CN114971097B - Soil moisture content data reconstruction method and prediction method - Google Patents
Soil moisture content data reconstruction method and prediction method Download PDFInfo
- Publication number
- CN114971097B CN114971097B CN202210925120.7A CN202210925120A CN114971097B CN 114971097 B CN114971097 B CN 114971097B CN 202210925120 A CN202210925120 A CN 202210925120A CN 114971097 B CN114971097 B CN 114971097B
- Authority
- CN
- China
- Prior art keywords
- soil moisture
- moisture content
- data
- content data
- historical
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 239000002689 soil Substances 0.000 title claims abstract description 779
- 238000000034 method Methods 0.000 title claims abstract description 103
- 238000005259 measurement Methods 0.000 claims abstract description 59
- 238000012549 training Methods 0.000 claims abstract description 50
- 238000013277 forecasting method Methods 0.000 claims abstract description 12
- 238000004590 computer program Methods 0.000 claims description 13
- 238000003860 storage Methods 0.000 claims description 11
- 238000012512 characterization method Methods 0.000 abstract description 3
- 238000004856 soil analysis Methods 0.000 abstract description 2
- 230000000875 corresponding effect Effects 0.000 description 59
- 238000003066 decision tree Methods 0.000 description 24
- 230000006870 function Effects 0.000 description 22
- 238000004422 calculation algorithm Methods 0.000 description 21
- 239000011159 matrix material Substances 0.000 description 13
- 230000008569 process Effects 0.000 description 11
- 238000012544 monitoring process Methods 0.000 description 10
- 238000005457 optimization Methods 0.000 description 9
- 238000009826 distribution Methods 0.000 description 7
- 238000012360 testing method Methods 0.000 description 7
- 238000004891 communication Methods 0.000 description 6
- 238000009825 accumulation Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000009434 installation Methods 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 239000000047 product Substances 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 3
- 241000287196 Asthenes Species 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 2
- 238000004220 aggregation Methods 0.000 description 2
- 239000003795 chemical substances by application Substances 0.000 description 2
- 230000002596 correlated effect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000003973 irrigation Methods 0.000 description 2
- 230000002262 irrigation Effects 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000002310 reflectometry Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000004927 clay Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001035 drying Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000004720 fertilization Effects 0.000 description 1
- 239000012467 final product Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000011065 in-situ storage Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000013138 pruning Methods 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/29—Geographical information databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/02—Agriculture; Fishing; Forestry; Mining
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- Databases & Information Systems (AREA)
- General Business, Economics & Management (AREA)
- Probability & Statistics with Applications (AREA)
- Quality & Reliability (AREA)
- Mathematical Analysis (AREA)
- Game Theory and Decision Science (AREA)
- Computational Mathematics (AREA)
- Algebra (AREA)
- Entrepreneurship & Innovation (AREA)
- Pure & Applied Mathematics (AREA)
- Operations Research (AREA)
- Development Economics (AREA)
- Mathematical Optimization (AREA)
- Medical Informatics (AREA)
- Agronomy & Crop Science (AREA)
- Animal Husbandry (AREA)
Abstract
The invention provides a soil moisture content data reconstruction method and a forecasting method, which relate to the technical field of soil analysis, and the method comprises the steps of acquiring first historical soil moisture content influence data of a target virtual station in a target area, wherein the target virtual station is used for representing the spatial position of the reconstructed soil moisture content data; and inputting the first historical soil moisture content influence data into a soil moisture content data reconstruction model to obtain a historical soil moisture content data reconstruction result of the target virtual station output by the soil moisture content data reconstruction model. According to the method, a soil moisture content data reconstruction model is obtained through data training obtained through measurement of a target real site, so that reconstruction of historical soil moisture content data of the target virtual site can be achieved, the historical soil moisture content data of the spatial position, which does not correspond to the real site, in a target area are supplemented, and characterization of the soil moisture content of the area can be achieved.
Description
Technical Field
The invention relates to the technical field of soil analysis, in particular to a soil moisture content data reconstruction method and a prediction method.
Background
The soil moisture content, namely the soil moisture content of the root zone of the crops, is the key for evaluating the water shortage state of the crops. At present, the monitoring mode of soil moisture content is developed from an early artificial soil drying judgment method to automatic measurement of a sensor. The method is characterized in that a Time Domain Reflectometry (TDR) measuring technology and a Frequency Domain Reflectometry (FDR) measuring technology are taken as representatives, the installation position of a sensor is taken as a site, and dielectric property parameters such as dielectric constant, dielectric loss and the like of the soil at different depths are measured through the sensor so as to be converted into the actual soil volume water content of the site. And the wireless communication networks such as 4G and 5G networks are matched, so that the soil moisture content Internet of things monitoring network for site in-situ monitoring can be realized.
However, the high heterogeneity of factors such as weather and soil physicochemical properties determines that the soil moisture content of a site is difficult to effectively characterize the soil moisture content of an area. Meanwhile, the installation and maintenance costs of the sensors are high, and accurate representation of soil moisture content of the region cannot be achieved through the mode that a large number of sensors are installed in the region.
Therefore, it is necessary to provide a method for reconstructing soil moisture data to realize the characterization of regional soil moisture.
Disclosure of Invention
The invention provides a soil moisture content data reconstruction method and a prediction method, which are used for overcoming the defects in the prior art.
The invention provides a soil moisture content data reconstruction method, which comprises the following steps:
acquiring first historical soil moisture content influence data of a target virtual site in a target area, wherein the target virtual site is used for representing the spatial position of the reconstructed soil moisture content data;
inputting the first historical soil moisture content influence data into a soil moisture content data reconstruction model to obtain a historical soil moisture content data reconstruction result of the target virtual site output by the soil moisture content data reconstruction model;
and the soil moisture content data reconstruction model is obtained by training based on second historical soil moisture content influence data of the target real site corresponding to the target virtual site and a historical soil moisture content data measurement result.
According to the soil moisture content data reconstruction method provided by the invention, the target virtual site and the target real site are determined based on the following methods:
acquiring time sequence satellite remote sensing data of the target area, and constructing a space-time cube based on the time sequence satellite remote sensing data; the time-space cube comprises a plurality of time columns, and each time column corresponds to time sequence data of one pixel position in the time sequence satellite remote sensing data;
selecting any pixel position in the time sequence satellite remote sensing data as a candidate virtual station, determining a target time column corresponding to the candidate virtual station, and calculating the time sequence data correlation between the target time column and other time columns in the space-time cube except the target time column;
and constructing a target plane area based on the pixel positions corresponding to other time columns with the time sequence data correlation larger than a first preset threshold, determining the number of real sites in the target plane area, and if the number is larger than a second preset threshold, determining the alternative virtual site as the target virtual site and determining the real site as the target real site.
According to the soil moisture content data reconstruction method provided by the invention, the acquiring of the first historical soil moisture content influence data of the target virtual site in the target area specifically comprises the following steps:
acquiring the first historical soil moisture content influence data every other first preset time period;
correspondingly, the inputting of the first historical soil moisture content influence data into a soil moisture content data reconstruction model to obtain a historical soil moisture content data reconstruction result of the target virtual site output by the soil moisture content data reconstruction model specifically includes:
determining a first acquisition moment for acquiring the first historical soil moisture content influence data each time;
inputting the first historical soil moisture content influence data acquired each time into the soil moisture content data reconstruction model to obtain a historical soil moisture content data reconstruction result of the target virtual site output by the soil moisture content data reconstruction model at the first acquisition moment.
According to the soil moisture content data reconstruction method provided by the invention, the first historical soil moisture content influence data acquired each time is input into the soil moisture content data reconstruction model, so as to obtain the historical soil moisture content data reconstruction result of the target virtual station output by the soil moisture content data reconstruction model at the first acquisition moment, and then the method further comprises the following steps:
updating the second historical soil moisture content influence data and the historical soil moisture content data measurement result every second preset time period, and updating the soil moisture content data reconstruction model based on the updated second historical soil moisture content influence data and the updated historical soil moisture content data measurement result to obtain an updated soil moisture content data reconstruction model;
and if the reconstruction error of the updated soil moisture content data reconstruction model is smaller than that of the soil moisture content data reconstruction model, inputting the first historical soil moisture content influence data acquired every time in a second preset time period after the soil moisture content data reconstruction model is updated every time into the updated soil moisture content data reconstruction model, and obtaining the historical soil moisture content data reconstruction result of the target virtual site output by the updated soil moisture content data reconstruction model at the first acquisition moment.
The invention also provides a soil moisture content data forecasting method, which comprises the following steps:
acquiring historical soil moisture content data, historical meteorological environment data and future meteorological environment data of a site to be forecasted;
inputting the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data into a soil moisture content data forecasting model to obtain the future soil moisture content data of the site to be forecasted, which is output by the soil moisture content data forecasting model;
the soil moisture content data forecasting model is obtained by training a future meteorological environment data sample of a sample site in a fourth preset time period and a second historical soil moisture content data sample of the sample site in the fourth preset time period based on a first historical soil moisture content data sample, a historical meteorological environment data sample and a forecast of the sample site in a third preset time period, wherein the sample site comprises a target virtual site or comprises the target virtual site and a real site, and the first historical soil moisture content data sample and the second historical soil moisture content data sample corresponding to the target virtual site are obtained based on the soil moisture content data reconstruction method.
According to the soil moisture content data forecasting method provided by the invention, the acquiring of the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data of the site to be forecasted specifically comprises the following steps:
acquiring the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data every fifth preset time period;
correspondingly, the inputting the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data into a soil moisture content data forecasting model to obtain the future soil moisture content data of the site to be forecasted, which is output by the soil moisture content data forecasting model, specifically includes:
determining a second acquisition time for acquiring the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data each time;
and inputting the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data which are acquired each time into the soil moisture content data forecasting model to obtain the future soil moisture content data of the site to be forecasted at the second acquisition moment, which is output by the soil moisture content data forecasting model.
The invention also provides a soil moisture content data reconstruction system, which comprises:
the system comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring first historical soil moisture content influence data of a target virtual site in a target area, and the target virtual site is used for representing the spatial position of reconstructed soil moisture content data;
the data reconstruction module is used for inputting the first historical soil moisture influence data into a soil moisture data reconstruction model to obtain a historical soil moisture data reconstruction result of the target virtual site output by the soil moisture data reconstruction model;
and the soil moisture content data reconstruction model is obtained by training based on second historical soil moisture content influence data of the target real site corresponding to the target virtual site and a historical soil moisture content data measurement result.
