, fuzzy set analysis is performed on the irrigation factors, a confidence matrix is ​​mined, and an irrigation decision model is supervised and trained, wherein the confidence matrix is ​​the support degree of each irrigation factor for the irrigation decision, and the irrigation decision model includes a multi-source fusion layer, an irrigation support layer, and a conflict balance layer, and the irrigation decision model is placed on the cloud computing end. Specifically, fuzzy set analysis is performed on the collected irrigation factor data to deal with the uncertainty and ambiguity in the data. Through fuzzy set analysis, the degree of influence of each irrigation factor on the irrigation decision is mined to form a confidence matrix. The confidence matrix and other relevant information (such as historical irrigation data, crop growth model, etc.) are used to supervise the training of the irrigation decision model. The model includes a multi-source fusion layer, an irrigation support layer, and a conflict balance layer, and is placed on the cloud computing end for efficient processing. It can receive multi-source information and output optimized irrigation strategies.

[0034] In a possible implementation, the irrigation decision model includes a multi-source fusion layer, and step S300 further includes step S310, determining a probability distribution matrix based on the irrigation factors, wherein the probability distribution matrix represents the basic probability of each irrigation factor for the irrigation decision. Specifically, the irrigation factor data is extracted from the multi-source information obtained from the data acquisition node, and the irrigation factor data is preprocessed, including data cleaning, missing value processing, outlier detection, etc., to ensure the accuracy and reliability of the data. Based on the preprocessed irrigation factor data, the basic probability of each irrigation factor for the irrigation decision is determined using a statistical method or a machine learning algorithm, that is, each irrigation factor is evaluated separately to determine its contribution to the irrigation demand. The basic probability of each irrigation factor is organized into a matrix form to form a probability distribution matrix, wherein each row of the matrix represents an irrigation factor, and each column represents a different irrigation decision result (such as irrigation, no irrigation, adjustment of irrigation amount, etc.), and the elements in the matrix represent the basic probability of the decision result corresponding to the irrigation factor. Step S320, based on the probability distribution matrix, a self-attention mechanism is set. Specifically, the self-attention mechanism is a deep learning technology that allows the model to dynamically focus on information at different positions in the input sequence when processing input data. This mechanism is achieved by calculating the correlation scores between different positions in the input sequence. In the irrigation decision model, a self-attention layer is designed, which receives the probability allocation matrix as input and outputs a weighted feature vector, which contains the comprehensive contribution of each irrigation factor to the irrigation decision. Through supervised training, the parameters in the self-attention layer are adjusted so that the model can accurately capture the interactions and relationships between the irrigation factors, thereby improving the accuracy of irrigation decisions. Step S330, based on the self-attention mechanism, supervise the training of the multi-source fusion layer. Specifically, in the irrigation decision model, a multi-source fusion layer is designed, which receives the weighted feature vector of the self-attention layer and fuses it with information from other sources (such as historical irrigation data, weather forecasts, etc.). In order to evaluate the performance of the model, a loss function is defined, which compares the irrigation decision output by the model with the actual irrigation demand. The model is supervised and trained using a labeled training dataset (i.e., a dataset containing the actual irrigation demand). During the training process, the parameters of the multi-source fusion layer and other layers are continuously adjusted to minimize the loss function. During the training process, the performance of the model is regularly evaluated using the validation dataset, and the parameters are tuned based on the evaluation results.This implementation method integrates information from different sources (such as soil moisture, crop physiological indicators, weather forecasts, etc.), so that the model can have a more comprehensive understanding of the conditions in the irrigation area and make more accurate irrigation decisions. The self-attention mechanism can dynamically focus on information at different positions in the input sequence and capture the interactions and relationships between various irrigation factors, which helps the model to understand irrigation needs more accurately and make more refined irrigation decisions, thereby achieving the technical effect of improving the accuracy and efficiency of irrigation decisions.

