CN-122020599-A - Farmland water-salt evolution rule digging system based on time sequence track clustering

CN122020599ACN 122020599 ACN122020599 ACN 122020599ACN-122020599-A

Abstract

The invention relates to the field of agricultural data analysis, in particular to a farmland water-salt evolution law mining system based on time sequence track clustering, which comprises the steps of acquiring and preprocessing farmland water and salt original data to obtain a two-dimensional time sequence track set; the method comprises the steps of obtaining a first optimization factor through water-salt lag cumulative effect analysis on a two-dimensional time sequence track set, obtaining a second optimization factor through local water-salt track fluctuation degree analysis on the two-dimensional time sequence track set, obtaining a final time sequence similarity distance through joint correction based on the first optimization factor and the second optimization factor on an original single-step distance, obtaining a farmland water-salt evolution rule through clustering division and rule extraction on the final time sequence similarity distance, and therefore the problem that the farmland water-salt asynchronous lag difference cannot be represented by an existing dynamic time-warping algorithm, and land plots with different water-salt evolution mechanisms are clustered in an error mode is solved.

Inventors

SUN YUNYUN
LIU FANGMING
GAO YUSHAN
LIU HUITAO
DOU JINGANG
HOU ZHONGHUA
SUN HAIQUAN
JIANG YECHENG
Deng Aoyan

Assignees

吉林省农业科学院(中国农业科技东北创新中心)

Dates

Publication Date: 20260512
Application Date: 20260408

Claims (10)

1. Farmland water and salt evolution law mining system based on time sequence track clustering is characterized in that the method comprises the following steps: step S1, acquiring a two-dimensional time sequence track set by collecting and preprocessing farmland moisture and salinity original data; S2, obtaining a first optimization factor by analyzing a water-salt lag cumulative effect of the two-dimensional time sequence track set; S3, obtaining a second optimization factor by analyzing the fluctuation degree of the local water-salt track of the two-dimensional time sequence track set; s4, acquiring a final time sequence similarity distance by carrying out joint correction based on a first optimization factor and a second optimization factor on the original single-step distance; And S5, obtaining a farmland water salt evolution rule by carrying out cluster division and rule extraction on the final time sequence similarity distance.
2. The farmland water and salt evolution law mining system based on time sequence track clustering according to claim 1, wherein the acquiring a two-dimensional time sequence track set by collecting and preprocessing farmland water and salt raw data comprises the following steps: Arranging a soil sensing sensor network in a target farmland area according to a preset space grid, arranging soil moisture sensor nodes for collecting soil moisture original data and soil conductivity sensor nodes for collecting soil salinity original data at monitoring plots corresponding to each space grid, setting a fixed sampling frequency, continuously monitoring and collecting each monitoring plot according to the fixed sampling frequency, and covering a complete irrigation and evaporation period for collecting time so as to obtain the soil moisture original data and the soil conductivity original data corresponding to each monitoring plot; Performing salinity conversion treatment on the original data of the soil conductivity of each monitoring land block based on a preset empirical conversion relation between the soil conductivity and the salinity, and obtaining the original data of the soil salinity corresponding to each monitoring land block, wherein the original data of the soil moisture is time series data of the soil volume moisture content, and the original data of the soil salinity is time series data of the soil total salinity; performing outlier rejection, missing value completion and time axis alignment treatment on the original soil moisture data and the original soil salinity data of each monitoring land, and fusing the treated original soil moisture data and the original soil salinity data to obtain a two-dimensional time sequence track set corresponding to each monitoring land.
3. The farmland water-salt evolution law mining system based on time-series trajectory clustering according to claim 1, wherein the obtaining the first optimization factor by performing water-salt lag cumulative effect analysis on the two-dimensional time-series trajectory set comprises: The method comprises the steps of obtaining a water-salt lag accumulation amount through carrying out association accumulation processing on water change data and salt change data in a two-dimensional time sequence track set; And obtaining a first optimization factor by performing difference punishment processing on the accumulated water-salt lag.
4. The farmland water and salt evolution law mining system based on time sequence track clustering according to claim 3, wherein the obtaining the water and salt lag accumulation amount by performing correlation accumulation processing on the water change data and the salt change data in the two-dimensional time sequence track set comprises: For any first target two-dimensional time sequence track to be clustered, extracting soil moisture original data and soil salinity original data corresponding to the first target two-dimensional time sequence track from a two-dimensional time sequence track set, and respectively carrying out adjacent moment first-order difference calculation on the soil moisture original data and the soil salinity original data corresponding to the first target two-dimensional time sequence track to acquire moisture change data and salinity change data corresponding to the first target two-dimensional time sequence track; For any first historical time node and any first current time node in the first target two-dimensional time sequence track, when the first current time node is not earlier than the first historical time node, taking the moisture change data corresponding to the first historical time node as first moisture driving evaluation; For any target time interval between the first historical time node and the first current time node, squaring the square of moisture change data and the square of salinity change data at adjacent moments in the target time interval, accumulating the squaring results corresponding to all adjacent moments, taking the accumulated results obtained correspondingly as the two-dimensional state space path length between the first historical time node and the first current time node, dividing the salinity response evaluation corresponding to each time node by the sum of a constant one and the two-dimensional state space path length, accumulating the corresponding obtained calculation results to obtain the salinity attenuation accumulated response evaluation, and taking the calculation result obtained by multiplying the first moisture driving evaluation and the salinity attenuation accumulated response evaluation as the hysteresis driving contribution corresponding to the first historical time node; For a first current time node, accumulating the hysteresis drive contribution amounts corresponding to all first historical time nodes before the first current time node to obtain a water-salt hysteresis accumulation amount corresponding to a first target two-dimensional time sequence track at the first current time node; And for any second target two-dimensional time sequence track for comparison and matching, acquiring the water-salt lag accumulation amount of the second target two-dimensional time sequence track corresponding to any second current time node by adopting the same processing mode as the first target two-dimensional time sequence track.
5. The farmland water salt evolution law mining system based on time-series trajectory clustering according to claim 3, wherein said obtaining a first optimization factor by performing a difference penalty process on a water salt lag accumulation amount comprises: For any first current time node in any first target two-dimensional time sequence track to be aligned and any second current time node in any second target two-dimensional time sequence track, acquiring a water-salt lag accumulated amount corresponding to the first target two-dimensional time sequence track at the first current time node and a water-salt lag accumulated amount corresponding to the second target two-dimensional time sequence track at the second current time node; Taking the absolute value of the difference value of the water-salt lag accumulation amount corresponding to the first target two-dimensional time sequence track at the first current time node and the water-salt lag accumulation amount corresponding to the second target two-dimensional time sequence track at the second current time node as the water-salt lag accumulation difference evaluation; Adding the absolute value of the water-salt lag accumulation amount corresponding to the first target two-dimensional time sequence track at the first current time node and the absolute value of the water-salt lag accumulation amount corresponding to the second target two-dimensional time sequence track at the second current time node, adding the absolute value of the water-salt lag accumulation amount to a preset minimum positive number constant on the addition result, and taking the corresponding obtained calculation result as a water-salt lag accumulation amount normalization reference; Taking the difference evaluation of the water-salt lag accumulation amount as a molecule, taking a normalization reference of the water-salt lag accumulation amount as a denominator, and taking a corresponding obtained partial formula as the relative difference evaluation of the water-salt lag accumulation amount; And carrying out exponential mapping processing based on natural constants on the relative difference evaluation of the water-salt lag accumulation amount, and obtaining a first optimization factor corresponding to the first target two-dimensional time sequence track at the first current time node and the second target two-dimensional time sequence track at the second current time node.
6. The farmland water and salt evolution law mining system based on time sequence track clustering according to claim 1, wherein the obtaining the second optimization factor by performing local water and salt track fluctuation degree analysis on the two-dimensional time sequence track set comprises: the method comprises the steps of obtaining local water-salt track fluctuation feature data by carrying out local fluctuation feature extraction processing on water data and salt data in a two-dimensional time sequence track set; carrying out normalization calculation processing on the local water-salt track fluctuation feature data to obtain local water-salt track fluctuation degree; And obtaining a second optimization factor by performing difference punishment processing on the local water-salt track fluctuation degree.
7. The farmland water and salt evolution law mining system based on time sequence track clustering according to claim 6, wherein the obtaining the local water and salt track fluctuation feature data by performing local fluctuation feature extraction processing on the water data and the salt data in the two-dimensional time sequence track set comprises: For any first target two-dimensional time sequence track to be clustered, extracting soil moisture original data and soil salinity original data corresponding to the first target two-dimensional time sequence track from a two-dimensional time sequence track set; For any first current time node in the first target two-dimensional time sequence track, when the first current time node is provided with the soil moisture original data and the soil salinity original data corresponding to the previous time node and the previous two time nodes, carrying out combined operation on the soil moisture original data corresponding to the first current time node, the twice of the soil moisture original data corresponding to the previous time node and the soil moisture original data corresponding to the previous two time nodes to obtain a moisture second-order fluctuation assessment corresponding to the first current time node; The square of the water second-order fluctuation estimation corresponding to the first current time node and the square of the salinity second-order fluctuation estimation corresponding to the first current time node are added and then squared, and the local fluctuation intensity estimation corresponding to the first current time node is obtained; accumulating all the obtained local fluctuation intensity evaluations for any first current time node in the first target two-dimensional time sequence track to obtain local water salt track fluctuation characteristic data corresponding to the first current time node of the first target two-dimensional time sequence track; And for any second target two-dimensional time sequence track for comparison and matching, adopting the same processing mode as the first target two-dimensional time sequence track to acquire local water salt track fluctuation characteristic data corresponding to the second target two-dimensional time sequence track at any second current time node.
8. The farmland water and salt evolution law mining system based on time sequence track clustering according to claim 6, wherein the obtaining the local water and salt track fluctuation degree by carrying out normalization calculation processing on the local water and salt track fluctuation feature data comprises the following steps: For any first current time node in any first target two-dimensional time sequence track to be aligned and any second current time node in any second target two-dimensional time sequence track, acquiring local water salt track fluctuation characteristic data corresponding to the first target two-dimensional time sequence track at the first current time node and local water salt track fluctuation characteristic data corresponding to the second target two-dimensional time sequence track at the second current time node; For any first current time node in the first target two-dimensional time sequence track, extracting soil moisture original data and soil salinity original data corresponding to the first current time node from the soil moisture original data and the soil salinity original data corresponding to the first target two-dimensional time sequence track, and extracting the soil moisture original data and the soil salinity original data corresponding to the starting time node; Taking the local water-salt track fluctuation characteristic data corresponding to the first target two-dimensional time sequence track at the first current time node as a numerator, taking the net displacement evaluation corresponding to the first target two-dimensional time sequence track at the first current time node as a denominator, and acquiring the local water-salt track fluctuation degree corresponding to the first target two-dimensional time sequence track at the first current time node; For any second current time node in the second target two-dimensional time sequence track, acquiring net displacement evaluation corresponding to the second target two-dimensional time sequence track at the second current time node by adopting the same processing mode as that of the first target two-dimensional time sequence track, taking local water salt track fluctuation characteristic data corresponding to the second target two-dimensional time sequence track at the second current time node as a molecule, taking the net displacement evaluation corresponding to the second target two-dimensional time sequence track at the second current time node as a denominator, and acquiring local water salt track fluctuation degree corresponding to the second target two-dimensional time sequence track at the second current time node.
9. The farmland water-salt evolution law mining system based on time-series trajectory clustering according to claim 6, wherein said obtaining a second optimization factor by performing a difference penalty process on local water-salt trajectory fluctuation degrees includes: For any first current time node in any first target two-dimensional time sequence track to be aligned and any second current time node in any second target two-dimensional time sequence track, acquiring the local water salt track fluctuation degree corresponding to the first target two-dimensional time sequence track at the first current time node and the local water salt track fluctuation degree corresponding to the second target two-dimensional time sequence track at the second current time node; taking the absolute value of the difference value of the local water-salt track fluctuation degree corresponding to the first target two-dimensional time sequence track at the first current time node and the local water-salt track fluctuation degree corresponding to the second target two-dimensional time sequence track at the second current time node as the local water-salt track fluctuation degree difference evaluation; adding a larger value of the local water-salt track fluctuation degree corresponding to the first target two-dimensional time sequence track at the first current time node and the local water-salt track fluctuation degree corresponding to the second target two-dimensional time sequence track at the second current time node with a preset minimum normal number constant, and taking a corresponding obtained calculation result as a local water-salt track fluctuation degree normalization reference; Taking the difference evaluation of the fluctuation degree of the local water-salt track as a molecule, taking a normalization reference of the fluctuation degree of the local water-salt track as a denominator, and taking a corresponding obtained partial formula as the relative difference evaluation of the fluctuation degree of the local water-salt track; and adding the relative difference evaluation of the fluctuation degree of the local water-salt track with a constant 1 to obtain a second optimization factor corresponding to the first target two-dimensional time sequence track at the first current time node and the second target two-dimensional time sequence track at the second current time node.
10. The farmland water-salt evolution law mining system based on time-series trajectory clustering according to claim 1, wherein said obtaining a final time-series similarity distance by performing joint correction based on a first optimization factor and a second optimization factor on an original single-step distance comprises: For any first current time node in any first target two-dimensional time sequence track to be aligned and any second current time node in any second target two-dimensional time sequence track, extracting soil moisture original data and soil salinity original data corresponding to the first current time node from soil moisture original data and soil salinity original data corresponding to the first target two-dimensional time sequence track, and extracting soil moisture original data and soil salinity original data corresponding to the second current time node from soil moisture original data and soil salinity original data corresponding to the second target two-dimensional time sequence track; The method comprises the steps of adding squares of differences of original soil moisture data corresponding to a first current time node and original soil moisture data corresponding to a second current time node, and then opening the squares to obtain an original single step distance between a first target two-dimensional time sequence track and a second target two-dimensional time sequence track corresponding to the first current time node and the second current time node; Taking the product of the original single step distance, the first optimization factor and the second optimization factor as a corrected single step distance corresponding to the first target two-dimensional time sequence track at the first current time node and the second target two-dimensional time sequence track at the second current time node; Adding the minimum accumulated distance corresponding to the correction single-step distance, the minimum accumulated distance corresponding to the previous time node of the first current time node and the previous time node of the second current time node, and the smaller value in the minimum accumulated distance corresponding to the previous time node of the first current time node and the previous time node of the second current time node to any one of the first current time node and any one of the second current time node in the second target two-dimensional time sequence track, so as to obtain the minimum accumulated distance corresponding to the first target two-dimensional time sequence track and the second target two-dimensional time sequence track in the second current time node; And taking the minimum accumulated distance corresponding to the end time node of the first target two-dimensional time sequence track and the end time node of the second target two-dimensional time sequence track as the final time sequence similarity distance between the first target two-dimensional time sequence track and the second target two-dimensional time sequence track.

Description

Farmland water-salt evolution rule digging system based on time sequence track clustering Technical Field The invention relates to the technical field of agricultural data analysis, in particular to a farmland water-salt evolution rule mining system based on time sequence track clustering. Background In the farmland water-salt regulation and control and saline-alkali soil treatment processes, the time sequence change process of soil moisture and salt directly reflects the water-salt migration rules under the actions of irrigation, rainfall, evaporation, infiltration and the like. In order to realize the fine management of a large-scale farmland plot, the soil moisture change condition and the salt change condition of different plots in a complete irrigation and evaporation period are required to be continuously monitored, and the water salt evolution types of the different plots are identified based on the monitoring results, so that data support is provided for irrigation regulation and control, salt washing and removal, soil improvement and regional treatment. In the prior art, aiming at the excavation of the water-salt evolution law of a plurality of plots in a farmland, a mode of combining time sequence track analysis and cluster recognition is generally adopted, the time sequence of the water and the salt of each plot is taken as an analysis object, the similarity between the plots is calculated by utilizing a time sequence similarity measurement method, and then the classification of the plots is completed by combining a clustering algorithm. In the existing time sequence similarity calculation method, the dynamic time warping algorithm can be widely applied to the alignment analysis process of the farmland water-salt two-dimensional time sequence track because the dynamic time warping algorithm can carry out nonlinear stretching and compression on a time axis and can overcome the problem that the whole time shifts exist at the irrigation starting moment, the rainfall response moment or the evaporation process of different plots to a certain extent. The dynamic time warping algorithm generally calculates local distances between different time nodes, builds an accumulated distance matrix, searches for a globally optimal alignment path among all possible matching paths, obtains a distance value between two time sequence tracks, and uses the distance value as a basis for subsequent clustering division. Based on the method, comparison and classification of water salt evolution processes among different monitored plots can be realized to a certain extent. However, in a real farmland scene, soil moisture changes and salinity changes do not occur synchronously, but obvious asynchronous hysteresis phenomenon generally exists. In general, irrigation or rainfall causes soil moisture to rise firstly, salt changes need to be gradually developed along with the processes of infiltration, leaching and redistribution of water, and different plots have great differences in hysteresis cycle and evolution intensity of water-salt response due to differences in soil texture, pore structure, plough layer state and initial distribution of salt. When the existing dynamic time warping algorithm calculates the local distance, the alignment is mainly carried out according to the instantaneous numerical value difference between time nodes, the hysteresis relationship of the water driving salinity change in different plots is difficult to reflect, and the intervention intensity difference corresponding to the local track fluctuation intensity is difficult to be described. Therefore, when the two-dimensional time sequence tracks of the water and salt of different plots are matched, the situation that the time axis is excessively stretched or compressed for pursuing the minimum numerical error easily occurs, and then the plots with obvious differences in the water and salt evolution mechanism are judged to be similar plots in error, so that the follow-up clustering result is inconsistent with the farmland real water and salt migration rule. Therefore, how to effectively represent the hysteresis difference and the fluctuation difference in the evolution of the farmland water salt in the time sequence track similarity calculation process, so that the accuracy of the mining result of the farmland water salt evolution law is improved, and the technical problem to be solved is urgent at present. Disclosure of Invention In view of the above, the invention aims to provide a farmland water salt evolution rule mining system based on time sequence track clustering, so as to solve the problem that the land plots with different water salt evolution mechanisms are clustered wrongly because the existing dynamic time warping algorithm cannot characterize the farmland water salt asynchronous hysteresis difference. In order to achieve the above purpose, the technical scheme of the invention is realized as follows: farmland water and salt evolution