A spatiotemporal analysis method of agricultural irrigation load based on clustering algorithm

Abstract

The present invention discloses a method for spatiotemporal analysis of agricultural irrigation load based on clustering algorithm, which relates to the technical field of load analysis. The method for spatiotemporal analysis of agricultural irrigation load includes the following steps: obtaining agricultural irrigation load data of a distribution station area, and preprocessing the agricultural irrigation load data; iteratively clustering the preprocessed agricultural irrigation load data, summarizing data points with similar load characteristics to generate agricultural irrigation load characteristic groups; sequentially performing time dimension analysis and space dimension analysis on the agricultural irrigation load characteristic groups, identifying agricultural irrigation load patterns corresponding to the agricultural irrigation load characteristic groups; and executing power dispatching and allocation of distribution stations based on the identified agricultural irrigation load patterns. The present invention can identify agricultural irrigation load patterns in a specific area by fusing the analysis results of time and space dimensions, thereby adopting more sophisticated irrigation strategies and power management strategies, which not only improves irrigation efficiency, but also helps save energy and reduce operating costs.

Description

Technical Field

[0001] The invention relates to the technical field of load analysis, and in particular to a spatiotemporal analysis method of agricultural irrigation load based on a clustering algorithm. Background Art

[0002] Agricultural irrigation load mainly involves the water and electricity demand required in the agricultural irrigation process, which includes the power consumption of pumping stations in the irrigation system, the water flow in the water pipeline, and the water distribution in the irrigation area. Accurate calculation and analysis of agricultural irrigation load is crucial because it can not only help plan and use water resources and electricity more effectively, but also ensure that these resources are allocated and used in the most reasonable way, thereby improving irrigation efficiency and reducing resource waste.

[0003] Agricultural irrigation usually relies on power supply from rural distribution networks. Due to the complexity of agricultural terrain and the wide range of areas that require electric irrigation, the centralized power demand for irrigation pumps increases significantly every time the irrigation cycle arrives, which may cause the distribution network to face the risk of insufficient power supply capacity. If this situation is not managed, it will not only affect the efficiency of agricultural production, but also threaten the sustainable development of agriculture.

[0004] Therefore, it is extremely important to conduct spatiotemporal analysis of agricultural irrigation load. Through this analysis, we can understand in detail the demand for electricity and water resources at different times and locations, so as to formulate more effective irrigation strategies, which can not only optimize the allocation of resources and reduce production interruptions caused by insufficient resources, but also adjust the load distribution of the power grid by predicting peak periods to avoid overload. Summary of the invention

[0005] In view of the problems in the related art, the present invention proposes a spatiotemporal analysis method of agricultural irrigation load based on a clustering algorithm to overcome the above technical problems existing in the existing related art.

[0006] To this end, the specific technical solution adopted by the present invention is as follows:

[0007] A method for spatiotemporal analysis of agricultural irrigation load based on clustering algorithm, the method for spatiotemporal analysis of agricultural irrigation load comprises the following steps:

[0008] S1. Obtain the agricultural irrigation load data of the distribution station area and pre-process the agricultural irrigation load data;

[0009] S2, iteratively clustering the preprocessed agricultural irrigation load data, and summarizing the data points with similar load characteristics to generate agricultural irrigation load characteristic groups;

[0010] S3, sequentially performing time dimension analysis and space dimension analysis on the agricultural irrigation load characteristic groups to identify the agricultural irrigation load patterns corresponding to the agricultural irrigation load characteristic groups;

[0011] S4. Execute power dispatch and allocation in the distribution station area based on the identified agricultural irrigation load pattern.

[0012] Preferably, obtaining the agricultural irrigation load data of the distribution station area and preprocessing the agricultural irrigation load data includes the following steps:

[0013] S11, obtaining the agricultural irrigation load data of the distribution station area, calculating and recording the Euclidean distance between the target sample and the remaining samples in the data sample space to obtain the nearest sample;

[0014] S12, calculating the weighted average of the most recent samples as filling data to achieve missing value processing of agricultural irrigation load data;

[0015] S13. Scale the agricultural irrigation load data to a preset range using a linear function to achieve normalization processing of the agricultural irrigation load data.

[0016] Preferably, iterative clustering of the preprocessed agricultural irrigation load data to summarize data points with similar load characteristics to generate agricultural irrigation load characteristic groups includes the following steps:

[0017] S21, setting an initial cluster center, and allocating each data point of the agricultural irrigation load data to the nearest cluster center based on the initial cluster center;

[0018] S22, dynamically optimizing the cluster center in combination with the quantum oscillator model to obtain an optimized cluster center, and iteratively performing cluster merging processing and cluster splitting processing based on the optimized cluster center;

[0019] S23. When the change of the cluster center is less than a preset threshold or reaches the maximum number of iterations, the iteration process is stopped to generate several agricultural irrigation load feature groups with similar characteristics.

[0020] Preferably, dynamically optimizing the cluster center by combining the quantum oscillator model to obtain the optimized cluster center includes the following steps:

[0021] Randomly select several data points of agricultural irrigation load data as initial cluster centers, and record the quantum energy of the initial cluster centers;

[0022] The mean and standard deviation of the Gaussian sampling function are set, and the quantum energy of the initial cluster center is formed into a sampling region representing each energy state to simulate the high energy state of the quantum oscillator;

[0023] A new cluster center is selected in each sampling area by Gaussian sampling, and the quantum energy of the new cluster center is compared with the quantum energy of the initial cluster center;

[0024] The Gaussian function is superimposed on the sampling region of the maximum quantum energy state to form the ground state of the quantum oscillator;

[0025] Based on the comparison results of the ground state of the quantum oscillator combined with the quantum energy, the cluster center of the state with the maximum quantum energy is selected from the sampling area as the optimized cluster center.

[0026] Preferably, performing time dimension analysis and space dimension analysis on the agricultural irrigation load feature group in sequence, and identifying the agricultural irrigation load mode corresponding to the agricultural irrigation load feature group comprises the following steps:

[0027] S31. Predict the amount of agricultural irrigation required for crops during the growth stage, and analyze the high-frequency cycle of electricity based on the required amount of agricultural irrigation to obtain a time dimension analysis result;

[0028] S32, establish the power grid topology structure between the distribution station area location, water pump marking point location and agricultural irrigation area to analyze the power high load area and obtain the spatial dimension analysis result;

[0029] S33. The time dimension analysis result and the space dimension analysis result are integrated to identify the agricultural irrigation load pattern corresponding to the agricultural irrigation load feature group.

[0030] Preferably, predicting the amount of agricultural irrigation required for crops during the growth stage, and analyzing the high-frequency cycle of electricity based on the required amount of agricultural irrigation to obtain the time dimension analysis result includes the following steps:

[0031] S311, obtaining agricultural irrigation influencing factors of crops in the growth stage, and converting the agricultural irrigation influencing factors into node embedding representations respectively, wherein the agricultural irrigation influencing factors at least include soil type, climate conditions and crop type;

[0032] S312, selecting similar node embedding representations as neighbors based on the cosine similarity between the node embedding representations, and obtaining neighbor node embedding representations;

[0033] S313. Construct a preliminary graph structure based on neighbor node embedding representation, and learn the graph adjacency matrix to optimize the graph structure;

[0034] S314, using the optimized graph structure to integrate the time series information represented by the neighbor node embedding into each layer, and generate the correlation relationship between the factors affecting agricultural irrigation;

[0035] S315, combining the periodic characteristics of the agricultural irrigation influencing factors based on the correlation between the agricultural irrigation influencing factors, and outputting the agricultural irrigation amount required for the crops in the growth stage;

[0036] S316. Based on the amount of agricultural irrigation required by crops during their growth stage, the high-frequency cycle of agricultural irrigation load and the corresponding high-frequency cycle of power load are revealed to obtain the analysis result of the time dimension.