[0035] In a possible implementation, the irrigation decision model includes an irrigation support layer, and step S300 further includes step S340, determining an irrigation index, which is a control factor of the irrigation equipment. Specifically, the irrigation factor data is extracted from the multi-source information obtained by the data acquisition node, and the specific irrigation demand of the irrigation area is analyzed according to factors such as the growth cycle of the crop, soil conditions, and climate conditions. Based on the irrigation demand analysis and the irrigation factor data, a series of irrigation indicators are determined, which are control factors of the irrigation equipment and are used to guide irrigation decisions. For example, the irrigation indicators may include irrigation volume, irrigation frequency, irrigation time, etc. Step S350, based on the distance function method, the similarity between the irrigation indicators is calculated to determine the indicator weight distribution. Specifically, according to the characteristics of the irrigation indicators, a distance function (such as Euclidean distance, Manhattan distance, etc.) is selected to calculate the similarity between the indicators. The higher the similarity, the stronger the correlation between the two irrigation indicators in the irrigation decision. Based on the similarity calculation result, a weighted average or entropy weight method is used to determine the weight distribution of each irrigation indicator, and the weight reflects the importance of each irrigation indicator in the irrigation decision. Step S360, combining the probability distribution matrix with the indicator weight distribution, constructing the confidence matrix, wherein the confidence matrix represents the support degree of each irrigation factor for the irrigation decision. Specifically, the probability distribution matrix obtained in step S310 is combined with the indicator weight distribution determined in step S350 to construct a confidence matrix. The elements in the confidence matrix represent the support degree of each irrigation factor for the irrigation decision, that is, the confidence of each factor in the decision. Step S370, based on the confidence matrix and combined with the irrigation decision basis, construct a decision support layer. Specifically, the basis for irrigation decision is analyzed according to factors such as crop growth requirements, soil conditions, and climate conditions. The confidence matrix is ​​combined with the irrigation decision basis to construct a decision support layer. The decision support layer is used to provide reference information for irrigation decisions and help make more accurate irrigation decisions. This implementation method clarifies the control elements of irrigation equipment by determining irrigation indicators, provides specific reference standards for irrigation decisions, and introduces a confidence matrix to enable the model to quantify the support degree of each irrigation factor for irrigation decisions, providing a more reliable basis for decision-making and improving the accuracy of irrigation decisions.

[0036] In a possible implementation, the irrigation decision basis is determined, and step S370 further includes step S371, for the target crop in the irrigation area, based on the correlation between the plant phenotypic characteristics and tuber development, determining the first association information. Specifically, the target crop in the irrigation area is regularly collected for plant phenotypic data, which includes but is not limited to the height, number of leaves, leaf area index, chlorophyll content, etc. of the plant. At the same time, the development of the tubers (such as potatoes, etc.) of the target crop is monitored, and key indicators such as the size, weight, and number of the tubers are recorded. The correlation between the plant phenotypic characteristics and the tuber development is analyzed by using statistical or machine learning methods. This correlation is manifested as a certain trend or pattern. For example, when the number of leaves increases to a certain threshold, the development speed of the tuber will accelerate. Based on the above analysis, the specific association information between the plant phenotypic characteristics and the tuber development is determined, that is, the first association information. Step S372, determine the fertilizer and water demand law of the growth stage, and determine the second association information. Specifically, according to the growth cycle of the target crop, it is divided into different growth stages, such as seedling stage, growth stage, flowering stage, fruiting stage, etc. In each growth stage, the fertilizer and water demand law of the target crop is analyzed, including the water demand, the type and quantity of fertilizer, etc. Based on the above analysis, the fertilizer and water demand law of the target crop at different growth stages is determined, that is, the second associated information. Step S373, fit the first associated information and the second associated information to determine the basis for irrigation decision. Specifically, the first associated information and the second associated information are integrated to form a complete data set. Using machine learning or statistical modeling methods, a model that can predict irrigation demand is constructed, and this model can comprehensively consider factors such as plant phenotypic characteristics, tuber development conditions, and fertilizer and water demand laws. Use historical data to train and verify the model to ensure the accuracy and reliability of its prediction results. Based on the trained model, the basis for irrigation decision is determined, that is, the irrigation strategy that should be adopted under different conditions (such as different plant phenotypic characteristics, tuber development conditions and growth stages). This implementation method makes irrigation decisions by correlating plant phenotypic characteristics with tuber development and analyzing fertilizer and water demand patterns at different growth stages. It avoids the problem of irrigation decision bias caused by a single factor and is conducive to more accurate guidance of irrigation operations.