The invention also provides a soil moisture content data forecasting system, which comprises:
the second acquisition module is used for acquiring historical soil moisture content data, historical meteorological environment data and future meteorological environment data of the site to be forecasted;
the data forecasting module is used for inputting the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data into a soil moisture content data forecasting model to obtain the future soil moisture content data of the site to be forecasted, which is output by the soil moisture content data forecasting model;
the soil moisture content data forecasting model is obtained based on a first historical soil moisture content data sample, a historical meteorological environment data sample and a future meteorological environment data sample of a sample site in a fourth preset time period, wherein the first historical soil moisture content data sample, the historical meteorological environment data sample and the future meteorological environment data sample are obtained through prediction, the future meteorological environment data sample is obtained through prediction, the sample site is obtained through training, the sample site comprises the target virtual site, or comprises the target virtual site and the real site, and the first historical soil moisture content data sample and the second historical soil moisture content data sample corresponding to the target virtual site are obtained through the soil moisture content data reconstruction method.
The invention also provides electronic equipment which comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, wherein the processor executes the program to realize the soil moisture content data reconstruction method or the soil moisture content data forecast method.
The present invention also provides a non-transitory computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the soil moisture data reconstruction method described above, or implements the soil moisture data prediction method described above.
The invention also provides a computer program product comprising a computer program which, when executed by a processor, implements the soil moisture data reconstruction method described above, or implements the soil moisture data forecasting method described above.
According to the soil moisture content data reconstruction method and the soil moisture content data prediction method, the soil moisture content data reconstruction model is obtained through data training obtained through measurement of the target real site, reconstruction of historical soil moisture content data of the target virtual site can be achieved, the historical soil moisture content data of the spatial position, which does not correspond to the real site, in the target area are supplemented, and representation of the regional soil moisture content can be achieved.
Drawings
In order to more clearly illustrate the present invention or the technical solutions in the prior art, the drawings needed for the description of the embodiments or the prior art will be briefly described below, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
FIG. 1 is a schematic flow chart of a soil moisture data reconstruction method provided by the present invention;
FIG. 2 is a schematic flow chart of a soil moisture content data forecasting method according to the present invention;
FIG. 3 is a schematic view of the complete process of the soil moisture data reconstruction and forecast method of the present invention;
FIG. 4 is a schematic structural diagram of a soil moisture data reconstruction system provided by the present invention;
FIG. 5 is a schematic structural diagram of a soil moisture content data forecasting system provided by the present invention;
fig. 6 is a schematic structural diagram of an electronic device provided in the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In the prior art, due to high heterogeneity of factors such as weather and soil physicochemical properties, soil moisture content of a real site is difficult to effectively represent soil moisture content of an area with a certain range. Meanwhile, the installation and maintenance costs of the sensors are high, and accurate representation of soil moisture content of the region cannot be achieved through the mode that a large number of sensors are installed in the region. The embodiment of the invention provides a soil moisture content data reconstruction method, which is used for realizing the representation of regional soil moisture content.
Fig. 1 is a schematic flow chart of a method for reconstructing soil moisture content data according to an embodiment of the present invention, as shown in fig. 1, the method includes:
s11, acquiring first historical soil moisture content influence data of a target virtual station in a target area, wherein the target virtual station is used for representing the spatial position of the reconstructed soil moisture content data;
s12, inputting the first historical soil moisture content influence data into a soil moisture content data reconstruction model to obtain a historical soil moisture content data reconstruction result of the target virtual site output by the soil moisture content data reconstruction model;
and the soil moisture content data reconstruction model is obtained by training based on second historical soil moisture content influence data of the target real site corresponding to the target virtual site and a historical soil moisture content data measurement result.
Specifically, in the soil moisture data reconstruction method provided in the embodiment of the present invention, the main execution body is a soil moisture data reconstruction device, the device may be configured in a server, the server may be a local server or a cloud server, and the local server may be a computer, which is not specifically limited in the embodiment of the present invention.
Step S11 is executed firstly, first historical soil moisture content influence data of a target virtual site in a target area are obtained, the target area is an area to be researched, and the range size of the target area can be set according to needs. The target virtual site is used for representing the spatial position of the reconstructed soil moisture content data, namely the spatial position of the soil moisture content data which can be reconstructed without a real site in the target area.
It can be understood that the target virtual site indicates that the spatial position does not have a real site in reality, and the soil moisture content data of the spatial position needs to be reconstructed through soil moisture content data measurement results obtained by actual measurement of real sites of other spatial positions in the target area.
The first historical soil moisture content influence data may be historical daily soil moisture content influence data in a first time period near and before the current time, and the historical daily soil moisture content influence data is historical soil moisture content influence data in the first time period in days. The length of the first period may be set as desired, and is not particularly limited herein. For example, the first time period may be the past week, the past year, the past years, etc. For example, it may be the last 7 years (2015-2021 years).
The data categories of the first historical soil moisture impact data may include at least one of meteorological environment data, soil property data, time series remote sensing monitoring data, and geographic information system data.
The meteorological environment data may include an average air temperature (c), a minimum air temperature (c), a maximum air temperature (c), a dew point temperature (c), a cumulative rainfall (mm), a surface air pressure (hPa), a sea level air pressure (hPa), and a wind speed (m/s) in units of days, and the like. Wherein the air temperature can be collected at a height of 2 m.
The soil property data may include soil property data for a particular depth, such as soil pH (dimensionless), soil organic carbon (g/kg), and soil volume weight (kg/m) 3 ) The water holding capacity (%) of the soil field, the sand content (%) (kg/kg) and the clay content (%) (kg/kg). The specific depth can be in the range of 0-20cm, 20-40cm, 40-60cm or 60-80cm underground. If the soil property measuring equipment cannot directly measure and obtain the soil property data in the soil depth range of 0-20cm, 20-40cm, 40-60cm or 60-80cm and the like, the soil property data at the depth nearest to the soil depth of 20cm, 40cm, 60cm or 80cm and capable of being directly measured by the soil property measuring equipment can be selected for replacement.
The time-series remote sensing monitoring data refers to time-series remote sensing data related to Soil Moisture data, and may include Soil Moisture Active and Passive (SMAP) data, normalized Difference reference Index (NDVI) data, and the like, for example, the SMAP data is satellite remote sensing data, and the SMAP data and the NDVI data are both in an Image File Format (TIFF).
The time-series remote sensing monitoring data can be day-by-day data in the first time period and an aggregation result of the day-by-day data, and the aggregation result is day-by-day historical mean distribution obtained by carrying out average processing on SMAP data of the same date in the first time period. If there is no SMAP data on the current day, the SMAP data on the current date can be taken instead. NDVI data may be processed in the same manner as SMAP data.
Geographic Information System (GIS) data may include longitude and latitude coordinates, digital Elevation Model (DEM) (m), and spatially interpolated surfaces. Wherein the longitude and latitude coordinates may be represented in decimal. The spatial interpolation surface may be in an image file format, and may be a soil moisture spatial interpolation surface at a specific depth position day by day generated by an Inverse Distance weight method (IDW) and a Kriging interpolation method (Kriging) based on an average value of soil moisture data measurement results at the specific depth position day by day of the real site.
And then, step S12 is executed, the first historical soil moisture content influence data are input into the soil moisture content data reconstruction model, and historical soil moisture content data reconstruction results of the target virtual site output by the soil moisture content data reconstruction model are obtained. The soil moisture content data reconstruction model is used for representing the corresponding relation between the first historical soil moisture content influence data of the target virtual site and the historical soil moisture content data reconstruction result.
The soil moisture content data reconstruction model can be obtained by training second historical soil moisture content influence data of a target real site corresponding to the target virtual site and a historical soil moisture content data measurement result.
The second historical soil moisture content impact data may include historical day-to-day soil moisture content impact data for a second time period adjacent to and prior to the current time, the historical soil moisture content data measurements including historical day-to-day soil moisture content data measurements for the second time period. The second time period may be selected according to needs, may be the same as the first time period, and may also be a longer time period longer than the first time period, and is not limited specifically herein. The second historical soil moisture content influence data are the same as the first historical soil moisture content influence data in data type, and can also comprise at least one of meteorological environment data, soil property data, time sequence remote sensing monitoring data and geographic information system data.
The historical soil moisture content data measurement result can be a historical soil moisture content data measurement result at a certain depth, and also can be a historical soil moisture content data measurement result at a plurality of different depths, and the historical soil moisture content data reconstruction result is the same as the depth information of the historical soil moisture content data measurement result.
When the second historical soil moisture content influence data simultaneously comprise meteorological environment data, soil property data, time sequence remote sensing monitoring data and geographic information system data, the second historical soil moisture content influence data and the measurement results of the historical soil moisture content data can form a multi-source data set. In the embodiment of the invention, a TDengine distributed database can be used as a core storage engine, and structured data and unstructured data in the multi-source data set are integrated in a time sequence data form, so that a basis is provided for quick acquisition of data in subsequent steps.
The infrastructure of the soil moisture content data reconstruction model may be selected according to needs, and may be, for example, a neural network model or a decision tree model, which is not specifically limited herein.
The soil moisture content data reconstruction method provided by the embodiment of the invention comprises the steps of firstly obtaining first historical soil moisture content influence data of a target virtual station in a target area, then inputting the first historical soil moisture content influence data into a soil moisture content data reconstruction model, and obtaining a historical soil moisture content data reconstruction result of the target virtual station output by the soil moisture content data reconstruction model. According to the method, a soil moisture content data reconstruction model is obtained through data training obtained through measurement of a target real site, so that reconstruction of historical soil moisture content data of the target virtual site can be achieved, the historical soil moisture content data of the spatial position, which does not correspond to the real site, in a target area are supplemented, and characterization of the soil moisture content of the area can be achieved.
On the basis of the above embodiment, in the soil moisture content data reconstruction method provided in the embodiment of the present invention, the soil moisture content data reconstruction model is determined based on the following method:
training a decision tree model based on second historical soil moisture content influence data of a target real site corresponding to the target virtual site and a measurement result of the historical soil moisture content data, and optimizing hyper-parameters of the decision tree model by adopting a Bayesian algorithm to obtain a soil moisture content data reconstruction model.
Specifically, in the embodiment of the invention, the adopted soil moisture data reconstruction model trains the decision tree model through the second historical soil moisture influence data of the target real site corresponding to the target virtual site and the measurement result of the historical soil moisture data, and the hyper-parameters of the decision tree model are optimized by adopting a Bayesian algorithm.
