Intelligent division method of sub-ecological zones of stored grain based on spatial analysis and clustering technology

Abstract

The application discloses a kind of based on spatial analysis and clustering technology's intelligent division method of sub-ecological zone of stored grain, comprising the following steps: step one, obtain the geographic position data of each sampling point in research area, meteorological data and grain condition data, as the sample data of each sampling point;The meteorological data and grain condition data of each sampling point are as the sample stored grain variable;Wherein, geographic position information includes the longitude, latitude and elevation of sampling point, meteorological data includes the air temperature, relative humidity, precipitation and sunshine duration of sampling point, and grain condition data includes the grain temperature of the corresponding grain depot of sampling point;Step two, calculate the global moran index of each stored grain variable in each sampling point, and according to global moran index, feature indicators in meteorological data and grain condition data are selected respectively;Step three, determine the number of clustering targets K , using K-means clustering algorithm to cluster the sampling points in research area, and divide the research area into K Sub-ecological region of stored grain.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of grain storage, and particularly relates to an intelligent division method of a sub-ecological zone of grain storage based on spatial analysis and clustering technology. BACKGROUND

[0002] The stability and efficient management of the ecological environment of grain storage are important links for guaranteeing the quality of grain storage and reducing post-production losses. Under the background of intensified climate change, diversified grain varieties and iterative storage technology, the traditional extensive division taking administrative regions or climate zones as units has failed, and its limitations are reflected in three aspects: first, the quantitative characterization of the microenvironment factors of grain storage is lacked, and the single-factor threshold division cannot capture the interactive effects of multiple factors; second, the empirical boundary is highly subjective and has poor repeatability, and lacks the dynamic modeling capability based on GIS spatial analysis, resulting in deviations between the division results and the actual grain storage conditions; and third, the technical scale still stays in the macro framework of "large zone and small zone", and neither the concept of "sub-ecological zone" nor the division standard is proposed, and the fine management technology of the micro-ecological unit is also lacked. 2

[0003] Grain safety storage is essentially the result of the synergistic effect of geographical environment, climate factors, grain factors and storage management. The existing research exposes three defects in this chain: in the index system, it presents single-dimension, and follows the "geographical location dominant" or "single meteorological element driven" mode, ignoring the shaping of the micro-ecology in the warehouse by the local microclimate circulation; in time, it uses static indicators such as annual mean temperature and annual precipitation, and lacks real-time coupling mechanism of meteorological factors and grain factors. In data correlation, there is a fault, and the causal chain of "meteorological driving-grain response" cannot be established. In data chain, there is a fault, the time and space resolution of the collection end is low, the data span is short, the processing end lacks outlier processing and feature selection, and the application end lacks dynamic updating mechanism of multi-source data fusion, which is difficult to cope with the frequent extreme climate and rapid iteration of quality.

[0004] Therefore, it is urgent to build a "meteorological driving-grain response-dynamic optimization" three-in-one grain storage ecological zoning system. The system can realize the precise configuration of grain storage environment regulation and control measures, improve the safety of grain storage and the efficiency of resource utilization, provide customized environment regulation and control strategies for different micro-ecological units, and finally form a new fine grain storage management mode that adapts to climate change and variety diversification. SUMMARY

[0005] The purpose of the present application is to overcome the defects of the prior art, and provide an intelligent division method of a sub-ecological zone of grain storage based on spatial analysis and clustering technology, which can realize the precise division of the sub-ecological zone of grain storage by deeply mining the internal correlation among meteorological data, grain data and geographical space.

[0006] The technical scheme provided by the present application is:​

[0007] An intelligent method for dividing sub-ecological zones of stored grain based on spatial analysis and clustering technology, comprising the following steps:

[0008] Step one, obtaining the geographic position data, meteorological data and grain condition data of each sampling point in the study area as the sample data of each sampling point; taking the meteorological data and grain condition data of each sampling point as the sample storage variables;

[0009] Wherein, the geographic position data includes the longitude, latitude and altitude of the sampling point, the meteorological data includes the air temperature, relative humidity, precipitation and sunshine duration of the sampling point, and the grain condition data includes the grain temperature of the corresponding grain warehouse of the sampling point;

[0010] Step two, calculating the global Moran's index of each storage variable in each sampling point, and screening out the characteristic indicators in the meteorological data and grain condition data according to the global Moran's index respectively;

[0011] Step three, determining the clustering target number based on the characteristic indicators K , using K-means clustering algorithm to cluster the sampling points in the study area, and dividing the study area into K sub-ecological zones of stored grain.