[0037] Step S400, the multi-source information is transmitted to the irrigation decision model, de-isomerization and matrix operation decision are performed, and the irrigation strategy is determined. Specifically, the collected multi-source information is input into the trained irrigation decision model, and the received multi-source information is de-isomerized inside the model, that is, data of different formats and different sources are converted into a unified format. Matrix operations are performed using the confidence matrix and other relevant information to output the irrigation strategy.

[0038] In a possible implementation, the irrigation strategy is determined, and step S400 further includes step S410, transmitting the multi-source information to the multi-source fusion layer, performing data de-heterogeneity, and determining homogeneous multi-source data. Specifically, the multi-source information from different data sources (such as soil moisture sensors, weather stations, crop physiological monitoring equipment, etc.) is transmitted to the multi-source fusion layer. Since different data sources use different data formats, units, and encoding methods, it is necessary to pre-process these data to eliminate data heterogeneity, including data format conversion, unit unification, missing value processing, outlier detection and processing, etc. After de-heterogeneity processing, homogeneous multi-source data with the same format, unit, and encoding method are obtained. Step S420, traverse the homogeneous multi-source data, and determine the initialization irrigation factor features based on the self-attention mechanism. Specifically, the homogeneous multi-source data is traversed to extract the irrigation factor features in each data point. Based on the self-attention mechanism, each irrigation factor feature is weighted to highlight the key features and suppress the secondary features. The self-attention mechanism can dynamically adjust the weights between different features, so as to more accurately reflect the actual impact of the irrigation factor. After being processed by the self-attention mechanism, the initialization irrigation factor features are obtained, which are the basic inputs of the irrigation support layer. Step S430, the initialization irrigation factor features are homologously fused, and the decision probability is allocated, and the features are transferred to the irrigation support layer, and the probability calculation is performed to determine the strategy point and optimize the strategy point based on the conflict balance layer to determine the irrigation strategy. Specifically, the initialization irrigation factor features are homologously fused, that is, the features with the same or similar attributes are merged or weighted to obtain a more accurate and comprehensive irrigation factor feature set. Based on the fused irrigation factor feature set, the decision probability is allocated to each irrigation factor, that is, the support degree or probability of each factor for the irrigation decision is calculated. The decision probability allocation result is transferred to the irrigation support layer, which performs probability calculation based on the irrigation factor feature set and the decision probability allocation result to determine the irrigation strategy point. The irrigation strategy points are optimized based on the conflict balance layer, which can identify and handle conflicts and contradictions between different irrigation indicators to ensure the accuracy and feasibility of the irrigation strategy. After the strategy points are optimized, the final irrigation strategy is obtained, which is used to guide the management and control of irrigation equipment. This implementation method eliminates the differences in format, unit and encoding between different data sources through data de-heterogeneity, dynamically adjusts the weights between different features through the self-attention mechanism, highlights key features and suppresses secondary features, accurately reflects the actual impact of irrigation factors, merges or weights features with the same or similar attributes, improves the accuracy and comprehensiveness of the irrigation factor feature set, and achieves the technical effect of improving the accuracy and reliability of irrigation decisions.