Through the construction of the decision tree model, the decision tree model can efficiently process a large amount of second historical soil moisture content influence data and historical soil moisture content data measurement results through three steps of feature selection, decision tree generation and decision tree pruning, efficient and accurate fitting performance can be realized, feature importance sequencing can be visually obtained, and collaborative drawing parameters can be conveniently selected.
On this basis, the decision tree model can also use Ensemble Learning (EL) as a theoretical basis, and complete a Learning task by constructing and combining multiple learners to obtain a more accurate, stable and robust final result.
Particularly, in the embodiment of the present invention, the Decision Tree model may be a CatBoost model, and the CatBoost model may solve the problems of Gradient deviation and prediction offset that often occur in a machine learning algorithm of a conventional Gradient Boosting Decision Tree (GBDT) framework, thereby reducing the occurrence of overfitting and improving the generalization capability of the algorithm. The Catboost model takes a symmetric decision tree as a base learning device, can efficiently and reasonably process the class type characteristics, and has excellent fitting performance and extremely high training speed. Compared with Deep Neural Networks (DNN) algorithms such as transformers and the like, the method greatly shortens the training time, and reduces the amount of data sets required by training.
The Catboost model is a machine learning framework based on a gradient lifting decision tree, the second historical soil moisture content influence data and the historical soil moisture content data measurement results of each target real site form a training sample, and the second historical soil moisture content influence data and the historical soil moisture content data measurement results of all the target real sites form a training sample set
Wherein n is the number of target real sites,
second historical soil moisture content influence data structure of ith target real siteThe second historical soil moisture content influence data comprises numerical characteristics and nominal characteristics,
and obtaining the measurement result of the historical soil moisture content data of the ith target real site.
Aiming at the processing flow of the training sample set, firstly all the obtained training samples are processed
Randomly ordering and then aiming at a certain one
And adding the priority and a weight coefficient of the priority. The calculation of the priority is generally by pair
And averaging to obtain the final product. Remember all
An arrangement obtained by random ordering
From
To
Sequentially traversing the random sequence by the first p traversed
The value of the nominal type feature is calculated. Here, p are
Each subscript of (a) is increased by p, respectively
To all with
Arrangement by random ordering
A distinction is made. At this time, p
The randomly ordered arrangement can be expressed as
，
Can be replaced by:
wherein a is a weight coefficient greater than 0, and p is a prior value.
The Catboost model obtains new features by combining the nominal features in the second historical soil moisture impact data so as to play a better role in soil moisture data forecasting. In addition, the Catboost model replaces a gradient estimation method in the traditional algorithm by adopting an Ordered boosting mode, so that the deviation of gradient estimation is reduced, and the generalization capability of the Catboost model is improved.
The hyper-parameters of the Catboost model may include iteration number (iterations), learning rate (learning _ rate), node number, random _ state, and the like. As the hyper-parameter optimization is a core step in the training process of the Catboost model, the fitting precision of the soil moisture content data reconstruction model obtained by training is greatly influenced. In the embodiment of the invention, a Bayesian algorithm is adopted to optimize the hyper-parameters of the Catboost model.
The Bayes algorithm is an approximate optimization algorithm based on a probability agent model, and the prior knowledge is used for approximating the posterior distribution of an unknown function so as to adjust the hyperparameter, so that the hyperparameter sampling efficiency is greatly improved, and meanwhile, the global optimal solution can be effectively obtained. The Bayesian algorithm is used for machine learning, and updates the posterior distribution of the objective function by continuously adding sample points on the premise of giving the optimized objective function.
Particularly, the bayesian algorithm adopted in the embodiment of the present invention may be a Tree-structured park Optimization (TPE) algorithm, the TPE algorithm is constructed Based on a Model-Based Sequential Optimization method (SMBO), and the TPE algorithm generates a proxy Model by using Kernel Density Estimation (KDE), so that the proxy Model has good global exploration capability and is not prone to fall into local Optimization.
The Bayes algorithm belongs to a global optimization algorithm, and the optimization process utilizes Bayes theorem, uses a probability agent model to fit a target function f, and selects the next evaluation point according to the preorder sampling result, thereby rapidly achieving the optimal solution. The bayesian algorithm can be expressed as:
in the formula,
is the prior probability distribution of the objective function f,
a set of observed collections is represented,
，
represents the observed value of the hyper-parameter x,
，
which represents the probability of the hyperparameter given the score of the objective function f, i.e. the likelihood distribution of the objective function f,
representing the collection of observed objects in a given set
The conditional probability distribution of the time objective function f, i.e. the posterior probability distribution. In the embodiment of the present invention, the objective function f is a loss function, and may be a Mean Square Error (MSE) function.
TPE as a probabilistic proxy model will
Is defined as
Then, there are:
wherein,
representing the optimal value of the objective function f on the observed collection;
the loss function for the hyperparameter x is less than
The density of the image to be measured is estimated,
a loss function representing the hyperparameter x is equal to or greater than
The density of (2) is estimated. TPE expected improvement (Expec)And (EI) as a sampling function, selecting a next evaluation point having an optimization effect on the objective function.
When in use
In that
The process integration is positive and setting the hyperparameter x for algorithmic modeling will yield better results than observing the suprathreshold optima.
Order to
And (3) constructing:
the above formula indicates that when the hyperparameter x has
And
then, the maximum EI value is obtained. TPE through
And
constructing a hyper-parameter set to
The form of (2) evaluates the hyperparameter x, and in each iteration process, the algorithm returns the hyperparameter value with the maximum EI value
。
Further comprising:
in the embodiment of the present invention, optimizing the hyperparameter by using the TPE is to find the best fitting accuracy of the castboost model on the test data set, and the process of hyperparameter optimization can be expressed as follows:
wherein,
an objective function f representing the castboost model,
is that
Hyper-parameters at which the best results are achieved.
Compared with other parameter adjusting methods, the TPE automatic super-parameter adjusting method can reduce the times of tests as much as possible and improve the efficiency of the tests when finding the combination of the optimal super-parameter values by forming the knowledge of the relation between the super-parameter values and the model performance and deducing the selection of the next group of super-parameters by using the prior knowledge. In the embodiment of the present invention, the number of iterations of the TPE is set to 200.
In addition, in the embodiment of the invention, when the decision tree model is trained, a leave-one cross validation method is also utilized to evaluate the fitting precision of the decision tree model so as to realize the high-precision soil moisture content data reconstruction model of the cooperative multi-element factor.
Therefore, the soil moisture content data reconstruction model actually applied in the embodiment of the invention can be represented as a TPE-Catboost model.
In the embodiment of the invention, the basic model of the soil moisture content data reconstruction model adopted by the method is a decision tree model, and compared with the existing deep neural network model, the method has the advantages that the quantity of training samples is smaller, the training time can be greatly shortened, and the model training efficiency is improved. In the training process of the decision tree model, the Bayesian algorithm is introduced to optimize the hyper-parameters of the decision tree model, so that the reconstruction accuracy of the soil moisture content data reconstruction model is better, the model performance is better, and the accuracy of the historical soil moisture content data reconstruction result can be further ensured.
On the basis of the above embodiment, in the soil moisture content data reconstruction method provided in the embodiment of the present invention, the target virtual site and the target real site are determined based on the following methods:
acquiring time sequence satellite remote sensing data of the target area, and constructing a space-time cube based on the time sequence satellite remote sensing data; the time-space cube comprises a plurality of time columns, and each time column corresponds to time sequence data of one pixel position in the time sequence satellite remote sensing data;
selecting any pixel position in the time sequence satellite remote sensing data as a candidate virtual station, determining a target time column corresponding to the candidate virtual station, and calculating the time sequence data correlation between the target time column and other time columns in the space-time cube except the target time column;
and constructing a target plane area based on the pixel positions corresponding to other time columns of which the time sequence data correlation is greater than a first preset threshold, determining the number of real sites in the target plane area, if the number is greater than a second preset threshold, determining that the alternative virtual site is the target virtual site, and determining that the real site is the target real site.
Specifically, in the embodiment of the present invention, the target virtual site is a spatial location for reconstructing soil moisture content data, and the reconstructed soil moisture content data requires an accurate soil moisture content data measurement result measured by the real site. Since the number of real sites and the soil moisture data measurement results corresponding thereto are limited, not all target virtual sites in the target area can perform soil moisture data reconstruction, depending on the soil moisture data measurement results of the target real sites corresponding to the target virtual sites.
Therefore, the embodiment of the invention provides a method for determining a target virtual site and a target real site by performing regional space-time clustering on time-series satellite remote sensing data, which will be described in detail below.
Firstly, time sequence satellite remote sensing data, namely time sequence SMAP data, of a target area are obtained, and the time sequence SMAP data can be obtained from second historical soil moisture content influence data.
Based on time sequence satellite remote sensing data, a space-time cube can be constructed, the two-dimensional coordinate axes (x, y) of the space-time cube can be used for representing the plane position of soil moisture content, and the one-dimensional time axis z is used for representing the change of SMAP data on the plane position along with time. If the time-series satellite remote sensing data are Level3 Level data, the time resolution is 7 days, and therefore the time resolution of the space-time cube is also 7 days. The spatio-temporal cube may include a plurality of time bins, each time bin corresponding to time series data for a pixel location in the time series satellite telemetry data.
And then, selecting any pixel position in the time sequence satellite remote sensing data as a candidate virtual station, and determining a target time column corresponding to the candidate virtual station. And calculating the time sequence data correlation between the target time column and other time columns except the target time column in the space-time cube. The time series data correlation can be characterized by a pearson correlation coefficient, for example, for any other time column a, the time series data corresponding to the other time column a can be expressed as
The time sequence data corresponding to the target time column b can be expressed as
Then, the time series data correlation p between the target time column b and any other time column a can be calculated by the following formula:
wherein,
and
the average values of n data are respectively; p has a value range of
And p is a positive value and a negative value, which respectively indicate that the two variables are positively correlated and negatively correlated with each other.
Thereafter, pixel positions corresponding to other time bins with time series data correlation greater than a first preset threshold may be selected to construct the target plane area. The value of the first preset threshold may be set as needed, and may be set to be greater than or equal to 0.3, for example. And a target plane area formed by the selected pixel positions is a homogeneous mode area of the alternative virtual station, and a real station in the target plane area is a similar mode station of the alternative virtual station.