[0012] Preferably, before the step two, it also includes standardizing the storage variables of the sample to remove the dimension.

[0013] Preferably, in the step two, for a single storage variable, the calculation formula of the global Moran's index is:

[0014] ;

[0015] Wherein, is the number of samples in the study area, is the spatial weight matrix, and are the standardized values of the storage variables of the th sample and the th sample respectively, is the mean value of the standardized values of the storage variables of all samples in the study area.

[0016] Preferably, the calculation formula of the spatial weight matrix is:

[0017] ;

[0018] Wherein, is the horizontal distance between the th sample and the th sample, is the spatial weight matrix. and samples The absolute value of the altitude difference between them. This is the attenuation coefficient due to altitude difference.

[0019] Preferably, in step two, meteorological data that meets the following criteria are selected. The grain storage variables and grain condition data satisfy Grain storage variables are used as characteristic indicators.

[0020] Preferably, in step three, the clustering target is determined using the elbow rule. ,include:

[0021] calculate Rate of change of the sum of squared errors when the value changes ;

[0022] ;

[0023] in, For the number of clusters is Sum of squared errors at time, For the number of clusters is Sum of squared errors at time, For the number of clusters is Sum of squared errors over time;

[0024] when At that time, determine the elbow point, and place the elbow point corresponding to the elbow point. Value as the clustering target;

[0025] in, The set threshold for the rate of change.

[0026] Preferably, in step three, the distance formula used for K-means clustering is:

[0027] ;

[0028] in, Indicates sample With Cluster Center The weighted distance, For the feature index weights, For geographical feature weights, For the sample The The standardized values ​​of each characteristic indicator Cluster center The Standardized values ​​of each characteristic indicator; For the sample With cluster center geographical distance; is a geographical attenuation function; is a latitude of the sample is a longitude of the sample is a latitude of the cluster center is a longitude of the cluster center is a latitude of the cluster center is a longitude of the cluster center is a longitude of the cluster center is a longitude of the cluster center

[0029] Preferably, .

[0030] Preferably, in the step three, further comprising boundary correction of the obtained sub-ecological region of stored grain according to topography, and the boundary correction satisfies the topography consistency principle.

[0031] The present application has the following advantages:

[0032] The method for intelligently dividing sub-ecological regions of stored grain based on spatial analysis and clustering technology provided by the present application can realize accurate division of sub-ecological regions of stored grain by deeply mining the internal correlation between meteorological data, grain condition data and geographical space, and overcomes the defects of single dimension of division index, decoupling of meteorological data and grain condition data correlation and three faults of data chain in the links of "collection-processing-application" in the prior art, resulting in insufficient division accuracy and poor adaptability. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 is a flowchart of the division of sub-ecological regions of stored grain according to the present application.

[0034] Figure 2 is a flowchart of the spatial correlation of meteorological data and grain condition data according to the present application.

[0035] Figure 3 is a flowchart of the determination of the number of sub-ecological regions of stored grain according to the present application.

[0036] Figure 4 is a schematic diagram of the interpolation result of the annual average grain temperature in the embodiment of the present application.

[0037] Figure 5 is a LISA cluster analysis diagram of an index (average value of winter balanced moisture) with high I value in the embodiment of the present application.

[0038] Figure 6 is a LISA cluster analysis diagram of an index (average humidity in summer) with low I value in the embodiment of the present application.

[0039] Figure 7 is a spatial cluster distribution diagram of sampling points in the embodiment of the present application. DETAILED DESCRIPTION

[0040] The present invention will now be described in further detail with reference to the accompanying drawings, so that those skilled in the art can implement it based on the description.

[0041] Grain storage sub-ecological zones are micro-units of grain storage that share similar meteorological driving and grain condition response characteristics in geographical space. For example... Figure 1 As shown, this invention provides a method for intelligently dividing sub-ecological zones of grain storage based on spatial analysis and clustering technology. The specific implementation process is as follows.

[0042] I. Data Collection and Preprocessing

[0043] Multiple sampling points were set up within the study area to obtain geographical location data and grain storage variables for each sampling point within the study area as samples. The obtained sample data were then preprocessed.

[0044] In one embodiment, for the study area, each meteorological station within the study area is used as a sampling point. First, the geographical location data of each sampling point within the study area is determined and acquired. Then, the grain storage variables for each sampling point are acquired, including meteorological observation data and grain condition data. Figure 2 As shown, in some areas, the number of grain warehouses is less than the number of weather stations. To ensure the integrity of grain condition data at each sampling point, it is necessary to spatially correlate meteorological and grain condition data. For each weather station, the area within 50 kilometers of the station is used as a buffer zone. If a grain warehouse exists within the buffer zone, the grain condition data of the closest grain warehouse to the weather station (sampling point) within the buffer zone is used as the grain condition data for that sampling point. If no grain warehouse exists within the buffer zone, the data is completed through interpolation. Geographic Information System (GIS) data includes longitude, latitude, and altitude; meteorological observation data includes temperature, relative humidity, precipitation, and sunshine duration; grain condition data includes grain temperature and equilibrium moisture content. By removing outliers, interpolating to complete missing values, and standardizing the data, data noise and dimensional differences are eliminated, providing a high-quality dataset for subsequent analysis.

[0045] For sampling points where there are no grain warehouses around the weather station, the grain temperature of the sampling points without grain warehouses is supplemented by spatial interpolation.

[0046] ;

[0047] in, Sampling points The grain temperature interpolation results The number of sample points (within a radius of 50 kilometers from the threshold) used in the interpolation calculation. Sample points involved in the calculation of the difference Grain temperature, This represents the spatial distance weight.

[0048] In one embodiment, the formula for calculating the spatial distance weight is:

[0049] ;

[0050] in, For the sample and samples The horizontal distance between them For the sample and samples The absolute value of the altitude difference between them. This is the attenuation coefficient due to altitude difference.

[0051] In one embodiment, the altitude difference attenuation coefficient is set to 0.005.

[0052] II. Screening Feature Indicators

[0053] The spatial clustering characteristics of grain storage variables are quantified using the global Moran's I index. The global Moran's I index is calculated for each grain storage variable individually, using the following formula:

[0054] ;

[0055] in, n The number of samples within the study area. This is the spatial weight matrix. and The first The first sample and the first Standardized values ​​of grain storage variables for each sample. This is the mean of the standardized values ​​of grain storage variables for all samples within the study area. For example, to calculate the global Moran's index for average winter temperature, then... and The first The first sample and the first Standardized winter average temperature values ​​for each sample. This is the mean of the standardized winter average temperature values ​​for all samples within the study area.

[0056] Calculate each grain storage variable (meteorological observation data and grain condition data) separately. value. ; I >0 indicates a positive correlation. I <0 indicates a negative correlation. I =0 indicates a random distribution.

[0057] By calculating the global Moran's index, the spatial clustering of grain storage variables such as air temperature, relative humidity, and grain temperature in the study area is quantified. IThe value of the feature index is high, which can directly reflect the regional differences of the grain storage environment due to its spatial aggregation characteristics.

[0058] III. Construction of the zoning criteria

[0059] The optimal clustering K value is determined by the elbow method. As shown in FIG. 1, the elbow point is determined. Figure 3 The specific process of determining the elbow point is as follows: based on the input feature matrix F, the initial clustering number K is initialized, and the search interval of K is [1, 10]. Then, the sum of squared errors (SSE) corresponding to each candidate K value is iteratively calculated. By calculating the curvature of the SSE corresponding to adjacent K values, the inflection point is identified by the change amount Δ(K). If Δ(K) is less than a set empirical threshold, it is determined that the current K value is the optimal solution; otherwise, the K value is increased and the above calculation process is repeated until the convergence condition is met. The change rate of the sum of squared errors SSE within the cluster is calculated, and the formula is as follows:

[0060] The change rate of the sum of squared errors SSE within the cluster is calculated, and the formula is as follows:

[0061]

[0062] K-

[0063]

[0064] In the formula, represents the i-th cluster, represents the standardized value of the feature index of the sampling point in the cluster, and represents the standardized value of the feature index of the center in the cluster.

[0065] When K is less than or equal to K, the elbow point is determined. In one embodiment, K is set to 2.

[0066] IV. Regional division and optimization​​​​​​​​​​​​​​​​​​​​​​​​​​​​

[0067] Based on the selected feature indicators and the established zoning criteria, a geographical distance weight matrix is constructed. First, a feature matrix is constructed based on the feature indicators obtained by the spatial autocorrelation method .

[0068] ;

[0069] wherein, n is the number of samples, m is the number of selected feature indicators, represents the n th feature indicator of the m th sample (here, the feature indicator is the standardized value after dimensionless); is the longitude of the n th sample, is the latitude of the n th sample.

[0070] Each sample (corresponding to a sampling point) is assigned to the nearest cluster center by the K-means algorithm, and the clustering result is the sub-ecological zoning result, i.e., each class corresponds to a sub-ecological zone. By clustering, the sampling points are divided into different sub-ecological zones.

[0071] In the K-means process, the distance of a sample to a cluster center is calculated using an improved distance formula, and the calculation formula is:

[0072] ;

[0073] wherein, is the weighted distance between the sample and the cluster center ; is the feature indicator weight, is the geographical feature weight, is the th feature indicator of the sample , is the th feature indicator of the cluster center ; is the geographical distance between the sample and the cluster center ; is the geographical attenuation function; is the latitude of the sample ; is the latitude of the sample ; is the latitude of the cluster center ; the longitude of the cluster center The characteristic index in the formula is a standardized value after de-dimensioning.

[0074] As a preferred, the optimal weight combination is determined by the profile coefficient method. α, β

[0075] To ensure the accuracy and reasonableness of the division results, the application also includes a geographic information system (GIS) based spatial analysis technique to overlay geographic feature data such as mountains and deserts to correct the division boundaries of the sub-ecological regions, and to analyze the influence of topography on meteorological data and grain condition data. The boundary correction needs to meet the topographic consistency principle.

[0076] V. Establishing a feature database

[0077] The geographic data, meteorological data, grain condition data, and division results of each grain storage sub-ecological region are integrated and stored in the database. The database uses a hybrid architecture, with a relational database storing static geographic data and a time series database storing dynamic meteorological and grain condition data.

[0078] VI. Correcting and publishing the division results

[0079] The division model is periodically iteratively optimized based on the latest data, and a scientific sub-ecological grain storage division standard and management mode are developed, and are published through a geographic information system platform on a regular basis. EMBODIMENT

[0080] This embodiment takes the Xinjiang Uygur Autonomous Region (east longitude 73°40′~96°18′, north latitude 34°25′~49°10′) as the research area, and performs sub-ecological division on the research area, which specifically includes the following processes.

[0081] I. Data collection and preprocessing

[0082] Data sources and collection specifications:

[0083] Geographic data: 30m resolution DEM data from the National Basic Geographic Information Center. Meteorological data: daily data from 105 national meteorological stations in Xinjiang (2021-2024), fields include: daily average temperature (℃), daily maximum / minimum temperature (℃), daily average relative humidity (%), daily precipitation (mm), and daily sunshine hours (h). Grain condition data: Internet of Things monitoring data from 30 grain stores (2022-2023), including: upper, middle, and lower layer temperatures (℃), and average grain temperature calculated from upper, middle, and lower layer grain temperatures.

[0084] Data cleaning and standardization process: outlier rejection (dynamic 3σ principle): [μ-3σ, μ+3σ], missing value interpolation.

[0085] ​Standardization process: Z-score standardizes all metrics to eliminate the influence of dimensions.

[0086] Missing data completion: Since the number of grain warehouses in Xinjiang is less than the number of weather stations, spatial correlation between meteorological and grain condition data is necessary to ensure the completeness of grain condition data at each sampling point. For each weather station, if a grain warehouse exists within 50 kilometers of the station, the grain condition data from the closest warehouse is used as the grain condition data for that sampling point. If no grain warehouse exists, grain temperature is supplemented using spatial interpolation. Based on the spatial coordinates of the weather stations, sparse grain condition data is mapped to the location of each weather station using interpolation.

[0087] ;

[0088] in, Sampling points The grain temperature interpolation results The number of sample points (within a radius of 50 kilometers from the threshold) used in the interpolation calculation. Sample points involved in the calculation of the difference Grain temperature, This represents the spatial distance weight.

[0089] In one embodiment, the formula for calculating the spatial distance weight is:

[0090] ;

[0091] in, For the first The first sample and the first The horizontal distance between samples For the first The first sample and the first The absolute value of the elevation difference for each sample. This is the altitude difference attenuation coefficient. The altitude difference attenuation coefficient is set to 0.005 in the formula.

[0092] like Figure 4 The figure shows the interpolation results of the average grain temperature. From this figure, it can be seen that spatial interpolation aligns the missing grain condition data with the meteorological data. The aligned station data is then used as the sample point data for analysis.

[0093] II. Feature Index Screening

[0094] Within the target region, spatial autocorrelation analysis is performed. Spatial clustering is assessed using the global Moran's I index, with the following formula:

[0095] ;

[0096] in, Let n be the spatial weight matrix, and n be the number of samples.

[0097] Molecular part: middle: This represents the difference between the temperature of a specific sample and the average temperature of the entire Xinjiang region. This means that if two neighboring properties are both at high or low temperatures, their product is positive (changing in the same direction); if one is at high temperature and the other at low temperature, their product is negative (changing in opposite directions).

[0098] Denominator part: Calculate the total deviation of temperature from the average value for all stations. This is equivalent to a "standardization factor": scaling the molecular result according to the overall fluctuation to avoid inflated values ​​due to the large temperature difference in Xinjiang.

[0099] Coefficient adjustment: The total number of all neighbor pairings.

[0100] The correlation calculation results between meteorological data, grain condition data and geospatial data are shown in Table 1.

[0101] Table 1 Spatial Autocorrelation Variables

[0102]

[0103] Based on the Moran's I index values ​​of various meteorological and grain condition indicators provided, the selection of reasonable spatial characteristic variables should comprehensively consider the spatial autocorrelation strength, indicator representativeness, and application objectives.

[0104] Selection principles and criteria:

[0105] Spatial autocorrelation intensity (Moran's I value) High autocorrelation ( I >0.7): This indicates that the indicator exhibits a significant spatial clustering pattern, making it suitable as a core basis for dividing partition boundaries; moderate autocorrelation (0.4 ≤ I ≤ 0.7): Its stability needs to be verified in conjunction with other indicators or domain knowledge; low autocorrelation ( I <0.4): Spatial distribution is close to random and not suitable as a separate partitioning indicator.

[0106] Practical significance of the indicators: Temperature and humidity: directly affect grain storage safety (e.g., mold, pests). Moisture balance and accumulated temperature: reflect the risk of moisture balance and heat accumulation in grain. Grain condition indicators (e.g., grain temperature): directly related to storage management needs.

[0107] Avoid redundancy: Among similar indicators, prioritize the variable with the strongest spatial autocorrelation.

[0108] In this embodiment, meteorological observation data is selected. IIndicators with a value greater than 0.7 were selected as feature indicators from grain condition data. I Indicators with a value greater than 0.6 were used as feature indicators, and the selected feature indicators are shown in Table 2.

[0109] Based on Moran's I value and its practical significance, feature analysis was performed on the selected feature indicators:

[0110] Table 2 Feature Indicators

[0111]

[0112] The spatial variability results of the selected characteristic indicators were analyzed, taking the winter equilibrium water as an example (e.g.) Figure 5 As shown in the figure, the map clearly presents a spatial clustering phenomenon: the red areas (HH, i.e., high-value clustering areas) are concentrated in northern Xinjiang (the region with high winter water balance), while the light red / white areas (LL / LH, i.e., low-value clustering areas) are mainly distributed in southern Xinjiang (the region with low winter water balance). This spatial differentiation characteristic indicates that the index is not randomly distributed, but follows a regional pattern of "high in the north and low in the south, with a transition to mountainous areas," exhibiting significant spatial heterogeneity. In contrast, the distribution of average summer humidity (such as...) Figure 6 As shown in the image, the red and white clusters exhibit an alternating distribution pattern, without significant spatial differentiation. Based on the image analysis results, it can be seen that variables with larger Moran's I index values ​​show significant spatial clustering, indicating that the Moran's I index is suitable for grain storage area delineation.

[0113] III. Establishing Zoning Criteria

[0114] Using the Elbow Method, calculate step by step. K =2~10 SSE rate of change, until Stop when Δ:

[0115] ;

[0116] when K When =4, (satisfy ).like Figure 7 As shown, Xinjiang is divided into four sub-ecological zones for grain storage.

[0117] IV. Regional Division and Optimization

[0118] Based on feature indicators and zoning criteria selected through the global Moran index, the Xinjiang grain storage sub-ecological zone is divided using the geographically weighted K-means clustering algorithm.

[0119] A feature matrix is ​​constructed by integrating key meteorological factors and geographic coordinates:

[0120] ;

[0121] wherein, n is the number of samples, m is the number of selected feature indicators, represents the n th feature indicator of the m th sample, where the feature indicator is the standardized value after de-dimensioning; is the longitude of the n th sample, is the latitude of the n th sample.

[0122] Randomly select 4 data points from the data set as initial cluster centers.

[0123] Assign each sample (corresponding to a sampling point) to the nearest cluster center by K-means algorithm, and the clustering result is the sub-ecoregion division result, i.e. each class corresponds to a sub-ecoregion. By clustering, the sampling points are divided into different sub-ecoregions.

[0124] In the K-means process, the distance between the sample and the cluster center is calculated using an improved distance formula, and the calculation formula is:

[0125] ;

[0126] wherein, is the weighted distance between the sample and the cluster center ; is the feature indicator weight, is the geographic feature weight, is the th feature indicator of the sample , is the th feature indicator of the cluster center ; is the geographic distance between the sample and the cluster center ; is the geographic attenuation function; is the latitude of the sample ; is the latitude of the sample ; is the latitude of the cluster center ; is the longitude of the cluster center . The feature indicator in the formula is the standardized value after de-dimensioning.

[0127] The optimal weight combination is determined by the silhouette coefficient method α, β , which satisfies:

[0128] ;

[0129] Silhouette coefficient (S): used to measure the accuracy of clustering results:

[0130] ;

[0131] wherein, represents the average feature distance between the sample corresponding sampling point and other sampling points in the same sub-ecological region; represents the average feature distance between the sample corresponding sampling point and the nearest sub-ecological region site. The combination of each ( α, β ) is calculated, and the corresponding silhouette coefficient is output as shown in Table 3.

[0132] Table 3 Parameter Response Surface Table

[0133]

[0134] The weight combination corresponding to the largest silhouette coefficient is selected, i.e. the weight α = 0.7 (feature factor), β = 0.3 (geographical feature).

[0135] The weight index is included in the determination process of the geographical weighted K-means clustering algorithm coefficient, and the winter average temperature, extreme high temperature, winter average humidity, winter equilibrium moisture, average grain temperature, and summer sunshine duration are selected as feature indexes. The corresponding meteorological stations and grain collection points are classified. The classification results are shown in Figure 7 , and then the classification results are converted to 5 km x 5 km grid data resolution, and the Kriging interpolation method (Kriging) is used to construct a continuous partition surface to realize the division of the sub-ecological region of Xinjiang region.

[0136] Subsequently, based on the spatial analysis technology of geographic information system (GIS), the boundaries of the sub-ecological region are corrected by superimposing geographical feature data such as mountains, deserts, and altitude. The boundary correction process strictly follows the principle of topographic consistency.

[0137] Five, establish a feature database

[0138] The geographical data, meteorological data, grain condition data, and division results of each sub-ecological region for grain storage are integrated and stored in the database. The database uses a hybrid architecture, with a relational database storing static geographical data and a time series database storing dynamic meteorological and grain condition data.

[0139] Six, periodic optimization and release of regionalization results

[0140] Periodically optimize the sub-ecological zoning based on the latest data, make scientific sub-ecological zoning standards and management mode, and release through the geographic information system platform.

[0141] Although embodiments of the application have been disclosed in connection with the above specification and drawings it will be understood that they are not limited to the specific details of the foregoing description, since various modifications can be made thereto without departing from the scope of the claims and equivalents thereof.

Claims

1. A method for intelligent division of sub-ecological zones of stored grain based on spatial analysis and clustering techniques, characterized in that, The method comprises the following steps: Step one, obtaining geographical position data, meteorological data and grain condition data of each sampling point in the research area as sample data of each sampling point; taking the meteorological data and the grain condition data of each sampling point as sample storage grain variables; The geographical position data comprises longitude, latitude and altitude of the sampling point, the meteorological data comprises air temperature, relative humidity, precipitation and sunshine duration of the sampling point, and the grain condition data comprises grain temperature of the warehouse corresponding to the sampling point; Step two, calculating the global Moran index of each storage grain variable in each sampling point, and screening feature indexes in the meteorological data and the grain condition data according to the global Moran index respectively; Selecting storage grain variables in the meteorological data satisfying I>0.7 and storage grain variables in the grain condition data satisfying I>0.6 as the feature indexes; Step three, determining a clustering target number K based on the feature indexes, adopting a K-means clustering algorithm to cluster the sampling points in the research area, and dividing the research area into K storage sub-ecological regions; When the K-means clustering is performed, the distance formula adopted is: where d(x j , μ i ) represents the weighted distance between sample j and cluster center μ i , α is the feature indicator weight, β is the geographic feature weight, x jk is the standardized value of the kth feature indicator of sample j, μ ik is the standardized value of the kth feature indicator of cluster center μ i ; d'(x j , μ i ) is the geographic distance between sample j and cluster center μ i ; is the geographic attenuation function; x j,lat is the latitude of sample j; x j,lon is the longitude of sample j; μ i,lat is the latitude of cluster center μ i ; μ i,lon is the longitude of cluster center μ i ; and m is the number of selected feature indicators.

2. The intelligent division method for grain storage sub-ecological zones based on spatial analysis and clustering technology according to claim 1, characterized in that, Before the step two, the method further comprises standardizing the sample storage grain variables to remove dimensions.

3. The intelligent division method for grain storage sub-ecological zones based on spatial analysis and clustering technology according to claim 2, characterized in that, In the step two, for a single storage grain variable, the calculation formula of the global Moran index is: where n is the number of samples in the study area, w ij is the spatial weight matrix, x i and x j are the normalized values of the grain storage variable of the i-th sample and the j-th sample, respectively, is the mean of the normalized values of the grain storage variable of all samples in the study area.

4. The method according to claim 3, wherein, The calculation formula of the spatial weight matrix is: where D ij is the horizontal distance between sample i and sample j, Ah ij is the absolute value of the elevation difference between sample i and sample j, and q is an elevation difference decay coefficient.

5. The method according to claim 1 or 4, wherein, In the step three, the elbow rule is adopted to determine the clustering target K, comprising: calculating the change rate Δ(K) of the error sum of squares when the K value changes; Wherein, SSE(K) is the error sum of squares when the clustering number is K, SSE(K+1) is the error sum of squares when the clustering number is K+1, and SSE(K-1) is the error sum of squares when the clustering number is K-1; When Δ(K)<θ, the elbow point is determined, and the K value corresponding to the elbow point is taken as the clustering target; Wherein, θ is a set change rate threshold.

6. The method of claim 5, wherein the method further comprises: α+β＝1。 7. The storage sub-ecological region intelligent division method based on spatial analysis and clustering technology according to claim 6, wherein in the step three, the method further comprises boundary correction of the storage sub-ecological regions obtained by clustering according to the topography, and the boundary correction satisfies the topography consistency principle.

Images