[0039] In a possible implementation, the probability calculation is performed, the strategy point is determined, and the strategy point is optimized based on the conflict balance layer to determine the irrigation strategy. Step S430 further includes step S431, based on the decision probability, by matching the irrigation index, making an indicator decision, and determining the strategy point, wherein the strategy point corresponds to the irrigation index one-to-one. Specifically, the decision probability corresponding to each irrigation factor is extracted from the data transferred from the multi-source fusion layer to the irrigation support layer, and these decision probabilities reflect the degree of support of each irrigation factor for the irrigation decision. The extracted decision probability is matched with the preset irrigation index, and based on the matching result, a decision is made for each irrigation index, for example, the decision probability is weighted to obtain a comprehensive score or priority for each irrigation index. According to the comprehensive score or priority, the irrigation strategy point is determined, and the strategy point corresponds to the irrigation index one-to-one, indicating the irrigation measures to be taken under specific conditions. Step S432, traverse the strategy points, and based on the indicator weight distribution, perform conflict determination and iterative optimization of the strategy point combination to determine the irrigation strategy, wherein the conflict iterative optimization is based on the normalization coefficient, and the priority is adjusted by the indicator weight distribution constraint, and the normalization coefficient is a measure of the conflict degree of the strategy point. Specifically, the strategy points determined in step S431 are traversed to check the potential conflicts between them. Based on the indicator weight distribution, conflict determination is performed on each strategy point combination, that is, the similarity or conflict degree between the strategy points is calculated to identify possible contradictions or inconsistencies. For the strategy point combination with conflicts, iterative optimization is performed, including adjusting the priority of the strategy point, reallocating irrigation indicators or introducing new irrigation indicators, etc., to eliminate conflicts and achieve the optimal irrigation strategy. After iterative optimization, the final irrigation strategy is obtained, which is the optimal solution obtained after comprehensively considering all irrigation factors, irrigation indicators, and the weight distribution and potential conflicts between them. Among them, the normalization coefficient is a value used to measure the conflict degree of the strategy point, which is obtained by calculating the similarity or difference between the strategy points, and is used to guide the iterative optimization process. This implementation method eliminates conflicts between strategic points through conflict determination and iterative optimization, ensuring the feasibility and effectiveness of the irrigation strategy.

[0040] Step S500, an intelligent irrigation decision system is installed to respond to the irrigation strategy and control the irrigation equipment. Specifically, the trained irrigation decision model is installed in the intelligent irrigation decision system, and the intelligent irrigation decision system controls the irrigation equipment (such as sprinkler equipment, drip irrigation equipment, etc.) in real time according to the output irrigation strategy, including adjusting the irrigation amount, irrigation time and other parameters to meet the irrigation needs of crops. The embodiment of the present application integrates multiple data sources, uses fuzzy set analysis and confidence matrix and other methods to explore the support degree of each irrigation factor for irrigation decision, and constructs an irrigation decision model including a multi-source fusion layer, an irrigation support layer and a conflict balance layer. The model is placed on the cloud computing end, and can process and analyze multi-source information in real time, determine the optimal irrigation strategy, and use the intelligent irrigation decision system to accurately control the irrigation equipment and other technical means, thereby achieving the technical effect of realizing intelligent and precise irrigation management.

[0041] In the above, refer to Figure 1 The irrigation decision optimization method based on multi-source information fusion according to an embodiment of the present invention is described in detail. Figure 2 An irrigation decision optimization platform based on multi-source information fusion according to an embodiment of the present invention is described.

[0042] The irrigation decision optimization platform of multi-source information fusion according to the embodiment of the present invention is used to solve the technical problems of lack of intelligence and low decision accuracy in existing irrigation decisions, and achieve the technical effect of realizing intelligent and precise irrigation management. The irrigation decision optimization platform of multi-source information fusion includes: a data acquisition node determination module 10, a data acquisition module 20, an irrigation decision model training module 30, an irrigation strategy determination module 40, and an irrigation control module 50.