And then, determining the number of real sites in the target plane area, and if the number is greater than a second preset threshold, indicating that the pixel position of the candidate virtual site has enough support data for reconstructing soil moisture content data, further determining that the candidate virtual site is the target virtual site, and simultaneously determining that all the real sites in the target plane area are the target real sites. The value of the second preset threshold may be set as needed, for example, may be set to be greater than or equal to 5.
In the embodiment of the invention, the target virtual site and the target real site in the target area are determined by constructing the space-time cube and calculating the time sequence data correlation between the target time column and other time columns in the space-time cube, so that the determination efficiency can be improved, and the reconstruction efficiency is further improved.
On the basis of the above embodiment, in the soil moisture content data reconstruction method provided in the embodiment of the present invention, the second historical soil moisture content influence data includes historical daily soil moisture content influence data, and the historical soil moisture content data measurement result includes multiple-depth historical daily soil moisture content data measurement results; the soil moisture content data reconstruction model comprises a reconstruction sub-model corresponding to each depth, and the reconstruction sub-model is obtained by training based on the historical day-by-day soil moisture content influence data and the historical day-by-day soil moisture content data measurement result corresponding to the depth;
correspondingly, the inputting the first historical soil moisture content influence data into a soil moisture content data reconstruction model to obtain a historical soil moisture content data reconstruction result of the target virtual site output by the soil moisture content data reconstruction model specifically includes:
and inputting the first historical soil moisture influence data into the reconstruction submodel to obtain a historical daily soil moisture data reconstruction result of the target virtual site at the corresponding depth output by the reconstruction submodel.
Specifically, in the embodiment of the present invention, the second historical soil moisture content influence data includes historical daily soil moisture content influence data, the historical daily soil moisture content influence data in the second time period as described in the above embodiment may be used, the historical soil moisture content data measurement result may include multiple-depth historical daily soil moisture content data measurement results, and the multiple-depth historical daily soil moisture content data measurement results in the second time period as described in the above embodiment may be used. The number and value of the depths can be set according to the needs, for example, the number of the depths can be 4, and the value ranges can be 0-20cm, 20-40cm, 40-60cm or 60-80cm underground respectively.
At this time, the soil property data included in the first and second historical soil moisture influence data may include soil attribute data of multiple depths, and the spatial interpolation surface of the GIS data included in the first and second historical soil moisture influence data may be a daily and multiple-depth soil moisture spatial interpolation surface generated by IDW and Kriging based on an average value of measurement results of the multiple-depth soil moisture data of the daily real site.
Furthermore, the soil moisture data reconstruction model may include a reconstruction sub-model corresponding to each depth, and the number of the reconstruction sub-models is the same as the number of the depths. The reconstruction submodel corresponding to each depth can train the decision tree model through historical daily soil moisture content influence data and the historical daily soil moisture content data measurement results of the corresponding depth, and the super parameters of the decision tree model are optimized through a Bayesian algorithm to obtain the reconstruction submodel. In the embodiment of the invention, historical daily soil moisture content influence data and historical daily soil moisture content data measurement results at different depths can be adopted for training to obtain the reconstruction submodels which can be suitable for reconstructing the historical soil moisture content data at different depths, and the soil moisture content data reconstruction model can be regarded as a model set formed by all the reconstruction submodels.
Correspondingly, when the first historical soil moisture content influence data is input into the soil moisture content data reconstruction model to obtain the historical soil moisture content data reconstruction result of the target virtual site output by the soil moisture content data reconstruction model, the first historical soil moisture content influence data can be input into each reconstruction sub-model to obtain the historical day-by-day soil moisture content data reconstruction result of the target virtual site output by each reconstruction sub-model at the corresponding depth of each reconstruction sub-model.
In the embodiment of the invention, the reconstruction submodels corresponding to a plurality of depths are obtained through training, so that the reconstruction of the historical daily soil moisture content data at the plurality of depths can be realized, the data size of the reconstructed historical daily soil moisture content data reconstruction result is ensured, and the subsequent prediction of the soil moisture content data at different depths in the whole target area in the future is facilitated.
On the basis of the foregoing embodiment, the method for reconstructing soil moisture content data according to the embodiment of the present invention, which acquires first historical soil moisture content influence data of a target virtual site in a target area, specifically includes:
acquiring the first historical soil moisture content influence data every other first preset time period;
correspondingly, the inputting of the first historical soil moisture content influence data into a soil moisture content data reconstruction model to obtain a historical soil moisture content data reconstruction result of the target virtual site output by the soil moisture content data reconstruction model specifically includes:
determining a first acquisition time for acquiring the first historical soil moisture content influence data each time;
and inputting the first historical soil moisture content influence data acquired each time into the soil moisture content data reconstruction model to obtain a historical soil moisture content data reconstruction result of the target virtual site at the first acquisition moment, which is output by the soil moisture content data reconstruction model.
Specifically, with the continuous update and accumulation of the data, in the embodiment of the present invention, the first historical soil moisture content influence data may be dynamically acquired, that is, the first historical soil moisture content influence data may be acquired according to a preset first acquisition frequency. The first acquiring frequency is a first preset time period, and the length of the first acquiring frequency can be set according to the data updating frequency and actual needs, and meanwhile, the acquiring cost and the storage cost are also considered. For example, the length of the first preset time period may be set to 1 day. The initial acquisition time may be when day 0, and the second acquisition time may be when day 0, so on.
It can be understood that, after the first historical soil moisture content influence data is obtained once, the first obtaining time for obtaining the first historical soil moisture content influence data every time needs to be determined, and the step S12 is executed, that is, the first historical soil moisture content influence data obtained every time is input into the soil moisture content data reconstruction model, so as to reconstruct the historical soil moisture content data, and the historical soil moisture content data reconstruction result of the target virtual site at each first obtaining time and output by the soil moisture content data reconstruction model is obtained. That is, the first preset time period is also the reconstruction frequency of the historical soil moisture content data.
In the embodiment of the invention, the first historical soil moisture content influence data is acquired every other first preset time period, and the historical soil moisture content data is further reconstructed, so that the target virtual station can be ensured to have a time sequence historical soil moisture content data reconstruction result.
On the basis of the foregoing embodiment, the method for reconstructing soil moisture content data according to an embodiment of the present invention includes that the first historical soil moisture content influence data acquired each time is input to the soil moisture content data reconstruction model, so as to obtain a historical soil moisture content data reconstruction result of the target virtual site output by the soil moisture content data reconstruction model at the first acquisition time, and then further includes:
updating the second historical soil moisture content influence data and the historical soil moisture content data measurement result every other second preset time period, and updating the soil moisture content data reconstruction model based on the updated second historical soil moisture content influence data and the updated historical soil moisture content data measurement result to obtain an updated soil moisture content data reconstruction model;
if the reconstruction error of the updated soil moisture content data reconstruction model is smaller than that of the soil moisture content data reconstruction model, inputting the first historical soil moisture content influence data acquired every time in a second preset time period after the soil moisture content data reconstruction model is updated every time into the updated soil moisture content data reconstruction model, and obtaining the historical soil moisture content data reconstruction result of the target virtual site at the first acquisition moment output by the updated soil moisture content data reconstruction model.
Specifically, with the continuous update and accumulation of data, in the embodiment of the present invention, the soil moisture content data reconstruction model may be dynamically updated, that is, the soil moisture content data reconstruction model is updated according to the preset first model update frequency, so as to determine the soil moisture content data reconstruction model with better reconstruction performance to reconstruct the subsequent historical soil moisture content data. The first model updating frequency is a second preset time period, the length of the first model updating frequency can be set according to the data updating frequency and actual needs, and meanwhile, the model training cost is also considered. For example, the length of the second preset time period may be set to 30 days, 60 days, or the like.
In the embodiment of the invention, when the soil moisture content data reconstruction model is updated, the second historical soil moisture content influence data and the historical soil moisture content data measurement result can be updated every second preset time period, so that the updated second historical soil moisture content influence data and the updated historical soil moisture content data measurement result are respectively obtained. And then updating the soil moisture content data reconstruction model according to the updated second historical soil moisture content influence data and the updated historical soil moisture content data measurement result, namely adjusting the model parameters of the soil moisture content data reconstruction model by adopting the updated second historical soil moisture content influence data and the updated historical soil moisture content data measurement result to obtain the updated soil moisture content data reconstruction model.
It should be noted that, although an updated soil moisture content data reconstruction model is obtained here, whether the updated soil moisture content data reconstruction model can be applied as a model for subsequent data reconstruction needs to be further judged through a reconstruction error, that is, a test set can be used to respectively test the soil moisture content data reconstruction model before updating and the updated soil moisture content data reconstruction model to obtain reconstruction errors of the soil moisture content data reconstruction model and the updated soil moisture content data reconstruction model, if the reconstruction error of the updated soil moisture content data reconstruction model is smaller than that of the soil moisture content data reconstruction model, the reconstruction performance of the updated soil moisture content data reconstruction model can be considered to be better than that of the updated soil moisture content data reconstruction model, and further, the first historical soil moisture content influence data acquired each time in a second preset time period after updating the soil moisture content data reconstruction model each time can be input into the updated soil moisture content data reconstruction model, and the historical soil moisture content data reconstruction is performed by using the updated soil moisture content data reconstruction model to obtain a soil moisture content data reconstruction result of a target site output by the updated soil moisture content data reconstruction model at each first acquisition time. Compared with a soil moisture content data reconstruction model before updating, the method can enable the obtained historical soil moisture content data reconstruction result to be more accurate.
If the reconstruction error of the updated soil moisture content data reconstruction model is larger than or equal to that of the soil moisture content data reconstruction model before updating, the reconstruction performance of the soil moisture content data reconstruction model before updating is considered to be superior to that of the updated soil moisture content data reconstruction model, so that the historical soil moisture content data is reconstructed by adopting the soil moisture content data reconstruction model before updating after the first historical soil moisture content influence data is acquired every time in the second preset time period after the soil moisture content data reconstruction model is updated every time, and the accuracy of the obtained historical soil moisture content data reconstruction result is ensured.
In the embodiment of the invention, all the soil moisture content data reconstruction models before and after updating and the corresponding reconstruction errors can be stored.
