Multi-objective optimization-based zoned irrigation control method and system for drip irrigation pipe network

Abstract

The invention relates to a drip irrigation pipe network zoning irrigation control method and system based on multi-objective optimization, and belongs to the technical field of artificial intelligence. The method comprises the following steps of collecting multi-source heterogeneous data and sampling point space coordinates, after abnormal value elimination and normalization processing, splicing features and normalized coordinates to form a joint feature vector, and obtaining irrigation partitions with continuous space and similar features through K-means clustering; constructing a multi-objective optimization model taking the zoning irrigation duration and the inlet target flow as decision variables; solving by adopting a hybrid strategy multi-objective evolutionary algorithm of chaotic mapping initialization and adaptive hybrid variation to obtain a Pareto optimal solution set; and finally, combining an entropy weight method and decision maker preference, determining an optimal irrigation parameter through a superior and inferior solution distance method, and introducing a real-time soil moisture monitoring feedback correction parameter. Partitioned precise irrigation of the drip irrigation pipe network can be achieved, multiple optimization objectives are balanced, the irrigation effect and the resource utilization efficiency are improved, and system energy consumption is reduced.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of artificial intelligence, and particularly relates to a drip irrigation pipe network partition irrigation control method and system based on multi-objective optimization. BACKGROUND

[0002] With the increasingly serious global water resource shortage and the increasingly intensified contradiction between supply and demand of agricultural water, developing efficient water-saving irrigation technology has become a strategic choice to guarantee food security and sustainable utilization of water resources. As an advanced irrigation technology, drip irrigation has been widely applied in arid and semi-arid areas and facility agriculture in China, which directly delivers water and nutrients to the vicinity of crop roots through a pipe system, and has the advantages of water saving, yield increasing and strong adaptability. However, in the actual operation process of the drip irrigation system, due to the spatial variability of factors such as soil type, terrain slope, crop variety and growth stage in the irrigation area, the traditional unified irrigation management method often uses fixed irrigation time and flow, which cannot provide differentiated water supply according to the actual water demand of different regions, resulting in poor irrigation uniformity, serious deep water seepage and prominent energy waste, which seriously restricts the full play of the water-saving potential of the drip irrigation system.

[0003] The prior art has the following defects: the conventional irrigation partition method either only performs geometric division based on spatial coordinates, resulting in large differences in characteristics such as soil moisture content and crop coefficient within the partition, and failing to achieve precise irrigation on demand, or only performs clustering based on feature similarity, resulting in fragmented and dispersed partitions in space, with the same color points scattered in different regions, causing difficulties in pipe network layout and confusion in water and fertilizer management; the conventional irrigation optimization model often ignores the coupling effect of pipe network pressure constraints on partition flow, does not consider the spatial variability of soil moisture, or only optimizes a single target, and cannot effectively balance the three conflicting targets of irrigation uniformity, system energy consumption and water resource utilization efficiency; the conventional multi-objective evolutionary algorithm either causes insufficient population diversity due to improper initialization, and easily falls into local optimum when solving high-dimensional complex optimization problems, or has a single search strategy, and cannot fully explore the global space in the early iteration stage and finely search the beneficial area in the later stage, resulting in low quality of the final solution set and deviation of the Pareto front from the ideal region; the existing irrigation control method is mostly open-loop control, lacks objective weights when selecting the optimal solution from the Pareto solution set, often relies on subjective experience, and cannot be adjusted in real time according to the change of soil moisture in the actual irrigation process, so that the irrigation accuracy is difficult to guarantee when encountering uncertain factors such as model error, weather change or sensor noise. SUMMARY

[0004] To achieve the above-mentioned purpose, the application realizes the following technical scheme: The application provides a drip irrigation pipe network partition irrigation control method based on multi-objective optimization, comprising the following steps: S1, uniformly arrange a plurality of sampling points in the target irrigation area, record the spatial coordinates of each sampling point, and collect multi-source heterogeneous farmland data of each sampling point; S2, performing outlier rejection and normalization processing on the multi-source heterogeneous farmland data to obtain normalized features; S3, clustering the normalized features and the spatial coordinates by constructing a joint feature vector and using a K-means clustering method to obtain a partition label of each sampling point; S4, constructing a multi-objective irrigation optimization model based on partition characteristics, taking the irrigation time of each partition and the inlet target flow as decision variables, correlating the partition flow and pressure through a pipe network hydraulic model, defining three objective functions of irrigation uniformity, system energy consumption and water resource utilization efficiency, and realizing modeling of multi-objective irrigation optimization; S5, using a hybrid strategy multi-objective evolutionary algorithm to obtain a Pareto optimal solution set through chaotic mapping initialization, adaptive hybrid mutation, fast non-dominated sorting and crowding distance selection; S6, using the distance method between superior and inferior solutions combined with entropy weight method and decision maker preference, introducing real-time feedback adjustment, and correcting irrigation parameters according to actual soil moisture monitoring.

[0005] Further, in step S1, the key features in the multi-source heterogeneous farmland data of each sampling point include soil moisture content, soil saturated hydraulic conductivity, terrain slope, crop coefficient and reference evapotranspiration.

[0006] Further, in step S2, for each key feature, the mean and standard deviation of all sampling points are calculated, if the feature value of a point exceeds the range of mean plus or minus three times the standard deviation, it is considered as an outlier, and the median of the feature in all points is replaced; the features of the data after removing outliers are linearly mapped to interval to eliminate the influence, for each feature dimension, the minimum and maximum values of the cleaned feature values of all sampling points in the dimension are calculated, then the minimum and maximum value normalization method is used to linearly transform the feature value of each sampling point, to obtain the normalized feature value, and then the normalized feature vector of each sampling point is obtained; define , which represents the normalized value of the th feature of the th sampling point, dimensionless, the value range is , the feature vector of the th sampling point after minimum and maximum value normalization is denoted as .

[0007] To realize spatial continuity and feature similarity of irrigation partition, the normalized features and the spatial coordinates are combined for clustering. The conventional clustering method only considers feature similarity and ignores spatial adjacency, which may easily lead to fragmented partitions or unreasonable boundaries.

[0008] Further, in step S3, the spatial coordinates of each sampling point are normalized to obtain normalized coordinates. The normalized feature vector is then concatenated with the normalized coordinates to form a joint feature vector: Define the first... The joint feature vector of the sampling points is The dimension is 7. , Indicates the first The normalized value of the x-coordinate of each sampling point. Indicates the first The normalized values ​​of the ordinates of each sampling point. This invention obtains spatially continuous and feature-uniform irrigation units by constructing a joint feature vector and employing K-means clustering.

[0009] Further, in step S4, the decision variables for each partition are irrigation duration and target inlet flow rate. The soil moisture state within the partition is described by the average soil moisture content and coefficient of variation. The partition irrigation uniformity is defined as the expected value of the soil moisture coefficient of variation after irrigation, and the overall system uniformity is the weighted average of the areas of each partition. The energy consumption of the drip irrigation system is determined by the power required for the pump to overcome the pipeline resistance and topographic elevation difference. The pressure required at the partition inlet is determined by the minimum working pressure, topographic elevation difference, and head loss along the pipeline. The total energy consumption of the system is calculated based on the total flow rate, pump head, and maximum irrigation duration. The partition water resource utilization efficiency is defined as the ratio of the increased crop evapotranspiration after irrigation to the irrigation water volume, and the overall system water resource utilization efficiency is the weighted average of the areas of each partition. Constraints are set, including flow balance, pressure constraints, irrigation duration constraints, and non-negativity constraints.

[0010] Further, in step S5, the decision variables of each partition are encoded as individuals, and a chaotic sequence is generated using logistic mapping and mapped to the decision space to obtain the initial population. Combining the ideas of differential evolution and particle swarm optimization, for the initial population, two mutation strategies are fused based on a hybrid strategy, and different velocity-position updates are selected according to the adaptive probability to generate an intermediate population. The intermediate population and the parent population are merged to obtain the merged population. In the zoned irrigation control of drip irrigation networks, there are three conflicting objectives (uniformity, energy consumption, and water resource efficiency). Non-dominated sorting can find a set of trade-off solutions (Pareto front), and crowding distance ensures that the solutions on the front are evenly distributed, avoiding clustering in a certain area, thus providing decision-makers with multiple choices. Specifically, a fast non-dominated sort is performed on the merged population, calculating the non-dominated level and crowding distance of each individual to provide a basis for selection. The top individuals are selected to form the next generation population in order of non-dominated level from low to high and crowding distance within the same level from large to small. After iteration, all non-dominated solutions are merged and deduplicated to obtain the Pareto optimal solution set.

[0011] Furthermore, in step S6, for each solution in the Pareto optimal solution set, its three objective function values ​​are standardized, entropy weights are calculated to determine objective weights, and comprehensive weights are obtained by combining them with decision-maker preferences; the weighted Euclidean distance from each solution to the ideal solution is calculated, and the solution with the highest relative closeness is selected as the final decision solution. The final decision solution Includes optimal irrigation parameters , Indicates the first Optimal ingress target traffic for each partition. Indicates the first The optimal irrigation duration for each zone is determined; the irrigation parameters corresponding to the final decision solution are converted into control commands for each zone, and feedback corrections are made based on real-time soil moisture monitoring during the actual irrigation process.

[0012] This invention also provides a multi-objective optimization-based zoned irrigation control system for drip irrigation networks, which executes the above-described multi-objective optimization-based zoned irrigation control method for drip irrigation networks, including: Data acquisition module: used to collect multi-source heterogeneous farmland data; Data preprocessing module: used to remove outliers and normalize multi-source heterogeneous farmland data to obtain normalized features; Sampling point partitioning module: This module is used to construct a joint feature vector and use the K-means clustering method to combine normalized features with spatial coordinates to perform clustering, thereby obtaining the partitioning label for each sampling point; Multi-objective irrigation optimization model construction module: used to construct a multi-objective irrigation optimization model based on zoning characteristics. The irrigation duration and inlet target flow rate of each zone are used as decision variables. The flow rate and pressure of each zone are associated through the pipeline hydraulic model. Three objective functions are defined: irrigation uniformity, system energy consumption and water resource utilization efficiency, so as to realize the modeling of multi-objective irrigation optimization. The optimal solution calculation module is used to obtain the Pareto optimal solution set by employing a hybrid strategy multi-objective evolutionary algorithm through chaotic mapping initialization, adaptive hybrid mutation, fast non-dominated sorting, and crowding distance selection. Irrigation parameter correction module: This module uses a combination of superior and inferior solution distance method, entropy weight method, and decision-maker preferences to introduce real-time feedback adjustments and correct irrigation parameters based on actual soil moisture monitoring.

[0013] The advantages of this invention are: The present application splices the agronomic characteristics such as soil properties, crop parameters and geographic location coordinates into a joint feature vector for clustering, so that the divided irrigation unit not only maintains the spatial continuous block structure, but also ensures that the internal characteristics of the unit are highly similar, solving the contradiction that the traditional partition method is either spatially fragmented or the characteristics are uneven; a multi-objective optimization model is constructed with the partition entrance target flow and irrigation duration as decision variables, considering three conflicting objectives of irrigation uniformity, system energy consumption and water resource utilization efficiency, and introducing pipe network hydraulic constraints, flow balance constraints and irrigation duration constraints, so that the optimization result is more suitable for practical engineering application; a multi-objective evolutionary algorithm with adaptive hybrid mutation strategy is adopted, the population diversity is ensured through chaotic mapping initialization, the differential evolution mutation step is dynamically adjusted according to the population diversity, and the differential evolution and particle swarm mutation probability are adaptively switched according to the iteration process, so as to realize the dynamic balance of global exploration and local development; a closed-loop control architecture combining offline optimization and online feedback is adopted, the optimal scheme is selected from the Pareto solution set by using the distance method combined with entropy weight method and decision maker preference, and the irrigation duration and flow are dynamically corrected according to real-time soil moisture monitoring, and the pipe network pressure constraint is checked, so as to effectively cope with the uncertainty in actual irrigation. BRIEF DESCRIPTION OF DRAWINGS

[0014] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, illustrate the present application together with the embodiments thereof, and explain the present application, and do not constitute a limitation of the present application.

[0015] Figure 1 The step flow chart of the method of the present application is shown in the figure; Figure 2 The irrigation partitioning effect of the feature clustering method is shown in the figure; Figure 3 The irrigation partitioning effect of the spatial coordinate clustering method is shown in the figure; Figure 4 The irrigation partitioning effect of the joint clustering method of the present application is shown in the figure; Figure 5 The performance comparison of the hybrid strategy multi-objective evolutionary algorithm of the present application and two conventional multi-objective algorithms in solving the irrigation optimization problem is shown in the figure; Figure 6 The box plot comparison of irrigation uniformity of the three irrigation control methods is shown in the figure; Figure 7 The box plot comparison of system energy consumption of the three irrigation control methods is shown in the figure; Figure 8 The box plot comparison of water resource utilization efficiency of the three irrigation control methods is shown in the figure; Figure 9 The influence of real-time feedback adjustment on irrigation accuracy is shown in the figure. DETAILED DESCRIPTION

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1 In this embodiment, as Figure 1 As shown, this invention provides a method for zoned irrigation control of drip irrigation networks based on multi-objective optimization, the specific steps of which include: S1. Multi-source heterogeneous farmland data collection To achieve precise control of zoned irrigation in drip irrigation networks, it is first necessary to collect multi-source heterogeneous farmland data covering the entire irrigation area. In one embodiment, the present invention evenly distributes 100 representative sampling points in the target irrigation area, and records the spatial geographical coordinates of each sampling point. The horizontal and vertical coordinate values ​​are obtained by measuring with high-precision GPS equipment to ensure the accuracy of spatial location.

[0018] For soil properties, soil moisture sensors were used to monitor soil moisture content in real time at each sampling point. Simultaneously, soil saturated hydraulic conductivity was determined through ring sampling and indoor infiltration tests. These parameters reflect the soil's water-holding capacity and water movement characteristics. Topographic slope data was extracted using UAV aerial photography or digital elevation models to obtain the slope percentage at each sampling point, characterizing the impact of surface undulation on irrigation uniformity. Crop coefficients were assigned based on the varieties and growth stages of the main crops in the irrigation area, referencing the FAO-56 manual or standard values ​​provided by local agricultural extension departments to ensure the reliability of crop water requirement estimation. Additionally, root zone depth was determined by consulting agricultural technical manuals based on crop type; this parameter is used to calculate the water storage capacity of the root zone. Reference evapotranspiration data was obtained from daily meteorological data at the nearest weather station to the irrigation area, including temperature, humidity, wind speed, and sunshine duration, and calculated using the Penman-Montis formula to reflect atmospheric evaporation capacity. In addition, to describe the soil moisture deficit limit, it is necessary to collect typical soil samples from the irrigation area for texture analysis and to determine the field water holding capacity and wilting coefficient. These two parameters are used to determine irrigation targets in the future.

[0019] Basic data of the pipeline system were collected, including the pipeline layout, pipe diameter, pipe length, sprinkler model and its minimum and maximum working pressure, and pump model and its efficiency curve for each zone. This data was obtained by consulting engineering design drawings and product manuals. Based on the pipe material and hydraulic characteristics, the irrigation water utilization coefficient was estimated, and the maximum total flow rate that the system could provide was determined according to the pump performance curve, serving as a constraint for subsequent optimization. Topographic elevation information was obtained through GIS analysis to obtain the elevation difference between the highest point and the water source point in each zone.

[0020] To facilitate subsequent feedback and adjustments, several soil moisture sensors are deployed in each zone to monitor changes in soil moisture content in real time during irrigation and at the end of the planned irrigation period, so as to compare with the actual target value and perform corrections.

[0021] S2, Multi-source heterogeneous farmland data preprocessing Optimizing farmland irrigation requires collecting multiple key parameters, but the raw data often contains outliers and dimensional differences, directly affecting the accuracy of subsequent analysis. This invention obtains normalized features through outlier removal and normalization. The specific steps are as follows: 1) Outlier removal Five key characteristics were collected for each sampling point: soil moisture content (unit: ), soil saturated hydraulic conductivity (unit: ), terrain slope (unit: ), crop coefficient (dimensionless) and reference evapotranspiration (unit: ), outliers were removed using statistical methods; Specifically, for each key feature, the mean and standard deviation of all sampling points are calculated. If a feature value at a point exceeds the range of the mean plus or minus three times the standard deviation, it is considered an outlier and replaced with the median of that feature across all points, as shown below: , In the formula, Indicates the first The sampling point The original acquired values ​​of the dimensional features; This is the sampling point index, with a value range of [value range missing]. ; This represents the total number of sampling points, which is set to 100 here. This is the feature dimension index, with a value range of [value range missing]. These correspond to soil moisture content, soil saturated hydraulic conductivity, topographic slope, crop coefficient, and reference evapotranspiration, respectively. Indicates the first The arithmetic mean of the original values ​​of all sampled points of a feature reflects the overall level of that feature. Indicates the first The standard deviation of the original values ​​of all sampled points in the dimension feature characterizes the degree of dispersion of the data; Indicates the first The median of the original values ​​of all sampled points for the dimensional feature is used to replace outliers to maintain the data distribution characteristics; Indicates the first The sampling point The value of the feature after the abnormal value is removed, the distribution characteristics of the original data are retained, but the abnormal value is removed, the median is replaced, so that the data is more robust, and the influence of extreme value on subsequent analysis is reduced.

[0022] 2) Min-max normalization The linear mapping of each feature of the data after removing the abnormal value to The interval is used to eliminate the influence of dimension, specifically, for each feature dimension, the minimum value and the maximum value of the cleaned feature value of all sampling points in the dimension are calculated, and then the minimum-maximum normalization method is used to linearly transform the value of each sampling point in the dimension, to obtain the normalized value; Definition The normalized value of the first dimension feature of the first sampling point, dimensionless, the value range is , the feature vector of the first sampling point after min-max normalization processing is denoted as , and the feature matrix of all sampling points after normalization processing is , the dimension is .

[0023] S3, irrigation unit division based on feature-space joint clustering In order to realize the irrigation partition with continuous space and similar features, the normalized features and the space coordinates are combined for clustering. The conventional clustering method only considers the feature similarity and ignores the spatial adjacency, which is easy to lead to fragmented partition or unreasonable boundary. The present application obtains the irrigation unit with continuous space and uniform features by constructing a joint feature vector and using K-means clustering, which is represented as: 1) Constructing a joint feature vector For each sampling point, the space coordinates need to be normalized to unify the dimension of the feature, and then the normalized feature vector and the normalized coordinates are spliced to form a joint feature vector; Definition of the joint feature vector of the first sampling point is , the dimension is 7, which integrates soil properties, crop parameters and geographical location information, i.e. ; Among them, The normalized value of the first sampling point is , The normalized value of the first sampling point is ,

[0024] 2) K-means clustering The K-means algorithm is used to cluster the joint feature vectors to obtain the partition label for each sampling point; The number of clusters is determined by the elbow rule: calculate the sum of squared intra-cluster errors for different numbers of clusters, and select the number of clusters corresponding to the inflection point where the rate of decrease slows down significantly. The K-means algorithm randomly initializes cluster centers and iteratively updates them until convergence. Finally, each point is assigned to the nearest cluster center, represented as: , In the formula, Indicates the first The partition label of each sampling point, with a value of arrive An integer between [a certain value] and [a certain value] identifies the irrigation unit to which the point belongs; The number of clusters, i.e., the total number of irrigation zones, is determined by the elbow rule; The index representing the cluster center is equivalent to the irrigation partition index, and its value range is... ; Indicates the first The cluster center vector, with dimension 7, is obtained through iterative optimization using the K-means algorithm, representing the _th _i_ cluster center vector. The characteristics and spatial location center of each partition; This represents the L2 norm, also known as the Euclidean norm.

[0025] It should be noted that the vector composed of the partition labels of all sampling points , dimension This is used to identify the irrigation zone to which each sampling point belongs, through... It can obtain the set of sampling points contained in each partition.

[0026] In one embodiment, the irrigation zoning effect based on feature-spatial joint clustering is compared. Specifically, two conventional zoning methods, feature-only clustering and spatial-only clustering, are compared. From 100 selected sampling points, the K-means clustering algorithm is used to divide the irrigation area into six zones. The x-axis represents the x-coordinate (meters), and the y-axis represents the y-coordinate (meters). Each sampling point is marked with a different color according to its zone. Figure 2 As a feature-based clustering method, the visible partitions are spatially fragmented, with the same color points scattered in different areas. Adjacent plots may be assigned to different partitions, which can lead to difficulties in pipeline layout and chaotic water and fertilizer management in actual irrigation. Figure 3 Based on spatial coordinate clustering, the partitions are presented as continuous geometric blocks, but the soil moisture content, crop coefficient and other characteristics within each partition are quite different, making it impossible to achieve precise irrigation on demand; Figure 4For the joint clustering method of the application, the partition not only maintains the continuous block structure in space, but also makes the characteristics inside each partition highly similar (the color points are concentrated and the boundary is relatively smooth in the figure), which lays a scientific foundation for subsequent differentiated irrigation of the partition.

[0027] S4, multi-objective irrigation optimization model construction based on partition characteristics The conventional irrigation optimization model ignores the coupling effect of pipe network pressure constraint on partition flow, and does not consider the spatial variation of soil moisture, and the application constructs a multi-objective irrigation optimization model based on partition characteristics, takes the irrigation time and inlet target flow of each partition as decision variables, correlates the partition flow and pressure through the pipe network hydraulic model, defines three objective functions of irrigation uniformity, system energy consumption and water resource utilization efficiency, realizes the modeling of multi-objective irrigation optimization, and is expressed as: 1) define the association between decision variables and partition parameters The decision variable of each partition is the irrigation time and the inlet target flow, and the decision variable is the variable that needs to be determined in the optimization process, the soil moisture state in the partition is described by the average soil moisture content and the coefficient of variation, which is expressed as: , , In the formula, represents the average soil moisture content of all sampling points in the i-th partition, the unit is , reflecting the average level of soil moisture in the partition; represents the coefficient of variation of soil moisture in the i-th partition, which is dimensionless, and represents the spatial variation degree of soil moisture in the partition; represents the sampling point set contained in the i-th partition, which is obtained by the partition label, that is , represents the number of sampling points in the i-th partition; represents the original soil moisture content of the i-th sampling point, the unit is . At the same time, the global soil parameters include field water capacity (unit: ) and wilting coefficient (unit: ) are obtained by soil texture determination.

[0028] 2) irrigation uniformity objective function The partition irrigation uniformity is defined as the expected value of the soil moisture variation coefficient after irrigation, and the overall uniformity of the system is the weighted average of the area of each partition, which is expressed as:

[0029] 2) irrigation uniformity objective function The partition irrigation uniformity is defined as the expected value of the soil moisture variation coefficient after irrigation, and the overall uniformity of the system is the weighted average of the area of each partition, which is expressed as: ​​​ , In the formula, represents the system overall irrigation uniformity, dimensionless, the smaller the value, the more uniform the spatial distribution of soil moisture after irrigation, which is the area-weighted average of the irrigation uniformity of each subarea; represents the area of the th subarea, with a unit of , which is calculated from the subarea boundary; represents the area of the th subarea, with a unit of , which is calculated from the subarea boundary; represents the index of the cluster center distinguished from , which is equivalent to the irrigation subarea index; represents the irrigation uniformity of the th subarea, which measures the deviation of the soil moisture content in the subarea from the target average moisture content after irrigation, dimensionless, and the calculation method is represented as ; represents the target average soil moisture content of the th subarea, with a unit of , and the value is , i.e. 90% of the field water holding capacity, as the expected soil moisture level after irrigation; represents the inlet target flow of the th subarea, with a unit of , which is a decision variable; represents the irrigation duration of the th subarea, with a unit of , which is a decision variable; represents the irrigation water use coefficient of the th subarea, dimensionless, reflecting the proportion of water loss (such as evaporation, seepage, etc.) during the irrigation process, which can be simulated through the loss during pipe water transportation or preset according to experience, and the default value can be taken - ; represents the root layer depth, with a unit of , which is determined by the crop type, for example, the value of for wheat, the value of for corn, etc., which can be consulted according to the crop variety in the agricultural manual or preset typical values.

[0030] It should be noted that in the calculation process of the irrigation uniformity , ​The item is the estimation of the soil moisture content at the point after irrigation (assuming that the water is uniformly distributed in the subarea and considering the root layer depth), minus the target average moisture content, and then squared sum, and then square root divided by the target value, to get the relative standard deviation (coefficient of variation), reflecting the uniformity of water distribution after irrigation. The smaller the value, the more uniform.

[0031] 3) System energy consumption objective function The energy consumption of the drip irrigation system is determined by the power required by the water pump to overcome the pipe network resistance and the terrain elevation difference. The required pressure at the subarea inlet is determined by the minimum working pressure, the terrain elevation difference, and the along-path water head loss. The total energy consumption of the system is calculated based on the total flow, the water pump lift, and the maximum irrigation time, and is represented as: , , In the formula, represents the total energy consumption of the system, with the unit of , representing the electrical energy consumed during the entire irrigation process; represents the density of water, which is a physical constant, and the value here is ; represents the acceleration due to gravity, which is a physical constant, and the value here is ; represents the total flow of the system, with the unit of , and the calculation method is represented as , representing the total flow demand when all subareas are irrigated at the same time, which is used to calculate the total power and energy consumption of the water pump; represents the required water pump lift, with the unit of , taking the maximum value of all subareas after conversion, and the conversion relationship is water column, to ensure that the water pump can provide sufficient lift to meet the pressure demand of the most unfavorable subarea; represents the pump efficiency, which is dimensionless and determined by the water pump model, and can take as a typical value; represents the maximum value of the irrigation time of all subareas, with the unit of , which determines the water pump operation time; represents the required pressure at the inlet of the th subarea, with the unit of , which is used to ensure that the irrigation emitters in the subarea can work normally; represents the minimum working pressure of the irrigation emitters, with the unit of , which is determined by the irrigation emitter model, for example, taking ; represents the elevation difference between the highest point in the th subarea and the water source, with the unit of , which is obtained from the terrain data and can be obtained through GIS or field measurement; represents the elevation difference between the highest point in the The head loss along the route for each zone, in units of The friction loss, representing the energy loss of water flowing through a pipe, is calculated using the Darcy-Weisbach formula, which is the standard method for calculating friction loss in fluid mechanics. ; This is the friction factor, dimensionless, and depends on the pipe material and flow conditions; an empirical value can be used. (For smooth pipes), or calculated based on the Reynolds number; For the first The pipe length of each zone, in units of The location is determined by the pipeline network layout and can be obtained from design drawings or measurements. Pipe diameter, unit: Determined by the pipeline network design, for example, taking ; Flow velocity, unit: That is, the flow rate divided by the cross-sectional area of ​​the pipe, expressed as .

[0032] It should be noted that in the total energy consumption of the computing system hour, The item is the power of the water pump (unit: ), divided by Convert to Multiply by the running time The total energy consumption is obtained because the water pump needs to overcome the resistance of the pipeline network to provide the head. The total flow rate determines the operating point of the water pump, the head is taken as the maximum value required to ensure that all zones can work, and the running time is taken as the maximum irrigation duration.

[0033] 4) Objective function for water resource utilization efficiency The water resource utilization efficiency of a zone is defined as the ratio of the increase in crop evapotranspiration after irrigation to the amount of irrigation water. The overall water resource utilization efficiency of the system is the weighted average of the areas of each zone, expressed as: , , In the formula, It represents the overall water resource utilization efficiency of the system, is dimensionless, and the larger the value, the higher the increase in crop evapotranspiration generated per unit of irrigation water. It is the area-weighted average of the water resource utilization efficiency of each zone, reflecting the economy of water use. Indicates the first Water resource utilization efficiency of each zone, dimensionless, quantifies the increase in crop evapotranspiration that can be brought about by a unit of net irrigation water volume, and characterizes the production efficiency of irrigation water in that zone. The larger the value, the more economical the water use, that is, the more crop evapotranspiration benefits are obtained with less water. Indicates the first The average crop coefficient within each zone is dimensionless and determined by crop type and growth stage. It can be preset by referring to the FAO-56 manual or local agricultural data. Indicates the first Average reference evapotranspiration within each zone, in units of It is obtained from meteorological data, such as calculations using the Penman formula or data from weather stations.

[0034] It should be noted that in the process of calculating the water resource utilization efficiency of each zone, The term represents the potential evapotranspiration rate of the crop (in units of...). ), multiplied by the water stress coefficient The actual increase in evapotranspiration (assuming that evapotranspiration decreases proportionally during water deficit) is obtained and then multiplied by the irrigation time. (Unit: days) The total increase in evapotranspiration during the irrigation period (unit: days) ), multiplied by area Obtain the volume (Note: Therefore, multiply by Will Convert to Finally, the increased evaporation volume is obtained, and, The term represents the amount of irrigation water (net irrigation water after considering losses), and the ratio is the increase in evapotranspiration per unit of irrigation water.

[0035] It should also be noted that when calculating the water resource utilization efficiency of each zone, The conversion factor from millimeters to meters is 24, which is then divided to convert days to hours.

[0036] 5) Constraints Optimization must satisfy constraints on flow rate, pressure, irrigation duration, and non-negativity to ensure safe system operation and effective decision variables, including: a) Flow balancing: ; b) Pressure constraint: , in, This indicates the maximum working pressure of the irrigation device, in units of... The model of the water dispenser determines the type of water dispenser used. For example, take... ; c) Irrigation duration constraints: , in, This indicates the minimum permissible irrigation duration, in units of... The minimum operating time of the irrigation system or the crop's needs are determined by factors such as the minimum operating time of the irrigation system. ; denotes the maximum allowed irrigation duration, unit is , considering the soil infiltration capacity or avoiding deep infiltration, for example , which can be set according to actual conditions; d) Non-negative constraint: .

[0037] S5, model solving based on hybrid strategy multi-objective evolutionary algorithm The conventional multi-objective evolutionary algorithm has problems of insufficient population diversity and unbalanced search ability, and the hybrid strategy multi-objective evolutionary algorithm is adopted, the initial population is initialized through chaotic mapping, adaptive hybrid mutation, fast non-dominated sorting and crowded distance selection, and the Pareto optimal solution set is obtained, and the specific steps are as follows: 1) Population initialization and chaotic mapping The decision variables (inlet target flow and irrigation duration) of each partition are coded into individuals, chaotic sequences are generated by using logistic mapping, and are mapped to the decision space to obtain the initial population, and are represented as: , In the formula, denotes the decision vector of the i-th initial individual, the dimension is , and the inlet target flow and irrigation duration of all partitions are included, which is the initial position of the individual in the decision space; is the individual index, and the value range is ; is the population size, and the value example is 100; denotes the inlet target flow of the i-th partition in the i-th individual, which is mapped to the interval by a chaotic variable, and is represented as ; denotes the irrigation duration of the i-th partition in the i-th individual, which is mapped to the interval by a chaotic variable, and is represented as ; ; is the chaotic variable of the i-th individual in the i-th dimension, which is used to generate the initial population, so that the individuals are uniformly distributed in the decision space, and the population diversity is improved, the chaotic sequence is generated by logistic mapping, and has ergodicity and randomness, that is , and the definition is a random number in the interval , and the chaotic mapping makes the initial population uniformly distributed in the decision space, and improves the diversity; , and the definition is a random number in the interval , and the chaotic mapping makes the initial population uniformly distributed in the decision space, and improves the diversity; ​​​For control parameter, the value is preferably 4 to ensure chaotic characteristics; represents the dimension index of chaotic variable, and the value range is Because each individual has decision variables, each variable corresponds to a chaotic variable.

[0038] 2) Adaptive hybrid mutation operator Fusion of differential evolution and particle swarm optimization, based on hybrid strategy, two mutation strategies are fused, and different speed-position updates are selected according to adaptive probability, so as to balance global exploration and local development ability, and generate intermediate population, which is represented as: , In the formula, represents the mutation vector of the th individual in the th generation, and the dimension is , which represents a candidate decision variable combination, which will be crossed with the original individual to generate a trial individual for constructing the next generation population; is the iteration index, and the value range is ; is the maximum iteration number, and the value example is 500; represents the th individual in the th generation, that is, a specific irrigation scheme decision vector, which contains the inlet target flow and irrigation time of all partitions; , , are three different individuals randomly selected from the current population, and the individual index is different from ; represents the scaling factor, which is dimensionless, and the calculation method is represented as When the population diversity decreases, The ratio of the term becomes smaller, Close to , reduce the mutation step, and strengthen local search, when the diversity is high, Close to , increase the mutation step, and promote global search; represents the mutation strategy selection probability, which is dimensionless, and the calculation method is represented as ; represents the inertia weight, which is dimensionless, and controls the degree of particle velocity inheritance, and the value example is 0.8; represents the mutation vector of the th individual in the th generation, which is initially zero; This represents the first learning factor, which controls the degree to which an individual moves toward its historical best; an example value is 2. This represents the second learning factor, which controls the degree to which an individual moves toward the global optimum; an example value is 2. Represents the first random number, which is Uniformly distributed random numbers within an interval; This represents the second random number. Uniformly distributed random numbers within an interval; Indicates the first The historical best position of each individual; This indicates the current global optimal position of the population. Indicates the first Generations of populations Indicates the initial population. Indicates the first The diversity measure of a generation of population is calculated by... The standard deviation of all individuals in the population across each decision variable dimension is calculated, and then the average of these standard deviations is taken. This average is used to measure the dispersion of the population in the decision space. A measure of the diversity of the initial population; This represents the hyperbolic tangent function, used to map the input to... Interval.

[0039] Furthermore, the intermediate population was obtained after the mutation. It is a new set of individuals generated through mutation and crossover, used to interact with the parent population. Merge, perform non-dominated sorting and selection to produce the next generation.

[0040] It should be noted that the mutation strategy selection probability During the iteration process, the proportions of the two mutation strategies are dynamically balanced. In the early stages, the probabilities of the two strategies are roughly equal, balancing exploration and development. In the later stages, the probability of differential mutation increases because the population tends to converge at this point, and the local search capability of differential mutation (through...) Adaptive step size reduction helps with fine optimization, while particle swarm mutation uses historical information to guide the search; the two work together to improve algorithm performance.

[0041] 3) Fast non-dominated sorting and crowding distance In the zoned irrigation control of drip irrigation networks, there are three conflicting objectives (uniformity, energy consumption, and water resource efficiency). By using non-dominated ordination, a set of trade-off solutions (Pareto front) can be found, and the crowding distance ensures that the solutions on the front are evenly distributed, avoiding clustering in a certain area, thus providing decision-makers with multiple options; Specifically, a fast non-dominated sort is performed on the merged population, and the non-dominated rank and crowding distance of each individual are calculated to provide a basis for the selection operation, represented as: , In the formula, Indicates the first The crowding distance of an individual is dimensionless and reflects the density around an individual in the same non-dominated layer. The larger the value, the sparser the solution in that region. Index of the objective function, Corresponding to the overall irrigation uniformity of the system Total system energy consumption and the overall water resource utilization efficiency of the system ; and Indicates the first On the first goal, with the second The objective function values ​​of two adjacent individuals; and Indicates the number of units in the current non-dominated layer. The maximum and minimum values ​​of each target.

[0042] In practical implementation, the non-dominant hierarchy This is obtained through a fast non-dominated sorting algorithm. Specifically, it compares the dominance relationships between individuals, groups those not dominated by any other individual into the first layer, removes them, and then groups the remaining non-dominated individuals into the second layer, and so on. Indicates the first The non-dominant hierarchy of an individual, the lower the hierarchy (e.g., ... The higher the level, the better. In the selection operation, individuals with lower levels are given priority. Within the same level, individuals with higher crowding are given priority.

[0043] It should be noted that the fast non-dominated sort is used to stratify individuals based on the three objective function values, grouping non-dominated individuals into the same stratum. The lower the stratum number (the smaller the rank), the better the individual (Pareto optimality). The crowding distance is used to measure the sparsity of individuals within the same stratum. The larger the distance, the sparser the solution in that region. When selecting, individuals with high crowding are preferred to be retained to maintain the diversity of solutions.

[0044] It should also be noted that when calculating the distance to congestion... During the process, and This is achieved by arranging individuals within the same non-dominated layer according to their position. After sorting the nth objective function values, take the nth value and... The objective function values ​​of two adjacent individuals. and It is the first in the current non-dominated layer The maximum and minimum values ​​of each target are obtained directly from the individuals in that layer.

[0045] It should also be noted that in the optimization of drip irrigation network zonal irrigation, the search space of decision variables is huge, and the objective functions (uniformity, energy consumption, water resource efficiency) are conflicting. Adaptive hybrid mutation can dynamically adjust the search step size and strategy according to the current state of the population: when the diversity is high in the early stage, large step size differential mutation helps to quickly explore new areas and avoid getting trapped in local optima; when the diversity decreases in the later stage, small step size particle swarm mutation uses the historical optimal and global optimal information of individuals to finely search for favorable areas and improve the quality of solutions. The adaptive mechanism enables the algorithm to efficiently approach the Pareto front, providing decision-makers with high-quality and diversified irrigation solutions.

[0046] It should also be noted that the distance to congestion The Pareto value measures the sparsity of the solutions around individuals within the same non-dominated layer. A larger value indicates a sparser solution in that area. In drip irrigation optimization, we aim to obtain a set of Pareto solutions that are evenly distributed in terms of uniformity, energy consumption, and water efficiency, allowing decision-makers to choose solutions based on different preferences. Individuals with large crowding distances represent fewer solutions in their respective areas. Prioritizing the retention of these individuals enriches the diversity of the solution set, avoids solutions concentrating near a certain objective extreme, and thus provides more diverse trade-offs, such as some solutions being more energy-efficient and others more uniform.

[0047] 4) Elite Preservation and Population Renewal Individuals are selected in ascending order of non-dominance level and descending order of crowding distance within the same level to form the next generation population, denoted as […]. ; After the iteration is complete, all non-dominated solutions are merged and duplicates are removed to obtain the Pareto optimal solution set, denoted as: : in, Indicates the first The generation population is obtained through selection operations; Represents the parent population (the first generation) The merged set of the first generation population and the intermediate population, i.e. ; Indicates population size; This indicates the selection based on non-dominance level and crowding distance. The operation of each individual; This is the Pareto optimal solution set, which contains all non-dominated solutions. It is the final output multi-objective optimization solution set for decision-makers to choose from. Indicates from the first The non-dominated solution set extracted from the population. The characterization process involves merging all non-dominated solutions generated in the iterations, removing duplicates, and obtaining the final Pareto optimal solution set.

[0048] It should be noted that the elite preservation strategy involves preserving the parent population. and intermediate population merged into Then, in order of non-dominance level from low to high, and within the same level, from largest to smallest crowding distance, select the top... Individuals constitute the next generation population. This ensures that excellent solutions are not lost and accelerates convergence. In drip irrigation optimization, it helps to retain solutions that perform well in terms of uniformity, energy consumption, and water resource efficiency, while maintaining diversity and avoiding getting trapped in local optima.

[0049] S6. Generation and Execution of Optimal Irrigation Parameters Based on Multi-Attribute Decision Making Conventional methods lack objective weights and feedback corrections when selecting the optimal solution from the Pareto solution set. This invention employs a method combining the distance between superior and inferior solutions with entropy weighting and decision-maker preferences, and introduces real-time feedback adjustments to correct irrigation parameters based on actual soil moisture monitoring, thus addressing uncertainties. The specific steps are as follows: 1) Multi-attribute decision-making for Pareto solutions For each solution in the Pareto optimal solution set, its three objective function values ​​are standardized, entropy weights are calculated to determine objective weights, and a comprehensive weight is obtained by combining these with the decision-maker's preferences. Then, the weighted Euclidean distance from each solution to the ideal solution is calculated, and the solution with the highest relative closeness is selected as the final decision, expressed as: In the formula, The decision solution represents the final irrigation scheme selected from the Pareto optimal solution set, which includes the optimal irrigation parameters. ; Indicates the first Optimal ingress target traffic for each partition; Indicates the first Optimal irrigation duration for each zone; Solution The weighted Euclidean distance to the positive ideal solution is calculated as follows: This is used to measure how close a solution is to the ideal solution; the smaller the distance, the better. Solution The weighted Euclidean distance to the negative ideal solution is calculated as follows: The negative ideal solution is a virtual solution where all objectives are at their worst. The larger this distance, the better the solution. The further away from the worst-case scenario, the better; Indicates the first The comprehensive weight of the target, combined with the objective entropy weight and the subjective preference weight, adopts the geometric mean calculation method to balance the influence of both, avoid a certain weight being too small or too large, and make the final weight reflect both the distinguishing degree of the data itself and the intention of the decision maker, denoted as ; denotes the normalized value of the th target of the solution , ranging from ; denotes the subjective preference weight of the decision maker for the three targets (uniformity, energy consumption, and water resource efficiency) according to actual needs, for example, if more emphasis is placed on energy saving, a higher weight can be given to the energy consumption target. It is used to adjust the entropy weight to make the finally selected scheme more in line with the actual production target. It is set by experience and the value example is ; denotes the entropy weight of the th target, which is calculated by information entropy.

[0050] In specific implementation, the positive ideal solution is defined as a virtual solution with all normalized targets being 1, and the negative ideal solution is defined as a virtual solution with all normalized targets being 0. They are used as reference benchmarks to calculate the closeness degree.

[0051] In specific implementation, the standardization processing method is as follows: for cost-type targets and , the standardized value is , and for benefit-type targets , the standardized value is , where is the th objective function value of the solution , and is the normalized value of the th target of the solution , ranging from , is the maximum value of the th target in the Pareto solution set, and is the minimum value of the th target in the Pareto solution set.

[0052] In specific implementation, the calculation method of is as follows: first, calculate the information entropy of the th target, then obtain the entropy weight , combine the decision maker's preference weight , and obtain the comprehensive weight and normalized to make wherein, represents the number of solutions in the Pareto optimal solution set.

[0053] 2) Irrigation control instruction generation and feedback adjustment Convert the optimal solution corresponding irrigation parameters into control instructions for each partition, and make feedback correction according to real-time soil moisture monitoring during actual irrigation to ensure that the target water content is reached, which is represented as: In the formula, represents the inlet target flow of the th partition after feedback adjustment, with the unit of , which is used to correct according to real-time monitoring in actual irrigation to make up for insufficient irrigation; represents the irrigation time that needs to be extended for the th partition, with the unit of , and the calculation method is represented as , which is used to correct the original plan. If , the actual water content is lower than the target, the irrigation is extended, and if , it is stopped in advance and does not need to be extended. represents the correction factor, which is dimensionless, reflecting the gap between the actual achieved water content and the target, and the calculation method is ; represents the average soil water content in the th partition at the end of the planned irrigation time , which is used to judge whether the irrigation needs to be extended, and is obtained by real-time monitoring of field sensors.

[0054] In specific implementation, if , the irrigation time is extended by , and the flow is adjusted to keep the total water unchanged. The adjusted inlet target flow needs to meet the pipe network pressure constraint, that is, , wherein, represents the pressure required by the partition inlet corresponding to the adjusted inlet target flow , if it exceeds , the boundary value is taken, and the adjusted instruction is finally executed to realize differentiated and accurate irrigation of the partition.

[0055] It should be noted that if the target is reached in advance within , the valve is closed in advance.

[0056] It should also be noted that conventional methods lack objective weights and feedback corrections when selecting the optimal solution from the Pareto solution set. This invention adopts the superior-inferior solution distance method combined with the entropy weight method and decision-maker preferences, and introduces real-time feedback adjustments to ensure precise irrigation. The entropy weight method objectively determines the weight of each objective based on the distribution of the solution set itself, avoiding subjective arbitrariness. Combined with the decision-maker preference weights, it can reflect actual needs. The superior-inferior solution distance method selects the comprehensive optimal solution by calculating the closeness to the ideal solution. Real-time feedback adjusts irrigation parameters based on actual soil moisture monitoring to cope with uncertainties and improve irrigation accuracy.

[0057] In one embodiment, such as Figure 5 As shown, a Pareto front comparison of multi-objective optimization algorithms is conducted, specifically comparing the performance of the proposed hybrid strategy multi-objective evolutionary algorithm with two conventional multi-objective algorithms (second-generation non-dominated sorting genetic algorithm and multi-objective particle swarm optimization) in solving irrigation optimization problems. The experiment selected 30 Pareto optimal solutions, each corresponding to a set of irrigation schemes. The three objectives were irrigation uniformity (dimensionless, smaller values ​​indicate greater uniformity), total system energy consumption (kilowatt-hours, smaller values ​​indicate greater energy efficiency), and water resource utilization efficiency (dimensionless, larger values ​​indicate greater efficiency). Blue dots represent the algorithm of this invention, while red triangles and green squares represent the two conventional algorithms, respectively. As can be seen from the figure, the solution set of the algorithm of this invention is distributed closer to the ideal region in three-dimensional space: in the dimension of irrigation uniformity, blue dots are concentrated in the smaller value range (corresponding to greater uniformity); in the dimension of energy consumption, blue dots are concentrated in the lower value range (corresponding to greater energy efficiency); and in the dimension of water resource utilization efficiency, blue dots are concentrated in the higher value range (corresponding to higher efficiency). In contrast, the solution set of the conventional algorithms deviates from the ideal region overall, exhibiting either poor uniformity, excessively high energy consumption, or low efficiency. The solution set of this invention forms a better Pareto front for the three objectives, indicating that the algorithm can better balance the three conflicting objectives and provide decision-makers with a higher quality and more evenly distributed set of solutions.

[0058] S7. Zoned irrigation control of drip irrigation networks based on multi-objective optimization Based on the aforementioned multi-source heterogeneous data collection, preprocessing, irrigation unit division, multi-objective optimization model construction, hybrid strategy evolutionary algorithm solving, and multi-attribute decision-making, a complete method for partitioned irrigation control of drip irrigation pipe networks is formed. First, through the soil moisture content, saturated hydraulic conductivity, terrain slope, crop coefficient, reference evapotranspiration, and spatial coordinate data of 100 sampling points collected by S1, the standardized feature matrix is obtained after abnormal value elimination and normalization preprocessing by S2. Then, in S3, the normalized features of each sampling point are spliced with the normalized spatial coordinates into a 7-dimensional joint feature vector, and the K-means clustering algorithm combined with the elbow rule is used to determine the optimal number of partitions, so as to divide the irrigation area into several spatially continuous and feature-similar irrigation units, each unit containing a group of sampling points and their corresponding partition labels. Based on the partition results, S4 constructs a multi-objective optimization model with the irrigation time length and inlet target flow of each partition as decision variables, considering the irrigation uniformity, system energy consumption, and water resource utilization efficiency as three conflicting objectives, and introducing pipe network hydraulic constraints, flow balance constraints, and irrigation time length constraints to form a mathematical optimization problem.

[0059] S5 uses a hybrid strategy multi-objective evolutionary algorithm to solve the problem: through chaotic mapping to initialize the population to ensure diversity, use adaptive hybrid mutation operators (combining differential evolution and particle swarm mutation) to dynamically balance global exploration and local development, combine fast non-dominated sorting and crowding distance for elite selection, and obtain a set of Pareto optimal solutions after iterative evolution, which contains multiple irrigation schemes with advantages in uniformity, energy consumption, and water efficiency.

[0060] S6 further uses the superior-inferior solution distance method combined with entropy weight method and decision maker preference to select the most comprehensive irrigation scheme from the Pareto solution set, generating the optimal inlet target flow and irrigation time length of each partition. In actual execution, the optimal parameters are converted into valve opening instructions and irrigation time length settings for each partition, and the water pump is started to supply water according to the partition flow; at the same time, the soil moisture sensors installed in the partition are used to monitor the soil moisture content in real time during irrigation, if the average moisture content at the end of the planned irrigation time does not reach the target value, the correction factor is calculated according to the deviation, the irrigation time is dynamically extended and the flow is adjusted to keep the total water volume unchanged, and the adjusted flow is checked to ensure that the pipe network pressure constraint is met to ensure safe operation of the system; if the target is reached in advance, the valve is automatically closed.

[0061] Based on this, through the closed-loop control mode of "offline optimization + online feedback", the partitioned differentiated precision irrigation of drip irrigation pipe networks is realized, effectively improving the irrigation uniformity, reducing energy consumption, and improving the water resource utilization efficiency.

[0062] In one embodiment, the box plot comparison of the three irrigation control methods on the three targets is carried out, and the performances of the three irrigation control methods: uniform irrigation, conventional zoning irrigation and the method of the present application are compared on three key performance indicators through 20 independent repeated experiments. Figure 6 、 Figure 7 、 Figure 8 The three box plots correspond to irrigation uniformity, system energy consumption and water resource utilization efficiency respectively. The horizontal axis marks the three methods, and the vertical axis is marked with specific units. From the irrigation uniformity box plot, it can be seen that the median and quartile range of the method of the present application are significantly smaller than those of the other two methods, indicating that the spatial distribution of soil moisture after irrigation is the most uniform and has high stability. In the system energy consumption box plot, the energy consumption value of the method of the present application is lower than that of uniform irrigation and conventional zoning as a whole, with a clear advantage in median, and the energy saving effect is outstanding. In the water resource utilization efficiency box plot, the median of the method of the present application is the highest, and the box body is located in the higher area as a whole, indicating that the crop evapotranspiration increment per unit of irrigation water is the largest, and the water utilization is more economical. The three box plots jointly verify that the method of the present application can achieve better performance on the three conflicting targets, and the results have statistical stability.

[0063] In one embodiment, as shown in Figure 9 , the influence of real-time feedback adjustment on irrigation accuracy is analyzed, the change of soil moisture content of a single zone during irrigation is selected, the target moisture content is set to 0.25 cubic meters per cubic meter, the horizontal coordinate is time (hours), and the vertical coordinate is soil moisture content (cubic meters per cubic meter). The blue curve represents conventional irrigation without feedback, and the moisture content is slightly lower than the target value at the end of four hours, and then remains constant, which cannot achieve the expected result; the red curve represents the irrigation with feedback of the present application, and after the moisture content is monitored to be insufficient for four hours, the irrigation time is dynamically extended according to the deviation, and the flow rate is adjusted to make the final moisture content accurately reach the target value (green horizontal dashed line). The experiment shows that the present application can effectively cope with the uncertainty in actual irrigation through the closed-loop control of "offline optimization + online feedback", and improve the irrigation accuracy.

[0064] Embodiment 2 The present embodiment provides a drip irrigation pipe network zoning irrigation control system based on multi-objective optimization, which executes the drip irrigation pipe network zoning irrigation control method based on multi-objective optimization described in embodiment 1, comprising: a data acquisition module for collecting multi-source heterogeneous farmland data; a data preprocessing module for performing outlier rejection and normalization processing on the multi-source heterogeneous farmland data to obtain normalized features; a sampling point zoning module for clustering the normalized features and spatial coordinates by constructing a joint feature vector and using a K-means clustering method to obtain the zoning label of each sampling point; A multi-objective irrigation optimization model construction module is configured to construct a multi-objective irrigation optimization model based on zoning characteristics, take irrigation time length and inlet target flow of each zone as decision variables, correlate zone flow and pressure through a pipe network hydraulic model, define three objective functions of irrigation uniformity, system energy consumption and water resource utilization efficiency, and realize modeling of multi-objective irrigation optimization; An optimal solution calculation module is configured to obtain a Pareto optimal solution set by using a hybrid strategy multi-objective evolutionary algorithm, through chaotic mapping initialization, adaptive hybrid mutation, fast non-dominated sorting and crowded distance selection; An irrigation parameter correction module is configured to introduce real-time feedback adjustment by using a good-bad solution distance method combined with an entropy weight method and a decision maker preference, and correct irrigation parameters according to actual soil moisture monitoring.

[0065] Finally, it should be noted that: the above only for the preferred embodiments of the present application, and not for the purpose of limiting the present application, although the foregoing detailed description of the present application is made with reference to the foregoing embodiments, for those skilled in the art, it still can be modified, or part of the technical features of the equivalent replacement. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A method for zoning irrigation control of a drip pipe network based on multi-objective optimization, characterized in that, The method comprises the following steps: S1, uniformly arranging a plurality of sampling points in a target irrigation area, recording the spatial coordinates of each sampling point, and collecting multi-source heterogeneous farmland data of each sampling point; S2, performing outlier rejection and normalization processing on the multi-source heterogeneous farmland data to obtain normalized features; S3, combining the normalized features and the spatial coordinates by constructing a joint feature vector and using a K-means clustering method to cluster, to obtain a partition label of each sampling point; S4, constructing a multi-objective irrigation optimization model based on partition characteristics, taking the irrigation time length and the inlet target flow of each partition as decision variables, associating the partition flow and pressure through a pipe network hydraulic model, defining three objective functions of irrigation uniformity, system energy consumption and water resource utilization efficiency, and realizing modeling of multi-objective irrigation optimization; S5, using a hybrid strategy multi-objective evolutionary algorithm, initializing through chaotic mapping, performing adaptive hybrid mutation, performing fast non-dominated sorting and selecting by crowding distance, and obtaining a Pareto optimal solution set; S6, using a superior-inferior solution distance method combined with an entropy weight method and a decision maker's preference, introducing real-time feedback adjustment, and correcting irrigation parameters according to actual soil moisture monitoring.

2. The multi-objective optimization based zoned irrigation control method of a drip pipe network according to claim 1, characterized in that, The key features in the multi-source heterogeneous farmland data of each sampling point in S1 include soil water content, soil saturated hydraulic conductivity, terrain slope, crop coefficient and reference evapotranspiration.

3. The multi-objective optimization based zoned irrigation control method of a drip pipe network according to claim 2, characterized in that, S2 is specifically: For each key feature, the mean and standard deviation of all sampling points are calculated. If a point feature value is beyond the range of mean plus or minus three times the standard deviation, it is considered an outlier and replaced with the median of the feature value of all points. Linearly map each feature of the data after removing outliers to the interval to eliminate the influence. For each feature dimension, calculate the minimum and maximum of the cleaned feature values of all sampling points in that dimension. Then, linearly transform each sampling point's feature value using the min-max normalization method to obtain the normalized feature value, and then obtain the normalized feature vector of each sampling point. Define as the normalized value of the th feature of the th sampling point, dimensionless, with a value range of . The feature vector of the th sampling point after min-max normalization is denoted as .

4. The multi-objective optimization based zoned irrigation control method of a drip pipe network according to claim 1, characterized in that, S3 is specifically: Normalize the spatial coordinates of each sampling point to obtain normalized coordinates. Concatenate the normalized feature vector with the normalized coordinates to form a joint feature vector: Define the first... The joint feature vector of the sampling points is The dimension is 7. , Indicates the first The normalized value of the x-coordinate of each sampling point. Indicates the first The normalized value of the ordinate of each sampling point.

5. The multi-objective optimization based zoned irrigation control method of a drip pipe network according to claim 1, characterized in that, S4 is specifically: The decision variables of each partition are the irrigation time length and the inlet target flow, the soil moisture state in the partition is described by the average soil water content and the coefficient of variation, the partition irrigation uniformity is defined as the expected value of the soil moisture coefficient of variation after irrigation, the overall uniformity of the system is the weighted average of the areas of the partitions, the energy consumption of the drip irrigation system is determined by the power required for the water pump to overcome the resistance of the pipe network and the terrain elevation difference, the required pressure at the partition inlet is determined by the minimum working pressure, the terrain elevation difference and the along-path water head loss, the total energy consumption of the system is calculated based on the total flow, the water pump lift and the maximum irrigation time length, the water resource utilization efficiency of the partition is defined as the ratio of the increased crop evapotranspiration after irrigation to the irrigation water volume, the overall water resource utilization efficiency of the system is the weighted average of the areas of the partitions, and the constraint conditions include flow balance, pressure constraint, irrigation time length constraint and non-negative constraint.

6. The multi-objective optimization based zoned irrigation control method of a drip pipe network according to claim 1, characterized in that, In S5, the chaotic mapping initialization is: The decision variables of each partition are encoded into individuals, a chaotic sequence is generated by using a logistic mapping, and the initial population is obtained by mapping to the decision space.

7. The multi-objective optimization based zone irrigation control method of a drip pipe network according to claim 6, characterized in that, In S5, the adaptive hybrid mutation is: For the initial population, two kinds of mutation strategies are fused based on the hybrid strategy, different speed-position updates are selected according to the adaptive probability, the intermediate population is obtained, and the intermediate population and the parent population are combined to obtain the combined population.

8. The multi-objective optimization based zone irrigation control method of a drip pipe network according to claim 7, characterized in that, In S5, the fast non-dominated sorting and selection by crowding distance are: The fast non-dominated sorting is performed on the combined population, the non-dominated level and the crowding distance of each individual are calculated, the first several individuals are selected in the order from low to high in the non-dominated level and from large to small in the crowding distance in the same level to form the next generation population, and after the iteration is ended, all non-dominated solutions are combined and de-duplicated to obtain the Pareto optimal solution set.

9. The multi-objective optimization based zoned irrigation control method of a drip pipe network according to claim 1, characterized in that, S6 is specifically: For each solution in the Pareto optimal solution set, its three objective function values ​​are standardized, entropy weights are calculated to determine objective weights, and comprehensive weights are obtained by combining these with decision-maker preferences. The weighted Euclidean distance from each solution to the ideal solution is calculated, and the solution with the highest relative closeness is selected as the final decision solution. The final decision solution Includes optimal irrigation parameters , Indicates the first Optimal ingress target traffic for each partition. Indicates the first Optimal irrigation duration for each zone; The final decision pair corresponding irrigation parameters are converted into control instructions of each partition, and feedback correction is made according to real-time soil moisture monitoring in the actual irrigation process.

10. A multi-objective optimization based zoned irrigation control system for drip pipe network, executing the multi-objective optimization based zoned irrigation control method for drip pipe network as claimed in claim 1, characterized in that, The method comprises the following steps: a data acquisition module is used to collect multi-source heterogeneous farmland data; a data preprocessing module is used to remove outliers and normalize the multi-source heterogeneous farmland data to obtain normalized features; a sampling point partition module is used to cluster the normalized features and spatial coordinates by constructing a joint feature vector and using a K-means clustering method to obtain the partition label of each sampling point; a multi-objective irrigation optimization model construction module is used to construct a multi-objective irrigation optimization model based on partition characteristics, take the irrigation time of each partition and the inlet target flow as decision variables, correlate the partition flow and pressure through a pipe network hydraulic model, define three objective functions of irrigation uniformity, system energy consumption and water resource utilization efficiency, and realize the modeling of multi-objective irrigation optimization; an optimal solution calculation module is used to obtain a Pareto optimal solution set by using a hybrid strategy multi-objective evolutionary algorithm, through chaotic mapping initialization, adaptive hybrid mutation, fast non-dominated sorting and crowded distance selection; an irrigation parameter correction module is used to introduce real-time feedback adjustment by using the superior-inferior solution distance method combined with the entropy weight method and the decision maker's preference, and correct the irrigation parameters according to the actual soil moisture monitoring.

Images