[0043] The data acquisition node determination module 10 is used to determine the data acquisition node, and the data acquisition node is determined based on the preset time interval and the interval irrigation factor characteristics as the compensation amount; the data acquisition module 20 is used to trigger the source end acquisition device based on the data acquisition node, synchronously collect data from the irrigation area, and determine the multi-source information; the irrigation decision model training module 30 is used to perform fuzzy set analysis on the irrigation factors, mine the confidence matrix, and supervise the training of the irrigation decision model, wherein the confidence matrix is ​​the support degree of each irrigation factor for the irrigation decision, and the irrigation decision model includes a multi-source fusion layer, an irrigation support layer and a conflict balance layer, and the irrigation decision model is placed on the cloud computing end; the irrigation strategy determination module 40 is used to transmit the multi-source information to the irrigation decision model, perform de-heterogeneity and matrix operation decisions, and determine the irrigation strategy; the irrigation control module 50 is used to carry the intelligent irrigation decision system, respond to the irrigation strategy, and control the irrigation equipment.

[0044] The specific configuration of the data acquisition node determination module 10 will be described in detail below. As described above, the data acquisition node is determined, and the data acquisition node determination module 10 may further include: a preset time interval setting unit is used to determine the growth stage of the target crop in the irrigation area, and initialize the preset time interval, wherein the preset interval settings of each growth stage are different; a lower-level acquisition node determination unit is used to determine the lower-level acquisition node based on the upper-level acquisition node and the preset time interval, and determine the variable trend of the interval irrigation factor feature; a lower-level acquisition node compensation unit is used to perform expected bias measurement on the variable trend, compensate the lower-level acquisition node, and determine the data acquisition node.

[0045] The specific configuration of the irrigation decision model training module 30 will be described in detail below. As described above, the irrigation decision model includes a multi-source fusion layer, and the irrigation decision model training module 30 may further include: a probability distribution matrix determination unit for determining a probability distribution matrix based on the irrigation factors, wherein the probability distribution matrix represents the basic probability of each irrigation factor for the irrigation decision; a self-attention mechanism setting unit for setting a self-attention mechanism based on the probability distribution matrix; and a multi-source fusion layer training unit for supervising the training of the multi-source fusion layer based on the self-attention mechanism.

[0046] Among them, the irrigation decision model includes an irrigation support layer, and the irrigation decision model training module 30 may further include: an irrigation index determination unit is used to determine the irrigation index, and the irrigation index is a control factor of the irrigation equipment; the index weight distribution determination unit is used to calculate the similarity between the irrigation indexes based on the distance function method, and determine the index weight distribution; the confidence matrix construction unit is used to combine the probability distribution matrix and the index weight distribution to construct the confidence matrix, and the confidence matrix represents the support degree of each irrigation factor for the irrigation decision; the decision support layer construction unit is used to construct the decision support layer based on the confidence matrix and the irrigation decision basis.

[0047] Among them, to determine the basis for irrigation decision, the decision support layer construction unit may further include: a first association information determination subunit is used to determine the first association information for the target crop in the irrigation area based on the correlation between the plant phenotypic characteristics and tuber development; a second association information determination subunit is used to determine the fertilizer and water demand law in the growth stage and determine the second association information; an irrigation decision basis determination subunit is used to fit the first association information and the second association information to determine the basis for irrigation decision.

[0048] The specific configuration of the irrigation strategy determination module 40 will be described in detail below. As described above, to determine the irrigation strategy, the irrigation strategy determination module 40 may further include: a data de-heterogeneity unit for transmitting the multi-source information to the multi-source fusion layer, performing data de-heterogeneity, and determining homogeneous multi-source data; an initialization irrigation factor feature determination unit for traversing the homogeneous multi-source data, and determining the initialization irrigation factor feature based on the self-attention mechanism; an irrigation strategy determination unit for homogenously fusing the initialization irrigation factor feature, and performing decision probability allocation, transferring to the irrigation support layer, performing probability calculation, determining the strategy point, and optimizing the strategy point based on the conflict balance layer, and determining the irrigation strategy.