On the basis of the above embodiments, the embodiment of the present invention further provides a method for forecasting soil moisture content data, as shown in fig. 2, the method includes:
s21, acquiring historical soil moisture content data, historical meteorological environment data and future meteorological environment data of a site to be forecasted;
s22, inputting the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data into a soil moisture content data forecasting model to obtain the future soil moisture content data of the site to be forecasted, which is output by the soil moisture content data forecasting model;
the soil moisture content data forecasting model is obtained based on a first historical soil moisture content data sample, a historical meteorological environment data sample and a future meteorological environment data sample of a sample site in a fourth preset time period, wherein the first historical soil moisture content data sample, the historical meteorological environment data sample and the future meteorological environment data sample are obtained through prediction, the future meteorological environment data sample is obtained through prediction, the sample site is obtained through training, the sample site comprises the target virtual site, or comprises the target virtual site and the real site, and the first historical soil moisture content data sample and the second historical soil moisture content data sample corresponding to the target virtual site are obtained through the soil moisture content data reconstruction method.
Specifically, in the soil moisture content data forecasting method provided in the embodiment of the present invention, the main execution body is a soil moisture content data forecasting device, the device may be configured in a server, the server may be a local server, and may also be a cloud server, and the local server may specifically be a computer, and the like, which is not specifically limited in the embodiment of the present invention.
Step S21 is executed first to obtain historical soil moisture content data, historical meteorological environment data, and future meteorological environment data of the site to be forecasted. The site to be forecasted is a site which needs to forecast soil moisture data of a spatial position where the site is located, and the site can be a real site or a target virtual site, and is not particularly limited herein. The historical soil moisture content data of the site to be forecasted can be historical day-by-day soil moisture content influence data in a third time period which is close to the current time and is before the current time, the length of the third time period can be set according to needs, and the third time period is not specifically limited. For example, the third time period may be the last three days, the last five days, the last week, the last two weeks, and so forth.
The historical weather environment data may be weather environment data for a third time period.
The future weather environment data may be weather environment data for a fourth time period adjacent to and after the current time. The length of the fourth time period may be set as needed, and is not particularly limited herein. For example, the fourth time period may be three days in the future, five days in the future, one week in the future, two weeks in the future, and so forth. It is to be understood that the third preset time period may be before and adjacent to the fourth preset time period.
And then, executing a step S22, and inputting the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data into the soil moisture content data forecasting model to obtain the future soil moisture content data of the site to be forecasted, which is output by the soil moisture content data forecasting model. The soil moisture content data forecasting model is used for representing historical soil moisture content data and historical meteorological environment data of a site to be forecasted and corresponding relations between the future meteorological environment data and the future soil moisture content data.
Before the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data are input into the soil moisture content data forecasting model, the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data can be spliced, and the splicing result is input into the soil moisture content data forecasting model.
The soil moisture content data forecasting model can be obtained by training a first historical soil moisture content data sample, a historical meteorological environment data sample, a future meteorological environment data sample and a second historical soil moisture content data sample of a sample site in a third preset time period. The third preset time period may have the same length as the third time period, the fourth preset time period may have the same length as the fourth time period, and the third preset time period may be before and adjacent to the fourth preset time period.
The historical meteorological environment data sample refers to meteorological environment data of the sample site in a third preset time period, and the future meteorological environment data sample refers to meteorological environment data of the sample site in a fourth preset time period, wherein the meteorological environment data is obtained through prediction in the third preset time period.
The sample site may include only a real site, or may include both a real site and a target virtual site. When the sample site is a real site, the first historical soil moisture content data sample and the second historical soil moisture content data sample can be obtained by directly measuring through the real site, and when the sample site is a target virtual site, the first historical soil moisture content data sample and the second historical soil moisture content data sample can be obtained by reconstructing through the soil moisture content data reconstruction method provided in the embodiments.
In the embodiment of the invention, the first historical soil moisture content data sample, the historical meteorological environment data sample and the future meteorological environment data sample can be used as the input of the initial model of the soil moisture content data forecasting model, and the second historical soil moisture content data sample can be used as the label corresponding to the input of the initial model. For example, given a certain historical date n, the third preset time period may be from the nth day to the n-6 th day, and the fourth preset time period may be from the n +1 th day to the n +7 th day. The first historical soil moisture data sample may be soil moisture data from the nth day to the nth-6 th day, and includes 4 indicators (4 depths) and 28 (4 × 7) data items, the historical meteorological environment data sample may be meteorological environment data from the nth day to the nth-6 th day, and includes 8 indicators and 56 (8 × 7) data items, and the future meteorological environment data sample may be meteorological environment data from the n +1 th day to the n +7 th day, and includes 4 indicators and 28 (4 × 7) data items.
Before the first historical soil moisture content data sample, the historical meteorological environment data sample and the future meteorological environment data sample are input into the initial model, the first historical soil moisture content data sample, the historical meteorological environment data sample and the future meteorological environment data sample can be spliced, and then the splicing result is input into the initial model. The splicing result can be a two-dimensional matrix of 14 × 16, wherein the two-dimensional matrix is longitudinally time and transversely different data items such as a first historical soil moisture content data sample and a future meteorological environment data sample.
The second historical soil moisture data sample may be the historical soil moisture data of the (n + 1) th to (n + 7) th days, and includes 4 indexes and 28 (4 × 7) data items. The second historical soil moisture data samples may constitute a 7 x 4 two-dimensional matrix. Thus, the output of the initial model is also a 7 × 4 two-dimensional matrix.
The initial model can output a soil moisture content data prediction value in the same time period as a future meteorological environment data sample, then a loss function value of the initial model is calculated through the output and input of the initial model, model parameters of the initial model are updated iteratively based on the loss function value, and further training of the initial model is achieved. In the training process, an Early Stopping method can be adopted to determine a termination point, namely when the prediction error MSE of the model obtained by a certain training is not the lowest for 10 times continuously, the training is terminated, and then the soil moisture content data prediction model is obtained.
The initial model adopted in the embodiment of the present invention may be a transform model based on a self-attention (self-attention) mechanism, and the initial model may obtain a weighted score by calculating correlations between elements in the input sequence and all other elements. The initial model can comprise an encoder and a decoder, wherein the encoder and the decoder are respectively formed by stacking 10 same layers, each layer of the encoder comprises a multi-head self-attention sublayer and a feedforward neural network sublayer, and each layer of the decoder comprises a shielding multi-head self-attention sublayer, a coding-decoder multi-head attention sublayer and a feedforward neural network sublayer.
The self-attention mechanism calculates the attention value of the feature matrix by adopting the zooming dot product attention, firstly calculates the weight coefficient by performing dot product and ReLU activation function value on the query matrix and the key matrix, and then performs weighted summation on the value matrix according to the weight coefficient, as shown in the following formula:
wherein,
in order to query the matrix, the matrix is,
in the form of a matrix of keys,
being a matrix of values, the three matrices being formed by the input feature matrices
Respectively corresponding weight matrix
、
、
The result of the multiplication is that,
is composed of
、
、
The dimension of (c).
The soil moisture content data forecasting method provided by the embodiment of the invention comprises the steps of firstly, acquiring historical soil moisture content data and future meteorological environment data of a site to be forecasted; and then inputting the historical soil moisture content data and the future meteorological environment data into a soil moisture content data forecasting model to obtain the future soil moisture content data of the site to be forecasted, which is output by the soil moisture content data forecasting model. The method can predict the future soil moisture content data of the site to be predicted in a future period of time, and is convenient for making water-saving irrigation, fertilization decision, drainage measures and the like in time so as to enable crops to be in an optimal growth state. The historical soil moisture content data samples of the target virtual site are used as training samples, so that the generalization capability of the obtained soil moisture content data forecasting model is stronger, and the introduction of the soil moisture content data forecasting model can make the obtained future soil moisture content data of the site to be forecasted more accurate.
On the basis of the above embodiments, in the soil moisture data forecasting method provided in the embodiments of the present invention, the historical soil moisture data includes historical day-by-day soil moisture data, the future meteorological environment data includes future day-by-day meteorological environment data, the first historical soil moisture data sample corresponding to the target virtual site includes a multi-depth first historical day-by-day soil moisture data sample, the future meteorological environment data sample includes a future day-by-day meteorological environment data sample, and the second historical soil moisture data sample corresponding to the target virtual site includes a multi-depth second historical day-by-day soil moisture data sample; the soil moisture content data forecasting model comprises a forecasting sub-model corresponding to each depth, and the forecasting sub-model is obtained by training based on a first historical day-by-day soil moisture content data sample, a future day-by-day meteorological environment data sample and a second historical day-by-day soil moisture content data sample of the corresponding depth;
correspondingly, the inputting the historical soil moisture content data and the future meteorological environment data into a soil moisture content data forecasting model to obtain the future soil moisture content data of the site to be forecasted, which is output by the soil moisture content data forecasting model, specifically includes:
and inputting the historical day-by-day soil moisture content data and the future day-by-day meteorological environment data into the forecasting submodel to obtain the future day-by-day soil moisture content data of the site to be forecasted at the corresponding depth, which is output by the forecasting submodel.
Specifically, in the embodiment of the present invention, the historical soil moisture data includes historical daily soil moisture data, such as historical daily soil moisture influence data in a third time period, the future meteorological environment data includes future daily meteorological environment data, such as daily meteorological environment data in a fourth time period, the first historical soil moisture data sample corresponding to the target virtual site may include a multi-depth first historical daily soil moisture data sample, that is, the multi-depth historical daily soil moisture influence data of the sample site in the third preset time period, and the second historical soil moisture data sample corresponding to the target virtual site includes a multi-depth second historical daily soil moisture data sample, that is, the multi-depth historical daily soil moisture influence data of the sample site in the fourth preset time period.
Furthermore, the soil moisture data forecasting model may include forecasting sub-models corresponding to each depth, and the number of the forecasting sub-models is the same as that of the depths. The forecasting sub-model corresponding to each depth can be obtained by training the initial model through the first historical daily soil moisture content data sample, the future meteorological environment data sample and the second historical daily soil moisture content data sample of each depth. In the embodiment of the invention, the first historical daily soil moisture content data sample, the future meteorological environment data sample and the second historical daily soil moisture content data sample of each depth can be adopted to train to obtain the forecasting submodel suitable for forecasting the future daily soil moisture content data at different depths, and at the moment, the soil moisture content data forecasting model can be regarded as a model set formed by all the forecasting submodels.
Accordingly, when the historical soil moisture content data and the future meteorological environment data are input into the soil moisture content data forecasting model to obtain the future soil moisture content data of the target virtual site output by the soil moisture content data forecasting model, the historical daily soil moisture content data and the future daily meteorological environment data can be input into each forecasting sub-model together to obtain the future daily soil moisture content data of the target virtual site output by each forecasting sub-model at the depth corresponding to each forecasting sub-model.
