A multidimensional clustering analysis method for agricultural irrigation load based on clustering algorithm

Abstract

The present invention discloses a multidimensional clustering analysis method for agricultural irrigation load based on a clustering algorithm, and relates to the technical field of agricultural irrigation load analysis. The multidimensional clustering analysis method for agricultural irrigation load includes the following steps: using a clustering algorithm to analyze power load data, and combining the power load data with agricultural irrigation load information to select agricultural irrigation fault load information; defining dimension categories to analyze fault load information according to characteristic attributes corresponding to the agricultural irrigation fault load information, and dividing agricultural irrigation load categories based on the fault load information; establishing a control model for agricultural irrigation load participation in power operation centered on the agricultural irrigation load category, and using the control model to allocate power resources to control irrigation water pumps; starting the irrigation water pump according to the allocation result to perform agricultural irrigation operations, and establishing an efficiency evaluation model to verify farmland irrigation efficiency. The present invention analyzes regional water usage data through a clustering algorithm, can accurately identify the characteristics of agricultural irrigation loads, and provide an accurate basis for subsequent analysis and control.

Description

Technical Field

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

[0002] With the improvement of the information level of power system, each power link such as power generation, transmission, distribution, transformation and power consumption will produce a large amount of load data. The load response is uneven and the load characteristics are seriously differentiated. It is urgent to carry out accurate statistics and classification of load data and convert it into valuable information. This is an important issue that the current power system needs to solve and deal with. Power load curve clustering is the key to screening and analyzing load data. Load data preprocessing, abnormal user detection, demand side management and energy efficiency management and other data all require power load curves for clustering analysis.

[0003] In view of the diverse load characteristics of the current power system, research on effective power data analysis and mining technology and load clustering algorithm is of great significance to the management, classification and conservation of power grid resources. Scientific and reasonable load clustering can help power supply units formulate reasonable electricity prices and achieve load peak control. At the same time, it can also discover problems and defects, adjust power service strategies in time, reduce power supply costs, and better improve energy utilization.

[0004] In the field of grain cultivation, rice is one of the most important farmland food crops. It requires a regular and sufficient supply of water. During the growth period, especially the seedling and filling periods, it needs a lot of water. However, due to frequent abnormal faults such as heavy overload, low voltage, and three-phase imbalance, the load caused by farmland irrigation problems accounts for a large proportion. Concentrated electricity use during the irrigation period often leads to a continuous increase in the power supply pressure of the local distribution network, and distribution transformers often have heavy overload faults. The disorder of farmland irrigation water will not only bring operational safety risks to the distribution network, but also increase users' electricity costs. Therefore, it is necessary to analyze the farmland irrigation load from multiple dimensions in order to manage and regulate the load in the irrigation area.

[0005] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0006] In order to solve the above problems, the present invention proposes a multidimensional clustering analysis method for agricultural irrigation load based on clustering algorithm, so as to optimize the operating state of irrigation pumps and improve their working efficiency and irrigation effect.

[0007] In order to achieve the above object, the present invention adopts the following technical solution:

[0008] The present invention provides a multidimensional clustering analysis method for agricultural irrigation load based on a clustering algorithm, and the multidimensional clustering analysis method for agricultural irrigation load includes the following steps:

[0009] The clustering algorithm is used to analyze the power load data, and the power load data is combined with the agricultural irrigation load information to determine the agricultural irrigation load characteristics, and the agricultural irrigation fault load information is selected from the agricultural irrigation load characteristics;

[0010] Analyze the fault load information by defining dimension categories according to the characteristic attributes corresponding to the agricultural irrigation fault load information, and classify the agricultural irrigation load categories based on the fault load information;

[0011] Establish a control model for agricultural irrigation loads participating in power operation, centered on agricultural irrigation load categories, and use the control model to allocate power resources to control irrigation pumps based on the assessed working capacity of irrigation pumps;

[0012] According to the allocation results, the irrigation pump is started to perform irrigation operations in the agricultural area, and an efficiency evaluation model is established based on the irrigation results to verify the farmland irrigation efficiency, and the control state of the irrigation pump is optimized based on the verification results.

[0013] Further, defining dimension categories to analyze the fault load information according to the characteristic attributes corresponding to the agricultural irrigation fault load information, and classifying the agricultural irrigation load categories based on the fault load information includes the following steps:

[0014] The agricultural irrigation fault load information is used as the data source, the final number of clusters is pre-set according to the K-means++ algorithm, and a group of data information points are randomly selected from the data source as the cluster center points;

[0015] Calculate the modified Euclidean distance between the remaining data information points, and select cluster centers from the remaining data information points to calculate the intra-class distance in each cluster of the data source;

[0016] Select new data information points as new cluster centers according to the calculation results and the weighted probability distribution algorithm, and repeatedly select new cluster centers until the distance centers of the preset number of clusters are selected;

[0017] According to the modified Euclidean distance algorithm, the remaining data information points are divided into new clusters with the closest cluster center, and the mean method is used to calculate the center point of each new cluster, and then the iterative process of the K-means clustering algorithm is performed to complete the clustering processing of the agricultural irrigation fault load information.

[0018] According to the clustering processing results, load fault duration, load heavy load rate and load overload rate are extracted as characteristic attributes, and the agricultural irrigation load categories are divided according to the characteristic attributes and K-means clustering algorithm.

[0019] Furthermore, the calculation formula for the intra-class distance in each cluster of the data source is:

[0020]

[0021] Where D in (k) represents the intra-class distance in each cluster of the data source, k represents the number of clusters, c inum represents the number of points in the i-th cluster, p i Indicates the selection of sampling points in the i-th cluster, p n represents the remaining points in the i-th cluster.

[0022] Furthermore, based on the clustering processing results, the load fault duration, load overload rate and load overload rate are extracted as characteristic attributes, and the agricultural irrigation load categories are divided according to the characteristic attributes and the K-means clustering algorithm, including the following steps:

[0023] The agricultural irrigation fault load information is divided into load heavy load situation and load overload situation, and coordinate systems are established to describe the relationship between load fault duration and load heavy load situation and load overload situation respectively;

[0024] The relationship description results are combined with the K-means clustering algorithm to divide the severity level of load overload by setting the number of clusters;

[0025] The load fault duration is obtained from the agricultural irrigation fault load information and combined with the K-means clustering algorithm to analyze the fault load duration in the flood season and the dry season;

[0026] Analyze the load failure concentration time and failure duration within a day based on load heavy load, load overload and load failure duration;

[0027] The severity level of load overload, duration of fault load, concentrated time of load faults within a day and duration of faults are used as agricultural irrigation load categories.

[0028] Furthermore, a control model for agricultural irrigation loads participating in power operation is established with agricultural irrigation load categories as the center, and the control model is used to allocate power resources to control irrigation pumps according to the working capacity of the irrigation pumps, including the following steps:

[0029] The distributed operation model is used to integrate the power generation equipment with the irrigation water pump to realize remote control and signal demodulation of the irrigation water pump, and the demodulated signal is integrated into the line to realize the unified management of the irrigation water pump;

[0030] A grid monitoring model is established using the double-layer modeling theory to achieve communication and monitoring between power equipment and irrigation pumps during power transmission;

[0031] The agricultural irrigation load category is combined with the generative adversarial network to generate the power operation control program to obtain the power distribution operation safety standard of the irrigation pump under the condition of solving the agricultural irrigation fault load;

[0032] The power generated by the irrigation water pump during use is obtained, and the time series simulation algorithm is used to analyze the power of the irrigation water pump to evaluate the working capacity of the irrigation water pump. Based on the evaluated working capacity of the irrigation water pump, the control model is used to allocate power resources to control the irrigation water pump.

[0033] Furthermore, combining the agricultural irrigation load category with the generative adversarial network to generate a power operation control program to obtain the power distribution operation safety standard of the irrigation water pump under the condition of solving the agricultural irrigation fault load includes the following steps:

[0034] The generative adversarial network algorithm is used to analyze the agricultural irrigation fault load information corresponding to the agricultural irrigation load category, and the agricultural irrigation fault load information is trained and analyzed through game theory.

[0035] According to the analysis results, the desired power distribution safety standard is digitized to obtain a safety expression, and based on the power distribution safety standard, a generative adversarial network algorithm is used to perform a safety analysis to obtain the optimal safety target for the interaction between power information and irrigation water pumps.

[0036] The interactive optimal safety target is used to represent the relationship between current information and power distribution safety standards. Based on the relationship results, a power operation control program is written using a microcomputer algorithm, and the power distribution safety standards are written into the power operation control program as a scheduling plan.

[0037] According to the agricultural irrigation fault load information, the probability that the initial scheduling scheme can solve the agricultural irrigation fault load situation is determined. If the probability result is less than the preset value, the safety expression is obtained again.

[0038] If the probability result is greater than the preset value, the distributed operation model is combined with the power operation control program to complete the operation value range setting of the irrigation water pump.

[0039] Furthermore, obtaining the power generated by the irrigation water pump during use, and using a time series simulation algorithm to analyze the power of the irrigation water pump to evaluate the working capacity of the irrigation water pump, and using a control model to allocate power resources to control the irrigation water pump according to the evaluated working capacity of the irrigation water pump includes the following steps:

[0040] Based on the irrigation history of the agricultural area, the power output of the irrigation pump during use is obtained, and the power fluctuation quantity curve of the irrigation pump at a preset sampling interval is generated according to the power output data condition fitting;

[0041] According to the power fluctuation curve, the proportion of active output fluctuation of the irrigation water pump power at different time scale sampling intervals is determined, and the maximum output of the irrigation water pump is defined according to the proportion of active output fluctuation;

[0042] The cluster effect coefficient is defined based on the ratio of the maximum output of the irrigation pump to the rated capacity of the agricultural area irrigation, and the fluctuation law of the cluster effect coefficient is analyzed to analyze the impact of the cluster effect on the fluctuation of the output of the irrigation pump;

[0043] The objective function is established based on the impact of output fluctuation and the operating cost of the irrigation pump, and the working capacity of the irrigation pump is evaluated and calculated by applying the time series simulation algorithm considering the balance constraints and irrigation speed constraints.

[0044] The working capacity of the irrigation water pump is matched with the operating value range, and power resources are allocated according to the corresponding results to control the irrigation water pump.

[0045] Furthermore, the irrigation pump is started to perform irrigation operation in the farmland according to the allocation result, and an efficiency evaluation model is established according to the irrigation result to verify the farmland irrigation efficiency. The control state of the irrigation pump is optimized based on the verification result, including the following steps:

[0046] According to the result of the power resource allocation, the irrigation water pump is started to perform the irrigation operation of the agricultural area, and the irrigation parameters are obtained after the irrigation operation of the agricultural area is completed. The irrigation parameters include the water flow and pressure of the irrigation water pump and the humidity of the irrigation agricultural area;

[0047] Generate a change curve based on irrigation parameters, and form characteristic index vectors of irrigation water flow, pressure and humidity in irrigation agricultural areas according to the extracted curve time domain and frequency domain characteristic indexes and modal decomposition singular value entropy;

[0048] The principal component analysis method is used to eliminate the redundancy of irrigation pump water flow, pressure and irrigation area humidity information, and to construct a corresponding characteristic index sample library;

[0049] The farmland irrigation status is divided based on the characteristic index sample library and the continuous hidden Markov model, and a comprehensive evaluation model of irrigation status based on the search algorithm and support vector machine is established according to the division results;

[0050] The comprehensive state evaluation model is used to evaluate and verify the farmland irrigation efficiency, and the control state of the irrigation pump is optimized based on the evaluation and verification results.

[0051] Furthermore, the farmland irrigation status is divided based on the characteristic index sample library and the continuous hidden Markov model, and the irrigation status comprehensive evaluation model of the search algorithm optimized support vector machine is established according to the division result, including the following steps:

[0052] The maximum expectation algorithm is used to train the continuous hidden Markov model, and after the training is completed, the characteristic indicator sample sequence is defined according to the characteristic indicator sample library, and the dynamic programming algorithm is used to find the hidden state sequence to explain the characteristic indicator sample sequence;

[0053] The forward-backward technique is used to analyze the likelihood probability value of the characteristic index sample sequence under the continuous hidden Markov model, and the state information corresponding to the characteristic index sample sequence is determined according to the likelihood probability value to obtain the farmland irrigation state;

[0054] The search algorithm is used to find the optimal parameters to optimize the support vector machine, and the farmland irrigation status is used as input and combined with the optimized support vector machine to establish a comprehensive irrigation status evaluation model.

[0055] Furthermore, the search algorithm is used to find the optimal parameters to optimize the support vector machine, and the farmland irrigation status is used as input to combine with the optimized support vector machine to establish an irrigation status comprehensive evaluation model, which includes the following steps:

[0056] The farmland irrigation status is preprocessed with grey correlation, and the farmland irrigation status is sorted according to the correlation degree based on the processing results, and the farmland irrigation status indicators are screened out in combination with the prior knowledge base;

[0057] After pre-initializing the search algorithm parameters, the fitness value of the search individual is calculated using the search algorithm implementation principle, and the optimal parameters are selected according to the optimal fitness value to establish the support vector regression model;

[0058] The particle swarm optimization algorithm is used to define the comprehensive fitness function and the linear decreasing inertia weight strategy to identify the hyperparameters of the support vector regression model to obtain the optimized support vector machine.

[0059] The farmland irrigation status was taken as input, and the farmland irrigation status indicators were used as labels and combined with the optimized support vector machine to establish a comprehensive irrigation status evaluation model.

[0060] The beneficial effects of the present invention are:

[0061] 1. The present invention analyzes regional water usage data through a clustering algorithm, and can accurately identify the characteristics of agricultural irrigation loads, providing an accurate basis for subsequent analysis and control. It also defines dimensional categories according to the characteristic attributes of agricultural irrigation fault load information, and can analyze fault loads more comprehensively and deeply, providing a scientific basis for classifying agricultural irrigation load categories. It also establishes a control model for agricultural irrigation loads participating in power operation centered on agricultural irrigation load categories, which can more effectively allocate power resources and ensure that irrigation water pumps obtain sufficient power support when needed. At the same time, based on the working capacity of the irrigation water pump, the control model is used to allocate power resources, which can optimize the operating state of the irrigation water pump and improve its working efficiency and irrigation effect. This provides a scientific and efficient management method for farmland irrigation by accurately identifying agricultural irrigation load characteristics, effectively screening fault information, multi-dimensional analysis, optimizing power resource allocation, improving work efficiency, realizing quantitative evaluation, and continuously optimizing the irrigation system.

[0062] 2. The present invention takes the main attributes of agricultural irrigation fault load as an example, and reflects the specific irrigation problems and demands of agricultural irrigation load after algorithm analysis, and uses a multidimensional analysis method for agricultural irrigation load to perform cluster analysis on the farmland irrigation load situation in a certain area from different dimensions. Through the multidimensional analysis method, the regional agricultural irrigation load can be divided into multiple categories, which helps the region to subdivide the categories of agricultural irrigation load according to load characteristics, so as to control the regional load in a targeted manner, so as to better alleviate the regional farmland irrigation problem, help to reasonably manage regional resources, guide the reasonable allocation of distribution resources and improve the utilization efficiency of distribution resources.

[0063] 3. The present invention establishes a control model for agricultural irrigation loads participating in power operation centered on the agricultural irrigation load category, thereby realizing remote control, monitoring, and optimal allocation of power resources for irrigation water pumps. Based on a distributed operation model, the power generation equipment and the irrigation water pumps are integrated and connected to realize remote control and signal demodulation of the irrigation water pumps. The grid monitoring model established using the double-layer modeling theory realizes communication and monitoring between power equipment and irrigation water pumps during power transmission, thereby ensuring stable operation, timely discovering and handling potential faults, and avoiding irrigation interruptions or power waste due to equipment failures.

[0064] 4. The present invention combines the agricultural irrigation load category with the generative adversarial network to generate an electric power operation control program, which can obtain the power distribution operation safety standard of the irrigation water pump under the condition of solving the agricultural irrigation fault load, which helps to reasonably allocate power resources while ensuring irrigation needs and avoid imbalance between power supply and demand. By obtaining the power data generated by the irrigation water pump during use and analyzing it using a timing simulation algorithm, the working capacity of the irrigation water pump can be accurately evaluated, which facilitates the precise allocation of power resources according to the actual working status of the water pump, ensuring that the water pump operates in the best working state and improving irrigation efficiency and effectiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0066] Figure 1 is a flow chart of a multidimensional clustering analysis method of agricultural irrigation load based on a clustering algorithm according to an embodiment of the present invention;

[0067] Figure 2 This is a flow chart of a K-means++ algorithm in a multidimensional clustering analysis method of agricultural irrigation load based on a clustering algorithm according to an embodiment of the present invention;

[0068] Figure 3 This is a flow chart of a multidimensional analysis method in a multidimensional clustering analysis method of agricultural irrigation load based on a clustering algorithm according to an embodiment of the present invention;

[0069] Figure 4 It is a grain yield statistical graph set in a multidimensional clustering analysis method of agricultural irrigation load based on a clustering algorithm according to an embodiment of the present invention;

[0070] Figure 5 A heavy load data group in a multidimensional clustering analysis method of agricultural irrigation load based on a clustering algorithm according to an embodiment of the present invention;

[0071] Figure 6 It is an overload data group in a multidimensional clustering analysis method of agricultural irrigation load based on a clustering algorithm according to an embodiment of the present invention;

[0072] Figure 7 It is a heavy load data group after clustering by the multidimensional analysis method when K=3 in a multidimensional clustering analysis method of agricultural irrigation load based on a clustering algorithm according to an embodiment of the present invention;

[0073] Figure 8 It is a heavy load data group after clustering by the multidimensional analysis method when K=7 in a multidimensional clustering analysis method of agricultural irrigation load based on a clustering algorithm according to an embodiment of the present invention;

[0074] Fig. 9 It is an overload load data group after clustering by the multidimensional analysis method when K=3 in a multidimensional clustering analysis method of agricultural irrigation load based on a clustering algorithm according to an embodiment of the present invention;

[0075] Fig.10 It is an overload load data group after clustering by the multidimensional analysis method when K=7 in a multidimensional clustering analysis method of agricultural irrigation load based on a clustering algorithm according to an embodiment of the present invention;

[0076] Fig.11 It is a dry season load data group after being clustered by a multidimensional analysis method in a multidimensional clustering analysis method of agricultural irrigation load based on a clustering algorithm according to an embodiment of the present invention;

[0077] Fig.12 It is a flood season load data group after being clustered by a multidimensional analysis method in a multidimensional clustering analysis method of agricultural irrigation load based on a clustering algorithm according to an embodiment of the present invention;

[0078] Fig.13 It is a 24-hour load data group of the substation after being clustered by the multidimensional analysis method in the multidimensional clustering analysis method of agricultural irrigation load based on a clustering algorithm according to an embodiment of the present invention. DETAILED DESCRIPTION

[0079] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0080] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0081] It should be noted that the terms used herein are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "include" and "have" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0082] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0083] See also Figure 1 The present invention provides a multidimensional clustering analysis method for agricultural irrigation load based on a clustering algorithm, and the multidimensional clustering analysis method for agricultural irrigation load includes the following steps:

[0084] Step S1, using a clustering algorithm to analyze power load data, and combining the power load data with agricultural irrigation load information to determine agricultural irrigation load characteristics, and selecting agricultural irrigation fault load information from the agricultural irrigation load characteristics.

[0085] In this embodiment, when using a clustering algorithm to analyze power load data, combining the power load data with agricultural irrigation load information to determine agricultural irrigation load characteristics, and selecting agricultural irrigation fault load information from agricultural irrigation load characteristics, relevant power load data and agricultural irrigation load data can be collected, including power consumption, load changes, time series and other information, and the data can be cleaned, including filling missing values, removing outliers, and standardizing or normalizing data; features are selected from the power and agricultural irrigation load data, which may specifically include selecting load peaks, average loads, load fluctuations and other features; clustering algorithms such as K-means are applied to group the power load data, and the purpose of clustering is to group data with similar load patterns together, which helps to identify different types of load patterns, including normal and abnormal loads; each clustering result is analyzed to identify load characteristics related to agricultural irrigation, and according to the clustering results, normal agricultural irrigation load patterns and possible fault load patterns are distinguished, including abnormal power consumption growth or irregular load fluctuations that may indicate a fault; key information is selected from the identified fault load characteristics as agricultural irrigation fault load information.

[0086] By using the specific principles of clustering algorithms (taking K-means++ algorithm as an example), the power load data is clustered and combined with the load in the agricultural irrigation area and applied to regional agricultural irrigation.

[0087] Step S2: define dimension categories to analyze the fault load information according to characteristic attributes corresponding to the agricultural irrigation fault load information, and classify the agricultural irrigation load categories based on the fault load information.

[0088] It should be explained that the current load situation in the irrigation area is studied, and the heavy overload situation of the agricultural irrigation load in the area is analyzed. The K-means++ algorithm is used to obtain different load center clustering situations, and the load scatter diagram is obtained by combining the multi-feature attributes of the load (load fault duration, load heavy load rate, load overload rate), and the fault load situation is analyzed from multiple dimensions (taking the severity of heavy overload, the severity of flood season and dry season, and the severity of 24 hours a day as examples);

[0089] Through the multidimensional analysis method of agricultural irrigation load, different agricultural irrigation load data points are clustered in a refined manner. By dividing the agricultural irrigation load data points into different load clustering centers, the specific classification of agricultural irrigation load can be obtained, and then different categories of agricultural irrigation loads can be managed and allocated, which can better alleviate regional agricultural irrigation problems, and is conducive to the reasonable allocation of electricity resources and improving the utilization rate of power resources.

[0090] Specifically, the K-means++ algorithm is mainly used to reduce the impact of the random position of the initial center point on the clustering results. The main criterion for selecting the cluster center point of the K-means++ algorithm is that the relative distance to the initial cluster center is as far as possible. The central idea of ​​the algorithm is to make the point farther away from the existing cluster center have a greater probability of being selected as the next cluster center, so as to improve the accuracy of the clustering results.

[0091] In this embodiment, when analyzing the fault load information according to the characteristic attributes corresponding to the agricultural irrigation fault load information and dividing the agricultural irrigation load categories based on the fault load information, the agricultural irrigation fault load information can be used as a data source, the final number of clusters can be pre-set according to the K-means++ algorithm, and a group of data information points can be randomly selected from the data source as cluster center points; the modified Euclidean distance between the remaining data information points is calculated, and the cluster center is selected from the remaining data information points to calculate the intra-class distance in each cluster of the data source; new data information points are selected as new cluster center points according to the calculation results and the weighted probability distribution algorithm, and new cluster centers are repeatedly selected until the distance center points of the preset number of clusters are selected; the remaining data information points are divided into new classes according to the modified Euclidean distance algorithm, and the center point of each new class is calculated using the mean method, and the iterative process of the K-means clustering algorithm is executed to complete the clustering processing of the agricultural irrigation fault load information; the load fault duration, load heavy load rate and load overload rate are extracted as characteristic attributes according to the clustering processing results, and the agricultural irrigation load categories are divided according to the characteristic attributes and the K-means clustering algorithm.

[0092] See also Figure 2 The operating process of the K-means++ algorithm is that the points farther from the existing cluster center have a greater probability of being selected as the next cluster center, so as to improve the accuracy of the clustering results and reduce the impact of the random position of the initial center point on the clustering results. The K-means++ algorithm uses the Euclidean distance between targets as the similarity evaluation criterion to find different numbers of clusters from a data set of n. The selection of the initial cluster center is to give the points farther from the currently selected initial cluster center a greater probability of becoming the next initial cluster center. The probability formula for the remaining points in the data set to be selected as the next initial cluster center is:

[0093]

[0094] Where P represents the data set, p represents the sampling point of the data set, p' represents the remaining points of the data set, D(p) represents the shortest distance to the initial cluster center currently selected by the sampling point, and D(p') represents the shortest distance to the initial cluster center selected by the remaining sampling points.

[0095] The intra-class distance in each cluster of the data set refers to the minimum value of the average distance between a sampling point and the remaining points in the same cluster. The maximum value of all intra-cluster distances is the intra-class distance in the data set. The specific expression formula is:

[0096]

[0097] Where D in(k) represents the intra-class distance in each cluster of the data source, k represents the number of clusters, c inum represents the number of points in the i-th cluster, p i Indicates the selection of sampling points in the i-th cluster, p n represents the remaining points in the i-th cluster.

[0098] The inter-class distance in each cluster of the dataset P refers to the minimum distance between the two nearest points in two different clusters. The specific expression formula is:

[0099]

[0100] Where D out (k) represents the distance between classes in each cluster of the dataset, c i With c j are two different cluster centers, p u With p v Belong to two cluster centers c i With c j The two points closest to each other.

[0101] In order to evaluate the clustering effect under different k values, a clustering effect evaluation function is constructed based on the intra-class distance and inter-class distance of the clustered data set. The clustering effect evaluation function for different k values ​​is:

[0102]

[0103] Where Q(k) represents the clustering effect evaluation function.

[0104] Specifically, when extracting load fault duration, load heavy load rate and load overload rate as characteristic attributes based on the clustering processing results, and dividing the agricultural irrigation load categories according to the characteristic attributes and the K-means clustering algorithm, the agricultural irrigation fault load information can be divided into load heavy load conditions and load overload conditions, and coordinate systems are established to describe the relationship between load fault duration and load heavy load conditions and load overload conditions; the relationship description results are combined with the K-means clustering algorithm to divide the load heavy overload severity level by setting the number of clusters; the load fault duration is obtained from the agricultural irrigation fault load information and combined with the K-means clustering algorithm to analyze the fault load duration in the flood season and the dry season; based on the load heavy load condition, load overload condition and load fault duration, the load fault concentrated time and fault duration within a day are analyzed; the load overload severity level, fault load duration, load fault concentrated time and fault duration within a day are used as agricultural irrigation load categories.

[0105] It should be explained that the multidimensional analysis method used in this embodiment is mainly divided into the following steps:

[0106] First, the clustered loads are processed and analyzed. For the clustered loads, their specific characteristics and influencing factors are analyzed to facilitate their classification and summary from multiple dimensions. After obtaining the different characteristic attributes and influencing factors of the loads, the load conditions are analyzed from different aspects, and a load classification table is established. Then, the loads obtained after algorithm clustering are refined and classified, and the load data groups are graded through the graphical data obtained by the clustering algorithm tool, which is helpful to classify the agricultural irrigation loads in the region from different aspects.

[0107] See also Figure 3 The multidimensional analysis method adopted in this embodiment is specific. This analysis method takes agricultural irrigation load as an example, mainly targeting the fault load in agricultural irrigation areas, and taking the main characteristic attributes of the fault load: load fault duration, load heavy load rate, and load overload rate as examples. The fault load is analyzed from three aspects: the severity of heavy overload, the severity of flood season and dry season, and the severity of 24 hours a day, and is specifically classified. This helps the region to subdivide the category of agricultural irrigation load through load characteristics, so as to control the regional load in a targeted manner, so as to better alleviate the problem of farmland irrigation in the region.

[0108] This example assumes that the farmland irrigation problem in a certain area is analyzed. First, the terrain and grain yield distribution of the area are set. The specific grain yield is as follows: Figure 4 As shown, the grain output in the region increases year by year. The grain in the region requires a large amount of water irrigation supply. Among them, rice, as one of the most important farmland food crops, requires regular and sufficient water supply, reaching one-third of the total grain output in the region.

[0109] With the annual increase in grain and rice production, the load pressure on the distribution network during the farmland irrigation period is also increasing. This requires a cluster analysis of the agricultural irrigation load in the region and the formulation of comprehensive measures for the region to reduce and alleviate the load pressure problems in more regions caused by farmland irrigation.

[0110] The faulty agricultural irrigation load in the area was selected as the data source, and a multidimensional method was used to analyze it. The agricultural irrigation load in the area was analyzed and discussed from the different severity levels of regional fault load (severity of heavy overload, severity of flood season and dry season, and severity of 24 hours a day).

[0111] (1) Severity of heavy overload;

[0112] Firstly, the load overload failure situation of a substation area in the region in different months is analyzed. 200 load data points of the substation area are selected as the load data set. Without using the K-means++ algorithm, a black and white scatter plot is obtained. For the specific distribution of the load data set, please refer to Figure 5 , 6 .

[0113] like Figure 5 , 6 As shown, the load data set is divided into two categories according to the heavy load situation and the overload situation, both of which are scattered load fault data groups. The horizontal axis in the figure is the load fault duration (hours), and the vertical axis is the load heavy load rate (%) and the overload rate (%). According to the duration of the load fault, it can be seen that most of the data points are distributed on the left side of the area, that is, most of the load faults last for 2-4 hours, and there are fewer data points on the right side of the area. According to the load heavy and overload rates, it can be seen that the data points in the upper and lower areas are distributed more evenly, and most of the load data points have a heavy load rate from 80% to 100% and an overload rate from 100% to 150% that are evenly distributed.

[0114] The K-means++ algorithm is used to cluster the load data points. By setting different K values ​​(K=3, K=7), the heavy load fault load data groups in the area can be divided into 3 categories and 7 categories. The load data points are divided into different load data groups, and the centroids of 200 load fault data points are found. Then the centroids are continuously updated. For the specific distribution of load data groups, please refer to Figure 7 , 8 .

[0115] Figure 7 , 8 The load and overload rate under K = 3 are divided into three categories according to the load and overload rate. The upper data group has the highest load rate, and the lower data group has the lowest load rate. The cross represents the centroid of the data group. The 200 load data points in the figure are divided into Figure 7 It is divided into three categories, as shown in Table 1:

[0116] Table 1 Load data point division

[0117] Load level 1 Load Level 2 Load Level 3 Heavy load rate 80％-87％ 87％-94％ 94％-100％ Overload rate 100％-118％ 118％-135％ 135％-150％

[0118] Table 2 Distribution of load data points

[0119] Load level 1 Load Level 2 Load Level 3 Heavy load rate 79 56 65 Overload rate 81 49 70

[0120] As can be seen from Table 1, the load levels are divided into three categories according to the severity of the heavy overload rate, among which load level 1 represents the load with a low degree of heavy load / overload severity, load level 2 represents the load with a high degree of heavy load / overload severity, and load level 3 represents the load with a very high degree of heavy load / overload severity.

[0121] It can be seen from Table 2 that when K=3, the heavy load rate and overload rate are the largest in load level 1, followed by load level 3, and finally load level 2. The two-level differentiation of heavy and overload loads in this region is quite serious, so the region should strengthen the distribution management of these two types of load levels and evenly distribute the distribution resources.

[0122] Then, cluster analysis is performed on the heavy overload data points when K=7, and we can get Fig. 9 , Fig.10 The above two figures divide the heavy and overloaded loads into 7 categories according to the load weight, overload rate and load fault duration. It can be seen that the 7 types of loads are distinguished more finely, and an additional classification indicator is added: the duration of the load fault. The loads are clustered according to the above two indicators. The severity of the load overload rate is divided into three categories: low, high, and very high. The duration of the load fault can be divided into three categories: short, long, and very long. Among them, it needs to be explained that short means the duration is 2-4 hours, long means the duration is 4-6 hours, and very long means the duration is more than 6 hours. Low means the heavy load rate of the fault load is 80%-88%, high means the heavy load rate of the fault load is 88%-96%, and very high means the heavy load rate of the fault load is 96%-100%.

[0123] The obtained table 3 is as follows:

[0124] Table 3 Heavy overload classification

[0125]

[0126] According to the load classification in Table 3 and Fig. 9 According to the heavy load distribution, heavy load data groups 1 and 7 can be divided into z-1 class loads, heavy load data groups 2 and 3 can be divided into z-2 class loads, heavy load data group 6 can be divided into z-3 class loads, heavy load data group 4 can be divided into z-6 class loads, and heavy load 5 can be divided into z-3 class loads.

[0127] According to the load classification in Table 3 and Fig.10 The overload data group 2 and 6 can be divided into g-1 load, the overload data group 4 and 7 can be divided into g-4 load, the overload data group 1 and 5 can be divided into g-7 load, the overload data group 3 can be divided into g-3 load, and the ...3 can be divided into g-3 load. Fig.10 It can be seen that the overload data groups 1, 2, and 5 occupy a larger area in the load fault duration. When the K value is larger, the load classification can be further refined. The clustering result is similar to the load overload situation. It can be seen that the larger the K value, the finer the clustering of the load data group, and the more obvious the distinction between load types.

[0128] (2) the severity of the wet and dry seasons;

[0129] The fault load in the region is clustered by the dry season (January) and the flood season (June).

[0130] See also Fig.11 , take K = 3, take January (days) as the horizontal axis, and the load fault duration (hours) as the vertical axis, and observe the clustering of the load in January (30 data points) of the substation. Fig.11 It can be seen that there are 3 centroids after load clustering, and the load clustering is relatively clear. The 3 centroids are located at the 4th, 19th and 25th time points respectively. The clustering is mainly classified according to time, which is divided into load level 1, load level 2 and load level 3. See Table 4 for the time period. According to the distribution law of load data points in the figure, it can be seen that the agricultural irrigation load failures in the region in January are mainly concentrated in the second half of January, but because there are not many overall load data points, the region needs to strengthen the relevant load control in the second half of January compared with the first half of January, and adopt decentralized irrigation for the region.

[0131] Table 4 Distribution of load data points during the dry season (January)

[0132] Load level 1 Load Level 2 Load Level 3 time No. 1-10 No. 11-22 No. 23-31

[0133] See also Fig.12 , take K = 3, take June as the representative of the flood season, take June (days) as the horizontal axis, and the load failure duration (hours) as the vertical axis, and observe the clustering of the load in June (200 data points) of the substation. Fig.12 It can be seen that there are 3 centroids after load clustering, and the 3 centroids are located at the time points of the 5th, 10th and 26th respectively. The clustering is mainly based on the horizontal axis time line for statistical classification. There are more data points for load 2 and load 3, which means that the agricultural irrigation load increased in the early June and the end of June. The region needs to strengthen the load control in these two time periods. In addition, since the load classification in the flood season is similar to that in the dry season, it will not be repeated here.

[0134] contrast Fig.11 , 12It can be seen that there are very few load data points in the dry season, with only 30 data points counted, while there are many load data points in the flood season, with a total of 200 data points counted. This is because the substation area is mainly used for farmland irrigation in the flood season. The continuous water use in June will cause frequent heavy overload failures in the regional load. At the same time, the K-means++ algorithm classifies the load according to the load duration and divides it into loads of different degrees, which is convenient for the later clustering analysis and statistics of the load data in the region, and also provides a basis for the effective management and control of regional power resources.

[0135] (3) 24-hour severity;

[0136] See also Fig.13 , statistics of agricultural irrigation load failures (50 data points) for 24 hours a day during the flood season.

[0137] The K-means++ algorithm is used to set the K value to 5, and the 50 load data points in the area are divided into 5 categories. The specific load clustering is as follows: Fig.13 As shown by Fig.13 It can be seen that the two load data groups on the left represent the load data groups with high load fault duration and low load fault duration during the early morning, respectively. The two load data groups in the middle represent the load groups in the early morning and morning, and the load data group on the right represents the load data group in the afternoon and evening. The clustering algorithm is used to divide the load into 4 categories (load 1, load 3, load 4, load 5) according to the overall distribution of the load in one day, and into 2 categories (load 2, load 4) according to the load duration distribution.

[0138] The clustered load data can be divided into 9 categories according to the load characteristics. According to the duration of the load fault, it can be divided into short (2-4 hours), long (4-6 hours), and very long (6-8 hours). According to the number of hours in a day, it can be divided into three time periods: early morning (0-8 o'clock), daytime (9-16 o'clock), and nighttime (17-23 o'clock). The fault load can be divided into 9 categories. Specifically, the fault load classification is shown in Table 5. It should be explained that the specific data definitions of short, long, and very long are consistent with Table 3. The early morning period indicates the time period of 0:00-8:59, the daytime period indicates the time period of 9:00-16:59, and the night period indicates the time period of 17:00-23:59.

[0139] Table 5 Fault load classification within one day

[0140]

[0141] It can be seen from Table 5 that Fig.13According to the above classification table, load 1 can be classified as load 2, load 2 can be classified as load 7, load 3 can be classified as load 3, load 4 can be classified as load 1, and load 5 can be classified as load 4. Fig.13 It can be seen that the agricultural irrigation load failure time in the area is concentrated in the early morning (0-8 o'clock), the load failure duration is 2-4h and 5-8h, and the load failure severity is relatively large; while the fault load during the daytime (9-16 o'clock) and nighttime (17-23 o'clock) is less, and the load failure duration is 2-5h, and the load failure severity is relatively small, reflecting that the substation mainly concentrates the power load for agricultural field irrigation at night. Due to the long duration of power consumption and the simultaneous power consumption in many places, the region has a heavy overload problem; in addition, it can be seen that the substation has fewer load data points during the day, and the load duration is shorter, and the severity is lower. The area needs to pay attention to time-sharing irrigation, shorten the night irrigation time and increase the daytime irrigation time, and also needs to reduce the centralized power supply of irrigation load.

[0142] Therefore, this embodiment uses the K-means++ algorithm to cluster the faulty agricultural irrigation loads in the substation, and then analyzes the faulty load characteristics from a multi-dimensional perspective. The faulty loads in the substation are described in three aspects (severity of heavy overload, severity of flood season and dry season, and severity of 24 hours a day). The main purpose of this multi-dimensional analysis method is to refine the clustering of different load data points. By dividing the load data group into different load clustering centers, the specific load classification can be obtained, and then the multi-attribute characteristics and multi-dimensionality of the load can be used to manage and allocate different types of loads, which can better alleviate the agricultural irrigation problem in the region, and is conducive to the reasonable allocation of electricity resources and improving the utilization rate of power resources.

[0143] Then, the distribution pattern of agricultural irrigation load in irrigation areas is analyzed from multiple dimensions (three aspects of fault load: severity of heavy overload, severity of flood season and dry season, severity of 24 hours a day), and the characteristics of agricultural irrigation load in the region are summarized and summarized; according to the agricultural irrigation load scatter plot under different dimensions selected by the algorithm, the above agricultural irrigation load scatter plot can be divided into several representative categories through multidimensional analysis method, and each category reflects the irrigation problems and needs specific to the agricultural irrigation load.

[0144] Combined with the load characteristics after load clustering analysis, through the above-mentioned multidimensional analysis method, taking multiple different parameter variables to cluster the load, the agricultural irrigation load resources in the region can be better allocated effectively, so as to reduce the waste of regional power resources. Through the multidimensional analysis method, the regional agricultural irrigation load can be divided into multiple categories, and the algorithm can be applied to the power field for load clustering. Cluster analysis of the farmland irrigation load situation in a certain area is carried out from different dimensions, which is expected to help the region alleviate farmland irrigation problems.

[0145] Step S3, establishing a control model for agricultural irrigation loads participating in power operation centered on the agricultural irrigation load category, and using the control model to allocate power resources to control the irrigation pumps based on the evaluated working capacity of the irrigation pumps.

[0146] In this embodiment, when establishing a control model for agricultural irrigation loads participating in power operation centered on agricultural irrigation load categories, and using the control model to allocate power resources to control the irrigation water pumps based on the evaluation of the working capacity of the irrigation water pumps, the distributed operation model can be used to integrate the power generation equipment and the irrigation water pumps to achieve remote control and signal demodulation of the irrigation water pumps, and the demodulated signals are integrated into lines to achieve unified management of the irrigation water pumps; a grid monitoring model is established using the two-layer modeling theory to achieve communication and monitoring between power equipment and irrigation water pumps during power transmission; the agricultural irrigation load categories are combined with the generative adversarial network to generate a power operation control program to obtain the power distribution operation safety standards of the irrigation water pumps under the condition of solving agricultural irrigation fault loads; the power generated by the irrigation water pumps during use is obtained, and the irrigation water pump power is analyzed using a timing simulation algorithm to evaluate the working capacity of the irrigation water pumps, and the control model is used to allocate power resources to control the irrigation water pumps based on the evaluated working capacity of the irrigation water pumps.

[0147] Specifically, when using the distributed operation model to integrate the power generation equipment with the irrigation water pump to realize remote control and signal demodulation of the irrigation water pump, and integrating the demodulated signal into lines to realize unified management of the irrigation water pump, the overall architecture can be designed, including the connection method, communication protocol and distributed control system between the power generation equipment and the irrigation water pump; determine the technical requirements of the system, such as hardware platform, software tools and network requirements; the design should consider the scalability, security and reliability of the system; select interfaces and communication protocols, such as using Modbus, CAN bus or Ethernet for physical connection between the power generation equipment and the irrigation water pump; use the distributed control model to develop a remote control system to manage the operation of the power generation equipment and the irrigation water pump; demodulate the signals received from the power generation equipment and the irrigation water pump, and process the signals with appropriate demodulation technology to ensure the accuracy and integrity of the data; integrate data from different devices to ensure information synchronization and consistency; integrate the demodulated data into lines to achieve unified data management and control.

[0148] Among them, when using the two-layer modeling theory to establish a grid monitoring model to realize the communication and monitoring of power equipment and irrigation water pumps during power transmission, the functional requirements, performance indicators and safety requirements of the monitoring system can be determined; the operating characteristics of power transmission and irrigation water pumps are analyzed to determine the key parameters that need to be monitored and the possible failure modes; according to the two-layer modeling theory, the monitoring model of power equipment and irrigation water pumps is designed. The first layer of the model is responsible for collecting and processing data from the power equipment, and the second layer of the model focuses on the operating data of the irrigation water pump. Effective integration and interaction of data need to be achieved between the two layers; design a grid structure that adapts to the characteristics of the power and irrigation systems, and ensure that the grid structure can carry data streams from different devices and realize efficient data communication and processing; select or develop protocols suitable for communication between power and irrigation equipment; integrate the monitoring system of power equipment and irrigation water pumps.

[0149] Specifically, when combining the agricultural irrigation load category with the generative adversarial network to generate the power operation control program to obtain the distribution operation safety standard of the irrigation water pump under the condition of solving the agricultural irrigation fault load, the generative adversarial network algorithm can be used to analyze the agricultural irrigation fault load information corresponding to the agricultural irrigation load category, and the agricultural irrigation fault load information can be trained and analyzed by game theory; according to the analysis results, the expected distribution safety standard is digitized to obtain a safety expression, and based on the distribution safety standard, the generative adversarial network algorithm is used to perform a safety analysis to obtain the interactive optimal safety goal between the power information and the irrigation water pump; the interactive optimal safety goal is used to represent the relationship between the current information and the distribution safety standard, and the power operation control program is written using a microcomputer algorithm based on the relationship result, and the distribution safety standard is written into the power operation control program as a scheduling plan; according to the agricultural irrigation fault load information, the probability that the initial scheduling plan can solve the agricultural irrigation fault load situation is judged, and if the probability result is less than the preset value, the safety expression is re-obtained; if the probability result is greater than the preset value, the distributed operation model is combined with the power operation control program to complete the setting of the operating value range of the irrigation water pump.

[0150] Specifically, when obtaining the power generated by the irrigation water pump during use, using the time series simulation algorithm to analyze the power of the irrigation water pump to evaluate the working capacity of the irrigation water pump, and using the control model to allocate power resources to control the irrigation water pump based on the evaluated working capacity of the irrigation water pump, the power output of the irrigation water pump during use can be obtained based on the irrigation history of the agricultural area, and the power fluctuation curve of the irrigation water pump at the preset sampling interval can be generated according to the power output data condition fitting; the proportion of the active output fluctuation of the irrigation water pump power at different time scale sampling intervals is determined according to the power fluctuation curve, and the maximum output of the irrigation water pump is defined according to the proportion of active output fluctuation; the cluster effect coefficient is defined based on the ratio of the maximum output of the irrigation water pump to the rated capacity of the agricultural area irrigation, and the fluctuation law of the cluster effect coefficient is analyzed to analyze the impact of the cluster effect on the output fluctuation of the irrigation water pump; an objective function is established based on the output fluctuation impact and the operating cost of the irrigation water pump, and the working capacity of the irrigation water pump is evaluated and calculated by applying the time series simulation algorithm considering the balance constraint and the irrigation speed constraint; the working capacity of the irrigation water pump is matched with the operating value range, and power resources are allocated according to the corresponding results to control the irrigation water pump.

[0151] Step S4, starting the irrigation water pump to perform agricultural irrigation operations according to the allocation results, and establishing an efficiency evaluation model based on the irrigation results to verify the farmland irrigation efficiency, and optimizing the control state of the irrigation water pump based on the verification results.

[0152] In this embodiment, when the irrigation water pump is started to perform the agricultural irrigation operation according to the allocation result, and the efficiency evaluation model is established according to the irrigation result to verify the farmland irrigation efficiency, and the control state of the irrigation water pump is optimized based on the verification result, the irrigation water pump can be started to perform the agricultural irrigation operation according to the power resource allocation result, and the irrigation parameters are obtained after the agricultural irrigation operation is completed, and the irrigation parameters include the water flow rate, pressure and humidity of the irrigation agricultural area of ​​the irrigation water pump; a change curve is generated based on the irrigation parameters, and the characteristic index vectors of the water flow rate, pressure and humidity of the irrigation agricultural area of ​​the irrigation water pump are respectively formed according to the time domain and frequency domain characteristic indicators of the extracted curve and the modal decomposition singular value entropy; the principal component analysis method is used to eliminate the redundancy of the information of the water flow rate, pressure and humidity of the irrigation agricultural area of ​​the irrigation water pump, and the corresponding characteristic index sample library is constructed; the farmland irrigation state is divided based on the characteristic index sample library and the continuous hidden Markov model, and a comprehensive irrigation state evaluation model of the search algorithm optimized support vector machine is established according to the division result; the comprehensive state evaluation model is used to evaluate and verify the farmland irrigation efficiency, and the control state of the irrigation water pump is optimized according to the evaluation and verification results.

[0153] Specifically, when the irrigation water pump is started according to the power resource allocation result to perform the agricultural irrigation operation, and the irrigation parameters are obtained after the agricultural irrigation operation is completed, the basic configuration of the irrigation system can be set, including the connection of the irrigation water pump and the allocation of power resources; according to the power resource allocation result, the power supply and control system of the water pump are configured to ensure that the system can start and run according to the predetermined plan; the irrigation water pump is started according to the power allocation and irrigation needs; the actual irrigation operation is carried out to irrigate the designated land area according to the needs of the agricultural area, and the running time and flow of the water pump are controlled to ensure uniform and effective irrigation; during and after the irrigation process, key irrigation parameters are collected; all key data in the irrigation process are recorded, and the recorded data include changes in water flow, pressure and humidity.

[0154] Specifically, when generating a change curve based on irrigation parameters, and forming characteristic index vectors of water flow, pressure of irrigation water pumps and humidity of irrigated agricultural areas according to the extracted time domain and frequency domain characteristic indicators and modal decomposition singular value entropy of the curve, the water flow, pressure and humidity data of irrigation water pumps and irrigated agricultural areas can be collected to ensure the integrity and accuracy of the data, and the data can be pre-processed, such as denoising and interpolation of missing values, to provide a clean and continuous data sequence for subsequent analysis; a time series change curve of water flow, pressure and humidity is generated according to the collected data; a time series curve is drawn using chart software or programming tools to intuitively display the time series curve. It shows the change of parameters over time; extracts key feature indicators from the time domain and frequency domain. Specific time domain features include mean, variance, peak and other statistics; frequency domain features include analyzing frequency composition through Fourier transform (FFT), extracting main frequency, spectrum energy, etc.; modal decomposition and singular value entropy calculation, modal decomposition of time series data, decomposition of data sequence, obtaining components of different modes, and calculating singular value entropy for each modal component; the extracted time domain, frequency domain features and singular value entropy are combined into a feature indicator vector, and for each parameter (water flow, pressure, humidity), a vector containing all relevant features is created.

[0155] Specifically, when the principal component analysis method is used to eliminate the redundancy of the water flow, pressure of the irrigation pump and the humidity information of the irrigated agricultural area and to construct the corresponding characteristic index sample library, the water flow, pressure and humidity data can be constructed into a characteristic matrix, in which each row represents a sample and each column represents a feature; the covariance matrix eigenvalues ​​and eigenvectors of the characteristic matrix are calculated, and the principal components are selected according to the size of the eigenvalues; the principal components are selected according to the proportion of explained variance; and the characteristic index sample library is constructed using the selected principal components, mainly by converting the original data into the principal component space to obtain a new characteristic index sample library. The features in this sample library are linearly independent, which reduces redundancy.

[0156] Specifically, when dividing the farmland irrigation status based on the feature indicator sample library and the continuous hidden Markov model, and establishing an irrigation status comprehensive evaluation model of the search algorithm-optimized support vector machine according to the division results, the maximum expectation algorithm can be used to train the continuous hidden Markov model, and after the training is completed, the feature indicator sample sequence is defined according to the feature indicator sample library, and the dynamic programming algorithm is used to find the hidden state sequence to explain the feature indicator sample sequence; the forward-backward technique is used to analyze the likelihood probability value of the feature indicator sample sequence under the continuous hidden Markov model, and the state information corresponding to the feature indicator sample sequence is determined according to the likelihood probability value to obtain the farmland irrigation status; the search algorithm is used to find the optimal parameters to optimize the support vector machine, and the farmland irrigation status is used as input and combined with the optimized support vector machine to establish a comprehensive irrigation status evaluation model.

[0157] Among them, when using the search algorithm to find the optimal parameters to optimize the support vector machine, and taking the farmland irrigation status as input and combining it with the optimized support vector machine to establish an irrigation status comprehensive evaluation model, the farmland irrigation status can be preprocessed with grey correlation, and the farmland irrigation status can be sorted according to the correlation degree based on the processing results, and the farmland irrigation status indicators can be screened out in combination with the prior knowledge base; after pre-initializing the search algorithm parameters, the fitness value of the search individual is calculated using the search algorithm implementation principle, and the optimal parameters are selected according to the optimal fitness value to establish a support vector regression model; the particle swarm optimization algorithm is used to define the comprehensive fitness function and the linear decreasing inertia weight strategy to identify the hyperparameters of the support vector regression model to obtain the optimized support vector machine; the farmland irrigation status is taken as input, and the farmland irrigation status indicators are used as labels and combined with the optimized support vector machine to establish an irrigation status comprehensive evaluation model.

[0158] To sum up, with the help of the above-mentioned technical scheme of the present invention, the present invention analyzes regional water usage data through a clustering algorithm, can accurately identify the characteristics of agricultural irrigation loads, and provide an accurate basis for subsequent analysis and control. The dimension categories are defined according to the characteristic attributes of agricultural irrigation fault load information, which can analyze the fault load more comprehensively and in-depth, provide a scientific basis for the classification of agricultural irrigation load categories, and establish a control model for agricultural irrigation loads participating in power operation centered on agricultural irrigation load categories, which can more effectively allocate power resources and ensure that irrigation water pumps obtain sufficient power support when needed. At the same time, based on the working capacity of the irrigation water pump, the control model is used to allocate power resources, which can optimize the operating state of the irrigation water pump and improve its working efficiency and irrigation effect. It provides a scientific and efficient management means for farmland irrigation by accurately identifying agricultural irrigation load characteristics, effectively screening fault information, multi-dimensional analysis, optimizing power resource allocation, improving work efficiency, realizing quantitative evaluation, and continuously optimizing irrigation systems. The present invention takes the main attributes of agricultural irrigation fault load as an example, and reflects the specific irrigation problems and demands of agricultural irrigation load after algorithm analysis, and uses a multidimensional analysis method of agricultural irrigation load to perform cluster analysis on the farmland irrigation load situation in a certain area from different dimensions. Through the multidimensional analysis method, the regional agricultural irrigation load can be divided into multiple categories, which helps the region to subdivide the categories of agricultural irrigation load according to load characteristics, so as to control the regional load in a targeted manner, so as to better alleviate the regional farmland irrigation problem, help to reasonably manage regional resources, guide the reasonable allocation of distribution resources and improve the utilization efficiency of distribution resources.

[0159] The present invention realizes remote control, monitoring and optimal allocation of power resources of irrigation water pumps by establishing a control model in which agricultural irrigation loads participate in power operation with agricultural irrigation load categories as the center. Based on the distributed operation model, the power generation equipment and the irrigation water pump are integrated and connected to realize remote control and signal demodulation of the irrigation water pump. The grid monitoring model established by the double-layer modeling theory is used to realize communication and monitoring between power equipment and irrigation water pumps during power transmission, thereby ensuring stable operation, timely discovering and handling potential faults, and avoiding irrigation interruption or power waste caused by equipment failure. At the same time, the agricultural irrigation load category is combined with the generative adversarial network to generate a power operation control program, which can obtain the power distribution operation safety standard of the irrigation water pump under the condition of solving agricultural irrigation fault load, which is helpful to reasonably allocate power resources while ensuring irrigation needs and avoid imbalance between power supply and demand. By obtaining the power data generated by the irrigation water pump during use and analyzing it using a timing simulation algorithm, the working capacity of the irrigation water pump can be accurately evaluated, which is convenient for accurate allocation of power resources according to the actual working state of the water pump, ensuring that the water pump operates in the best working state, and improving irrigation efficiency and effect.

[0160] Those skilled in the art will appreciate that the units, i.e., algorithm steps, of the various examples described in conjunction with this embodiment can be implemented in electronic hardware or in a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0161] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A multidimensional clustering analysis method for agricultural irrigation load based on clustering algorithm, characterized in that: The multidimensional cluster analysis method of agricultural irrigation load includes the following steps: The power load data is analyzed by clustering algorithm, and the power load data is combined with the agricultural irrigation load information to determine the agricultural irrigation load characteristics, and the agricultural irrigation fault load information is selected from the agricultural irrigation load characteristics; Analyze the fault load information by defining dimension categories according to the characteristic attributes corresponding to the agricultural irrigation fault load information, and classify the agricultural irrigation load categories based on the fault load information; Establish a control model for agricultural irrigation loads participating in power operation, centered on agricultural irrigation load categories, and use the control model to allocate power resources to control irrigation pumps based on the assessed working capacity of irrigation pumps; According to the allocation results, the irrigation pump is started to perform irrigation operations in the agricultural area, and an efficiency evaluation model is established based on the irrigation results to verify the efficiency of farmland irrigation, and the control state of the irrigation pump is optimized based on the verification results; The step of defining dimension categories according to characteristic attributes corresponding to the agricultural irrigation fault load information to analyze the fault load information and classifying the agricultural irrigation load categories based on the fault load information includes the following steps: The agricultural irrigation fault load information is used as the data source, the final number of clusters is pre-set according to the K-means++ algorithm, and a group of data information points are randomly selected from the data source as the cluster center points; Calculate the modified Euclidean distance between the remaining data information points, and select cluster centers from the remaining data information points to calculate the intra-class distance in each cluster of the data source; Select new data information points as new cluster centers according to the calculation results and the weighted probability distribution algorithm, and repeatedly select new cluster centers until the distance centers of the preset number of clusters are selected; According to the modified Euclidean distance algorithm, the remaining data information points are divided into new clusters with the closest cluster center, and the mean method is used to calculate the center point of each new cluster, and then the iterative process of the K-means clustering algorithm is performed to complete the clustering processing of the agricultural irrigation fault load information. According to the clustering processing results, load fault duration, load overload rate and load overload rate are extracted as characteristic attributes, and agricultural irrigation load categories are divided according to the characteristic attributes and K-means clustering algorithm; The method of extracting load fault duration, load overload rate and load overload rate as characteristic attributes according to the clustering processing results, and classifying agricultural irrigation load categories according to the characteristic attributes and the K-means clustering algorithm includes the following steps: The agricultural irrigation fault load information is divided into load heavy load situation and load overload situation, and coordinate systems are established to describe the relationship between load fault duration and load heavy load situation and load overload situation respectively; The relationship description results are combined with the K-means clustering algorithm to divide the severity level of load overload by setting the number of clusters; The load fault duration is obtained from the agricultural irrigation fault load information and combined with the K-means clustering algorithm to analyze the fault load duration in the flood season and the dry season; Analyze the load failure concentration time and failure duration within a day based on load heavy load, load overload and load failure duration; The severity level of load overload, duration of fault load, concentrated time of load faults within a day and duration of faults are used as agricultural irrigation load categories.

2. The multidimensional clustering analysis method of agricultural irrigation load based on clustering algorithm according to claim 1 is characterized in that: The calculation formula of the intra-class distance in each cluster of the data source is: Where D in (k) represents the intra-class distance in each cluster of the data source, k represents the number of clusters, c inum represents the number of points in the i-th cluster, p i Indicates the selection of sampling points in the i-th cluster, p n represents the remaining points in the i-th cluster.

3. The multidimensional clustering analysis method of agricultural irrigation load based on clustering algorithm according to claim 2 is characterized in that: The method of establishing a control model for agricultural irrigation loads participating in power operation centered on the agricultural irrigation load category and using the control model to allocate power resources to control the irrigation pumps according to the working capacity of the irrigation pumps includes the following steps: The distributed operation model is used to integrate the power generation equipment with the irrigation water pump to realize remote control and signal demodulation of the irrigation water pump, and the demodulated signal is integrated into the line to realize the unified management of the irrigation water pump; A grid monitoring model is established using the double-layer modeling theory to achieve communication and monitoring between power equipment and irrigation pumps during power transmission; The agricultural irrigation load category is combined with the generative adversarial network to generate the power operation control program to obtain the power distribution operation safety standard of the irrigation pump under the condition of solving the agricultural irrigation fault load; The power generated by the irrigation water pump during use is obtained, and the time series simulation algorithm is used to analyze the power of the irrigation water pump to evaluate the working capacity of the irrigation water pump. Based on the evaluated working capacity of the irrigation water pump, the control model is used to allocate power resources to control the irrigation water pump.

4. The multidimensional clustering analysis method of agricultural irrigation load based on clustering algorithm according to claim 3 is characterized in that: The method of combining the agricultural irrigation load category with the generative adversarial network to generate a power operation control program to obtain the power distribution operation safety standard of the irrigation water pump under the condition of solving the agricultural irrigation fault load includes the following steps: The generative adversarial network algorithm is used to analyze the agricultural irrigation fault load information corresponding to the agricultural irrigation load category, and the agricultural irrigation fault load information is trained and analyzed through game theory. According to the analysis results, the desired power distribution safety standard is digitized to obtain a safety expression, and based on the power distribution safety standard, a generative adversarial network algorithm is used to perform a safety analysis to obtain the optimal safety target for the interaction between power information and irrigation water pumps. The interactive optimal safety target is used to represent the relationship between current information and power distribution safety standards. Based on the relationship results, a power operation control program is written using a microcomputer algorithm, and the power distribution safety standards are written into the power operation control program as a scheduling plan. According to the agricultural irrigation fault load information, the probability that the initial scheduling scheme can solve the agricultural irrigation fault load situation is determined. If the probability result is less than the preset value, the safety expression is obtained again. If the probability result is greater than the preset value, the distributed operation model is combined with the power operation control program to complete the operation value range setting of the irrigation water pump.

5. The method for multidimensional clustering analysis of agricultural irrigation load based on clustering algorithm according to claim 4 is characterized in that: The method of obtaining the power generated by the irrigation water pump during use, analyzing the power of the irrigation water pump using a time series simulation algorithm to evaluate the working capacity of the irrigation water pump, and allocating power resources to control the irrigation water pump using a control model according to the evaluated working capacity of the irrigation water pump includes the following steps: Based on the irrigation history of the agricultural area, the power output of the irrigation pump during use is obtained, and the power fluctuation quantity curve of the irrigation pump at a preset sampling interval is generated according to the power output data condition fitting; According to the power fluctuation curve, the proportion of active output fluctuation of the irrigation water pump power at different time scale sampling intervals is determined, and the maximum output of the irrigation water pump is defined according to the proportion of active output fluctuation; The cluster effect coefficient is defined based on the ratio of the maximum output of the irrigation pump to the rated capacity of the agricultural area irrigation, and the fluctuation law of the cluster effect coefficient is analyzed to analyze the impact of the cluster effect on the fluctuation of the output of the irrigation pump; The objective function is established based on the impact of output fluctuation and the operating cost of the irrigation pump, and the working capacity of the irrigation pump is evaluated and calculated by applying the time series simulation algorithm considering the balance constraints and irrigation speed constraints. The working capacity of the irrigation water pump is matched with the operating value range, and power resources are allocated according to the corresponding results to control the irrigation water pump.

6. The method for multidimensional clustering analysis of agricultural irrigation load based on clustering algorithm according to claim 1, characterized in that: The method of starting the irrigation water pump according to the allocation result to perform the irrigation operation in the farm area, establishing an efficiency evaluation model according to the irrigation result to verify the farmland irrigation efficiency, and optimizing the control state of the irrigation water pump based on the verification result includes the following steps: Starting the irrigation water pump to perform the irrigation operation of the agricultural area according to the power resource allocation result, and obtaining the irrigation parameters after the irrigation operation of the agricultural area is completed, wherein the irrigation parameters include the water flow and pressure of the irrigation water pump and the humidity of the irrigation agricultural area; Generate a change curve based on irrigation parameters, and form characteristic index vectors of irrigation water flow, pressure and humidity in irrigation agricultural areas according to the extracted curve time domain and frequency domain characteristic indexes and modal decomposition singular value entropy; The principal component analysis method is used to eliminate the redundancy of irrigation pump water flow, pressure and irrigation area humidity information, and to construct a corresponding characteristic index sample library; The farmland irrigation status is divided based on the characteristic index sample library and the continuous hidden Markov model, and a comprehensive evaluation model of irrigation status based on the search algorithm and support vector machine is established according to the division results; The comprehensive state evaluation model is used to evaluate and verify the farmland irrigation efficiency, and the control state of the irrigation pump is optimized based on the evaluation and verification results.

7. The method for multidimensional clustering analysis of agricultural irrigation load based on clustering algorithm according to claim 6, characterized in that: The method of dividing the farmland irrigation status based on the characteristic index sample library and the continuous hidden Markov model, and establishing the irrigation status comprehensive evaluation model of the search algorithm optimized support vector machine according to the division result includes the following steps: The maximum expectation algorithm is used to train the continuous hidden Markov model, and after the training is completed, the characteristic indicator sample sequence is defined according to the characteristic indicator sample library, and the dynamic programming algorithm is used to find the hidden state sequence to explain the characteristic indicator sample sequence; The forward-backward technique is used to analyze the likelihood probability value of the characteristic index sample sequence under the continuous hidden Markov model, and the state information corresponding to the characteristic index sample sequence is determined according to the likelihood probability value to obtain the farmland irrigation state; The search algorithm is used to find the optimal parameters to optimize the support vector machine, and the farmland irrigation status is used as input and combined with the optimized support vector machine to establish a comprehensive irrigation status evaluation model.

8. The method for multidimensional clustering analysis of agricultural irrigation load based on clustering algorithm according to claim 7, characterized in that: The method of using a search algorithm to find the optimal parameters to optimize the support vector machine, and using the farmland irrigation status as input and combining it with the optimized support vector machine to establish an irrigation status comprehensive evaluation model includes the following steps: The farmland irrigation status is preprocessed with grey correlation, and the farmland irrigation status is sorted according to the correlation degree based on the processing results, and the farmland irrigation status indicators are screened out in combination with the prior knowledge base; After pre-initializing the search algorithm parameters, the fitness value of the search individual is calculated using the search algorithm implementation principle, and the optimal parameters are selected according to the optimal fitness value to establish the support vector regression model; The particle swarm optimization algorithm is used to define the comprehensive fitness function and the linear decreasing inertia weight strategy to identify the hyperparameters of the support vector regression model to obtain the optimized support vector machine. The farmland irrigation status was taken as input, and the farmland irrigation status indicators were used as labels and combined with the optimized support vector machine to establish a comprehensive irrigation status evaluation model.

Images