[0049] Among them, the probability calculation is performed, the strategy points are determined, and the strategy points are optimized based on the conflict balance layer to determine the irrigation strategy. The irrigation strategy determination unit may further include: a strategy point determination subunit is used to make an indicator decision based on the decision probability and by matching the irrigation indicators to determine the strategy points, wherein the strategy points correspond to the irrigation indicators one by one; the irrigation strategy determination subunit is used to traverse the strategy points, and based on the indicator weight distribution, perform conflict judgment and iterative optimization of the strategy point combination to determine the irrigation strategy, wherein the conflict iterative optimization is based on the normalization coefficient, and the priority is adjusted according to the indicator weight distribution constraint, and the normalization coefficient is a measure of the conflict degree of the strategy point.

[0050] The irrigation decision optimization platform based on multi-source information fusion provided in the embodiment of the present invention can execute the irrigation decision optimization method based on multi-source information fusion provided in any embodiment of the present invention, and has the functional modules and beneficial effects corresponding to the execution method.

[0051] Although the present application makes various references to certain modules in the platform according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be implemented; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0052] The above specific implementation manner does not constitute a limitation to the protection scope of the present application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be performed in an order different from that in the embodiment and can still achieve the desired results. In addition, the process depicted in the accompanying drawings does not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. The irrigation decision optimization method based on multi-source information fusion is characterized by: The method comprises: Determine a data collection node, wherein the data collection node is determined based on a preset time interval as a reference and an interval irrigation factor characteristic as a compensation amount. Specifically, according to the growth cycle of the crop and the irrigation demand factor, a preset time interval is determined as a data collection cycle. Then, according to the change law and demand of the irrigation factor characteristic, the preset time interval is adjusted to form a final collection time interval. Based on the adjusted collection time interval, determine the data collection node, wherein the irrigation factor characteristic includes soil moisture, temperature, rainfall, and crop physiology; Based on the data collection node, the source end collection device is triggered to synchronously collect data for the irrigation area to determine multi-source information; Perform fuzzy set analysis on irrigation factors, mine confidence matrix, and supervise the training of irrigation decision model, wherein the confidence matrix is ​​the support degree of each irrigation factor for irrigation decision, the irrigation decision model includes multi-source fusion layer, irrigation support layer and conflict balance layer, and the irrigation decision model is placed on the cloud computing end; The multi-source information is transmitted to the irrigation decision model, de-heterogeneity and matrix operation decision are performed, and the irrigation strategy is determined, wherein the de-heterogeneity includes data format conversion, unit unification, missing value processing, and outlier detection and processing; Equipped with an intelligent irrigation decision-making system, it responds to the irrigation strategy and manages the irrigation equipment; Among them, determining the data collection nodes includes: Determine the growth stage of the target crop in the irrigation area, and initialize and set a preset time interval, wherein the preset interval settings for each growth stage are different; According to the upper acquisition node, based on the preset time interval, the lower acquisition node is determined, and the variable trend of the interval irrigation factor characteristic is determined. The upper acquisition node refers to the time point of the last data collection and is the basis for determining the next data collection time; The expected bias measurement is performed on the variable trend, the lower-level acquisition nodes are compensated, and the data acquisition nodes are determined. Specifically, the expected change of the irrigation factor characteristics is measured according to the variable trend and the irrigation demand of the crop, including considering the influence of the crop growth stage, weather conditions, and soil type factors on the change of the irrigation factor characteristics, and based on the result of the expected bias measurement, the data acquisition time of the lower-level acquisition nodes is compensated and adjusted; The irrigation decision model contains a multi-source fusion layer, including: Based on the irrigation factors, a probability distribution matrix is ​​determined, wherein each row of the matrix represents an irrigation factor, each column represents a different irrigation decision result, and the elements in the matrix represent the probability of the decision result corresponding to the irrigation factor; Based on the probability allocation matrix, a self-attention mechanism is set. Specifically, in the irrigation decision model, a self-attention layer is designed to receive the probability allocation matrix as input and output a weighted feature vector, which includes the comprehensive contribution of each irrigation factor to the irrigation decision; Based on the self-attention mechanism, supervise the training of the multi-source fusion layer. Specifically, in the irrigation decision model, design a multi-source fusion layer, receive the weighted feature vector of the self-attention layer, and fuse it with the historical irrigation data and weather forecast, define a loss function, which compares the irrigation decision output by the model with the actual irrigation demand, and use the labeled training data set to supervise the model. During the training process, continuously adjust the parameters of the multi-source fusion layer and other layers to minimize the loss function; The irrigation decision model comprises an irrigation support layer, including: Determining an irrigation index, wherein the irrigation index is a control factor of the irrigation equipment; Based on the distance function method, the similarity between the irrigation indicators is calculated to determine the indicator weight distribution; Combining the probability distribution matrix with the indicator weight distribution, constructing the confidence matrix, wherein the confidence matrix represents the support degree of each irrigation factor for the irrigation decision; Based on the confidence matrix and combined with the irrigation decision basis, a decision support layer is constructed; Determining the irrigation strategy includes: Transmitting the multi-source information to the multi-source fusion layer to perform data de-heterogeneity and determine homogeneous multi-source data; Traversing the homogeneous multi-source data, and determining the initialization irrigation factor features based on the self-attention mechanism; The initialized irrigation factor features are homogenously fused and decision probability is allocated, which are transferred to the irrigation support layer for probability calculation, the strategy points are determined, and the strategy points are optimized based on the conflict balance layer to determine the irrigation strategy; The method of performing probability calculation, determining the strategy points, optimizing the strategy points based on the conflict balance layer, and determining the irrigation strategy includes: Based on the decision probability, by matching the irrigation index, an index decision is made to determine a strategy point, wherein the strategy point corresponds to the irrigation index one by one; Traversing the strategy points, performing conflict determination and iterative optimization of strategy point combinations based on the indicator weight distribution, and determining the irrigation strategy; The conflict iterative optimization is based on the normalization coefficient and the priority is adjusted based on the indicator weight distribution constraint. The normalization coefficient is a measure of the conflict degree of the strategy point.