In the embodiment of the invention, the forecasting sub-models corresponding to a plurality of depths are obtained through training, so that the future day-by-day soil moisture content data at the plurality of depths can be forecasted, and a basis is provided for agricultural irrigation in the follow-up process.
On the basis of the above embodiment, the method for forecasting soil moisture content data provided in the embodiment of the present invention specifically includes the following steps:
acquiring the historical soil moisture content data and the future meteorological environment data every fifth preset time period;
correspondingly, the inputting the historical soil moisture content data and the future meteorological environment data into a soil moisture content data forecasting model to obtain the future soil moisture content data of the site to be forecasted, which is output by the soil moisture content data forecasting model, specifically includes:
and inputting the historical soil moisture content data and the future meteorological environment data acquired each time into the soil moisture content data forecasting model to obtain the future soil moisture content data which is output by the soil moisture content data forecasting model and is obtained by the station to be forecasted at the acquisition time of the historical soil moisture content data and the future meteorological environment data each time.
Specifically, with the continuous update and accumulation of the data, the historical soil moisture content data and the future meteorological environment data can be dynamically acquired, that is, the historical soil moisture content data and the future meteorological environment data are acquired according to the preset second acquisition frequency. The second acquiring frequency is a fifth preset time period, and the length of the second acquiring frequency can be set according to the data updating frequency and actual needs, and meanwhile, the acquiring cost and the storage cost are also considered. The length of the fifth preset time period may be the same as or different from the length of the first preset time period.
It can be understood that, after acquiring the historical soil moisture content data and the future meteorological environment data each time, the step S22 is executed, that is, the historical soil moisture content data and the future meteorological environment data acquired each time are input into the soil moisture content data forecasting model to forecast the future soil moisture content data, so as to obtain the future soil moisture content data output by the soil moisture content data forecasting model and obtained at the time of acquiring the historical soil moisture content data and the future meteorological environment data by the station to be forecasted each time. That is, the fifth preset time period is also the forecast frequency of the future soil moisture data.
In the embodiment of the invention, the historical soil moisture content data and the future meteorological environment data are acquired every fifth preset time period, and the future soil moisture content data is further forecasted, so that the site to be forecasted can be ensured to have the time-series future soil moisture content data.
On the basis of the above embodiments, the method for forecasting soil moisture content data according to the embodiments of the present invention includes that the historical soil moisture content data and the future meteorological environment data acquired each time are input to the soil moisture content data forecasting model, and the future soil moisture content data output by the soil moisture content data forecasting model and obtained by the station to be forecasted at the time of acquiring the historical soil moisture content data and the future meteorological environment data each time is obtained, and then the method further includes:
updating the first historical soil moisture content data sample, the future meteorological environment data sample and the second historical soil moisture content data sample every sixth preset time period, and updating the soil moisture content data forecasting model based on the updated first historical soil moisture content data sample, the updated future meteorological environment data sample and the updated second historical soil moisture content data sample to obtain an updated soil moisture content data forecasting model;
if the prediction error of the updated soil moisture content data prediction model is smaller than that of the soil moisture content data prediction model, inputting historical soil moisture content data and future meteorological environment data acquired every time within a sixth preset time period after the soil moisture content data prediction model is updated every time into the updated soil moisture content data reconstruction model, and obtaining future soil moisture content data of the station to be predicted at the acquisition time, which is output by the updated soil moisture content data reconstruction model.
Specifically, with the continuous update and accumulation of data, the soil moisture content data forecasting model can be dynamically updated in the embodiment of the invention, that is, the soil moisture content data forecasting model is updated according to the preset second model updating frequency, so that the soil moisture content data forecasting model with better forecasting performance is determined to forecast the subsequent future soil moisture content data. The second model updating frequency is a sixth preset time period, the length of the second model updating frequency can be set according to the data updating frequency and actual needs, and meanwhile, the model training cost is also considered. The length of the sixth preset time period may be the same as or different from the length of the second preset time period.
In the embodiment of the present invention, when the soil moisture content data prediction model is updated, the first historical soil moisture content data sample, the future meteorological environment data sample, and the second historical soil moisture content data sample may be updated every sixth preset time period, so as to obtain the updated first historical soil moisture content data sample, the updated future meteorological environment data sample, and the updated second historical soil moisture content data sample, respectively. And then, updating the soil moisture content data forecasting model according to the updated first historical soil moisture content data sample, the updated future meteorological environment data sample and the updated second historical soil moisture content data sample, namely, adjusting the model parameters of the soil moisture content data forecasting model by adopting the updated first historical soil moisture content data sample, the updated future meteorological environment data sample and the updated second historical soil moisture content data sample to obtain the updated soil moisture content data forecasting model.
It should be noted that, although the updated soil moisture content data prediction model is obtained here, whether the updated soil moisture content data prediction model can be applied as a model for subsequent data prediction needs to be further judged through prediction errors, that is, the test set can be used to respectively test the soil moisture content data prediction model before updating and the updated soil moisture content data prediction model to obtain prediction errors of the two models, if the prediction error of the updated soil moisture content data prediction model is smaller than that of the soil moisture content data prediction model, the prediction performance of the updated soil moisture content data prediction model can be considered to be better than that of the soil moisture content data prediction model before updating, and further, historical soil moisture content data and future meteorological environment data acquired each time in a sixth preset time period after updating the soil moisture content data prediction model each time can be input into the updated soil moisture content data prediction model, and future soil moisture content data to be predicted at the time of acquiring the future soil moisture content data output by the updated soil moisture content data prediction model can be predicted by using the updated soil moisture content data prediction model. Compared with the method adopting the soil moisture content data forecasting model before updating, the obtained future soil moisture content data can be more accurate.
If the prediction error of the updated soil moisture content data prediction model is larger than or equal to that of the soil moisture content data prediction model before updating, the prediction performance of the soil moisture content data prediction model before updating is considered to be superior to that of the updated soil moisture content data prediction model, so that the historical soil moisture content data and the future meteorological environment data acquired in the sixth preset time period after updating the soil moisture content data prediction model each time are still adopted to predict the future soil moisture content data, and the accuracy of the obtained future soil moisture content data is ensured.
In the embodiment of the invention, all the soil moisture content data forecasting models before and after updating and the corresponding forecasting errors can be stored.
Fig. 3 is a schematic view of a completion flow of the soil moisture data reconstruction prediction method provided in the embodiment of the present invention, as shown in fig. 3, the method includes:
1) The method comprises the steps of constructing a multi-source data set, namely acquiring second historical soil moisture content influence data and measurement results of the historical soil moisture content data, wherein the second historical soil moisture content influence data are multivariate cooperative data and can comprise data categories such as meteorological environment data, soil property data, time sequence remote sensing monitoring data, geographic information system data and the like, and the measurement results of the historical soil moisture content data can comprise multi-depth historical day-by-day soil moisture content data measurement results.
2) The soil moisture content data reconstruction model and the soil moisture content data forecasting model can be obtained by respectively training by adopting a multi-source data set, and can be dynamically updated respectively.
3) The obtained soil moisture data reconstruction model can be applied to reconstruction of historical soil moisture data of a target virtual site in a target area, and the obtained soil moisture data prediction model can be applied to prediction of future soil moisture data of the target virtual site or a real site in the target area.
4) And (3) fusing the reconstruction result of the historical soil moisture content data of the target virtual site, the future soil moisture content data of the target virtual site, the measurement result of the historical soil moisture content data of the real site and the future soil moisture content data of the real site, so as to obtain the reconstruction and forecast service of the complete regional soil moisture content data of the target region.
In the aspect of data acquisition, the three types of remote sensing auxiliary data, namely multi-depth soil property data, historical day-by-day meteorological environment data and a digital elevation model, related in the embodiment of the invention can be acquired through a Google Earth Engine (GEE) platform. Soil property data the data set provided by EnvirometriX Ltd in 2018 was used. Specifically, the ERA5 Daily Aggregates dataset may be employed by the historical diurnal meteorological environment data. The DEM data can adopt a Multi-Error-Removed Improved-Terrain DEM data set. The huge data resources and online IDE provided by GEE are benefited, the shape layer of a real site is uploaded to a cluster asset, codes are written in a GEE Code Editor by utilizing Python language to realize extraction and export of data, and each kind of data is sequentially exported to a Google Drive platform in a csv format to be stored and downloaded to the local. Further, soil moisture content and auxiliary drawing elements are sequentially introduced into an SQL Server 2016 database by using a Navicat tool, correlation query is carried out on multi-dimensional data by taking a monitering station code as a main key, daily data of different depths are formed, and training of initial models of a decision tree model and a soil moisture content data forecasting model is carried out on the basis of the daily data. In addition, future day-by-day meteorological environment data may be obtained through an Application Programming Interface (API).
In terms of model building, in an embodiment of the present invention, the training environment is a graphics workstation configured as a CPU: intel (R) Xeon (R) CPU E5-1620 v4 @3.50GHz, GPU: NVIDIA Quadro K2200 and RAM:32GB. Model training adopts an Ananconda platform as a model training basic platform, adopts CatBOOST1.0.6 as a model framework, TPE can be realized through a pyOpt toolkit, IDW and other spatial interpolation can be realized through a PySAL toolkit, and the bottom Python version is 3.7.
As shown in fig. 4, on the basis of the above embodiments, an embodiment of the present invention provides a soil moisture data reconstruction system, including:
a first obtaining module 41, configured to obtain first historical soil moisture content influence data of a target virtual site in a target area, where the target virtual site is used to represent a spatial position of reconstructed soil moisture content data;
a data reconstruction module 42, configured to input the first historical soil moisture content influence data into a soil moisture content data reconstruction model, and obtain a historical soil moisture content data reconstruction result of the target virtual station output by the soil moisture content data reconstruction model;
and the soil moisture content data reconstruction model is obtained by training based on second historical soil moisture content influence data of the target real site corresponding to the target virtual site and a historical soil moisture content data measurement result.
On the basis of the above embodiment, in the soil moisture data reconstruction system provided in the embodiment of the present invention, the soil moisture data reconstruction model is determined based on the following method:
training a decision tree model based on second historical soil moisture content influence data of a target real site corresponding to the target virtual site and a measurement result of the historical soil moisture content data, and optimizing hyper-parameters of the decision tree model by adopting a Bayesian algorithm to obtain a soil moisture content data reconstruction model.
