Kiwi fruit intelligent irrigation method and system adopting distributed optimization control

Abstract

The application provides a kiwi fruit intelligent irrigation method and system adopting distributed optimization control, and belongs to the field of intelligent control. The method comprises the following steps: obtaining weather prediction data of a future time zone of a target area, initial irrigation parameters and kiwi fruit expected growth; performing environment zoning and kiwi fruit growth zoning to obtain a first zoning result and a second zoning result, and intersecting the first zoning result and the second zoning result to obtain a target area zoning result; performing growth prediction to obtain a kiwi fruit growth prediction result; extracting a zoning where the kiwi fruit growth prediction result is inconsistent with the kiwi fruit expected growth, performing irrigation parameter optimization, obtaining distributed irrigation parameters, and performing kiwi fruit distributed irrigation control. The application solves the technical problem that the existing technology has low kiwi fruit irrigation intelligence, and cannot accurately irrigate according to different regional environments and growth differences, and achieves the technical effects of improving the kiwi fruit irrigation intelligence level and realizing distributed accurate irrigation based on regional environment and growth difference.

Description

Technical Field

[0001] This invention relates to the field of intelligent control, and in particular to a smart irrigation method and system for kiwifruit using distributed optimization control. Background Technology

[0002] Kiwifruit has a fleshy root system, with the root system mainly concentrated in the shallow soil layer of 20-40 cm. This makes it quite sensitive to changes in soil moisture and soil structure, requiring precise irrigation management to ensure normal growth and high-quality fruit. Currently, kiwifruit irrigation uses a traditional, area-wide irrigation model, which involves fixed irrigation according to the phenological patterns of the growing plants. When the weather changes, growers adjust irrigation schedules based on experience and the growth of the kiwifruit trees.

[0003] However, this traditional irrigation method has a low level of intelligence, lacking the prediction and analysis of environmental factors and fruit tree growth, making it difficult to achieve precise water management. Secondly, it relies heavily on the expertise of growers, with irrigation decisions primarily based on manual experience and lacking objective data support. Furthermore, traditional methods cannot differentiate irrigation based on regional environmental conditions and variations in kiwifruit growth, potentially leading to over- or under-irrigation in some areas, negatively impacting kiwifruit growth quality and fruit formation. Summary of the Invention

[0004] This invention addresses the technical problem of low intelligence in kiwifruit irrigation in existing technologies, which prevents precise irrigation based on different regional environments and growth differences. It provides a smart irrigation method and system for kiwifruit using distributed optimization control to solve this problem.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0006] In a first aspect, this invention provides a smart irrigation method for kiwifruit using distributed optimization control, applied to a distributed irrigation controller for kiwifruit. The method includes: obtaining weather forecast data for the future time zone of a target area, initial irrigation parameters, and expected kiwifruit growth; dividing the target area into environmental partitions to obtain a first partition result, and dividing the target area into kiwifruit growth partitions to obtain a second partition result; intersecting the first and second partition results to obtain a target area partition result; traversing the target area partition results, extracting the monitored kiwifruit growth in each partition, and combining the weather forecast data and the initial irrigation parameters to predict the growth, obtaining a kiwifruit growth prediction result, wherein the kiwifruit growth prediction result corresponds one-to-one with the target area partition result; extracting partitions where the kiwifruit growth prediction result does not match the expected kiwifruit growth, optimizing the irrigation parameters to obtain distributed irrigation parameters, and performing distributed irrigation control for kiwifruit.

[0007] Secondly, this invention provides a smart irrigation system for kiwifruit using distributed optimization control, applied to a distributed irrigation controller for kiwifruit. The system includes: a data acquisition module for acquiring weather forecast data for the future time zone of a target area, initial irrigation parameters, and expected kiwifruit growth; a dual-partitioning module for partitioning the target area into environmental partitions to obtain a first partition result, and partitioning the target area into kiwifruit growth partitions to obtain a second partition result; a partition intersection module for intersecting the first and second partition results to obtain a target area partition result; a growth prediction module for traversing the target area partition results, extracting the monitored kiwifruit growth in each partition, and combining the weather forecast data and the initial irrigation parameters to predict the growth, obtaining a kiwifruit growth prediction result, wherein the kiwifruit growth prediction result corresponds one-to-one with the target area partition result; and an optimization control module for extracting partitions where the kiwifruit growth prediction result does not match the expected kiwifruit growth, performing irrigation parameter optimization to obtain distributed irrigation parameters, and performing distributed irrigation control for kiwifruit.

[0008] The beneficial effects of this invention are:

[0009] The process involves obtaining weather forecast data for the target area in the future time zone, initial irrigation parameters, and expected kiwifruit growth, providing foundational data support for subsequent distributed irrigation control. The target area is divided into environmental zones (first zone result) and kiwifruit growth zones (second zone result), achieving a refined division of the target area through dual zoning. The first and second zone results are intersected to obtain the target area zoning results, ensuring that each zone considers both environmental factors and differences in kiwifruit growth. The target area zoning results are iterated through, and the monitored kiwifruit growth of each zone is extracted. Combined with weather forecast data and initial irrigation parameters, growth prediction is performed to obtain kiwifruit growth prediction results. These kiwifruit growth prediction results correspond one-to-one with the target area zoning results, enabling prediction of future growth in each zone. Zones where the kiwifruit growth prediction results do not match the expected kiwifruit growth are extracted. Irrigation parameters are optimized to obtain distributed irrigation parameters, and distributed irrigation control of kiwifruit is implemented, achieving precise distributed optimization control of the zones requiring adjustment.

[0010] The above technical solution realizes intelligent irrigation of kiwifruit based on distributed optimization control, which changes the fixed mode of traditional whole-area irrigation and adopts non-periodic, zoned distributed irrigation method, improving the level of intelligence of kiwifruit irrigation. It can carry out differentiated and precise irrigation control according to the environmental conditions and kiwifruit growth differences in different areas, effectively solving the technical problems of low intelligence and inability to accurately irrigate in existing technologies. Attached Figure Description

[0011] Figure 1 A flowchart illustrating a smart irrigation method for kiwifruit using distributed optimization control provided by this invention;

[0012] Figure 2 This invention provides a schematic diagram of a smart irrigation system for kiwifruit using distributed optimization control.

[0013] In the attached diagram, the components represented by each number are as follows:

[0014] Data acquisition module 11, dual partitioning module 12, partition intersection module 13, growth prediction module 14, and optimization control module 15. Detailed Implementation

[0015] 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.

[0016] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0017] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0018] Example 1, as Figure 1 As shown, this embodiment of the invention provides a smart irrigation method for kiwifruit using distributed optimization control, applied to a distributed irrigation controller for kiwifruit, including:

[0019] S1. Obtain weather forecast data for the future time zone of the target area, initial irrigation parameters, and expected growth of kiwifruit.

[0020] Specifically, the target area refers to the kiwifruit planting area that requires intelligent irrigation control. This area can be a single orchard, multiple contiguous orchards, or a large-scale kiwifruit planting base. After determining the target area, the system obtains weather forecast data for the target area's future time zone, initial irrigation parameters, and expected kiwifruit growth.

[0021] The weather forecast data for the future time zone includes, but is not limited to, meteorological elements such as rainfall, temperature, humidity, wind speed, and sunshine duration for the next 7-15 days. This weather forecast data is obtained through a data interface with meteorological departments or third-party weather forecasting service platforms and stored in time series format. The time resolution can be set to hourly or daily levels according to actual needs. The scope of the weather forecast data acquisition should cover the target area and its surrounding affected areas to ensure the accuracy and representativeness of the forecast.

[0022] Initial irrigation parameters refer to the basic irrigation control parameters set before the kiwifruit distributed irrigation controller starts operating. These parameters include irrigation frequency, irrigation duration, irrigation intensity, and irrigation water temperature, and are directly uploaded and input through the user terminal. These initial irrigation parameters can be set based on traditional experience or by referring to historical irrigation data from similar kiwifruit growing areas, providing basic data support for optimizing subsequent irrigation parameters.

[0023] The expected growth status of kiwifruit is preset by the user based on planting goals and variety characteristics, including expected values ​​for growth indicators such as leaf number, internode length, shoot thickness, leaf shape index, and fruit development status. The expected growth status should be determined in conjunction with the kiwifruit's growth cycle characteristics, setting corresponding expected parameter ranges for different growth stages. Expected growth status data is stored in numerical form for easy quantitative comparison with predicted growth status later.

[0024] By acquiring weather forecast data for the future time zone of the target area, initial irrigation parameters, and expected growth of kiwifruit, a data foundation was laid for distributed irrigation control.

[0025] S2. Divide the target area into environmental zones to obtain the first zone result, and divide the target area into kiwi fruit growth zones to obtain the second zone result.

[0026] Specifically, the target area was divided into environmental zones and kiwi fruit growth zones, and the first zone results and the second zone results were obtained respectively.

[0027] First, the target area is divided into environmental zones. Specifically, based on the differences in environmental attributes at different locations within the target area, it is divided into several environmentally similar sub-regions. Environmental zoning primarily considers key environmental factors affecting kiwifruit growth, including but not limited to soil moisture, soil temperature, soil pH, light intensity, altitude, slope, and wind speed. By performing cluster analysis on the distribution information of these environmental attribute monitoring values, regions with similar environmental characteristics are identified and merged to obtain the first zoning result. Each zone in the first zoning result has relatively uniform environmental conditions, providing an environmental foundation for subsequent precision irrigation control.

[0028] Simultaneously, the target area was divided into kiwifruit growth zones. Specifically, based on the differences in the actual growth status of kiwifruit within the target area, it was divided into several sub-regions with similar growth. Kiwifruit growth zoning considered key indicators reflecting the growth status of kiwifruit, including but not limited to leaf number, internode length, shoot thickness, leaf shape index, and fruit development status. This zoning process was the same as the environmental zoning process; by performing cluster analysis on the distribution information of kiwifruit growth monitoring values, regions with similar growth characteristics were identified and merged to obtain the second zoning results. Each zone in the second zoning results has a relatively consistent kiwifruit growth level, providing a growth basis for subsequent growth prediction and irrigation optimization.

[0029] By dividing the target area into environmental zones and kiwi fruit growth zones, the first and second zone results were obtained, laying the foundation for subsequent comprehensive zoning.

[0030] S3. Intersect the first partitioning result and the second partitioning result to obtain the target area partitioning result.

[0031] Specifically, the spatial intersection operation is performed between the first partition result (environmental partition result) and the second partition result (kiwi fruit growth partition result) to form a comprehensive target area partition result.

[0032] Based on the intersection of the first and second partitioning results, spatial overlay analysis is used to perform geometric intersection operations between the boundaries of each environmental partition in the first partitioning result and the boundaries of the kiwifruit growth partitions in the second partitioning result. During the intersection process, overlapping areas between the two partitioning results are identified, and new partition boundaries are generated based on the overlap. Each new partition generated by the intersection simultaneously possesses relatively uniform environmental conditions and similar kiwifruit growth characteristics, forming a comprehensive partitioning unit under the dual constraints of environment and growth.

[0033] The target area zoning result is the final zoning scheme formed after intersection. Each zone in this result has a clear spatial boundary and dual attribute characteristics. Specifically, each zone includes environmental parameters (such as soil moisture, soil temperature, and light intensity) and growth parameters (such as leaf number, internode length, and shoot thickness). These parameters are relatively consistent within a zone, but show significant differences between different zones. The target area zoning result provides precise spatial units for subsequent zone growth prediction and personalized irrigation parameter optimization.

[0034] By intersecting the results of the first and second partitions, a target area partitioning result that takes into account both environmental conditions and kiwi fruit growth was obtained, thus achieving a refined division of the target area.

[0035] S4. Traverse the target area partitioning results, extract the kiwi fruit monitoring growth of each partition, combine the weather forecast data and the initial irrigation parameters to make a growth prediction, and obtain the kiwi fruit growth prediction results, wherein the kiwi fruit growth prediction results correspond one-to-one with the target area partitioning results.

[0036] Specifically, by processing each partition in the target area partitioning results one by one, the kiwi fruit monitoring growth of each partition is extracted, and the growth is predicted by combining weather forecast data and initial irrigation parameters to form the kiwi fruit growth prediction results.

[0037] First, each partition in the target area partitioning results is visited sequentially, and the same growth prediction process is performed on each partition. During the traversal, the current kiwifruit growth monitoring data for the current partition is extracted, including the current monitored values ​​of growth parameters such as leaf number, internode length, shoot thickness, leaf shape index, and fruit development status. Then, based on the partition's kiwifruit growth monitoring, weather forecast data, and initial irrigation parameters, the expected growth changes of kiwifruit in that partition in the future are predicted. During the prediction process, the partition's environmental parameters, current growth parameters, weather forecast data, and initial irrigation parameters are combined and calculated using the target area representative sample growth prediction model to obtain the predicted kiwifruit growth value for that partition in the future. Finally, the predicted kiwifruit growth values ​​for each partition in the future are summarized to form the kiwifruit growth prediction result. This kiwifruit growth prediction result corresponds one-to-one with the target area partitioning results, meaning that each partition has a corresponding growth prediction result, including predicted values ​​for growth parameters such as leaf number, internode length, and shoot thickness in the future.

[0038] By traversing the target area partitioning results and performing growth prediction, the kiwifruit growth prediction results corresponding to each partition were obtained, providing predictive data support for subsequent irrigation parameter optimization.

[0039] S5. Extract the partitions where the predicted growth of kiwifruit does not match the expected growth of kiwifruit, perform irrigation parameter optimization, obtain distributed irrigation parameters, and perform distributed irrigation control for kiwifruit.

[0040] Specifically, by comparing the predicted growth of kiwifruit in each zone with the expected growth of kiwifruit, zones with substandard growth are identified, and irrigation parameters are optimized for these zones to achieve distributed irrigation control.

[0041] First, the predicted kiwifruit growth results for each zone are compared with the expected kiwifruit growth, and the deviation between the two is calculated. When the deviation is greater than or equal to a preset growth parameter deviation threshold, the zone is identified as having inconsistent growth and requires irrigation parameter optimization. When the deviation is less than the threshold, the zone is identified as having consistent growth, and the current irrigation parameters are maintained. Then, an irrigation parameter optimization process is performed on the identified inconsistent growth zones. This process initializes multiple sets of irrigation parameters, calculates the predicted growth results for each set, compares them with the expected kiwifruit growth, iterates, and gradually converges to the optimal irrigation parameters. Afterward, the corresponding optimal irrigation parameters are determined for each inconsistent growth zone, and combined with the maintenance parameters for consistent growth zones, forming a distributed irrigation parameter set. This distributed irrigation parameter set includes personalized irrigation control parameters for each zone, and differentiated irrigation control operations are performed on the corresponding zones based on these parameters.

[0042] By extracting regions with inconsistent growth and optimizing irrigation parameters, targeted distributed irrigation parameters were obtained, enabling distributed irrigation control of kiwifruit based on growth prediction.

[0043] Furthermore, the target area is divided into environmental partitions to obtain the first partition result, including:

[0044] S21. Collect the distribution information of the first environmental attribute monitoring values ​​of the target area, perform cluster analysis, and obtain the first environmental attribute partitioning results;

[0045] S22. Continue collecting the distribution information of the Nth environmental attribute monitoring value of the target area, perform cluster analysis, and obtain the Nth environmental attribute partitioning result;

[0046] S23. Intersect the first environment attribute partitioning results up to the Nth environment attribute partitioning results to obtain the first partitioning result.

[0047] In a preferred embodiment, when dividing the target area into environmental zones to obtain the first zone result, the distribution information of the first environmental attribute monitoring values ​​of the target area is first collected. The first environmental attribute can be any one of the environmental factors affecting kiwifruit growth, such as soil moisture, soil temperature, soil pH, light intensity, altitude, slope, aspect, and wind speed. Real-time monitoring values ​​of this first environmental attribute at different spatial locations are acquired through an environmental monitoring sensor network deployed within the target area, forming the first environmental attribute monitoring value distribution information. This first environmental attribute monitoring value distribution information is stored in the form of coordinate-value pairs, recording the spatial location of each monitoring point within the target area and the corresponding monitoring value of the first environmental attribute. Subsequently, clustering analysis methods such as K-means clustering, hierarchical clustering, or DBSCAN clustering are used to process the first environmental attribute monitoring value distribution information, grouping spatial locations with similar environmental attribute monitoring values ​​into the same category, and forming continuous zone boundaries based on the principle of spatial continuity, thereby obtaining the first environmental attribute zoning result.

[0048] Then, following the same procedure, the monitoring value distribution information of the second, third, and up to the Nth environmental attribute in the target area was collected. The collection process for each environmental attribute was completed using corresponding dedicated sensors, such as soil moisture sensors, temperature sensors, pH meters, and light meters. Cluster analysis was performed on the monitoring value distribution information of each environmental attribute, using the same algorithm parameter settings to ensure consistency in the partitioning results. Each environmental attribute cluster analysis generated corresponding partitioning results, ultimately yielding the partitioning results for the second, third, and up to the Nth environmental attribute. Here, N represents the total number of environmental attributes affecting kiwifruit growth.

[0049] Subsequently, spatial intersection operations are performed on the first environmental attribute partitioning results up to the Nth environmental attribute partitioning results. The intersection process employs GIS spatial analysis technology, overlaying the geometric boundaries of each environmental attribute partitioning result. Specifically, the partition boundary of the first environmental attribute partitioning result is intersected with the partition boundary of the second environmental attribute partitioning result to generate a new partition boundary. This new partition boundary is then intersected with the third environmental attribute partitioning result, and so on, until the intersection with the Nth environmental attribute partitioning result is achieved. After this progressive intersection process, the original environmental attribute partitions are re-divided, forming new partitioning units. Each new partitioning unit simultaneously inherits the partitioning characteristics of multiple environmental attributes; that is, multiple environmental parameters such as soil moisture, temperature, pH value, and light intensity are relatively uniform within each partition. The resulting first partitioning result contains several sub-regions with relatively uniform overall environmental conditions.

[0050] Through stepwise clustering analysis and spatial intersection of multiple environmental attributes, the first zoning result was obtained, which comprehensively considers multiple environmental factors. This result can accurately reflect the spatial distribution characteristics of environmental conditions within the target area, providing an accurate environmental zoning basis for subsequent growth prediction and irrigation control.

[0051] The process of zoning kiwifruit growth in the target area is the same as the environmental zoning process. First, the distribution information of the first growth attribute monitoring value is collected in the target area. This first growth attribute can be any one of the indicators reflecting the growth status of kiwifruit, such as leaf number, internode length, shoot thickness, leaf shape index, and fruit development status. The monitoring values ​​of this first growth attribute at different spatial locations are acquired using image recognition technology or dedicated sensors, forming spatial distribution information. Then, a clustering analysis algorithm is used to process the monitoring value distribution information, grouping spatial locations with similar growth attribute monitoring values ​​into the same category, obtaining the first growth attribute zoning result. The monitoring value distribution information of the second, third, up to the Mth growth attribute is then collected in the target area, and clustering analysis is performed on each growth attribute to obtain the second, third, up to the Mth growth attribute zoning result. Here, M is the total number of growth attributes reflecting the growth status of kiwifruit. Subsequently, spatial intersection operations are performed on the first growth attribute partitioning results up to the Mth growth attribute partitioning results. The geometric boundaries of each growth attribute partitioning result are superimposed and analyzed. After stepwise intersection operations, new partitioning units are formed. Each new partitioning unit inherits the partitioning characteristics of multiple growth attributes. That is, multiple growth parameters such as the number of leaves, internode length, and new shoot thickness are relatively uniform within each partition, and finally, the second partitioning results are obtained.

[0052] Furthermore, the distribution information of the first environmental attribute monitoring values ​​of the target area is collected, and cluster analysis is performed to obtain the first environmental attribute partitioning results, including:

[0053] S211. Extract the environmental attribute monitoring value at the first location and the environmental attribute monitoring value at the second location from the first environmental attribute monitoring value distribution information, wherein the first location and the second location are adjacent locations;

[0054] S212. Calculate the environmental attribute monitoring value deviation between the environmental attribute monitoring value of the first location and the environmental attribute monitoring value of the second location;

[0055] S213. When the deviation of the environmental attribute monitoring value is less than the first environmental attribute deviation threshold, the first position and the second position are merged into the third position. At the same time, the average value of the environmental attribute monitoring value of the first position and the environmental attribute monitoring value of the second position is calculated and set as the environmental attribute monitoring value of the third position. The first environmental attribute deviation threshold is a user preset value.

[0056] S214. When the deviation of the environmental attribute monitoring value is greater than or equal to the first environmental attribute deviation threshold, the first position and the second position are regarded as different positions.

[0057] S215. When the deviation of the environmental attribute monitoring values ​​of any two adjacent locations is greater than or equal to the first environmental attribute deviation threshold, output the first environmental attribute partitioning result.

[0058] In a preferred embodiment, when performing cluster analysis on the distribution information of the first environmental attribute monitoring values, iterative clustering based on neighborhood merging is adopted. By gradually comparing the differences in environmental attribute monitoring values ​​of adjacent locations, similar locations are merged into the same partition until merging can no longer continue.

[0059] First, adjacent position pairs are extracted from the first environmental attribute monitoring value distribution information according to a predetermined traversal order. The first environmental attribute monitoring value distribution information is stored in the form of spatial coordinates and corresponding monitoring values, containing the location information of all monitoring points within the target area and the first environmental attribute values. Each pair of adjacent positions, namely the first position and the second position, is extracted sequentially, and their corresponding environmental attribute monitoring values ​​are obtained as comparison objects, namely, the environmental attribute monitoring values ​​of the first position and the second position.

[0060] Then, the degree of difference between the environmental attribute monitoring values ​​at the first location and the second location is calculated to obtain the environmental attribute monitoring value deviation. Specifically, the environmental attribute monitoring value deviation is calculated using the absolute value difference method, i.e., Environmental attribute monitoring value deviation = |Environmental attribute monitoring value at the first location - Environmental attribute monitoring value at the second location|. This environmental attribute monitoring value deviation reflects the degree of difference in environmental conditions between adjacent locations and is an indicator for determining whether to merge locations.

[0061] Subsequently, the calculated deviation of the environmental attribute monitoring values ​​is compared with a first environmental attribute deviation threshold. This first environmental attribute deviation threshold is a numerical parameter preset by the user based on the natural variation range of specific environmental attributes, the measurement accuracy of the monitoring equipment, the desired level of zoning fineness, and actual application requirements. When the deviation of the environmental attribute monitoring values ​​is less than the first environmental attribute deviation threshold, it indicates that the environmental attributes of the two adjacent locations are highly similar, meeting the merging condition. At this time, a location merging operation is performed, geometrically merging the spatial regions of the first and second locations to generate a new third location. Simultaneously, the arithmetic mean of the environmental attribute monitoring values ​​of the first and second locations is calculated as the environmental attribute monitoring value of the third location, representing the overall environmental characteristics of the merged area. When the deviation of the environmental attribute monitoring values ​​is greater than or equal to the first environmental attribute deviation threshold, the first and second locations are considered different locations, and no merging operation is performed. This indicates that the environmental attributes of the two adjacent locations are significantly different and should belong to different zones, maintaining the independence of the two locations, and the next pair of adjacent locations is processed.

[0062] The comparison and merging process described above is repeated continuously, traversing all possible pairs of adjacent locations. After completing one round of traversal, it is checked whether there are still pairs of adjacent locations with a deviation less than a threshold. If they still exist, the next round of comparison and merging operations is performed. When the deviation of the environmental attribute monitoring values ​​of any two adjacent locations is greater than or equal to the first environmental attribute deviation threshold, it indicates that all locations meeting the merging conditions have been merged, and the clustering process ends. At this point, the first environmental attribute partitioning result is output. This first environmental attribute partitioning result contains several spatially contiguous partitions. The environmental attribute monitoring values ​​within each partition are relatively uniform, while there are significant differences in environmental attributes between partitions.

[0063] By using cluster analysis based on the stepwise merging of adjacent locations, the target area can be divided into different partitions according to the spatial similarity of environmental attributes, providing an accurate environmental partitioning basis for subsequent multi-attribute partition intersection and irrigation control.

[0064] Furthermore, the target area is traversed through the partitioning results to extract the kiwifruit growth monitoring data for each partition. Combined with the weather forecast data and the initial irrigation parameters, growth prediction is performed to obtain the kiwifruit growth prediction results, including:

[0065] S41. Obtain the kiwi fruit planting duration, add the future time zones to obtain the kiwi fruit planting duration sequence;

[0066] S42. Extract the first kiwifruit planting duration up to the Qth kiwifruit planting duration from the kiwifruit planting duration sequence;

[0067] S43. Extract the first partition from the target area partitioning result, wherein the first partition has first partition environmental parameters and first partition growth parameters;

[0068] S44. Traverse the first kiwifruit planting duration up to the Qth kiwifruit planting duration, and combine them with the first partition environmental parameters, the first partition growth parameters and the initial irrigation parameters respectively to construct growth constraints, perform representative sample statistics, obtain representative sample growth parameters, and set them as the first partition kiwifruit predicted growth sequence.

[0069] S45. Add the predicted growth sequence of kiwifruit in the first partition to the predicted growth result of kiwifruit.

[0070] In a preferred embodiment, firstly, the kiwifruit planting duration is obtained, and then accumulated with future time zones to obtain a kiwifruit planting duration sequence. The kiwifruit planting duration refers to the cumulative time from the start of the kiwifruit's budding stage in the current year to the current moment, usually expressed in days or weeks. The current kiwifruit planting duration is then accumulated with the time length of the future time zone to generate a continuous time series, i.e., the kiwifruit planting duration sequence. This kiwifruit planting duration sequence includes all time points from the current moment to the end of the predicted future time period, providing a time basis for subsequent growth prediction.

[0071] Then, prediction time points are extracted from the kiwifruit planting duration sequence. Specifically, the first kiwifruit planting duration up to the Qth kiwifruit planting duration is extracted from the kiwifruit planting duration sequence at preset time intervals, where Q is the total number of prediction time points. These extracted time points constitute the time nodes for growth prediction, with each time point corresponding to a specific kiwifruit growth stage.

[0072] Subsequently, the information of the zones to be predicted is extracted from the target area zoning results. Taking the first zone as an example, this first zone is any one of the multiple zones in the target area zoning results. First, the first zone is selected as the prediction object. This first zone has first zone environmental parameters and first zone growth parameters. The first zone environmental parameters include environmental characteristics such as soil moisture, temperature, and light intensity. The first zone growth parameters include current growth monitoring values ​​of kiwifruit in this zone, such as the number of leaves, internode length, and new shoot thickness, providing zone-specific basic data for growth prediction.

[0073] Next, the planting duration of kiwifruit from the first time point to the Qth time point is iterated, and growth constraints are constructed by combining these constraints with the environmental parameters, growth parameters, and initial irrigation parameters of the first region. For each time point, the corresponding planting duration of kiwifruit is combined with the environmental parameters, growth parameters, and initial irrigation parameters of the first region to form the growth constraint input for that time point. Subsequently, representative sample statistical analysis is performed on these growth constraints using a growth prediction model for representative samples of the target area to calculate the representative sample growth parameters for the corresponding time points. By iterating through all time points and performing corresponding prediction calculations, the predicted growth sequence of kiwifruit in the first region is obtained, which contains the predicted growth values ​​for the first region at future time points.

[0074] Next, the predicted kiwifruit growth sequence for the first partition was added to the kiwifruit growth prediction results. The above prediction process was repeated for each partition in the target area partitioning results, gradually constructing a complete kiwifruit growth prediction result. This kiwifruit growth prediction result includes predictions of growth changes in each partition over future time periods, providing predictive data support for subsequent irrigation parameter optimization and distributed irrigation control.

[0075] By using time-series-based growth forecasting, we can accurately predict the growth changes of kiwifruit in each region over future periods, laying a predictive foundation for achieving precise distributed irrigation control.

[0076] Furthermore, by iterating through the first kiwifruit planting duration up to the Qth kiwifruit planting duration, and combining these with the first partition environmental parameters, the first partition growth parameters, and the initial irrigation parameters, growth constraints are constructed. Representative sample statistics are then performed to obtain representative sample growth parameters, which are set as the predicted growth sequence for the first partition kiwifruit, including:

[0077] S441. Collect time-series information on the growth of the first kiwifruit without varietal differences;

[0078] S442. Based on the segmentation time step configured by the user, traverse the first kiwifruit growth monitoring time series information set and randomly extract multiple kiwifruit growth monitoring time series information.

[0079] S443. Traverse the multiple kiwifruit growth monitoring time series information, extract the kiwifruit growth monitoring information at the end time and set it as the supervisory data, extract the kiwifruit growth monitoring information at the start time, and combine the planting environment record data, the planting duration at the end time and the irrigation parameter record data as the input data to train the initial representative sample growth prediction model.

[0080] S444. Collect the second kiwifruit growth monitoring time series information set of the kiwifruit varieties in the target area, train the initial representative sample growth prediction model to obtain the target area representative sample growth prediction model, and embed it into the kiwifruit distributed irrigation controller.

[0081] In a preferred embodiment, firstly, a time-series data set of kiwifruit growth monitoring data without varietal differences is collected. This time-series data set eliminates varietal differences, meaning all kiwifruit growth monitoring data in the dataset comes from the same variety, thus avoiding interference from differences in growth characteristics between different varieties on model training. This time-series data set is stored in time series format, recording the changes in growth parameters such as leaf number, internode length, shoot thickness, and leaf shape index at different growth stages, covering the complete growth cycle from initial planting to maturity. Each time-series data point includes the growth parameter value at a specific time and the corresponding time identifier, forming a continuous time-series dataset. This varietal-neutral data collection method ensures that the initial representative sample growth prediction model can learn pure growth change patterns without being affected by variety-specific factors, providing reliable basic data support for subsequent transfer learning training.

[0082] Then, based on the segmentation time step configured on the user's end, the first kiwifruit growth monitoring time series information set is traversed and randomly segmented to obtain multiple kiwifruit growth monitoring time series information. The segmentation time step is a time interval parameter preset by the user according to the prediction accuracy requirements and computing resource limitations, usually set to several days to several weeks. The complete first kiwifruit growth monitoring time series information set is segmented according to this segmentation time step, and each segment contains a complete input-output data pair. The random segmentation process ensures the diversity and representativeness of the training data and avoids the time bias problem of time series data.

[0083] Subsequently, a training dataset was constructed by traversing multiple time-series monitoring data points on kiwifruit growth. Specifically, for each time-series segment, the kiwifruit growth monitoring information at the end time was extracted as supervisory data, representing the target output that the model needs to predict. Simultaneously, the kiwifruit growth monitoring information at the start time was extracted and combined with corresponding planting environment records, planting duration records, and irrigation parameter records at the end time to form complete input data, resulting in the training dataset. Through this input-output data pair construction method, a growth prediction model for the initial representative samples was trained using supervised learning methods.

[0084] Specifically, the initial representative sample growth prediction model adopts a fully connected neural network structure. The number of nodes in the input layer of this network equals the total dimension of the input parameters, specifically including the sum of the dimensions of kiwifruit growth parameters at the starting time (such as leaf number, internode length, and shoot thickness), planting environment recorded data (such as soil moisture, temperature, and light intensity), planting duration at the ending time, and irrigation parameter recorded data (such as irrigation frequency and intensity). The number of nodes in the output layer of the network equals the dimension of the kiwifruit growth parameters to be predicted, corresponding to the number of growth indicators such as leaf number, internode length, and shoot thickness at the ending time. The constructed training dataset is divided into a training set and a validation set according to a preset ratio, usually 8:2 or 7:3. The training process uses the batch gradient descent algorithm, inputting the training data into the neural network in batches for forward propagation calculation. During forward propagation, the input data passes through each layer of the network sequentially, undergoes weight matrix operations and activation function processing, and then generates predicted growth parameter values ​​at the output layer. The mean squared error loss function between the predicted values ​​and the actual supervised data is calculated, reflecting the current prediction error level of the model. Then, the gradient of the loss function with respect to the weight parameters of each layer is calculated using the backpropagation algorithm, and the network weights are updated using an optimization algorithm (such as the Adam optimizer or SGD optimizer). During optimization, the weight parameters are adjusted according to the calculated gradient information and the set learning rate, so that the loss function value gradually decreases. This process of forward propagation-error calculation-backpropagation-weight update constitutes a complete training iteration cycle. Multiple training iteration cycles are repeated, and the model performance is evaluated on the validation set after each cycle, monitoring the trend of the loss function and the prediction accuracy. When the loss function on the validation set converges and the prediction accuracy reaches the preset requirement, or when the validation performance shows no significant improvement over several consecutive cycles, the training process is stopped, and the initial representative sample growth prediction model is obtained.

[0085] Subsequently, a second set of time-series data on the growth of kiwifruit varieties in the target area was collected, and the initial representative sample growth prediction model was trained using transfer learning. This second set of data specifically collected growth monitoring data for particular kiwifruit varieties within the target area, exhibiting both regional and variety specificity, reflecting the environmental conditions of the target area and the growth characteristics of specific varieties. During transfer learning, the weight parameters of the pre-trained fully connected layers in the initial representative sample growth prediction model were kept unchanged, i.e., the original fully connected layer structure was frozen to retain the general growth change patterns learned from the first set of time-series data. Specifically, the transfer learning employed a progressive neuron expansion method for model adaptive training. First, a new neuron layer was added after the output layer of the initial representative sample growth prediction model. The input dimension of this new neuron layer was the same as the output dimension of the initial representative sample growth prediction model, while the output dimension remained consistent with the growth parameter dimension. During training, the second set of time-series data on kiwifruit growth was segmented according to the same time step, constructing input-output data pairs for specific varieties in the target area. The input data is first processed through a frozen fully connected layer to obtain an intermediate feature representation, which is then used as the input to a newly added neuron layer. This new neuron layer learns the growth variation characteristics of a specific variety in the target region, correcting and optimizing the general prediction results of the initial representative sample growth prediction model. During training, the mean squared error loss function is used to calculate the prediction error between the output of the new neuron layer and the actual growth data of the target region. The backpropagation algorithm is used to update only the weight parameters of the new neuron layer, while keeping the weights of the original fully connected layers unchanged. This training method retains the general growth prediction capability while incorporating the specific characteristics of the target region. After the training of the new neuron layer converges, the model's prediction accuracy is evaluated on the validation set. Subsequently, the performance of the current model on the target region growth prediction task is evaluated. If the prediction accuracy does not meet the preset requirements, a progressive expansion strategy is continued, adding another neuron layer after the current one. The second newly added neuron layer uses the output of the first newly added neuron layer as input to continue learning deeper region- and variety-specific characteristics. The same method of freezing the weights of the preceding layers and training only the weights of the new layer is used for training. Repeat the above progressive expansion and training process until the model's prediction accuracy on the validation set reaches the preset requirement or the accuracy improvement no longer becomes significant after multiple expansions. The trained model is then designated as the target region's representative sample growth prediction model. This model can accurately predict the growth changes of a specific kiwifruit variety in the target region under given environmental conditions, planting duration, and irrigation parameters. This target region representative sample growth prediction model is then embedded in the kiwifruit distributed irrigation controller as the core computational module for subsequent growth prediction and irrigation parameter optimization.

[0086] Through progressive neuron expansion and transfer learning training, the general growth prediction rules are preserved while incorporating the environmental and variety specificities of the target area. This results in a high-precision growth prediction model applicable to the target area, providing reliable prediction support for achieving precise distributed irrigation control.

[0087] Furthermore, regions where the predicted kiwifruit growth does not match the expected kiwifruit growth are extracted; irrigation parameters are optimized to obtain distributed irrigation parameters; and distributed irrigation control of the kiwifruit is implemented, including:

[0088] S511. Extract the predicted growth of kiwifruit in the first region from the predicted growth results of kiwifruit;

[0089] S512. Compare the predicted growth of kiwifruit in the first partition with the expected growth of kiwifruit to obtain the kiwifruit growth deviation, wherein the kiwifruit growth deviation is equal to the Euclidean distance of the normalized values ​​of the kiwifruit growth parameters.

[0090] S513. When the growth deviation of the kiwifruit is greater than or equal to the growth parameter deviation threshold, it is considered that the predicted growth of the kiwifruit in the first zone is inconsistent with the expected growth of the kiwifruit; otherwise, it is considered that they are consistent.

[0091] In a preferred embodiment, firstly, prediction data for a specific zone is extracted from the kiwifruit growth prediction results. Taking the first zone as an example, this first zone is any one of multiple zones in the target area zoning results. The predicted growth of kiwifruit in the first zone is extracted from the kiwifruit growth prediction results. This predicted growth of kiwifruit in the first zone includes predicted values ​​of growth parameters such as the number of leaves, internode length, shoot thickness, leaf shape index, and fruit development status in the first zone over a future time period. These predicted values ​​are stored in numerical form, providing basic data for subsequent growth deviation calculations.

[0092] Then, the predicted growth of kiwifruit in the first zone is compared with the expected growth, and the degree of difference between the two is calculated. Specifically, the predicted and expected growth of kiwifruit in the first zone are normalized to eliminate dimensional and numerical range differences between different growth parameters. The normalization process uses the min-max normalization method, mapping the values ​​of each growth parameter to the [0,1] interval. The calculation formula is: Normalized value = (Original value - Minimum value) / (Maximum value - Minimum value). Here, the minimum and maximum values ​​are the minimum and maximum monitored values ​​of the growth parameter in historical data, respectively. After normalization, both the predicted and expected growth of kiwifruit in the first zone are converted into normalized vectors of the same dimension.

[0093] Subsequently, the Euclidean distance between the two normalized vectors was calculated to obtain the kiwifruit growth deviation. This kiwifruit growth deviation reflects the overall difference between the predicted and expected growth of kiwifruit in the first zone; a larger value indicates a more significant difference. Next, the calculated kiwifruit growth deviation was compared with a preset growth parameter deviation threshold to determine growth consistency. The growth parameter deviation threshold is a pre-set numerical parameter based on kiwifruit planting management requirements, growth control precision needs, and actual production experience. When the kiwifruit growth deviation is greater than or equal to the growth parameter deviation threshold, it indicates a significant difference between the predicted and expected growth of kiwifruit in the first zone, and is considered inconsistent. In this case, irrigation parameters need to be optimized for this zone; adjusting the irrigation strategy can improve the growth prediction results and make them closer to the expected growth. When the deviation in kiwifruit growth is less than the growth parameter deviation threshold, it indicates that the difference between the predicted and expected kiwifruit growth in the first zone is within an acceptable range, and is considered consistent with the expected kiwifruit growth in the first zone. At this point, the zone can maintain its current irrigation parameter settings without requiring additional parameter optimization.

[0094] By calculating the growth deviation based on Euclidean distance and comparing thresholds, we can accurately identify the zones where the predicted growth results do not match the expected growth, providing a clear optimization target for subsequent irrigation parameter optimization and ensuring the accuracy and effectiveness of distributed irrigation control.

[0095] Furthermore, regions where the predicted kiwifruit growth does not match the expected kiwifruit growth are extracted; irrigation parameters are optimized to obtain distributed irrigation parameters; and distributed irrigation control of the kiwifruit is implemented, including:

[0096] S521. Extract the first partition where the kiwifruit growth prediction result is inconsistent with the expected kiwifruit growth.

[0097] S522. Initialize the irrigation parameters to obtain several irrigation parameters;

[0098] S523. Traverse the several irrigation parameters and predict the growth of the first zone to obtain several predicted growth of the first zone.

[0099] S524. Traverse the predicted growth of the several first zones and compare it with the expected growth of the kiwifruit to obtain the growth deviation of the kiwifruit in the several first zones.

[0100] S525. When the growth deviation of the kiwifruit in the several first zones is greater than or equal to the growth parameter deviation threshold, based on the growth deviation of the kiwifruit in the several first zones, the minimum growth deviation irrigation parameter, the median growth deviation irrigation parameter, and the maximum growth deviation irrigation parameter are extracted from the several irrigation parameters. The median growth deviation irrigation parameter is one or two sets, and the minimum growth deviation irrigation parameter and the maximum growth deviation irrigation parameter are both one set.

[0101] S526. Using the minimum deviation irrigation parameter of the growth as the target point and the median deviation irrigation parameter of the growth as the starting point, construct a first search axis.

[0102] S527. Using the irrigation parameter with the minimum deviation of growth as the target point and the irrigation parameter with the maximum deviation of growth as the starting point, construct a second search axis.

[0103] S528. Using the median deviation irrigation parameter of the growth as the target point and the maximum deviation irrigation parameter of the growth as the starting point, construct a third search axis.

[0104] S529. Perform a zigzag search along the first search axis, the second search axis and the third search axis respectively to obtain updated irrigation parameters, and execute the loop.

[0105] In a preferred embodiment, firstly, all zones requiring irrigation parameter optimization are selected from the kiwifruit growth prediction results, and these zones are then updated and optimized with their irrigation parameters. From all zones requiring optimization, one zone is selected sequentially for processing, designated as the first zone. Here, the first zone represents any zone requiring irrigation parameter optimization. Then, the irrigation parameters for the first zone are initialized, generating several sets of candidate irrigation parameters for subsequent optimization, resulting in several irrigation parameters. To improve the efficiency and accuracy of irrigation parameter optimization, the initialization process employs a constraint generation method based on healthy kiwifruit samples of the same variety. Specifically, firstly, through historical data retrieval and sample screening, irrigation parameter records for healthy kiwifruit samples of the same variety and in good growth condition as those in the target area are obtained. These healthy samples should meet standards such as excellent growth indicators, absence of pests and diseases, and stable yield, ensuring that their irrigation parameters have reference value. Then, based on the current kiwifruit growth stage of the first zone, irrigation parameter records for the same growth stage are extracted from the healthy sample database, including historical values ​​of key parameters such as irrigation frequency, irrigation duration, irrigation intensity, and irrigation water temperature. Next, a central tendency analysis was performed on the irrigation parameters of the selected healthy samples at the same growth stage. Statistical methods were used to calculate the mean, standard deviation, quantiles, and other statistical characteristics of each irrigation parameter, analyzing the distribution patterns and central tendency of the parameter values. Based on the statistical analysis results, a reasonable distribution range for each irrigation parameter was determined, typically using the mean ± 2 standard deviations or the 10th percentile to the 90th percentile as the parameter distribution range. This parameter distribution range reflects the irrigation parameter characteristics of the healthy samples while avoiding the interference of extreme values, thus narrowing the optimization search space for irrigation parameters. Based on the determined parameter distribution range, several irrigation parameters were generated within the constraints of each parameter using random sampling. Each set of irrigation parameters included complete irrigation control parameter settings, and all parameter values ​​were within the constraints of the healthy samples, ensuring the rationality and feasibility of the initial parameters. This constraint initialization method effectively improves the convergence speed and optimization effect of irrigation parameter optimization.

[0106] Subsequently, the growth prediction and evaluation of the generated irrigation parameters were performed one by one. Specifically, each set of irrigation parameters was selected in turn and combined with the environmental parameters, current growth parameters, and weather forecast data of the first region. The growth prediction model of the representative sample of the target area was used to calculate the growth prediction. During the prediction process, the initial irrigation parameters of the first region were replaced with the current irrigation parameters, keeping other input conditions unchanged. The growth prediction model of the representative sample of the target area was used to calculate the changes in kiwifruit growth in the first region under the given irrigation parameters over a future time period. Each set of irrigation parameters corresponds to an independent growth prediction result, which includes predicted values ​​of key growth indicators such as leaf number, internode length, and shoot thickness. By traversing all initialized irrigation parameters, several predicted growth values ​​for the first region were obtained, forming the basis for subsequent parameter optimization evaluation.

[0107] Next, the growth prediction effect corresponding to each group of irrigation parameters was evaluated. Specifically, the predicted growth of each first zone was selected in turn, and its quantitative comparison with the expected growth of kiwifruit was performed. Using the Euclidean distance calculation method, the growth parameters were normalized, and the deviation value between the predicted growth and the expected growth was calculated. This deviation, representing the degree of closeness between the predicted growth result and the expected target under the current irrigation parameter settings, indicates a better irrigation effect. By traversing all the predicted growth of the first zones, several first-zone kiwifruit growth deviations were obtained, providing a quantitative basis for subsequent parameter selection and search direction determination.

[0108] When the growth deviations corresponding to all irrigation parameters fail to meet the expected results, further parameter optimization search is required. Specifically, when the growth deviations of several kiwifruit in the first zone are all greater than or equal to the growth parameter deviation threshold, it indicates that the currently initialized irrigation parameters cannot meet the growth requirements, and more in-depth parameter optimization is needed. At this time, based on the order of the growth deviations of several kiwifruit in the first zone, key reference points are selected from the corresponding irrigation parameters, namely, the irrigation parameter with the minimum growth deviation, the irrigation parameter with the median growth deviation, and the irrigation parameter with the maximum growth deviation. Specifically, firstly, all irrigation parameters are sorted according to the growth deviation value from smallest to largest, and the irrigation parameter with the smallest deviation is extracted as the irrigation parameter with the minimum growth deviation, which represents the relatively optimal solution under the current conditions. Then, the irrigation parameter with the largest deviation is extracted as the irrigation parameter with the maximum growth deviation, which represents the worst solution under the current conditions. Next, the irrigation parameter with the median growth deviation is determined. When the total number of several irrigation parameters is odd, the median corresponds to a unique middle position, and the irrigation parameter with the median growth deviation is a group. When the total number of several irrigation parameters is even, the median corresponds to the irrigation parameters in the middle two positions. In this case, there are two sets of irrigation parameters representing the median deviation of growth. The median deviation irrigation parameter represents the moderate level of the current parameter distribution, providing a balancing reference point for subsequent searches. Through parameter screening based on deviation ranking, the irrigation parameters with the smallest growth deviation, the median growth deviation irrigation parameter, and the largest growth deviation irrigation parameter were determined, laying the foundation for constructing subsequent parameter search strategies.

[0109] Subsequently, multi-directional search axes were constructed based on the selected key reference points to provide directional guidance for subsequent parameter space searches. When constructing the first search axis, the irrigation parameter with the minimum deviation in growth was used as the target point, and the irrigation parameter with the median deviation in growth was used as the starting point. The direction vector of this first search axis points from the average level towards the current optimal solution in the parameter space, representing an improvement path from average to better performance. When constructing the second search axis, the irrigation parameter with the minimum deviation in growth was used as the target point, and the irrigation parameter with the maximum deviation in growth was used as the starting point. The direction vector of this second search axis points from the worst solution towards the current optimal solution in the parameter space, representing a significant improvement path from poor to good performance. When constructing the third search axis, the irrigation parameter with the median deviation in growth was used as the target point, and the irrigation parameter with the maximum deviation in growth was used as the starting point. The direction vector of this third search axis points from the worst solution towards the average level in the parameter space, representing a gradual improvement path from poor to average performance. Each search axis defines a specific search direction in the multidimensional irrigation parameter space. These axes cover multiple possible improvement directions starting from the currently known parameter points, providing a search strategy for comprehensive parameter space exploration.

[0110] Subsequently, a zigzag search strategy is executed along the three constructed search axes (i.e., the first search axis, the second search axis, and the third search axis) to achieve efficient exploration and optimization of the parameter space. The zigzag search is a non-strict axis search method; compared to a strict linear search along the axis, it can cover a larger search range and increase the probability of finding the global optimum. Specifically, the zigzag search uses a broken-line search path on each axis. Taking the first search axis as an example, the search process starts from the starting point of the median deviation irrigation parameter, advances a certain step along the axis direction, then shifts a certain distance perpendicular to the axis, and continues to advance along the axis direction, thus repeatedly forming a zigzag search trajectory. This search method maintains the overall trend along the axis direction while expanding the exploration to the surrounding neighborhood space. During the search process, each search position corresponds to a new set of irrigation parameters. For each new set of irrigation parameters, a growth prediction model representing the target area is used to predict growth and calculate the corresponding growth deviation value. The optimal irrigation parameters found during the search process, i.e., the parameter combination corresponding to the minimum growth deviation, are recorded. Simultaneously, the same zigzag search process is performed on the second and third search axes. Parallel searching along the three axes ensures comprehensive coverage of the parameter space, avoiding local optima caused by a single search direction. All new parameter combinations obtained during the search process constitute the updated irrigation parameters. Using the updated irrigation parameters as input for the next round of optimization, the iterative process from S523 to S529 is repeated. In each iteration, the search axis and search strategy are redefined based on the current optimal parameter distribution, gradually approaching the globally optimal irrigation parameter combination. The iterative process terminates when the improvement in growth deviation is less than a preset convergence threshold after multiple iterations, or when the search reaches a preset maximum number of iterations. The optimal irrigation parameters obtained at this point are the distributed irrigation parameters for the first partition, ensuring that the growth prediction results for the first partition are closest to the desired growth target.

[0111] By using a parameter optimization method based on multi-axis zigzag search, the optimal combination of irrigation parameters can be efficiently found in a high-dimensional parameter space, providing reliable parameter optimization support for achieving precise distributed irrigation control.

[0112] Example 2, as Figure 2 As shown, based on the same inventive concept as the kiwifruit smart irrigation method using distributed optimization control provided in Embodiment 1, this embodiment of the invention also provides a kiwifruit smart irrigation system using distributed optimization control, comprising:

[0113] Data acquisition module 11 is used to obtain weather forecast data for the future time zone of the target area, initial irrigation parameters, and expected growth of kiwifruit.

[0114] The dual partitioning module 12 is used to partition the target area into environmental partitions to obtain the first partition result, and to partition the target area into kiwi fruit growth partitions to obtain the second partition result.

[0115] The partition intersection module 13 is used to intersect the first partition result and the second partition result to obtain the target area partition result;

[0116] The growth prediction module 14 is used to traverse the target area partitioning results, extract the kiwi fruit monitoring growth of the partition, combine the weather prediction data and the initial irrigation parameters to make growth predictions and obtain kiwi fruit growth prediction results, wherein the kiwi fruit growth prediction results correspond one-to-one with the target area partitioning results.

[0117] The optimization control module 15 is used to extract the partitions where the predicted growth of kiwifruit is inconsistent with the expected growth of kiwifruit, perform irrigation parameter optimization, obtain distributed irrigation parameters, and perform distributed irrigation control of kiwifruit.

[0118] Furthermore, the execution steps of the dual partitioning module 12 include:

[0119] The distribution information of the first environmental attribute monitoring values ​​of the target area is collected, and cluster analysis is performed to obtain the first environmental attribute partitioning results;

[0120] Until the distribution information of the Nth environmental attribute monitoring value of the target area is collected, cluster analysis is performed to obtain the Nth environmental attribute partitioning result;

[0121] The first environment attribute partitioning result is intersected up to the Nth environment attribute partitioning result to obtain the first partitioning result.

[0122] Furthermore, the execution steps of the dual partitioning module 12 also include:

[0123] From the distribution information of the first environmental attribute monitoring values, extract the environmental attribute monitoring values ​​at the first location and the environmental attribute monitoring values ​​at the second location, wherein the first location and the second location are adjacent locations;

[0124] Calculate the environmental attribute monitoring value deviation between the environmental attribute monitoring value at the first location and the environmental attribute monitoring value at the second location;

[0125] When the deviation of the environmental attribute monitoring value is less than the first environmental attribute deviation threshold, the first position and the second position are merged into the third position. At the same time, the average value of the environmental attribute monitoring value of the first position and the environmental attribute monitoring value of the second position is calculated and set as the environmental attribute monitoring value of the third position. The first environmental attribute deviation threshold is a user preset value.

[0126] When the deviation of the environmental attribute monitoring value is greater than or equal to the first environmental attribute deviation threshold, the first position and the second position are considered to be different positions.

[0127] When the deviation of the environmental attribute monitoring values ​​of any two adjacent locations is greater than or equal to the first environmental attribute deviation threshold, the first environmental attribute partitioning result is output.

[0128] Furthermore, the execution steps of the growth forecasting module 14 include:

[0129] Obtain the kiwifruit planting duration, add the future time zones to obtain the kiwifruit planting duration sequence;

[0130] Extract the first kiwifruit planting duration up to the Qth kiwifruit planting duration from the kiwifruit planting duration sequence.

[0131] From the target area partitioning results, a first partition is extracted, wherein the first partition has first partition environmental parameters and first partition growth parameters;

[0132] Traverse the first kiwifruit planting duration up to the Qth kiwifruit planting duration, and combine them with the first partition environmental parameters, the first partition growth parameters and the initial irrigation parameters respectively to construct growth constraints. Perform representative sample statistics to obtain representative sample growth parameters, and set them as the first partition kiwifruit predicted growth sequence.

[0133] The predicted growth sequence of kiwifruit in the first partition is added to the predicted growth result of kiwifruit.

[0134] Furthermore, the execution steps of the growth forecasting module 14 also include:

[0135] Collect time-series data on the growth monitoring of the first kiwifruit variety without varietal differences;

[0136] Based on the segmentation time step configured on the user end, the first kiwifruit growth monitoring time series information set is traversed and randomly extracted to obtain multiple kiwifruit growth monitoring time series information.

[0137] The growth monitoring time series information of multiple kiwifruit trees is traversed, the growth monitoring information of kiwifruit trees at the end time is extracted and set as the supervisory data, the growth monitoring information of kiwifruit trees at the start time is extracted, and the planting environment record data, the planting duration and irrigation parameter record data at the end time are combined as the input data to train the growth prediction model of the initial representative sample.

[0138] The second set of kiwifruit growth monitoring time series information of kiwifruit varieties in the target area is collected, and the growth prediction model of the initial representative sample is trained to obtain the growth prediction model of the representative sample in the target area, which is embedded in the kiwifruit distributed irrigation controller.

[0139] Furthermore, the execution steps of the optimized control module 15 include:

[0140] Extract the predicted growth of kiwifruit in the first region from the predicted growth results of kiwifruit;

[0141] By comparing the predicted growth of kiwifruit in the first partition with the expected growth of kiwifruit, the kiwifruit growth deviation is obtained, wherein the kiwifruit growth deviation is equal to the Euclidean distance of the normalized values ​​of the kiwifruit growth parameters.

[0142] If the deviation in the growth of the kiwifruit is greater than or equal to the growth parameter deviation threshold, it is considered that the predicted growth of the kiwifruit in the first zone is inconsistent with the expected growth of the kiwifruit; otherwise, it is considered that they are consistent.

[0143] Furthermore, the execution steps of the optimization control module 15 also include:

[0144] Extract the first partition where the kiwifruit growth prediction result is inconsistent with the expected kiwifruit growth;

[0145] The irrigation parameters are initialized to obtain several irrigation parameters;

[0146] By iterating through the irrigation parameters, the growth of the first zone is predicted, and the predicted growth of the first zone is obtained.

[0147] The predicted growth of the kiwifruit in the first partition is traversed and compared with the expected growth of the kiwifruit to obtain the growth deviation of the kiwifruit in the first partition.

[0148] When the growth deviation of the kiwifruit in the first zone is greater than or equal to the growth parameter deviation threshold, based on the growth deviation of the kiwifruit in the first zone, the minimum growth deviation irrigation parameter, the median growth deviation irrigation parameter, and the maximum growth deviation irrigation parameter are extracted from the irrigation parameters. The median growth deviation irrigation parameter is one or two sets, and the minimum growth deviation irrigation parameter and the maximum growth deviation irrigation parameter are both one set.

[0149] Using the minimum deviation irrigation parameter of the growth as the target point and the median deviation irrigation parameter of the growth as the starting point, a first search axis is constructed;

[0150] Using the irrigation parameter with the minimum deviation in growth as the target point and the irrigation parameter with the maximum deviation in growth as the starting point, a second search axis is constructed.

[0151] Using the median deviation irrigation parameter of the growth as the target point and the maximum deviation irrigation parameter of the growth as the starting point, a third search axis is constructed.

[0152] Perform a zigzag search along the first search axis, the second search axis, and the third search axis respectively to obtain updated irrigation parameters, and then repeat the process.

[0153] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0154] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0155] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0156] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0157] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0158] Although preferred embodiments of the invention have been described, those skilled in the art, once they have learned the basic inventive concept, can make other changes and modifications to these embodiments.

[0159] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of this invention and its equivalents, this invention also intends to include these modifications and variations.

Claims

1. A kiwi fruit intelligent irrigation method using distributed optimization control, characterized by, The application is applied to a kiwi distributed irrigation controller, and comprises the following steps: obtaining weather prediction data of a future time zone of a target area, initial irrigation parameters and kiwi expected growth; environmental zoning of the target area is performed to obtain a first zoning result, and kiwi growth zoning of the target area is performed to obtain a second zoning result; the first zoning result and the second zoning result are intersected to obtain a target area zoning result; the target area zoning result is traversed to extract zoning kiwi monitoring growth, and the weather prediction data and the initial irrigation parameters are combined to perform growth prediction to obtain a kiwi growth prediction result, wherein the kiwi growth prediction result corresponds to the target area zoning result one by one; zones where the kiwi growth prediction result is inconsistent with the kiwi expected growth are extracted, irrigation parameter optimization is performed to obtain distributed irrigation parameters, and kiwi distributed irrigation control is performed. Wherein, extracting the zoning where the kiwi growth prediction result is inconsistent with the kiwi expected growth, performing irrigation parameter optimization to obtain distributed irrigation parameters, and performing kiwi distributed irrigation control, comprises: extracting the first zoning where the kiwi growth prediction result is inconsistent with the kiwi expected growth; initializing irrigation parameters to obtain a plurality of irrigation parameters; traversing the plurality of irrigation parameters, the first zoning is subjected to growth prediction to obtain a plurality of first zoning predicted growths; traversing the plurality of first zoning predicted growths, and comparing with the kiwi expected growth to obtain a plurality of first zoning kiwi growth deviations; when the plurality of first zoning kiwi growth deviations are all greater than or equal to a growth parameter deviation threshold, based on the plurality of first zoning kiwi growth deviations, the minimum deviation growth irrigation parameter, the median deviation growth irrigation parameter and the maximum deviation growth irrigation parameter are extracted from the plurality of irrigation parameters, the median deviation growth irrigation parameter is one or two groups, and the minimum deviation growth irrigation parameter and the maximum deviation growth irrigation parameter are both one group; taking the minimum deviation growth irrigation parameter as a target point, and taking the median deviation growth irrigation parameter as a starting point to construct a first search axis; taking the minimum deviation growth irrigation parameter as a target point, and taking the maximum deviation growth irrigation parameter as a starting point to construct a second search axis; taking the median deviation growth irrigation parameter as a target point, and taking the maximum deviation growth irrigation parameter as a starting point to construct a third search axis; Z-shaped search is performed along the first search axis, the second search axis and the third search axis respectively to obtain updated irrigation parameters, and a loop is performed.

2. The method of claim 1, wherein, The target area is subjected to environmental zoning to obtain a first zoning result, which comprises the following steps: first environmental attribute monitoring value distribution information of the target area is collected for cluster analysis to obtain a first environmental attribute zoning result; until the Nth environmental attribute monitoring value distribution information of the target area is collected for cluster analysis to obtain an Nth environmental attribute zoning result; the first environmental attribute zoning result to the Nth environmental attribute zoning result are intersected to obtain the first zoning result.

3. The method of claim 2, wherein, Collecting a first environmental attribute monitoring value distribution information of the target area, performing cluster analysis to obtain a first environmental attribute partition result, comprising: From the first environmental attribute monitoring value distribution information, extracting a first location environmental attribute monitoring value and a second location environmental attribute monitoring value, wherein the first location and the second location are adjacent locations; Calculating the environmental attribute monitoring value deviation of the first location environmental attribute monitoring value and the second location environmental attribute monitoring value; When the environmental attribute monitoring value deviation is less than a first environmental attribute deviation threshold, merging the first location and the second location into a third location, and calculating the mean of the first location environmental attribute monitoring value and the second location environmental attribute monitoring value as the third location environmental attribute monitoring value, the first environmental attribute deviation threshold is a user preset value; When the environmental attribute monitoring value deviation is greater than or equal to the first environmental attribute deviation threshold, the first location and the second location are considered as different locations; When the environmental attribute monitoring value deviation of any two adjacent locations is greater than or equal to the first environmental attribute deviation threshold, outputting the first environmental attribute partition result.

4. The method of claim 1, wherein, Traversing the target area partition result, extracting a partition kiwi monitoring growth, combining the weather prediction data and the initial irrigation parameter to perform growth prediction, and obtaining a kiwi growth prediction result, comprising: Obtaining a kiwi planting time length, adding the future time zone to obtain a kiwi planting time length sequence; From the kiwi planting time length sequence, extracting a first kiwi planting time length to a Qth kiwi planting time length; From the target area partition result, extracting a first partition, wherein the first partition has a first partition environmental parameter and a first partition growth parameter; Traversing the first kiwi planting time length to the Qth kiwi planting time length, combining the first partition environmental parameter, the first partition growth parameter and the initial irrigation parameter respectively to construct a growth constraint, performing representative sample statistics to obtain a representative sample growth parameter, and setting it as a first partition kiwi predicted growth sequence; Adding the first partition kiwi predicted growth sequence to the kiwi growth prediction result.

5. The method of claim 4, wherein, Traversing the first kiwi planting time length to the Qth kiwi planting time length, combining the first partition environmental parameter, the first partition growth parameter and the initial irrigation parameter respectively to construct a growth constraint, performing representative sample statistics to obtain a representative sample growth parameter, and setting it as a first partition kiwi predicted growth sequence, comprising: Collecting a first kiwi growth monitoring time sequence information set without variety difference; Based on the segmentation time step configured by the user end, traversing the first kiwi growth monitoring time sequence information set to obtain multiple kiwi growth monitoring time sequence information by random interception; Traversing the multiple kiwi growth monitoring time sequence information, extracting end time kiwi growth monitoring information as supervision data, and extracting start time kiwi growth monitoring information as input data combined with planting environment record data, end time planting time length and irrigation parameter record data to train an initial representative sample growth prediction model; Collecting a second kiwi growth monitoring time series information set of kiwi varieties in the target area, training the initial representative sample growth prediction model to obtain a target area representative sample growth prediction model, and embedding the kiwi distributed irrigation controller.

6. The method of claim 1, wherein, Extracting a subarea where the kiwi growth prediction result is inconsistent with the expected kiwi growth, performing irrigation parameter optimization to obtain distributed irrigation parameters, and performing kiwi distributed irrigation control, including: Extracting a first subarea kiwi predicted growth from the kiwi growth prediction result; Comparing the first subarea kiwi predicted growth with the expected kiwi growth to obtain a kiwi growth deviation, wherein the kiwi growth deviation is equal to the Euclidean distance of the normalized value of the kiwi growth parameter; When the kiwi growth deviation is greater than or equal to the growth parameter deviation threshold, it is considered that the first subarea kiwi predicted growth is inconsistent with the expected kiwi growth, otherwise it is considered to be consistent.

7. A kiwifruit smart irrigation system employing distributed optimal control, characterized in that, For implementing the method as claimed in any one of claims 1 to 6, applied to a kiwi distributed irrigation controller, comprising: A data acquisition module for obtaining weather prediction data of a future time zone in a target area, initial irrigation parameters and expected kiwi growth; A dual partition module for performing environmental partitioning on the target area to obtain a first partitioning result, and performing kiwi growth partitioning on the target area to obtain a second partitioning result; A partition intersection module for intersecting the first partitioning result and the second partitioning result to obtain a target area partitioning result; A growth prediction module for traversing the target area partitioning result, extracting subarea kiwi monitoring growth, combining the weather prediction data and the initial irrigation parameters, and performing growth prediction to obtain a kiwi growth prediction result, wherein the kiwi growth prediction result corresponds one-to-one to the target area partitioning result; An optimization control module for extracting a subarea where the kiwi growth prediction result is inconsistent with the expected kiwi growth, performing irrigation parameter optimization to obtain distributed irrigation parameters, and performing kiwi distributed irrigation control.

Images