2. The irrigation decision optimization method based on multi-source information fusion according to claim 1, characterized in that: Determine the basis for irrigation decisions, including: For the target crop in the irrigation area, determining first correlation information based on the correlation between plant phenotypic characteristics and tuber development; Determine the fertilizer and water demand pattern during the growth stage and determine the second related information; The first association information and the second association information are fitted to determine an irrigation decision basis.

3. The irrigation decision optimization system based on multi-source information fusion is characterized by: The system is used to implement the irrigation decision optimization method of multi-source information fusion according to any one of claims 1-2, and the system comprises: A data collection node determination module, the data collection node determination module is used to determine the data collection node, the data collection node is determined based on a preset time interval as a benchmark and the interval irrigation factor characteristics as a compensation amount, specifically, according to the crop growth cycle and irrigation demand factors, a preset time interval is determined as the data collection cycle, then, according to the change law and demand of the irrigation factor characteristics, the preset time interval is adjusted to form a final collection time interval, and the data collection node is determined based on the adjusted collection time interval, wherein the irrigation factor characteristics include soil moisture, temperature, rainfall, and crop physiology; A data acquisition module, the data acquisition module is used to trigger the source end acquisition device based on the data acquisition node, synchronously collect data for the irrigation area, and determine multi-source information; An irrigation decision model training module, which is used to perform fuzzy set analysis on irrigation factors, mine confidence matrices, and supervise the training of irrigation decision models, wherein the confidence matrix is ​​the degree of support of each irrigation factor for irrigation decision-making. The irrigation decision model includes a multi-source fusion layer, an irrigation support layer, and a conflict balance layer. The irrigation decision model is placed on the cloud computing end; An irrigation strategy determination module, the irrigation strategy determination module is used to transmit the multi-source information to the irrigation decision model, perform de-heterogeneity and matrix operation decision, and determine the irrigation strategy, wherein de-heterogeneity includes data format conversion, unit unification, missing value processing, and outlier detection and processing; The irrigation control module is used to carry an intelligent irrigation decision-making system, respond to the irrigation strategy, and control the irrigation equipment.

Images