On the basis of the above embodiment, the soil moisture data reconstruction system provided in the embodiment of the present invention further includes a site determination module, configured to:
acquiring time sequence satellite remote sensing data of the target area, and constructing a space-time cube based on the time sequence satellite remote sensing data; the time-space cube comprises a plurality of time columns, and each time column corresponds to time sequence data of one pixel position in the time sequence satellite remote sensing data;
selecting any pixel position in the time sequence satellite remote sensing data as a candidate virtual station, determining a target time column corresponding to the candidate virtual station, and calculating the time sequence data correlation between the target time column and other time columns in the space-time cube except the target time column;
and constructing a target plane area based on the pixel positions corresponding to other time columns of which the time sequence data correlation is greater than a first preset threshold, determining the number of real sites in the target plane area, if the number is greater than a second preset threshold, determining that the alternative virtual site is the target virtual site, and determining that the real site is the target real site.
On the basis of the above embodiment, in the soil moisture content data reconstruction system provided in the embodiment of the present invention, the second historical soil moisture content influence data includes historical daily soil moisture content influence data, and the historical soil moisture content data measurement result includes multiple-depth historical daily soil moisture content data measurement results; the soil moisture content data reconstruction model comprises a reconstruction sub-model corresponding to each depth, and the reconstruction sub-model is obtained by training based on the historical day-by-day soil moisture content influence data and the historical day-by-day soil moisture content data measurement result corresponding to the depth;
accordingly, the data reconstruction module is specifically configured to:
and inputting the first historical soil moisture influence data into the reconstruction submodel to obtain historical day-by-day soil moisture data reconstruction results of the target virtual site at the corresponding depth output by the reconstruction submodel.
On the basis of the foregoing embodiment, in the soil moisture data reconstruction system provided in the embodiment of the present invention, the first obtaining module is specifically configured to:
acquiring the first historical soil moisture content influence data every other first preset time period;
accordingly, the data reconstruction module is specifically configured to:
determining a first acquisition moment for acquiring the first historical soil moisture content influence data each time;
and inputting the first historical soil moisture content influence data acquired each time into the soil moisture content data reconstruction model to obtain a historical soil moisture content data reconstruction result of the target virtual site at the first acquisition moment, which is output by the soil moisture content data reconstruction model.
On the basis of the foregoing embodiment, the soil moisture content data reconstruction system provided in the embodiment of the present invention further includes a first model updating module, configured to:
updating the second historical soil moisture content influence data and the historical soil moisture content data measurement result every other second preset time period, and updating the soil moisture content data reconstruction model based on the updated second historical soil moisture content influence data and the updated historical soil moisture content data measurement result to obtain an updated soil moisture content data reconstruction model;
if the reconstruction error of the updated soil moisture content data reconstruction model is smaller than that of the soil moisture content data reconstruction model, the data reconstruction module is specifically used for: inputting the first historical soil moisture influence data acquired each time within a second preset time period after the soil moisture data reconstruction model is updated each time into the updated soil moisture data reconstruction model, and obtaining a historical soil moisture data reconstruction result of the target virtual site at the first acquisition moment output by the updated soil moisture data reconstruction model.
On the basis of the above embodiment, in the soil moisture content data reconstruction system provided in the embodiment of the present invention, the data category of the first historical soil moisture content influence data includes at least one of meteorological environment data, soil property data, time-series remote sensing monitoring data, and geographic information system data.
Specifically, the functions of the modules in the soil moisture data reconstruction system provided in the embodiment of the present invention correspond to the operation flows of the steps in the method embodiments one to one, and the implementation effect is also consistent.
As shown in fig. 5, on the basis of the above embodiments, in an embodiment of the present invention, there is provided a soil moisture content data forecasting system, including:
the second acquiring module 51 is configured to acquire historical soil moisture content data, historical meteorological environment data, and future meteorological environment data of a site to be forecasted;
a data forecasting module 52, configured to input the historical soil moisture content data, the historical meteorological environment data, and the future meteorological environment data into a soil moisture content data forecasting model, so as to obtain future soil moisture content data of the site to be forecasted, which is output by the soil moisture content data forecasting model;
the soil moisture content data forecasting model is obtained by training a future meteorological environment data sample of a sample site in a fourth preset time period and a second historical soil moisture content data sample of the sample site in the fourth preset time period based on a first historical soil moisture content data sample, a historical meteorological environment data sample and a forecast of the sample site in a third preset time period, wherein the sample site comprises a target virtual site or comprises the target virtual site and a real site, and the first historical soil moisture content data sample and the second historical soil moisture content data sample corresponding to the target virtual site are obtained based on the soil moisture content data reconstruction method.
On the basis of the above embodiments, the soil moisture content data forecasting system provided in the embodiments of the present invention includes that the historical soil moisture content data includes historical day-by-day soil moisture content data, the historical meteorological environment data samples include historical day-by-day meteorological environment data samples, the future meteorological environment data includes future day-by-day meteorological environment data, the first historical soil moisture content data sample corresponding to the target virtual site includes a multi-depth first historical day-by-day soil moisture content data sample, the future meteorological environment data sample includes a future day-by-day meteorological environment data sample, and the second historical soil moisture content data sample corresponding to the target virtual site includes a multi-depth second historical day-by-day soil moisture content data sample; the soil moisture content data forecasting model comprises a forecasting sub-model corresponding to each depth, and the forecasting sub-model is obtained by training based on a first historical day-by-day soil moisture content data sample, a future day-by-day meteorological environment data sample and a second historical day-by-day soil moisture content data sample of the corresponding depth;
correspondingly, the data forecasting module is specifically configured to:
and inputting the historical day-by-day soil moisture content data, the historical day-by-day meteorological environment data and the future day-by-day meteorological environment data into the forecasting submodel to obtain the future day-by-day soil moisture content data, at the corresponding depth, of the station to be forecasted, which is output by the forecasting submodel.
On the basis of the foregoing embodiment, in the soil moisture content data forecasting system provided in the embodiment of the present invention, the second obtaining module is specifically configured to:
acquiring the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data every fifth preset time period;
accordingly, the data forecasting module is specifically configured to:
determining a second acquisition moment for acquiring the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data each time;
and inputting the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data which are acquired each time into the soil moisture content data forecasting model to obtain the future soil moisture content data of the site to be forecasted at the second acquisition moment, which is output by the soil moisture content data forecasting model.
On the basis of the above embodiment, the soil moisture content data forecasting system provided in the embodiment of the present invention further includes a second model updating module, configured to:
updating the first historical soil moisture content data sample, the future meteorological environment data sample and the second historical soil moisture content data sample every sixth preset time period, and updating the soil moisture content data forecasting model based on the updated first historical soil moisture content data sample, the updated future meteorological environment data sample and the updated second historical soil moisture content data sample to obtain an updated soil moisture content data forecasting model;
if the prediction error of the updated soil moisture content data prediction model is smaller than that of the soil moisture content data prediction model, the data prediction module is specifically used for: inputting the historical soil moisture content data and the future meteorological environment data acquired each time within a sixth preset time period after the soil moisture content data forecasting model is updated each time into the updated soil moisture content data reconstruction model, and acquiring the future soil moisture content data of the site to be forecasted at the second acquisition time, which is output by the updated soil moisture content data reconstruction model.
Specifically, the functions of the modules in the soil moisture content data forecasting system provided in the embodiment of the present invention correspond to the operation flows of the steps in the embodiments of the methods one to one, and the implementation effects are also consistent.
Fig. 6 illustrates a physical structure diagram of an electronic device, which may include, as shown in fig. 6: a Processor (Processor) 610, a communication Interface (Communications Interface) 620, a Memory (Memory) 630 and a communication bus 640, wherein the Processor 610, the communication Interface 620 and the Memory 630 communicate with each other via the communication bus 640. The processor 610 may invoke logic instructions in the memory 630 to perform the soil moisture data reconstruction methods, or soil moisture data forecasting methods provided in the various embodiments described above.
In addition, the logic instructions in the memory 630 may be implemented in the form of software functional units and stored in a computer readable storage medium when the logic instructions are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, an optical disk, or other various media capable of storing program codes.
In another aspect, the present invention also provides a computer program product, which includes a computer program, the computer program being stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is capable of executing the soil moisture data reconstruction method or the soil moisture data prediction method provided by the above methods.
In yet another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored, the computer program being implemented by a processor to perform the soil moisture data reconstruction method or the soil moisture data forecasting method provided by the above methods.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one position, or may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment may be implemented by software plus a necessary general hardware platform, and may also be implemented by hardware. Based on the understanding, the above technical solutions substantially or otherwise contributing to the prior art may be embodied in the form of a software product, which may be stored in a computer-readable storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method according to the various embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (9)
1. A soil moisture content data reconstruction method is characterized by comprising the following steps:
acquiring first historical soil moisture content influence data of a target virtual site in a target area, wherein the target virtual site is used for representing the spatial position of the reconstructed soil moisture content data;
inputting the first historical soil moisture content influence data into a soil moisture content data reconstruction model to obtain a historical soil moisture content data reconstruction result of the target virtual site output by the soil moisture content data reconstruction model;
the soil moisture content data reconstruction model is obtained by training based on second historical soil moisture content influence data of a target real site corresponding to the target virtual site and a measurement result of the historical soil moisture content data;
the target virtual site and the target real site are determined based on the following method:
acquiring time sequence satellite remote sensing data of the target area, and constructing a space-time cube based on the time sequence satellite remote sensing data; the time-space cube comprises a plurality of time columns, and each time column corresponds to time sequence data of one pixel position in the time sequence satellite remote sensing data;
selecting any pixel position in the time sequence satellite remote sensing data as an alternative virtual site, determining a target time column corresponding to the alternative virtual site, and calculating the time sequence data correlation between the target time column and other time columns in the space-time cube except the target time column;
and constructing a target plane area based on the pixel positions corresponding to other time columns with the time sequence data correlation larger than a first preset threshold, determining the number of real sites in the target plane area, and if the number is larger than a second preset threshold, determining the alternative virtual site as the target virtual site and determining the real site as the target real site.