[0037] Preferably, establishing the power grid topology structure between the distribution station area location, the water pump marking point and the agricultural irrigation area to analyze the high power load area and obtaining the spatial dimension analysis result includes the following steps:

[0038] S321, respectively collecting the spatial position coordinates of the distribution station area, the water pump marking point and the agricultural irrigation area, and constructing a three-dimensional power grid model based on the spatial position coordinates;

[0039] S322, connecting the distribution station area location, the water pump mark point location and the associated area of ​​the agricultural irrigation area in the three-dimensional power grid model to obtain the power grid topology structure;

[0040] S323, dynamically simulate the agricultural irrigation behavior under different flow conditions in combination with the power grid topology, and generate several agricultural irrigation paths;

[0041] S324. Select the optimal agricultural irrigation path from the agricultural irrigation paths, and reveal the agricultural irrigation high load area and the corresponding power high load area based on the optimal agricultural irrigation path to obtain the spatial dimension analysis result.

[0042] Preferably, dynamically simulating the agricultural irrigation behavior under different flow conditions in combination with the power grid topology structure and generating a number of agricultural irrigation paths includes the following steps:

[0043] S3231, determining the topological nodes and topological edges in the power grid topological structure, the topological nodes corresponding to the distribution station area location, the water pump marking point location and the agricultural irrigation area, and the topological edges are the connections between the topological nodes;

[0044] S3232, setting the flow of each topological edge in the power grid topology structure as the initial flow;

[0045] S3233, taking the water pump mark point as the source point and the agricultural irrigation area as the sink point, searching for the augmenting path from the source point to the sink point, and there is residual flow on each augmenting path;

[0046] S3234, calculating the minimum residual flow on each augmenting path, and updating the flow of the topological edges on each augmenting path based on the minimum residual flow;

[0047] S3235. Repeat steps S3233-S3234 until no new augmenting path can be found and then stop searching; use the augmenting path corresponding to the flow of the topological edge within the preset range as the agricultural irrigation path to obtain several agricultural irrigation paths.

[0048] Preferably, selecting the optimal agricultural irrigation path from the agricultural irrigation paths comprises the following steps:

[0049] The distribution station area, water pump and agricultural irrigation area are respectively regarded as intelligent entities. The status of each intelligent entity includes the geographical location, current flow and power status, and the initial parameters are set for each intelligent entity.

[0050] Initialize the agent configuration, obtain the time step of each agent when the game round reaches the specified number of times, and generate the global and local situations of each agent;

[0051] Select the next action and reward for each agent based on the global and local situation;

[0052] Assign action behavior reward coefficients to the irrigation paths, and select the optimal irrigation path based on the action behavior reward coefficients of each irrigation path.

[0053] Preferably, the calculation formula for filling data is:

[0054]

[0055] In the formula, y represents the filling data; k represents the number of recent sample amounts; W i Represents the weight value of the i-th sample.

[0056] The beneficial effects of the present invention are:

[0057] 1. The present invention iteratively clusters the agricultural irrigation load through a cluster analysis algorithm to obtain distribution station areas and agricultural irrigation load feature groups with similar operating conditions, and uses a spatiotemporal analysis method to analyze the clustering results. By comparing the agricultural irrigation load analysis results at different times and spaces, it provides effective technical support and decision-making basis for solving distribution network problems.

[0058] 2. The present invention can more accurately predict the amount of irrigation required for various crops at different growth stages by acquiring and analyzing the influencing factors of the crop growth stage, using node embedding representation and graph structure optimization, and analyzing the high-frequency cycle of electricity based on the predicted crop irrigation demand, so as to effectively plan and adjust the allocation of power resources, thereby helping to prevent irrigation interruptions caused by insufficient power supply.

[0059] 3. The present invention can effectively identify high-load areas by accurately simulating the power grid topology and agricultural irrigation paths, thereby achieving optimal allocation of water and power resources, which helps to reduce resource waste and improve resource utilization efficiency. It can also dynamically simulate agricultural irrigation behavior under different flow conditions, which helps to predict and manage load changes in the power grid, thereby enhancing the stability and reliability of the power grid, and selecting the optimal agricultural irrigation path to ensure reasonable distribution of irrigation levels.

[0060] 4. By integrating the analysis results of time and space dimensions, the present invention can identify the agricultural irrigation load pattern in a specific area, so as to adopt more sophisticated irrigation strategies and power management strategies, which not only improves irrigation efficiency, but also helps to save energy and reduce operating costs. Accurate spatiotemporal analysis helps to achieve sustainable management of agricultural irrigation and power resources, reduces negative impacts on the environment, and can identify potential risks and challenges in advance. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0062] Figure 1 is a flow chart of a method for spatiotemporal analysis of agricultural irrigation load based on a clustering algorithm according to an embodiment of the present invention;

[0063] Figure 2 It is a schematic diagram of cluster centers of agricultural irrigation load in a certain area in one year in a method for spatiotemporal analysis of agricultural irrigation load based on a clustering algorithm according to an embodiment of the present invention;

[0064] Figure 3 This is a schematic diagram of clustering results of agricultural irrigation load in a certain area in one year in a method for spatiotemporal analysis of agricultural irrigation load based on a clustering algorithm according to an embodiment of the present invention;

[0065] Figure 4 This is a schematic diagram of cluster centers of agricultural irrigation loads in different substations in the same month in a method for spatiotemporal analysis of agricultural irrigation loads based on a clustering algorithm according to an embodiment of the present invention;

[0066] Figure 5 This is a schematic diagram of clustering results of agricultural irrigation loads in different substations in the same month in a method for spatiotemporal analysis of agricultural irrigation loads based on a clustering algorithm according to an embodiment of the present invention;

[0067] Figure 6 This is a schematic diagram of clustering centers of agricultural irrigation loads in different months in the same substation in a method for spatiotemporal analysis of agricultural irrigation loads based on a clustering algorithm according to an embodiment of the present invention;

[0068] Figure 7 It is a schematic diagram of clustering results of agricultural irrigation loads in different months in the same substation in a method for spatiotemporal analysis of agricultural irrigation loads based on a clustering algorithm according to an embodiment of the present invention. DETAILED DESCRIPTION

[0069] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention and are mainly used to illustrate the embodiments. They can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these contents, ordinary technicians in this field should be able to understand other possible implementation methods and advantages of the present invention.

[0070] According to an embodiment of the present invention, a method for spatiotemporal analysis of agricultural irrigation load based on a clustering algorithm is provided.

[0071] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to the agricultural irrigation load spatiotemporal analysis method based on clustering algorithm according to an embodiment of the present invention, the agricultural irrigation load spatiotemporal analysis method includes the following steps:

[0072] S1. Obtain the agricultural irrigation load data of the distribution station area and pre-process the agricultural irrigation load data.

[0073] The steps of obtaining the agricultural irrigation load data of the distribution station area and preprocessing the agricultural irrigation load data include:

[0074] S11, obtaining the agricultural irrigation load data of the distribution station area, calculating and recording the Euclidean distance between the target sample and the remaining samples in the data sample space to obtain the nearest sample;

[0075] S12, calculating the weighted average of the most recent samples as filling data to achieve missing value processing of agricultural irrigation load data;

[0076] S13. Scale the agricultural irrigation load data to a preset range using a linear function to achieve normalization processing of the agricultural irrigation load data.

[0077] It should be noted that the preprocessing in the present invention includes processing of missing values ​​based on the K nearest neighbor filling algorithm and data normalization, wherein the K nearest neighbor filling algorithm specifically includes:

[0078] For the sample that needs to be filled with data, first find the K samples closest to it in the sample space according to the given distance metric, and record the distances d1, d2, ..., d between these K samples and it. k ;

[0079] The distance metric selected by the present invention is the Euclidean distance, where two n-dimensional vectors x i =(x i1 ,x i2 ,...,x in ) and x j =(x j1 ,x j2 ,...,x jn ) is:

[0080]

[0081] The weighted average of the K nearest samples is calculated according to the distance value as the filling data y. The relationship between the filling data y and the K nearest samples is as follows:

[0082]

[0083] In the formula, y represents the filling data; k represents the number of recent sample amounts; W i represents the weight value of the i-th sample; W i The expression is:

[0084]

[0085] Linear function normalization is to scale the original data to the range of [0, 1] through linear transformation. The calculation formula is:

[0086]

[0087] In the formula, X represents the original data sequence, X min Represents the minimum value in the original data, X max Indicates the maximum value in the original data.

[0088] It should be noted that for load data, due to the variety of load curves, it may not conform to the Gaussian distribution assumption, and the normalization of load data is mainly used for the subsequent load curve clustering task. The normal property obtained after z-score standardization is not very helpful for the clustering algorithm. Therefore, for load data, z-score standardization is not an excellent choice, and linear function normalization is more appropriate.

[0089] S2. Iteratively cluster the preprocessed agricultural irrigation load data, and summarize the data points with similar load characteristics to generate agricultural irrigation load characteristic groups.

[0090] The iterative clustering of the preprocessed agricultural irrigation load data and the generation of agricultural irrigation load feature groups by summarizing the data points with similar load characteristics include the following steps:

[0091] S21. Set an initial cluster center, and assign each data point of the agricultural irrigation load data to the nearest cluster center based on the initial cluster center.

[0092] It should be noted that the clustering algorithm used in the present invention is the L-ISODATA algorithm (iterative self-organizing data analysis technology algorithm), which classifies data with similar load characteristics into the same category. The principle of the L-ISODATA algorithm is:

[0093] Regarding the problem of randomly selecting initial cluster centers, the optimization strategy adopted by the present invention is as follows: When selecting the first initial cluster center, it is still selected randomly. When selecting subsequent cluster centers, assuming that n initial cluster centers have been selected (0 < n < k), then when selecting the (n + 1)-th initial cluster center, the points farther away from the current n cluster centers have a higher probability of being selected. The main idea of this optimization idea is to make the k initial cluster centers as far apart as possible: When the randomly selected initial cluster centers are all close to each other, they should all be in the same cluster, and thus more iterations are required to disperse them into different clusters. However, when the initial cluster centers are far apart, the possibility that they belong to different clusters is greater, the convergence speed of the clustering algorithm is faster, and the clustering effect is also better.

[0094] The optimization algorithm for selecting initial cluster centers includes:

[0095] 1) Randomly select a sample from the sample set as the first initial cluster center;

[0096] 2) For each remaining sample x i , first calculate the distances x0 to each of the existing cluster centers, and record the shortest distance d(x i ), and its calculation formula is:

[0097] d(x i ) = min(x i - x0) 2 , i ∈ [1, n];

[0098] 3) For each sample x i , the probability that it is selected as the next cluster center is:

[0099]

[0100] Select the cluster center of this round according to the probability. The larger the d(x i ) of the sample, the higher the probability that it is selected;

[0101] 4) Return to step 2) until k initial cluster centers are selected.

[0102] Through this selection strategy of initial cluster centers, the distances between the selected initial cluster centers are farther, the possibility that they belong to different clusters is greater, and they are more likely to converge.

[0103] In order to solve the problem that the Euclidean distance in the original load curve input space is difficult to obtain the high-dimensional features of the load curve, the kernel method can be used for optimization. The core idea of ​​the kernel method is to map the data points in the original input space, that is, the load curve, to the high-dimensional feature space through nonlinear mapping, and then cluster them in the new high-dimensional feature space. The nonlinear mapping increases the probability that the data points are linearly separable. The clustering algorithm can calculate the distance between samples according to the high-dimensional features of the load curve to perform clustering, and the clustering effect is correspondingly improved.

[0104] However, if a mapping function is explicitly defined to directly map low-dimensional data to a high-dimensional space, solving the problem in the high-dimensional space will result in a cumbersome process, as it is necessary to complete the mapping of each data point, and both data sparsity and computational complexity will increase when performing calculations in the high-dimensional space.

[0105] The kernel function provides a convenient solution to the mapping relationship of the kernel method. Assume that X is the input space and H is a high-dimensional space. There exists a mapping function Φ(X) that maps a point x in the X space to a point h in the H space: h = Φ(x). If there exists a function K(x, z) such that for all x, z∈X, the condition K(x, z) = Φ(x) × Φ(z) is satisfied, then K(x, z) is called a kernel function, where Φ(x) × Φ(z) is the inner product of Φ(x) and Φ(z). Therefore, if the kernel function is found, the inner product of the two points after mapping can be directly solved in the original space.

[0106] The kernel method can bypass the mapping function Φ and solve it through the kernel function K(x, z), which makes the calculation much easier. If you first map it to a high-dimensional space and then perform the algorithm calculation, the amount of calculation is much greater than directly using the kernel function to calculate in the input space. In clustering algorithms, it is necessary to calculate the similarity of two samples, which can be represented by the inner product of the two samples or by the kernel function. Therefore, the kernel method can be perfectly applied to clustering algorithms.

[0107] In clustering algorithms, it is necessary to calculate the distance between two samples. The distance between two samples in high-dimensional space can be expressed by a kernel function:

[0108] d(x i ,x j )=||Φ(x i )-Φ(x j )|| 2

[0109] =Φ(x i ) 2 -2Φ(x i )Φ(x j )+Φ(xj ) 2

[0110] =K(x i ,x i )-2K(x i ,x j )+K(x j ,x j );

[0111] In addition, we need to calculate each sample x i To each cluster center μ j The distance in high-dimensional space is:

[0112] d(x i ,μ j )=||Φ(x i )-Φ(μ j )|| 2

[0113] =Φ(x i ) 2 -2Φ(x i )Φ(μ j )+Φ(μ j ) 2 ;

[0114] The cluster center in high-dimensional space can be expressed as:

[0115]

[0116] In the formula, C is the result of the cluster after division; x i represents the i-th sample; Φ represents the mapping function; μ j Represents the cluster center.

[0117] Sample x i and cluster centers Φ(μ j ) can be further solved as:

[0118]

[0119] From this derivation process, we can also see that in the calculation process of the clustering algorithm, we do not need to explicitly know the expression of the mapping function Φ(x), and we can use the kernel function to calculate it completely. In order to speed up the algorithm calculation and avoid repeated calculations, we can calculate the kernel function between each sample in the sample set in advance and store it in the kernel function matrix. When using it, we can directly obtain it by looking up the table. Commonly used kernel functions include:

[0120] Linear kernel function:

[0121] K(x,z)=xT z+c;

[0122] Polynomial kernel function:

[0123] K(x,z)=(ax T z+c) d ;

[0124] Gaussian kernel function:

[0125] ;

[0126] Using different kernel functions will result in different high-dimensional feature spaces after mapping, and the clustering results will also be different.

[0127] S22. Dynamically optimize the cluster center by combining the quantum oscillator model to obtain the optimized cluster center, and iteratively perform cluster merging and cluster splitting based on the optimized cluster center.

[0128] The method of dynamically optimizing the cluster center by combining the quantum oscillator model to obtain the optimized cluster center includes the following steps:

[0129] Randomly select several data points of agricultural irrigation load data as initial cluster centers, and record the quantum energy of the initial cluster centers;

[0130] The mean and standard deviation of the Gaussian sampling function are set, and the quantum energy of the initial cluster center is formed into a sampling region representing each energy state to simulate the high energy state of the quantum oscillator.

[0131] It should be noted that the quantum oscillator is a basic concept in quantum mechanics, which is used to describe a basic vibration system whose behavior conforms to the rules of quantum mechanics. The quantum oscillator model is a theoretical model that is often used to understand and study the quantum behavior of particles in potential energy wells. The quantum oscillator model is a quantum version of the oscillator concept in classical mechanics, the simplest example of which is the quantum simple harmonic oscillator.

[0132] It should be noted that the mean and standard deviation of the Gaussian sampling function are set, and the quantum energy of the initial cluster center is formed into a sampling area representing each energy state to simulate the high energy state of the quantum oscillator, including:

[0133] Each initial cluster center is assigned a quantum energy value that reflects its position and importance in the data set, where the quantum energy can be determined based on multiple factors, such as the data density of the cluster center, the distribution of surrounding data points, etc.

[0134] It is set as the quantum energy of each initial cluster center, which means that sampling will be performed around these energy values, and a suitable standard deviation value is selected to control the discreteness of the sampling. A smaller standard deviation means that the sampling is more concentrated on the mean, and a larger standard deviation makes the sampling more dispersed.

[0135] The quantum energy of the initial cluster center and the set Gaussian sampling function are used to form sampling regions of each energy state, which simulate the probability distribution of the quantum oscillator in different energy states.

[0136] Within each sampling region, Gaussian sampling is used to select new cluster center candidates. The simulation of high-energy states means that sampling is not limited to the lowest energy state (ground state) but also includes excited states, which helps to explore areas in the data space that may be neglected.

[0137] New cluster centers are selected in each sampling region by Gaussian sampling, and the quantum energies of the new cluster centers are compared with the quantum energies of the initial cluster centers.

[0138] The Gaussian function is superimposed on the sampling region of the maximum quantum energy state to form the ground state of the quantum oscillator.

[0139] It should be noted that in order to form the ground state of the quantum oscillator, one or more Gaussian functions need to be set. The parameters (mean, variance) of these functions will be adjusted based on the previously determined maximum quantum energy state. Gaussian functions are superimposed in the state region with the maximum quantum energy. This step simulates the formation process of the quantum ground state, in which the Gaussian function simulates the probability density distribution of quantum particles in the ground state. By superimposing Gaussian functions, the energy representation in this region will be enhanced, further emphasizing the potential of this region as a clustering center.

[0140] The formation of the ground state means that the system has reached the lowest possible energy configuration, which corresponds to the most stable and representative data clustering point in the cluster center optimization process. After superimposing the Gaussian function, the quantum energy distribution will be reconfigured, and the final determined ground state position represents the final selection of the cluster center.

[0141] Based on the comparison results of the ground state of the quantum oscillator combined with the quantum energy, the cluster center of the state with the maximum quantum energy is selected from the sampling area as the optimized cluster center.

[0142] It should be noted that the dynamic optimization of the cluster centers in the L-ISODATA clustering algorithm combined with the quantum oscillator model helps to determine the cluster centers more accurately. By simulating the ground state and excited state of the quantum state, the true distribution center of the data is found. This method can identify more precise clustering points in multimodal data sets, thereby improving the accuracy of the clustering results.

[0143] With the help of the quantum oscillator model, the updating process of the cluster center is more stable, reducing the problems caused by improper initial center selection or local optimal solutions. This method ensures that the algorithm converges to the global optimal or approximate optimal solution by dynamically adjusting the cluster center. The introduction of the quantum model can find the optimized cluster center faster in each iteration, because the change of the quantum state directly points to the state with the lowest energy (most stable). This direct energy guidance reduces invalid iterations and explorations, thereby speeding up the clustering process.

[0144] S23. When the change of the cluster center is less than a preset threshold or reaches the maximum number of iterations, the iteration process is stopped to generate several agricultural irrigation load feature groups with similar characteristics.

[0145] S3. Perform time dimension analysis and space dimension analysis on the agricultural irrigation load characteristic groups in turn to identify the agricultural irrigation load patterns corresponding to the agricultural irrigation load characteristic groups.

[0146] The time dimension analysis and the space dimension analysis are performed on the agricultural irrigation load characteristic groups in turn, and the agricultural irrigation load mode corresponding to the agricultural irrigation load characteristic groups is identified, which includes the following steps:

[0147] S31. Predict the amount of agricultural irrigation required for crops during the growth stage, and analyze the high-frequency cycle of electricity based on the required amount of agricultural irrigation to obtain the analysis result of the time dimension.

[0148] Among them, predicting the amount of agricultural irrigation required for crops during the growth stage, and analyzing the high-frequency cycle of electricity based on the required amount of agricultural irrigation to obtain the time dimension analysis results includes the following steps:

[0149] S311. Obtain the agricultural irrigation influencing factors of crops during their growth stage, and convert the agricultural irrigation influencing factors into node embedding representations respectively, wherein the agricultural irrigation influencing factors at least include soil type, climate conditions and crop type.

[0150] It should be noted that soil type is obtained through soil sample analysis, which usually involves collecting soil samples from different areas and testing them for chemical and physical properties, such as pH, organic matter content, texture, etc.

[0151] Climate conditions are collected through weather stations or satellite data. Key parameters include rainfall, temperature, humidity, sunshine duration, etc. These data can be obtained from local meteorological bureaus.

[0152] Crop types are based on agricultural statistics or field surveys and include the types of crops grown and their growth cycles, drought tolerance, water and nutrient requirements, etc.

[0153] It should be noted that node embedding includes selecting embedding technology and generating node embedding including:

[0154] Select appropriate node embedding technology according to the characteristics of the data and its intended use, such as graph embedding algorithm; input the characteristic data of factors affecting agricultural irrigation into the embedding model to generate an embedding representation of each factor. These embeddings can capture the complex relationships and interactions between factors and provide information-rich representation for subsequent analysis and prediction.

[0155] S312: Select similar node embedding representations as neighbors based on the cosine similarity between the node embedding representations to obtain neighbor node embedding representations.

[0156] It should be noted that similar node embedding representations are selected as neighbors based on the cosine similarity between node embedding representations, and the neighbor node embedding representations are obtained including:

[0157] For each node embedding representation, calculate the cosine similarity between it and all other node embedding representations. Cosine similarity is a method to measure the directional similarity of two vectors. Its value range is between -1 and 1. The closer the value is to 1, the more similar the vectors are.

[0158] A certain number of the most similar nodes are selected as neighbors for each node based on the calculated cosine similarity. Usually, the nodes with the highest similarity are selected, or a similarity threshold is set to screen neighbors. This selection process can help capture the local structure of each node and provide contextual information for subsequent analysis and modeling.

[0159] For each node, the embedding representations of its neighboring nodes are collected according to the similarity sorting results to form a set of neighboring nodes. This set will be used for subsequent clustering, classification or prediction tasks to enhance the model's understanding of local structures and relationships. In some cases, it may be necessary to integrate the embedding representations of neighboring nodes, for example, by taking the average or using a weighted average to form a more representative node embedding.

[0160] S313. Construct a preliminary graph structure based on the neighbor node embedding representation, and learn the graph adjacency matrix to optimize the graph structure.

[0161] It should be noted that the process of constructing a preliminary graph structure and optimizing the graph structure by learning the graph adjacency matrix is ​​aimed at optimizing the graph structure by using the embedded representation of neighbor nodes so that it can better represent the relationship between nodes.

[0162] S314. The time series information represented by the embedding of neighbor nodes is integrated into each layer by using the optimized graph structure to generate the correlation between the factors affecting agricultural irrigation.

[0163] It should be noted that the time series information represented by neighbor node embedding is integrated into each layer by optimizing the graph structure, and the correlation between the factors affecting agricultural irrigation is generated, including:

[0164] Assign a unique identifier to each node in the graph, create edges based on the relationships between nodes, and the creation of preliminary edges can be based on existing data relationships, such as co-occurrence, high similarity, etc.; use an adjacency matrix or adjacency list to represent the structure of the entire graph, and use selected node embedding technology to learn a low-dimensional vector representation of each node.

[0165] An initial graph is constructed based on the static attributes of the nodes (such as soil type, climate conditions, and crop type). In this graph, the nodes represent different factors affecting agricultural irrigation, and the edges represent the basic associations between them. For each node, the time series information represented by its neighbor nodes is embedded. This can be done in various ways, such as aggregating the time series features of neighbor nodes (average, weighted average, maximum, minimum, etc.). Through this fusion, each node not only reflects its own temporal dynamic characteristics, but also integrates the temporal dynamics of the nodes directly connected to it.

[0166] The initial graph structure is decomposed into layers or multiple layers to process and represent information at different time scales. For example, short-term climate change and long-term soil change can be processed separately. Each layer can specialize in processing different types of time series data to ensure effective stratification and integration of information. New edges are generated based on the correlation and dynamic changes of time series data.

[0167] It should be noted that we are assuming that we are processing a dataset involving three main factors affecting agricultural irrigation: soil type (sandy, clay, loam), climate conditions (dry, wet, mild), and crop type (wheat, corn, soybean).

[0168] Initial graph construction:

[0169] Nodes: represent different soil types, climate conditions and crop types.

[0170] Bian: Based on existing research and data analysis, such as the impact of sandy soil and dry climate on corn growth.

[0171] Time Series Data Fusion:

[0172] Soil Moisture: Collect historical moisture data for each soil type.

[0173] Climate Change: Time series analysis of temperature and rainfall data for each climate condition.

[0174] Crop growth data: Monitor the growth rate and yield of different crops under various conditions.

[0175] Multi-layer graph structure optimization:

[0176] Short-term layer: deals with the impacts of daily or weekly meteorological changes.

[0177] Long-term layer: Analyze the long-term effects of soil type changes on the crop growing season.

[0178] Dynamic edge generation and association analysis:

[0179] Temporal correlation analysis: Calculate the correlation between node features in different time windows and dynamically adjust the weights of edges.

[0180] Critical path identification: Analyze the paths that have the greatest impact on crop yields, such as the effects of long-term wet climate and clay on corn growth.

[0181] S315, combining the periodic characteristics of the agricultural irrigation influencing factors based on the correlation between the agricultural irrigation influencing factors, and outputting the agricultural irrigation amount required for the crops in the growth stage;

[0182] S316. Based on the amount of agricultural irrigation required by crops during their growth stage, the high-frequency cycle of agricultural irrigation load and the corresponding high-frequency cycle of power load are revealed to obtain the analysis result of the time dimension.

[0183] S32. Establish the power grid topology structure between the distribution station area location, water pump marking points and agricultural irrigation areas to analyze the high power load area and obtain the spatial dimension analysis results.

[0184] It should be noted that there is a significant correlation between the high-frequency cycle of agricultural irrigation load and the corresponding high-frequency cycle of power load, which is mainly reflected in the following aspects:

[0185] Farm irrigation is an electricity-intensive activity, especially when using groundwater or distant water sources. During peak irrigation demand periods, such as dry seasons or critical crop growth periods, the electricity load increases significantly because irrigation pumps and other irrigation equipment require more electricity to operate.

[0186] Many crops have clear growth cycles and irrigation demand cycles, which usually correspond to seasonal changes. For example, in the dry season, in order to meet the needs of crop growth, agricultural irrigation activities increase, and the power load will also increase accordingly.

[0187] In some areas, irrigation may tend to be carried out at night to take advantage of lower electricity prices and reduce water evaporation. This pattern will also form a specific high-frequency cycle in electricity consumption data, and peak irrigation periods often require special dispatching of the power grid to meet the peak electricity demand. Understanding this periodic association can help power companies optimize their resource allocation and energy management, such as managing peak loads by implementing demand response measures or adjusting electricity price policies.

[0188] Among them, establishing the power distribution station area location, water pump marking point location and the power grid topology between the agricultural irrigation area to analyze the high power load area and obtain the spatial dimension analysis results includes the following steps:

[0189] S321. Collect the spatial coordinates of the distribution station area, water pump marking points and agricultural irrigation areas respectively, and construct a three-dimensional power grid model based on the spatial coordinates.

[0190] It should be noted that the steps of collecting the spatial coordinates of the distribution station area, water pump marking points and agricultural irrigation areas, and constructing a three-dimensional power grid model based on the spatial coordinates are as follows:

[0191] Collect spatial coordinates of distribution substation locations, water pump locations, and irrigation areas from Geographic Information Systems (GIS), on-site GPS surveys, or existing map databases; verify the accuracy and completeness of the collected coordinate data to ensure that all important locations are recorded and the coordinates are accurate.

[0192] Ensure that all location data use the same coordinate system and units, convert the location data into a format suitable for processing by the modeling software, and select software for power grid model design, such as AutoCAD, SketchUp and other power grid design software.

[0193] Use terrain data to create the basic terrain of the model, including terrain undulations and major geomorphic features, and place distribution stations, water pumps, and irrigation areas in three-dimensional space, ensuring that their locations are consistent with actual coordinates, and draw wires and pipes to connect the various structures to form a complete power and water network layout.

[0194] Add other data layers, such as power flow, water flow velocity, etc., for simulation and analysis, and use the model for flow simulation and network analysis to evaluate the efficiency and problem points of the power and water networks, and analyze the efficiency and potential bottleneck problems of the network based on the simulation results.

[0195] S322, connecting the distribution station area location, the water pump mark point location and the associated area of ​​the agricultural irrigation area in the three-dimensional power grid model to obtain a power grid topology structure;

[0196] S323. Combined with the grid topology, dynamically simulate the agricultural irrigation behavior under different flow conditions and generate several agricultural irrigation paths.

[0197] Among them, combining the power grid topology to dynamically simulate the agricultural irrigation behavior under different flow conditions, generating several agricultural irrigation paths includes the following steps:

[0198] S3231, determining the topological nodes and topological edges in the power grid topological structure, the topological nodes corresponding to the distribution station area location, the water pump marking point location and the agricultural irrigation area, and the topological edges are the connections between the topological nodes;

[0199] S3232, setting the flow of each topological edge in the power grid topology structure as the initial flow;

[0200] S3233, taking the water pump mark point as the source point and the agricultural irrigation area as the sink point, searching for the augmenting path from the source point to the sink point, and there is residual flow on each augmenting path;

[0201] S3234, calculating the minimum residual flow on each augmenting path, and updating the flow of the topological edges on each augmenting path based on the minimum residual flow;

[0202] S3235. Repeat steps S3233-S3234 until no new augmenting path can be found and then stop searching; use the augmenting path corresponding to the flow of the topological edge within the preset range as the agricultural irrigation path to obtain several agricultural irrigation paths.

[0203] It should be noted that the agricultural irrigation behavior under different flow conditions is dynamically simulated in combination with the power grid topology structure, and several agricultural irrigation paths are generated in specific applications, including:

[0204] Create a three-dimensional space in the modeling software and mark the collected coordinate points as nodes in the space; draw edges between nodes according to the actual connection conditions (such as transmission lines, water pipes, etc.) to form a power grid topology; evaluate the initial flow of each edge based on historical data or simulation results; set the initial flow parameters of each edge to the evaluation value in the model.

[0205] Set the source point at the water pump mark point and the sink point in the agricultural irrigation area; use the search algorithm in graph theory (such as depth-first search DFS or breadth-first search BFS) to search all possible paths from the source point to the sink point; for each path found, check whether each edge on the path has enough residual flow to support irrigation.

[0206] For each augmenting path, calculate the minimum residual flow of each edge on the path; based on the calculated minimum residual flow, update the flow value of each edge, repeat the above steps until no new augmenting path can be found, and determine the augmenting path corresponding to the flow of all edges in the preset range as the agricultural irrigation path.

[0207] It should be noted that a grid topology consisting of three distribution substations, two water pumps, and two agricultural irrigation areas is assumed.

[0208] Determine the grid topology:

[0209] Distribution station areas A, B, and C are located at coordinates (1, 2, 3), (4, 5, 6), and (7, 8, 9), respectively.

[0210] Pumps P1 and P2 are located at coordinates (2, 3, 4) and (5, 6, 7) respectively.

[0211] The irrigation areas F1 and F2 are located at coordinates (3, 4, 5) and (6, 7, 8) respectively.

[0212] Connecting these nodes in three-dimensional space forms the grid topology.

[0213] Set the initial flow:

[0214] According to historical data, the initial flow rate from distribution substation A to B is set to 100 units, the initial flow rate from B to C is set to 150 units, the initial flow rate from P1 to F1 is set to 200 units, and the initial flow rate from P2 to F2 is set to 250 units.

[0215] Search for augmenting paths:

[0216] Set P1 and P2 as source points, and F1 and F2 as sink points.

[0217] Search all possible paths from P1 to F1 and P2 to F2, ensuring that there is enough residual flow on the paths.

[0218] Update topology edge traffic:

[0219] For the found path, calculate the minimum residual flow for each edge on the path.

[0220] Update the flow values ​​of the corresponding edges. For example, if the minimum remaining flow on the path from P1 to F1 is 180 units, update the flow from P1 to F1 to 180 units.

[0221] Iterative search and update:

[0222] The search and update process is repeated until no new augmenting paths are found.

[0223] The augmented paths corresponding to the flow of all edges that meet the preset range are determined as the irrigation paths.

[0224] S324. Select the optimal agricultural irrigation path from the agricultural irrigation paths, and reveal the agricultural irrigation high load area and the corresponding power high load area based on the optimal agricultural irrigation path to obtain the spatial dimension analysis result.

[0225] Among them, selecting the optimal agricultural irrigation path from the agricultural irrigation paths includes the following steps:

[0226] The distribution station area, water pump and agricultural irrigation area are respectively regarded as intelligent entities. The status of each intelligent entity includes the geographical location, current flow and power status, and the initial parameters are set for each intelligent entity.

[0227] Initialize the agent configuration, obtain the time step of each agent when the game round reaches the specified number of times, and generate the global and local situations of each agent.

[0228] It should be noted that initializing the agent configuration, obtaining the time step of each agent when the game round reaches the specified number of times, and generating the global situation and local situation of each agent include:

[0229] Define the various agents participating in the game, including their roles, capabilities, strategies and other characteristics; set initial parameters for each agent, such as speed, perception range, decision logic, etc.; create a configuration file to store the initialization settings of all agents to ensure that each simulation starts from the same state.

[0230] Build or select a game environment, including terrain, obstacles, environmental conditions, etc.; set the dynamic changing parameters of the environment, such as weather changes, lighting conditions, etc., to affect the behavior and perception of the agent; start the game simulation according to the configuration file, and all agents start to act according to the set parameters.

[0231] Record the state and behavior of each agent at each time step, such as position, speed, etc.; set a trigger condition, such as reaching a specific number of game rounds or time steps, to trigger data collection and analysis; automatically collect key data of each agent when the game reaches the specified number of rounds; analyze the collected data to generate global and local situation reports, which may include environmental changes and key events, etc.; integrate the data of all agents and environmental factors to generate a global situation map representing the space of agent interactions.

[0232] Use graphs and charts to display the distribution of agents, action paths and key events; the global situation map should support dynamic updates to reflect real-time changes in the spatial situation of agent interactions; generate more detailed local situation maps for specific areas or specific agents.

[0233] Based on the global and local situations, the next action of each agent and the reward for each action are selected.

[0234] It should be noted that the greedy strategy is used to select the next action of each agent based on the global situation and local situation, and the rewards for each action include:

[0235] Define the greediness value. In the exploration phase, the agent chooses non-optimal or random actions with a certain probability to accumulate more environmental information and avoid falling into the local optimum. In the early stage, the greediness value is high to encourage exploration. As time goes by, the greediness value gradually decreases, making the agent more inclined to use the currently known optimal strategy;

[0236] An action space is defined for each agent. These actions should reflect their possible behaviors in the agent interaction space, and the actions are divided into low-level actions and high-level actions. High-level actions are composed of multiple low-level actions.

[0237] Analyze the global situation map and the local situation map to obtain the dynamic information of the interaction space of intelligent agents. For example, when analyzing the global situation, if a certain area is the key to selecting the optimal target, the intelligent agent needs to sacrifice local interests to cover the area, thereby creating conditions for the optimal target.

[0238] Based on the greedy strategy, the agent randomly selects an action (exploration) with a probability of the greedy value in each decision step, or selects the known optimal action; for each optional action, its expected reward is evaluated, usually calculated based on historical data or a prediction model; the action behavior reward is calculated based on the direct result of the action; the potential impact of the action on the future state is evaluated, such as improving the stability of the agent interaction space or creating favorable conditions for the next action; the agent executes the selected action, updates its state and the agent interaction space environment in real time, and the agent interaction space environment provides feedback based on the agent's actions.

[0239] Assign action behavior reward coefficients to the irrigation paths, and select the optimal irrigation path based on the action behavior reward coefficients of each irrigation path.

[0240] S33. The time dimension analysis result and the space dimension analysis result are integrated to identify the agricultural irrigation load pattern corresponding to the agricultural irrigation load feature group.

[0241] S4. Execute power dispatch and allocation in the distribution station area based on the identified agricultural irrigation load pattern.

[0242] It should be noted that the power dispatching and allocation of the distribution station area based on the identified agricultural irrigation load mode includes:

[0243] Check the existing power resources in the distribution station area, including power generation, power supply capacity and backup power supply, and analyze whether the resources can meet the predicted agricultural irrigation needs and identify possible areas of power shortage.

[0244] Design power dispatching plans based on the cyclical demand of agricultural irrigation loads to ensure power supply during critical periods, and plan the startup of backup power sources and emergency power dispatching to cope with unexpected power demand peaks.

[0245] Convert the designed power dispatching plan into operation instructions, and manually control the operation instructions to relevant departments and equipment, monitor the execution of power dispatching in real time, and ensure that all operations are carried out as planned.

[0246] Implement demand response strategies, such as adjusting irrigation schedules to disperse loads and reduce grid pressure, and dynamically adjust power distribution and scheduling strategies based on actual power usage and grid load, collect usage feedback, and adjust future power scheduling plans based on actual conditions.

[0247] For example, during peak hours in a certain agricultural irrigation area (such as mornings and evenings in summer), by predicting the high demand period for electricity, the distribution substation in the area needs to adjust the power output in advance to ensure sufficient power during these high demand periods. At the same time, emergency generators are equipped as backup power sources to start when the main power grid is insufficient to ensure uninterrupted power supply. In addition, by adjusting the irrigation time or implementing a time-of-use electricity price policy to guide farmers to disperse their electricity use, the load peak of the entire power grid can be reduced.

[0248] In practical applications, we first conduct a cluster analysis on the load data of a certain area for one year to understand the agricultural irrigation load situation in the area for one year and conduct an overall analysis; secondly, we conduct a comparative analysis on the load data of different regions in the same month to understand the load situation of the two places and divide the load urgency; finally, we conduct a cluster analysis on the load data of the same power supply station in different months to understand the peak period of irrigation load in the area.

[0249] Figure 2 It is the cluster center of agricultural irrigation load data of a certain area in one year; at this time, it is considered that when the normalized per-unit value of load data is greater than 0.5, it is the peak load. As can be seen from the figure, the peak of irrigation for the first type of load is from 7:00 to 18:00, lasting for 11 hours, and the peak point occurs at 16:00; the peak of irrigation for the second type of load is from 8:00 to 24:00, lasting for 16 hours, and the peak points appear at 11:00 and 19:00; the peak of irrigation for the third type of load is from 7:00 to 16:00, lasting for 9 hours, and the peak point appears at 9-10:00; the peak of irrigation for the fourth type of load is from 6:00 to 16:00, lasting for 10 hours, and the peak point appears at 8-9:00; the peak of irrigation for the fifth type of load is from 7:00 to 14:00, lasting for 7 hours, and the peak point appears at 10-11:00; the peak of irrigation for the sixth type of load is from 8:00 to 18:00, lasting for 10 hours, and the peak point appears at 10:00; from the above data, it can be seen that in this area, the peak load lasts for up to 16 hours in a year, and most of the peak load points appear at 9:00 to 10:00.

[0250] Figure 3 This is the clustering result of load data for a certain region in one year, and the agricultural irrigation load in the region is divided into 6 categories. From the peak of the curve in the figure, it can be seen that the peak values ​​of the second, fourth and fifth categories can reach 5kw, the peak values ​​of the third and sixth categories are slightly lower, reaching 4kw, and the peak value of the first category is the lowest, reaching 2kw. The density of the curve in the figure is that the third category is the most dense, followed by the fourth category, followed by the first category, the second category, the sixth category and the fifth category. It can be seen that most of the load in the region is concentrated in the third category, that is, most of the load peaks in this year are around 3kw.

[0251] comprehensive Figure 2 and Figure 3The data shows that the peak load of most loads in the region is 3kw in a year, and the irrigation peak is between 7:00 and 16:00, lasting for 9 hours. The agricultural irrigation area in the fourth category has a large power consumption and a long duration, and is more likely to be overloaded, followed by the second category. Therefore, during the peak period of agricultural irrigation, the areas where the agricultural irrigation loads of the fourth and second categories are located should be paid special attention.

[0252] Figure 4 The three figures on the left are the agricultural irrigation load data of area A in May, and the three figures on the right are the agricultural irrigation load data of area B in May. The agricultural irrigation load data of each area in May are clustered into three categories. At this time, it is considered that the load peak is when the normalized per-unit value of the load data is greater than 0.5. As can be seen from the figure, the peak of the first type of load irrigation in the A area is from 6:00 to 16:00, lasting for 10 hours, and the peak point occurs at 10:00; the peak of the second type of load irrigation is from 6:00 to 13:00, lasting for 7 hours, and the peak point occurs at 8:00; the peak of the third type of load irrigation is from 6:00 to 11:00, lasting for 5 hours, and the peak point occurs at 10:00; the peak of the first type of load irrigation in the B area is from 6:00 to 12:00, lasting for 6 hours, and the peak point occurs at 8:00; the peak of the second type of load irrigation is from 6:00 to 12:00, lasting for 6 hours, and the peak point occurs at 7:00; the peak of the third type of load irrigation is from 6:00 to 20:00, lasting for 12 hours, and the peak point occurs at 7:00. From the above data, it can be seen that in May, the peak load of the A area lasted for up to 10 hours, and most of the peak load points occurred at 10:00, and the peak load of the B area lasted for up to 12 hours, and most of the peak load points occurred at 7:00.

[0253] Figure 5 The clustering results of agricultural irrigation loads in different areas in the same month, of which the three figures on the left are agricultural irrigation load data of Area A in May, and the three figures on the right are agricultural irrigation load data of Area B in May. The agricultural irrigation load data of each area in May are clustered into three categories. It can be seen from the peak of the curve in the figure that the peak of the first category of Area A can reach 5kw, the peak of the second category is slightly lower, which can reach 4kw, and the peak of the third category is the lowest, which can reach 3kw; the peak of the first category of Area B can reach 2kw, the peak of the third category is slightly lower, which can reach 1.5kw, and the peak of the second category is the lowest, which can reach 1kw. The density of curves in substation A is that the second category is the most dense, followed by the first category, and the third category is the last. It can be seen that most of the load in this substation is concentrated in the second category, that is, most of the load peaks in this month are around 4kw; the density of curves in substation B is that the second category is the most dense, followed by the first category, and the third category is the last. It can be seen that most of the load in this substation is concentrated in the second category, that is, most of the load peaks in this month are around 1kw.

[0254] comprehensive Figure 4 and Figure 5The data shows that in May, the peak load of most loads in Area A was 4kw, the irrigation peak was between 6:00 and 13:00, the duration was 7 hours, and the peak time was generally concentrated at 10:00; the peak load of most loads in Area B was 1kw, the irrigation peak was between 6:00 and 12:00, the duration was 6 hours, and the peak time was generally concentrated at 10:00. In summary, Area A is more likely to be overloaded than Area B, so Area A should be paid special attention to during the agricultural irrigation period in May.

[0255] Figure 6 The cluster centers of agricultural irrigation loads in different months in the same substation. The three figures on the left are the agricultural irrigation load data of a substation in May, and the three figures on the right are the agricultural irrigation load data of the substation in September. The agricultural irrigation load data of each substation in May are clustered into three categories. At this time, it is considered that the load peak is when the normalized per-unit value of the load data is greater than 0.5. As can be seen from the figure, the peak of the first type of load irrigation in May in this area is from 6:00 to 16:00, lasting for 10 hours, and the peak point appears at 10:00; the peak of the second type of load irrigation is from 6:00 to 13:00, lasting for 7 hours, and the peak point appears at 8:00; the peak of the third type of load irrigation is from 6:00 to 11:00, lasting for 5 hours, and the peak point appears at 10:00; the peak of the first type of load irrigation in September is from 7:00 to 13:00, lasting for 6 hours, and the peak points appear at 8:00 and 12:00; the peak of the second type of load irrigation is from 8:00 to 20:00, lasting for 12 hours, and the peak point appears at 12:00; the peak of the third type of load irrigation is from 8:00 to 18:00, lasting for 10 hours, and the peak points appear at 11:00 and 14:00. From the above data, it can be seen that the peak load duration of this area in May is up to 10 hours, and most of the load peak points appear at 10:00; the peak load duration in September is up to 12 hours, and most of the load peak points appear at 11:00-12:00.

[0256] Figure 7The clustering results of agricultural irrigation load in different months in the same substation, where the three figures on the left are the agricultural irrigation load data of a substation in May, and the three figures on the right are the agricultural irrigation load data of the substation in September. The monthly agricultural irrigation load data of the substation are clustered into 3 categories. It can be seen from the peak of the curve in the figure that the peak of the first category in May can reach 5kw, the peak of the second category is slightly lower, reaching 4kw, and the peak of the third category is the lowest, reaching 3kw; the peak of the first category in September is the lowest, reaching 0.8kw, the peak of the second category is the highest, reaching 3.5kw, and the peak of the third category is slightly lower, reaching 2kw. In May, the density of the load curve is that the second category is the most dense, followed by the first category, and the third category is the last. It can be seen that most of the load in this area is concentrated in the second category, that is, most of the load peaks in this month are around 4kw; in September, the density of the load curve is that the second category is the most dense, followed by the third category, and the first category is the last. It can be seen that most of the load in this area is concentrated in the second category, that is, most of the load peaks in this month are around 1kw, but compared with May, the curve density of this month is much sparser, which shows that the number of irrigation days in May is more than that in September.

[0257] comprehensive Figure 6 and Figure 7 From the data, we can see that in May, the peak load of most loads in this area was 4kw, the irrigation peak was from 6:00 to 13:00, the duration was 7 hours, and the peak time was generally concentrated at 10:00; in September, the peak load of most loads was 3.5kw, the irrigation peak was from 8:00 to 20:00, the duration was 12 hours, and the peak time was generally concentrated at 11:00 to 12:00. However, compared with May, there were fewer irrigation days in September, the load peak was lower, and the possibility of heavy overload was lower. Therefore, we should focus on the agricultural irrigation period of this area in May.

[0258] It should be noted that through the above analysis, firstly, the clustering results of a certain area were analyzed on a yearly time scale to understand the overall irrigation characteristics of the area; secondly, the clustering results were compared and analyzed in different regions and months on a monthly time scale to gain an in-depth understanding of the load characteristics of agricultural irrigation, which provided technical support and decision-making basis for solving the distribution network problems and allocation of distribution resources caused by agricultural irrigation.

[0259] In summary, with the help of the above technical scheme of the present invention, the present invention iteratively clusters the agricultural irrigation load through a cluster analysis algorithm to obtain distribution station areas and agricultural irrigation load feature groups with similar operating states, and uses the spatiotemporal analysis method to analyze the clustering results. By comparing the agricultural irrigation load analysis results at different times and spaces, effective technical support and decision-making basis are provided for solving distribution network problems; the present invention obtains and analyzes the influencing factors of the crop growth stage, and uses node embedding representation and graph structure optimization to more accurately predict the irrigation amount required for various crops at different growth stages, and analyzes the high-frequency cycle of electricity according to the predicted crop irrigation demand, so as to effectively plan and adjust the allocation of power resources, thereby helping to prevent irrigation interruptions caused by insufficient power supply; the present invention accurately simulates the topology of the power grid The structure and agricultural irrigation path can effectively identify high-load areas, thereby achieving optimal allocation of water and power resources, which helps to reduce resource waste and improve resource utilization efficiency, and dynamically simulate agricultural irrigation behavior under different flow conditions, which helps to predict and manage load changes in the power grid, thereby enhancing the stability and reliability of the power grid, and selecting the optimal agricultural irrigation path to ensure reasonable irrigation level distribution; the present invention can identify the agricultural irrigation load pattern in a specific area by integrating the analysis results of the time and space dimensions, so as to adopt more sophisticated irrigation strategies and power management strategies, which not only improves irrigation efficiency, but also helps to save energy and reduce operating costs. Accurate spatiotemporal analysis helps to achieve sustainable management of agricultural irrigation and power resources, reduces negative impacts on the environment, and can identify potential risks and challenges in advance.

[0260] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for spatiotemporal analysis of agricultural irrigation load based on clustering algorithm, characterized in that: The spatial-temporal analysis method of agricultural irrigation load includes the following steps: S1. Obtain the agricultural irrigation load data of the distribution station area and pre-process the agricultural irrigation load data; S2, iteratively clustering the preprocessed agricultural irrigation load data, and summarizing the data points with similar load characteristics to generate agricultural irrigation load characteristic groups; S3, sequentially performing time dimension analysis and space dimension analysis on the agricultural irrigation load characteristic groups to identify the agricultural irrigation load patterns corresponding to the agricultural irrigation load characteristic groups; S4. Execute power dispatch and allocation in the distribution area based on the identified agricultural irrigation load mode; The step of sequentially performing time dimension analysis and space dimension analysis on the agricultural irrigation load feature group to identify the agricultural irrigation load mode corresponding to the agricultural irrigation load feature group includes the following steps: S31. Predict the amount of agricultural irrigation required for crops during the growth stage, and analyze the high-frequency cycle of electricity based on the required amount of agricultural irrigation to obtain a time dimension analysis result; S32, establish the power grid topology structure between the distribution station area location, water pump marking point location and agricultural irrigation area to analyze the power high load area and obtain the spatial dimension analysis result; S33, integrating the time dimension analysis results and the space dimension analysis results to identify the agricultural irrigation load pattern corresponding to the agricultural irrigation load feature group; The method of predicting the amount of agricultural irrigation required for crops during the growth stage and analyzing the high-frequency cycle of electricity based on the required amount of agricultural irrigation to obtain the time dimension analysis result includes the following steps: S311, obtaining agricultural irrigation influencing factors of crops in the growth stage, and converting the agricultural irrigation influencing factors into node embedding representations respectively, wherein the agricultural irrigation influencing factors at least include soil type, climate conditions and crop type; S312, selecting similar node embedding representations as neighbors based on the cosine similarity between the node embedding representations, and obtaining neighbor node embedding representations; S313. Construct a preliminary graph structure based on neighbor node embedding representation, and learn the graph adjacency matrix to optimize the graph structure; S314, using the optimized graph structure to integrate the time series information represented by the neighbor node embedding into each layer, and generate the correlation relationship between the factors affecting agricultural irrigation; S315, combining the periodic characteristics of the agricultural irrigation influencing factors based on the correlation between the agricultural irrigation influencing factors, and outputting the agricultural irrigation amount required for the crops in the growth stage; S316, revealing the high-frequency cycle of agricultural irrigation load and the corresponding high-frequency cycle of power load according to the amount of agricultural irrigation required by crops during the growth stage, and obtaining the time dimension analysis result; The establishment of the power grid topology structure between the distribution station area location, the water pump mark point and the agricultural irrigation area to analyze the power high load area and obtain the spatial dimension analysis result includes the following steps: S321, respectively collecting the spatial position coordinates of the distribution station area, the water pump marking point and the agricultural irrigation area, and constructing a three-dimensional power grid model based on the spatial position coordinates; S322, connecting the distribution station area location, the water pump mark point location and the associated area of ​​the agricultural irrigation area in the three-dimensional power grid model to obtain a power grid topology structure; S323, dynamically simulate the agricultural irrigation behavior under different flow conditions in combination with the power grid topology, and generate several agricultural irrigation paths; S324, selecting an optimal agricultural irrigation path from the agricultural irrigation paths, and revealing agricultural irrigation high load areas and corresponding power high load areas based on the optimal agricultural irrigation path, to obtain spatial dimension analysis results; The method of dynamically simulating the agricultural irrigation behavior under different flow conditions in combination with the power grid topology structure to generate a number of agricultural irrigation paths includes the following steps: S3231, determining topological nodes and topological edges in the power grid topology structure, wherein the topological nodes correspond to the locations of distribution substations, water pump marking points and agricultural irrigation areas, and the topological edges are connections between the topological nodes; S3232, setting the flow of each topological edge in the power grid topology structure as the initial flow; S3233, taking the water pump mark point as the source point and the agricultural irrigation area as the sink point, searching for the augmenting path from the source point to the sink point, and there is residual flow on each augmenting path; S3234, calculating the minimum residual flow on each augmenting path, and updating the flow of the topological edges on each augmenting path based on the minimum residual flow; S3235, repeating steps S3233-S3234 until no new augmenting path can be found, then stopping the search; taking the augmenting path corresponding to the flow of the topological edge within the preset range as the agricultural irrigation path, and obtaining a plurality of agricultural irrigation paths; The selecting of the optimal agricultural irrigation path from the agricultural irrigation paths comprises the following steps: The distribution station area, water pump and agricultural irrigation area are respectively regarded as intelligent entities. The status of each intelligent entity includes the geographical location, current flow and power status, and the initial parameters are set for each intelligent entity. Initialize the agent configuration, obtain the time step of each agent when the game round reaches the specified number of times, and generate the global and local situations of each agent; Select the next action and reward for each agent based on the global and local situation; Assign action behavior reward coefficients to the irrigation paths, and select the optimal irrigation path based on the action behavior reward coefficients of each irrigation path.

2. The method for spatiotemporal analysis of agricultural irrigation load based on clustering algorithm according to claim 1, characterized in that: The obtaining of agricultural irrigation load data of the distribution station area and preprocessing of the agricultural irrigation load data comprises the following steps: S11, obtaining the agricultural irrigation load data of the distribution station area, calculating and recording the Euclidean distance between the target sample and the remaining samples in the data sample space to obtain the nearest sample; S12, calculating the weighted average of the most recent samples as filling data to achieve missing value processing of agricultural irrigation load data; S13. Scale the agricultural irrigation load data to a preset range using a linear function to achieve normalization processing of the agricultural irrigation load data.

3. The method for spatiotemporal analysis of agricultural irrigation load based on clustering algorithm according to claim 2, characterized in that: The iterative clustering of the pre-processed agricultural irrigation load data to generate agricultural irrigation load feature groups by summarizing data points with similar load characteristics comprises the following steps: S21, setting an initial cluster center, and allocating each data point of the agricultural irrigation load data to the nearest cluster center based on the initial cluster center; S22, dynamically optimizing the cluster center in combination with the quantum oscillator model to obtain an optimized cluster center, and iteratively performing cluster merging processing and cluster splitting processing based on the optimized cluster center; S23. When the change of the cluster center is less than a preset threshold or reaches the maximum number of iterations, the iteration process is stopped to generate several agricultural irrigation load feature groups with similar characteristics.

4. The method for spatiotemporal analysis of agricultural irrigation load based on clustering algorithm according to claim 3 is characterized in that: The method of dynamically optimizing the cluster center by combining the quantum oscillator model to obtain the optimized cluster center includes the following steps: Randomly select several data points of agricultural irrigation load data as initial cluster centers, and record the quantum energy of the initial cluster centers; The mean and standard deviation of the Gaussian sampling function are set, and the quantum energy of the initial cluster center is formed into a sampling region representing each energy state to simulate the high energy state of the quantum oscillator; A new cluster center is selected in each sampling area by Gaussian sampling, and the quantum energy of the new cluster center is compared with the quantum energy of the initial cluster center; The Gaussian function is superimposed on the sampling region of the maximum quantum energy state to form the ground state of the quantum oscillator; Based on the comparison results of the ground state of the quantum oscillator combined with the quantum energy, the cluster center of the state with the maximum quantum energy is selected from the sampling area as the optimized cluster center.

5. The method for spatiotemporal analysis of agricultural irrigation load based on clustering algorithm according to claim 2, characterized in that: The calculation formula of the filling data is: In the formula, y represents the filling data; k represents the number of recent sample amounts; W i Represents the weight value of the i-th sample.

Images