2. The soil moisture content data reconstruction method according to claim 1, wherein the obtaining of the first historical soil moisture content influence data of the target virtual site in the target area specifically includes:
acquiring the first historical soil moisture content influence data every other first preset time period;
correspondingly, the inputting the first historical soil moisture content influence data into a soil moisture content data reconstruction model to obtain a historical soil moisture content data reconstruction result of the target virtual site output by the soil moisture content data reconstruction model specifically includes:
determining a first acquisition moment for acquiring the first historical soil moisture content influence data each time;
inputting the first historical soil moisture content influence data acquired each time into the soil moisture content data reconstruction model to obtain a historical soil moisture content data reconstruction result of the target virtual site output by the soil moisture content data reconstruction model at the first acquisition moment.
3. The method for reconstructing soil moisture content data according to claim 2, wherein the step of inputting the first historical soil moisture content influence data acquired each time into the soil moisture content data reconstruction model to obtain a historical soil moisture content data reconstruction result of the target virtual site output by the soil moisture content data reconstruction model at the first acquisition time further comprises:
updating the second historical soil moisture content influence data and the historical soil moisture content data measurement result every second preset time period, and updating the soil moisture content data reconstruction model based on the updated second historical soil moisture content influence data and the updated historical soil moisture content data measurement result to obtain an updated soil moisture content data reconstruction model;
and if the reconstruction error of the updated soil moisture content data reconstruction model is smaller than that of the soil moisture content data reconstruction model, inputting the first historical soil moisture content influence data acquired every time in a second preset time period after the soil moisture content data reconstruction model is updated every time into the updated soil moisture content data reconstruction model, and obtaining the historical soil moisture content data reconstruction result of the target virtual site output by the updated soil moisture content data reconstruction model at the first acquisition moment.
4. A soil moisture content data forecasting method is characterized by comprising the following steps:
acquiring historical soil moisture content data, historical meteorological environment data and future meteorological environment data of a site to be forecasted;
inputting the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data into a soil moisture content data forecasting model to obtain the future soil moisture content data of the site to be forecasted, which is output by the soil moisture content data forecasting model;
the soil moisture content data forecasting model is obtained based on a first historical soil moisture content data sample, a historical meteorological environment data sample, a predicted future meteorological environment data sample of a sample site in a fourth preset time period and a second historical soil moisture content data sample of the sample site in the fourth preset time period, wherein the sample site comprises the target virtual site or comprises the target virtual site and the real site, and the first historical soil moisture content data sample and the second historical soil moisture content data sample corresponding to the target virtual site are obtained based on the soil moisture content data reconstruction method as claimed in any one of claims 1 to 3.
5. The method for forecasting soil moisture content data according to claim 4, wherein the acquiring historical soil moisture content data, historical meteorological environment data and future meteorological environment data of the site to be forecasted specifically comprises:
acquiring the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data every fifth preset time period;
correspondingly, the inputting the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data into a soil moisture content data forecasting model to obtain the future soil moisture content data of the site to be forecasted, which is output by the soil moisture content data forecasting model, specifically includes:
determining a second acquisition moment for acquiring the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data each time;
and inputting the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data which are acquired each time into the soil moisture content data forecasting model to obtain the future soil moisture content data of the site to be forecasted at the second acquisition moment, which is output by the soil moisture content data forecasting model.
6. A soil moisture data reconstruction system, comprising:
the system comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring first historical soil moisture content influence data of a target virtual station in a target area, and the target virtual station is used for representing the spatial position of reconstructed soil moisture content data;
the data reconstruction module is used for inputting the first historical soil moisture content influence data into a soil moisture content data reconstruction model to obtain a historical soil moisture content data reconstruction result of the target virtual station output by the soil moisture content data reconstruction model;
the soil moisture content data reconstruction model is obtained by training based on second historical soil moisture content influence data of a target real site corresponding to the target virtual site and a measurement result of the historical soil moisture content data;
the target virtual site and the target real site are determined based on the following method:
acquiring time sequence satellite remote sensing data of the target area, and constructing a space-time cube based on the time sequence satellite remote sensing data; the time-space cube comprises a plurality of time columns, and each time column corresponds to time sequence data of one pixel position in the time sequence satellite remote sensing data;
selecting any pixel position in the time sequence satellite remote sensing data as an alternative virtual site, determining a target time column corresponding to the alternative virtual site, and calculating the time sequence data correlation between the target time column and other time columns in the space-time cube except the target time column;
and constructing a target plane area based on the pixel positions corresponding to other time columns with the time sequence data correlation larger than a first preset threshold, determining the number of real sites in the target plane area, and if the number is larger than a second preset threshold, determining the alternative virtual site as the target virtual site and determining the real site as the target real site.
7. A soil moisture content data forecasting system, comprising:
the second acquisition module is used for acquiring historical soil moisture content data, historical meteorological environment data and future meteorological environment data of the site to be forecasted;
the data forecasting module is used for inputting the historical soil moisture content data, the historical meteorological environment data and the future meteorological environment data into a soil moisture content data forecasting model to obtain the future soil moisture content data of the site to be forecasted, which is output by the soil moisture content data forecasting model;
the soil moisture content data forecasting model is obtained based on a first historical soil moisture content data sample, a historical meteorological environment data sample, a predicted future meteorological environment data sample of a sample site in a fourth preset time period and a second historical soil moisture content data sample of the sample site in the fourth preset time period, wherein the sample site comprises the target virtual site or comprises the target virtual site and the real site, and the first historical soil moisture content data sample and the second historical soil moisture content data sample corresponding to the target virtual site are obtained based on the soil moisture content data reconstruction method as claimed in any one of claims 1 to 3.
8. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the program implements a soil moisture data reconstruction method according to any one of claims 1 to 3 or implements a soil moisture data prediction method according to any one of claims 4 to 5.
9. A non-transitory computer-readable storage medium on which a computer program is stored, wherein the computer program, when executed by a processor, implements the soil moisture data reconstruction method of any one of claims 1 to 3, or implements the soil moisture data forecasting method of any one of claims 4 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210925120.7A CN114971097B (en) | 2022-08-03 | 2022-08-03 | Soil moisture content data reconstruction method and prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210925120.7A CN114971097B (en) | 2022-08-03 | 2022-08-03 | Soil moisture content data reconstruction method and prediction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114971097A CN114971097A (en) | 2022-08-30 |
| CN114971097B true CN114971097B (en) | 2022-11-29 |
Family
ID=82969603
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210925120.7A Active CN114971097B (en) | 2022-08-03 | 2022-08-03 | Soil moisture content data reconstruction method and prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114971097B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105825433A (en) * | 2016-04-01 | 2016-08-03 | 北京邮电大学 | Soil moisture status determining method and apparatus |
| CN110456026A (en) * | 2019-08-13 | 2019-11-15 | 北京农业信息技术研究中心 | A kind of soil moisture content monitoring method and device |
| CN113533695A (en) * | 2021-07-26 | 2021-10-22 | 山东省农业机械科学研究院 | Farmland soil moisture content data estimation method and system |
| CN113591288A (en) * | 2021-07-19 | 2021-11-02 | 杭州领见数字农业科技有限公司 | Soil humidity data prediction method and device based on kriging interpolation |
| CN114723149A (en) * | 2022-04-14 | 2022-07-08 | 北京市农林科学院信息技术研究中心 | Soil moisture content prediction method and device, electronic equipment and storage medium |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018061255A1 (en) * | 2016-09-30 | 2018-04-05 | 日本電気株式会社 | Soil estimation device, soil estimation method, and computer-readable recording medium |
-
2022
- 2022-08-03 CN CN202210925120.7A patent/CN114971097B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105825433A (en) * | 2016-04-01 | 2016-08-03 | 北京邮电大学 | Soil moisture status determining method and apparatus |
| CN110456026A (en) * | 2019-08-13 | 2019-11-15 | 北京农业信息技术研究中心 | A kind of soil moisture content monitoring method and device |
| CN113591288A (en) * | 2021-07-19 | 2021-11-02 | 杭州领见数字农业科技有限公司 | Soil humidity data prediction method and device based on kriging interpolation |
| CN113533695A (en) * | 2021-07-26 | 2021-10-22 | 山东省农业机械科学研究院 | Farmland soil moisture content data estimation method and system |
| CN114723149A (en) * | 2022-04-14 | 2022-07-08 | 北京市农林科学院信息技术研究中心 | Soil moisture content prediction method and device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114971097A (en) | 2022-08-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110751094B (en) | Crop yield estimation method based on GEE comprehensive remote sensing image and deep learning method | |
| CN112446559B (en) | Large-range ground subsidence space-time prediction method and system based on deep learning | |
| CN114254561A (en) | Waterlogging prediction method, waterlogging prediction system and storage medium | |
| CN115077474A (en) | Ground settlement trend prediction method and system based on machine learning | |
| CN115730684A (en) | Air quality detection system based on LSTM-CNN model | |
| CN112668606B (en) | Step type landslide displacement prediction method based on gradient elevator and quadratic programming | |
| CN113902580A (en) | Historical farmland distribution reconstruction method based on random forest model | |
| CN112200354A (en) | Landslide prediction method, device, equipment and storage medium | |
| CN114399073A (en) | Ocean surface temperature field prediction method based on deep learning | |
| CN108733952B (en) | Three-dimensional characterization method for spatial variability of soil water content based on sequential simulation | |
| CN108764527B (en) | Screening method for soil organic carbon library time-space dynamic prediction optimal environment variables | |
| CN114970302A (en) | Regional underground water condition prediction method based on underground water monitoring system | |
| CN116680548A (en) | Time sequence drought causal analysis method for multi-source observation data | |
| CN116738822A (en) | Drainage pipeline maximum corrosion depth prediction method based on LightGBM | |
| CN115345069A (en) | Lake water volume estimation method based on maximum water depth record and machine learning | |
| CN112926251B (en) | Landslide displacement high-precision prediction method based on machine learning | |
| CN114971097B (en) | Soil moisture content data reconstruction method and prediction method | |
| CN111275072B (en) | Mountain area soil thickness prediction method based on clustering sampling | |
| Cao et al. | Probabilistic runoff forecasting considering stepwise decomposition framework and external factor integration structure | |
| CN110909492B (en) | Sewage treatment process soft measurement method based on extreme gradient lifting algorithm | |
| CN117332702A (en) | Waterlogging water depth prediction and multi-factor time sequence analysis method based on ILSTM | |
| CN117405175A (en) | Intelligent marine environment monitoring system | |
| CN116401962A (en) | Method for pushing optimal characteristic scheme of water quality model | |
| CN116796291A (en) | LSTM-MEA-SVR-based air quality forecasting system | |
| CN116183868A (en) | Remote sensing estimation method and system for organic carbon in soil of complex ecological system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |