CN116523673A - Intelligent agricultural system for digital tobacco field - Google Patents
Intelligent agricultural system for digital tobacco field Download PDFInfo
- Publication number
- CN116523673A CN116523673A CN202310415073.6A CN202310415073A CN116523673A CN 116523673 A CN116523673 A CN 116523673A CN 202310415073 A CN202310415073 A CN 202310415073A CN 116523673 A CN116523673 A CN 116523673A
- Authority
- CN
- China
- Prior art keywords
- tobacco
- flue
- module
- sub
- weather
- 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.)
- Pending
Links
- 235000002637 Nicotiana tabacum Nutrition 0.000 title claims abstract description 226
- 241000208125 Nicotiana Species 0.000 title claims abstract description 219
- 238000000034 method Methods 0.000 claims abstract description 33
- 238000004519 manufacturing process Methods 0.000 claims abstract description 28
- 239000002689 soil Substances 0.000 claims abstract description 27
- 230000035558 fertility Effects 0.000 claims abstract description 17
- 238000009313 farming Methods 0.000 claims abstract description 10
- 230000004044 response Effects 0.000 claims abstract description 9
- 244000061176 Nicotiana tabacum Species 0.000 claims abstract description 8
- 238000013528 artificial neural network Methods 0.000 claims description 51
- 238000005516 engineering process Methods 0.000 claims description 22
- IJGRMHOSHXDMSA-UHFFFAOYSA-N Atomic nitrogen Chemical compound N#N IJGRMHOSHXDMSA-UHFFFAOYSA-N 0.000 claims description 18
- 238000001556 precipitation Methods 0.000 claims description 14
- ZLMJMSJWJFRBEC-UHFFFAOYSA-N Potassium Chemical compound [K] ZLMJMSJWJFRBEC-UHFFFAOYSA-N 0.000 claims description 9
- 230000001419 dependent effect Effects 0.000 claims description 9
- 229910052757 nitrogen Inorganic materials 0.000 claims description 9
- 239000011591 potassium Substances 0.000 claims description 9
- 229910052700 potassium Inorganic materials 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 9
- OAICVXFJPJFONN-UHFFFAOYSA-N Phosphorus Chemical compound [P] OAICVXFJPJFONN-UHFFFAOYSA-N 0.000 claims description 7
- 229910052698 phosphorus Inorganic materials 0.000 claims description 7
- 239000011574 phosphorus Substances 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 7
- 230000007774 longterm Effects 0.000 claims description 6
- 238000005904 alkaline hydrolysis reaction Methods 0.000 claims description 4
- 230000004720 fertilization Effects 0.000 claims description 4
- 230000000246 remedial effect Effects 0.000 claims description 4
- 230000009418 agronomic effect Effects 0.000 claims description 3
- 239000005416 organic matter Substances 0.000 claims 1
- 230000006870 function Effects 0.000 abstract description 11
- 230000008569 process Effects 0.000 abstract description 7
- 238000012417 linear regression Methods 0.000 description 16
- 238000007726 management method Methods 0.000 description 16
- 238000007637 random forest analysis Methods 0.000 description 12
- 238000004364 calculation method Methods 0.000 description 8
- 238000004458 analytical method Methods 0.000 description 6
- 238000003066 decision tree Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 239000003337 fertilizer Substances 0.000 description 5
- 238000007689 inspection Methods 0.000 description 5
- 238000012544 monitoring process Methods 0.000 description 5
- 241000196324 Embryophyta Species 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 4
- 238000004140 cleaning Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000003062 neural network model Methods 0.000 description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 4
- 238000010276 construction Methods 0.000 description 3
- 230000007423 decrease Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 201000010099 disease Diseases 0.000 description 2
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 2
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012876 topography Methods 0.000 description 2
- 241000238631 Hexapoda Species 0.000 description 1
- 241000377350 Puan Species 0.000 description 1
- 239000008619 Xingren Substances 0.000 description 1
- 241000607479 Yersinia pestis Species 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000004027 cell Anatomy 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000007599 discharging Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000003306 harvesting Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 235000015097 nutrients Nutrition 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 238000013138 pruning Methods 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 238000004171 remote diagnosis Methods 0.000 description 1
- 239000002881 soil fertilizer Substances 0.000 description 1
- 238000005527 soil sampling Methods 0.000 description 1
- 238000000528 statistical test Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
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
- 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
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- 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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- General Health & Medical Sciences (AREA)
- Development Economics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Game Theory and Decision Science (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Agronomy & Crop Science (AREA)
- Marine Sciences & Fisheries (AREA)
- Mining & Mineral Resources (AREA)
- Artificial Intelligence (AREA)
- Primary Health Care (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Animal Husbandry (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Educational Administration (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a digital tobacco field intelligent agricultural system, which comprises a flue-cured tobacco planting information module, a meteorological big data module and a production suggestion module, wherein the flue-cured tobacco planting information module is used for collecting flue-cured tobacco plots, flue-cured tobacco plant numbers, house owner information, curing barn information and seedling shed information, and simultaneously collecting soil fertility data of each flue-cured tobacco plot according to grids, and superposing the soil fertility data on a grid digital base map of flue-cured tobacco planting layout; the weather big data module comprises a tobacco weather live sub-module, a real-time tobacco weather early warning sub-module and a tobacco weather forecast sub-module, wherein the tobacco weather live sub-module is used for displaying tobacco weather live data of each grid point on the gridding digital base map; the production suggestion module comprises a disaster response sub-module, a production technical scheme sub-module, a farming reminding sub-module and a tobacco yield forecasting sub-module. The invention can realize the accurate matching of meteorological and soil data to land parcels in the tobacco planting process, and is convenient for realizing and managing various functions in production.
Description
Technical Field
The invention relates to a digital tobacco field intelligent agricultural system, and belongs to the technical field of tobacco field management.
Background
The intelligent agriculture combines modern science and technology with agricultural planting, thereby realizing unmanned, automatic and intelligent management, for example, combining integrated application computer and network technology, internet of things technology, audio and video technology, 3S technology, wireless communication technology and expert intelligence with crop planting, and realizing intelligent management means such as agricultural visual remote diagnosis, remote control, disaster early warning and the like.
In recent years, the intelligent weather service becomes a development trend of the weather service in China, the concept is that the weather service system is 'based on people, everywhere and everywhere', and is combined with the geospatial information technology represented by the 3s technology, so that the weather service system is a system with self-perception, judgment, selection, action, innovation and self-adaptation capability, and the effect of the weather information in guiding people to generate life and promoting economic and social development is better exerted, so that the weather service is refined, specialized and personalized.
However, the flue-cured tobacco planting areas are widely and dispersedly widespread, such as Guizhou, and the flue-cured tobacco planting areas are complicated and changeable in climate environment due to unique karst landforms, broken land plots, large relief fluctuation and planting dispersion, so that the accuracy and timeliness of tobacco field management according to common climate are poor, and the fine management requirement of the tobacco field is difficult to meet. For example, timely and accurate early warning of tobacco weather, production advice, and the like cannot be accurately achieved.
Disclosure of Invention
Based on the above, the invention provides a digital tobacco field intelligent agricultural system, which aims at the problems of low refinement degree, low specificity, low pertinence and the like of the special meteorological service of tobacco leaves, realizes the accurate matching of meteorological and soil data to land parcels in the planting process, reasonably plans the operation starting time of key agriculture according to the current meteorological conditions, effectively utilizes the climate resources, avoids the influence of adverse meteorological conditions, and scientifically organizes and formulates flue-cured tobacco production.
The technical scheme of the invention is as follows: the intelligent digital tobacco field agriculture system comprises a flue-cured tobacco planting information module, a meteorological big data module and a production suggestion module, wherein,
the flue-cured tobacco planting information module is used for collecting flue-cured tobacco plots, flue-cured tobacco plants, house owner information, curing barn information and seedling shed information, and simultaneously collecting soil fertility data of each flue-cured tobacco plot according to grids, and superposing the soil fertility data on a grid digital base map of flue-cured tobacco planting layout;
the weather big data module comprises a tobacco weather live sub-module, a real-time tobacco weather early-warning sub-module and a tobacco weather forecast sub-module, wherein the tobacco weather live sub-module is used for displaying tobacco weather live data of each grid point on the gridding digital base map, the tobacco weather early-warning sub-module is used for carrying out weather disaster early warning on each flue-cured tobacco plot according to the tobacco weather live data, and the tobacco weather forecast sub-module is used for displaying future weather data of each grid point on the gridding digital base map;
The production suggestion module comprises a disaster response sub-module, a production technology scheme sub-module, a farming reminding sub-module and a tobacco yield forecasting sub-module, wherein the disaster response sub-module is used for providing corresponding treatment measures according to weather disaster early warning, the production technology scheme sub-module is used for providing production suggestions according to soil fertility data and tobacco weather live data, the farming reminding sub-module is used for displaying the current key farming stage, and the tobacco yield forecasting sub-module is used for forecasting the yield of a flue-cured tobacco land according to weather data.
Preferably, the grid digital base map is divided according to a map of a flue-cured tobacco planting area with the size of 1 multiplied by 1, and the soil fertility data comprises organic matters, alkaline hydrolysis nitrogen, quick-acting phosphorus, quick-acting potassium and soil pH information.
Preferably, the method for displaying the tobacco weather live data and the future weather data of each grid point on the gridding digital base chart is as follows:
acquiring original meteorological element data to form a meteorological element product with 5km resolution;
interpolation is carried out on weather element products with 5km resolution to grids with 1km resolution by using an inverse distance weight method, and then regression models are utilized to correct the forecast values after interpolation, so that weather elements with 1km resolution are obtained;
And (3) interpolating the grid data with the resolution of 1km to the geometric center position of each grid on the grid digital base map by using a bilinear interpolation method, and accurately matching the meteorological elements to each grid.
Preferably, the disaster response submodule comprises a disaster grade determining submodule and a disaster matching measure submodule, wherein the disaster grade determining submodule is used for reading meteorological elements according to each flue-cured tobacco land block and determining whether a tobacco field is in the disasters and the grades of the disasters according to the related standards for drought, hail disaster, wind disaster and waterlogging in the standard of QX/T363-2016 flue-cured tobacco meteorological disaster grade; the disaster matching measure submodule is used for matching disaster countermeasures and remedial measures after the flue-cured tobacco land disaster and the grade are determined.
Preferably, the production technical scheme submodule comprises a fertilization recommendation submodule, a transplanting period recommendation submodule, a topdressing time recommendation submodule, a worthless foot leaf processing time submodule, a topping time recommendation submodule and a picking and baking period recommendation submodule.
Preferably, the processing steps of the agricultural reminding sub-module are as follows: and reading the current time, and comparing the current time with the time period in the production overall technical scheme to display the current key agronomic stage.
Preferably, the processing steps of the tobacco yield forecasting submodule are as follows:
calling a corresponding flue-cured tobacco yield prediction model according to the position information of the flue-cured tobacco land block;
and acquiring meteorological data corresponding to the flue-cured tobacco land, inputting the meteorological data into the flue-cured tobacco yield prediction model, and calculating the tobacco yield corresponding to the flue-cured tobacco land.
Preferably, the flue-cured tobacco yield prediction model comprises a first BP neural network prediction model and a second BP neural network prediction model, when the flue-cured tobacco plot is in a first preset area, the first BP neural network prediction model is called as the flue-cured tobacco yield prediction model, and when the flue-cured tobacco plot is in a second preset area, the second BP neural network prediction model is called as the flue-cured tobacco yield prediction model.
Preferably, the method for constructing the first BP neural network prediction model is as follows:
collecting the historical tobacco leaf yield of the flue-cured tobacco plots as a first dependent variable;
collecting first historical meteorological data of a flue-cured tobacco land as a first independent variable, wherein the first historical meteorological data comprises average air temperature, average highest air temperature, average lowest air temperature, sunshine hours and precipitation amount from 3 to 9 months to ten days;
And training the BP neural network by taking the first independent variable as a network input value and taking the first independent variable as a network predicted value to obtain a first BP neural network predicted model.
Preferably, the construction method of the second BP neural network prediction model is as follows:
collecting the historical tobacco leaf yield of the flue-cured tobacco plots as a second dependent variable;
collecting second historical meteorological data of a flue-cured tobacco plot as a second independent variable, wherein the second historical meteorological data comprises air temperature in a tobacco maturity stage, rainfall capacity in a vigorous and long-term period, a time of Tian Qiri hours and available time in a field growth period;
and training the BP neural network by taking the second independent variable as a network input value and taking the second independent variable as a network predicted value to obtain a second BP neural network predicted model.
The invention has the beneficial effects that: according to the invention, the flue-cured tobacco planting layout is divided into the grid digital base map, the collected flue-cured tobacco plots, flue-cured tobacco plants, house owner information, curing barn information and seedling shed information and soil fertility data are superimposed on the grid digital base map, and then the meteorological data are interpolated into each grid point of the grid digital base map through a downscaling technology, so that the meteorological and soil data are accurately matched to the plots in the tobacco planting process, and the functions of subsequent disaster coping, production technical scheme, agricultural reminding, tobacco yield forecasting and the like are accurately and timely realized.
Drawings
FIG. 1 is a functional block diagram of a digital tobacco field intelligent agricultural system;
the spatial calculation schematic of fig. 2 z;
FIG. 3 is a schematic diagram of the adjustment of weighted exponentiations;
FIG. 4 w (d) j ) Is a schematic image of (1);
fig. 5 is a schematic diagram of the principle of bilinear interpolation.
FIG. 6BP neural network topology;
FIG. 7 is a schematic diagram of a random forest algorithm;
FIG. 8 Guizhou province yield (left), and meteorological element (right) change from year to year;
FIG. 9 shows a ten-day-by-ten maximum positive (negative) correlation meteorological element profile for each leaf yield (upper: west, lower: middle east, 0 indicating no significant verification);
FIG. 10 regression model yield forecast accuracy for each leaf position;
FIG. 11 linear regression model yield forecast accuracy (green: highest, red: lowest) for each leaf position;
FIG. 12 shows the yield forecast accuracy of each leaf position of the BP neural network model;
FIG. 13BP neural network model each leaf position yield forecast accuracy (green: highest, red: lowest);
FIG. 14 shows the yield forecast accuracy of each leaf position of the random forest model;
FIG. 15 random forest model leaf yield forecast accuracy (green: highest, red: lowest);
the error of each predictive model of FIG. 16 is compared to the percent error (A, B: middle east, C, D: west, A, C: error, B, D: percent error).
Detailed Description
In order that the above objects, features and advantages of the invention will be readily understood, a more particular description of the invention will be rendered by reference to the appended drawings. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. The invention may be embodied in many other forms than described herein and similarly modified by those skilled in the art without departing from the spirit or scope of the invention, which is therefore not limited to the specific embodiments disclosed below.
Referring to fig. 1, the digital tobacco field intelligent agriculture system according to the embodiment of the invention comprises a flue-cured tobacco planting information module, a meteorological big data module, a production suggestion module and a background management module.
The flue-cured tobacco planting information module is used for collecting flue-cured tobacco plots, flue-cured tobacco plant numbers, house owner information, curing barn information and seedling shed information, and simultaneously collecting soil fertility data of each flue-cured tobacco plot according to grids, and superposing the soil fertility data on a grid digital base map of the flue-cured tobacco planting layout.
Flue-cured tobacco plots: actually measuring the number, area, distribution information, nitrogen, phosphorus and potassium content of each tobacco field, soil nutrient index and other information of three-stage flue-cured tobacco plots in province, city and county. Flue-cured tobacco plant number: and actually measuring the variety, plant number, plant spacing and other information of each tobacco field. Home owner information: and collecting basic information of the householder for planting the flue-cured tobacco. Curing barn information: actually measured flue-cured tobacco group and flue-cured tobacco house number information. Seedling shed information: actual measurement of the number of seedling raising sheds, the in-use area, the idle area, the damaged area and the like
Specifically, a remote sensing technology is combined with a geographic information technology to collect data such as flue-cured tobacco plots, flue-cured tobacco plants, house owner information, curing barn information, seedling shed information and the like, establish a perfect flue-cured tobacco planting digital layout and form a grid digital base map. In the embodiment, the grid digital base map is divided into a map of a flue-cured tobacco planting area according to 1×1, and soil fertility data comprises organic matters, alkaline hydrolysis nitrogen, quick-acting phosphorus, quick-acting potassium and soil pH information. Soil sampling is carried out according to grids of 1X 1 km, soil fertilizer data such as organic matters, alkaline hydrolysis nitrogen, quick-acting phosphorus, quick-acting potassium, soil pH and the like are measured, a tobacco planting soil fertility database is further constructed, and the tobacco planting soil fertility database is superimposed on a grid digital base map in a flue-cured tobacco planting layout.
The crop monitoring technology based on the remote sensing technology is a mature technology, and has early development and wide application in foreign countries. In large-area and flaky planting areas, the information monitoring and information management of the distribution condition, growth vigor, area and the like of crops by means of a remote sensing technology can be basically realized. In northern areas with flatter domestic topography, the method has more applications in remote sensing and monitoring of crops, and utilizes remote sensing to monitor a planting database for building crops and digital plots. The Guizhou is difficult to analyze the crop planting condition by remote sensing technology alone because of unique karst landform, broken land, large relief fluctuation and scattered planting, and the application is not wide. In consideration of the specificity of Guizhou terrain, a tobacco field basic database is established through early acquisition and correction in a mode of field acquisition, remote sensing correction and information system management, and is supplemented and modified on the basis of the database in the later stage, and management and inquiry are carried out through app and webpage application. Is an effective attempt and innovation of combining remote sensing monitoring with informatization and Guizhou terrain characteristics.
The weather big data module comprises a tobacco weather live sub-module, a real-time tobacco weather early-warning sub-module and a tobacco weather forecast sub-module, wherein the tobacco weather live sub-module is used for displaying tobacco weather live data of each grid point on the gridding digital base map, the tobacco weather early-warning sub-module is used for carrying out weather disaster early warning on each flue-cured tobacco plot according to the tobacco weather live data, and the tobacco weather forecast sub-module is used for displaying future weather data of each grid point on the gridding digital base map.
Specifically, the method of displaying the tobacco weather live data and future weather data for each grid point on the gridded digital base map is as follows:
step 1: the method comprises the steps of obtaining data of original meteorological elements, such as air temperature, rainfall, sunshine hours and the like of a monitoring station of a meteorological department, carrying out quality control and data cleaning on the data of the original meteorological elements, fusing station meteorological element products and satellite data, and correcting grid point data products inverted by high-resolution satellites by using the station data to form 5KM resolution meteorological element products.
Step 2: and (3) interpolating the weather element product with the resolution of 5KM formed in the step (1) onto a grid with the resolution of 1KM in Guizhou province by using an inverse distance weight method, and correcting the interpolated forecast value by using a regression model to obtain the weather element with the resolution of 1 KM. The formula is as follows:
1) Inverse distance weighting method
The inverse distance weighting method is also called 'inverse distance weighted interpolation' or 'shebard method'.
Is provided with n points, and the plane coordinate is (x i ,y i ) The vertical height is z i I=1, 2, …, n, the interpolation function of reciprocal distance weighted interpolation is:
wherein, the liquid crystal display device comprises a liquid crystal display device,is (x, y) point to (x) j ,y j ) Horizontal distance of points, j=1, 2, …, n. p is a constant greater than 0 and is referred to as a weighted power exponent.
It is easy to see that,is z 1 ,z 2 ,…,z n Is schematically shown in figure 2.
f (x, y) is expressed in a piecewise expression, appears discontinuous, and is virtually everywhere continuous.
Therefore, f (x, y) is in (x) i ,y i ) Continuous.
The weighted power exponent p may adjust the shape of the interpolation function surface. The larger p is, the flatter the function surface is at the node; the smaller p, the sharper the function surface at the node. A schematic of the adjustment of the weighted exponentiation is shown in fig. 3.
The reciprocal distance weighted interpolation has the advantages that: the formula is simpler, and is particularly suitable for the problem that nodes are scattered and are not grid points. Its disadvantages are: the maximum and minimum values of the function can only be taken at the nodes (since this interpolation is a weighted average of the values at each node).
When the number of nodes is relatively large, the calculation workload of reciprocal distance weighted interpolation is relatively large, and the interpolation formula can be simplified as follows:
Wherein the method comprises the steps of
As shown in fig. 4, whenWhen w (d) j ) Is a hyperbola (i.e., the reciprocal distance weighted interpolation formula where p=1 is the original); when->When w (d) j ) Is a segment of a parabola; when d j At > R, w (d j ) For nodes with distance greater than R, calculation is unnecessary, and the calculation workload is greatly reduced.
The weather service works, i.e. it is assumed that n known sample points all have a certain influence on the prediction of the predicted point value, and that their influence decreases with increasing distance. The principle is that the attribute value of the point to be interpolated is a weighted average of the attribute values of the influence area of the point to be interpolated, and the weight is related to the distance between the point to be interpolated and the point in the influence area. The formula is as follows:
wherein: f (x, y) is interpolation, w (d) j ) In order to calculate the weight of the lattice point/site, Z is the element value of the estimated point, zi is the element value on the ith site, di is the distance from the interpolation point to the ith point, and n is the total number of the interpolation lattice points.
2) The formula is corrected.
And establishing a corresponding relation between a forecast value for forecasting time and an observation value of the measuring station on each station after interpolation, and correcting errors caused by mode interpolation. The specific operation is as follows:
Y i ＝aX i +b
wherein X is i Is a time sequence after mode forecast interpolation, Y i Is a time series of observations of the corresponding station. And determining regression coefficients a and b in the training period, so as to establish a statistical downscaling relation from the forecasting mode to the forecasting of the observation station.
Step 3: and (3) interpolating the grid data with the resolution of 1KM obtained in the step (2) to the geometric center position of the tobacco land by using a linear interpolation mode, and accurately matching the meteorological elements to the tobacco land.
The formula for bilinear interpolation is as follows:
R 1 ＝(x,y 1 )
R 2 ＝(x,y 2 )
as shown in fig. 5, let interpolation to P point be assumed, where x is the longitude of P point, y is the latitude of P point, x1, x2 are the longitudes of grid points around P point, y1, y2 are the latitudes of grid points around P point, f (Q) 11 )，f(Q 21 )，f(Q 12 )，f(Q 22 ) Is the meteorological element value of four surrounding grid points, R 1 And R is 2 The weather element values of two points on the grid line are calculated as intermediate variables f (R1) and f (R2) on the same longitude as the point P, and are used for calculating f (P).
After the live data of the tobacco weather are obtained, tobacco weather disasters such as drought, hail, wind disasters, waterlogging and the like can be alarmed on each flue-cured tobacco land.
The production suggestion module comprises a disaster response sub-module, a production technical scheme sub-module, a farming reminding sub-module and a tobacco yield forecasting sub-module, wherein the disaster response sub-module is used for providing corresponding treatment measures according to weather disaster early warning, the production technical scheme sub-module is used for providing production suggestions according to soil fertility data and tobacco weather live data, the farming reminding sub-module is used for displaying the current key farming stage, and the tobacco yield forecasting sub-module is used for forecasting the yield of the flue-cured tobacco land according to weather data.
Specifically, the disaster response submodule comprises a disaster grade determining submodule and a disaster matching measure submodule, wherein the disaster grade determining submodule is used for reading meteorological elements according to each flue-cured tobacco land block and determining whether a tobacco field is in the disasters and the grades of the disasters according to the related standards for drought, hail disaster, wind disaster and waterlogging in QX/T363-2016 flue-cured tobacco meteorological disaster grade standards; the disaster matching measure submodule is used for matching disaster countermeasures and remedial measures after the flue-cured tobacco land disaster and the grade are determined. Countermeasures such as waterlogging are: firstly, tobacco field planning is performed, moderate land circulation is performed, and flue-cured tobacco is forbidden to be planted in low-lying and waterlogged fields; secondly, cleaning and deep digging of flood discharging channels of large tobacco fields are made in advance; thirdly, ridging along the height direction of the field topography, paying attention to the cleaning of furrows, and ensuring that rainwater can be directly discharged outwards along the furrows; fourthly, the surrounding ditch of each tobacco field is required to be lower than the waist ditch, and the waist ditch is lower than the furrow, so that smooth drainage in the field is ensured, and water is stopped as much as possible; fifthly, the cleaning of weeds in the ditches in the waterlogging period is paid attention to, and smooth drainage in the ditches is ensured. The water and soil loss problem is to be noted in the drainage of the tobacco field in the sloping field, and the ridge direction or the ditch direction of the tobacco field with the steep slope should be perpendicular to the slope. The remedial measures are: grasping, draining, cultivating and ridging in time, dressing fertilizer and preventing and controlling rhizome diseases
The production technology scheme sub-module comprises a fertilization recommendation sub-module, a transplanting period recommendation sub-module, a topdressing time recommendation sub-module, a worthless foot leaf processing time sub-module, a topping time recommendation sub-module and a picking and baking period recommendation sub-module. The calculation of each sub-module is as follows:
and (3) recommending fertilization scheme calculation, namely firstly calculating the average nitrogen, phosphorus and potassium content of the county where the tobacco field is located, displaying the average nitrogen, phosphorus and potassium content at the left upper corner of the page of the county in the district, and for a single land, increasing or decreasing the data amount of the base fertilizer by = (the nitrogen content of the current land-the nitrogen average value of the current county)/0.1. Data amount of additional fertilizer increase or decrease= (current potassium content of land block-current average value of county potassium)/0.3, and after calculating the increase and decrease amount of base fertilizer and additional fertilizer, displaying on a webpage.
Calculating a recommended transplanting period: for the first day, in which the average air temperature is greater than 15 ℃ for two consecutive days, the average low temperature at night is not less than 10 ℃ and the relative humidity is greater than 50%, calculating the recommended transplanting period in the future 3 days and 9 days respectively. The latter time was calculated over 3 days.
And (5) calculating recommended topdressing time: 10 days after the recommended transplanting period.
The treatment time of the worthless foot leaves; 45-55 days after the recommended transplanting period.
Recommending topping time: the transplanting period is recommended to be 65-72 days later, and the period is 2-3 days later without rain.
Recommended harvest period: 72-79 days after the transplanting period and no-rain date are recommended.
The processing steps of the agricultural reminding sub-module are as follows: and reading the current time, and comparing the current time with the time period in the production overall technical scheme to display the current key agronomic stage. And when the soil humidity is more than 60%, the rest suitable time periods are days without rain in the future 7-day forecast. And mounting the latest plant diseases and insect pests forecasting products on the page.
The tobacco yield forecasting sub-module comprises the following processing steps:
calling a corresponding flue-cured tobacco yield prediction model according to the position information of the flue-cured tobacco land block;
and acquiring meteorological data corresponding to the flue-cured tobacco land, inputting the meteorological data into the flue-cured tobacco yield prediction model, and calculating the tobacco yield corresponding to the flue-cured tobacco land.
Specifically, on the basis that the meteorological data is reduced in scale and matched with tobacco field plots, the corresponding meteorological data is input into a flue-cured tobacco yield prediction model to predict tobacco yield.
The flue-cured tobacco yield prediction model comprises a first BP neural network prediction model and a second BP neural network prediction model. And when the tobacco region is in the first preset region, calling a first BP neural network prediction model as a flue-cured tobacco yield prediction model. And when the tobacco region is in the second preset region, calling a second BP neural network prediction model as a flue-cured tobacco yield prediction model.
Specifically, the construction method of the first BP neural network prediction model is as follows:
collecting the historical tobacco leaf yield of the flue-cured tobacco plots as a first dependent variable;
collecting first historical meteorological data of a flue-cured tobacco land as a first independent variable, wherein the first historical meteorological data comprises average air temperature, average highest air temperature, average lowest air temperature, sunshine hours and precipitation amount from 3 to 9 months to ten days;
and training the BP neural network by taking the first independent variable as a network input value and taking the first independent variable as a network predicted value to obtain a first BP neural network predicted model.
During prediction, the average air temperature, the average highest air temperature, the average lowest air temperature, the sunshine hours and the precipitation amount of the flue-cured tobacco plots are obtained as inputs from 3 to 9 months of the current year, and the predicted tobacco yield can be obtained.
The construction method of the second BP neural network prediction model comprises the following steps:
collecting the historical tobacco leaf yield of the flue-cured tobacco plots as a second dependent variable;
collecting second historical meteorological data of the flue-cured tobacco land as a second independent variable, wherein the second historical meteorological data comprises the air temperature in the tobacco maturity stage, the rainfall capacity in the vigorous long-term, the time of the large Tian Qiri hours and the available time in the field growth period;
and training the BP neural network by taking the second independent variable as a network input value and taking the second independent variable as a network predicted value to obtain a second BP neural network predicted model.
In the prediction process, the temperature of the tobacco in the current year of the flue-cured tobacco field in the maturity stage, the rainfall in the vigorous and long-term stage, the time of the large Tian Qiri hours and the available time in the field growth period are obtained as inputs, and the predicted tobacco yield can be obtained.
The following details the research process of the above prediction model:
1. data material
The yield data is the average single leaf weight of the lower leaf, the middle leaf and the upper leaf of the region where tobacco leaves are planted in Guizhou province from 2010-2021 year district county. The weather data is selected from the average air temperature, the average highest air temperature, the average lowest air temperature, the sunshine hours and the precipitation amount of the weather data from 2010 to 2021 in 3 to 9 months. The tobacco single leaf weight data and the meteorological data of the same county and city are corresponding, the meteorological element is taken as an independent variable, and the tobacco single leaf weight data is taken as an independent variable. And simultaneously, four tobacco climate suitability evaluation indexes in China tobacco planting division are used as a second group of independent variables, wherein the indexes are respectively the mature period air temperature (7, 8 months average air temperature), the vigorous long-term rainfall (6 months rainfall), the accumulated sunshine hours in a period of Tian Qiri months (5, 6, 7, 8) and the available time in the field growth period (the days between the initial days when the average day after transplanting is more than 13 ℃ and the final days when the average day is more than or equal to 18 ℃).
The Guizhou provinces can also be divided into western (Anlong, chapter, nayon, panzhou, puan, qinglong, shuiyuan, weining, xingyi, xingren, zhenfeng) and middle eastern (all tobacco planting counties) from climate conditions and the type of tobacco planted. Analysis was performed from two major divisions and nine local states.
When a forecast model is built, 60-70% of data are randomly screened to be used for building the model, and the rest data are used for inspection.
2. Model building algorithm
(1) BP neural network
The BP neural network has self-learning and self-organizing nonlinear mapping capability, is suitable for model establishment of problems of unclear knowledge background, complex information and ambiguous reasoning rules, is not only a method for artificial intelligence research, but also a mathematical model, can be simulated by a computer program, can be effectively applied to the identification and control of a nonlinear system, and does not depend on functions of the model. The BP neural network is a multilayer feedforward neural network, and the main characteristic of the network is that signals are transmitted forward and errors are propagated backward. In forward pass, the input signal is processed layer by layer from the input layer through the hidden layer to the output layer. The neuron state value of each layer affects the next layer of neuron state. If the output layer does not need to output, the back propagation is carried out, and the network weight and the threshold value are adjusted according to the prediction error, so that the predicted output of the BP neural network is enabled to be continuously approximate to the expected output.
The topology of the BP neural network is shown in FIG. 6, in which X 1 ,X 2 ，…，X n Is the input value of BP neural network, Y 1 ,Y 2 ，…，Y m Is the predicted value of BP neural network, and ωij and ωjk are the weight of BP neural network. From the figure, the BP neural network can be seen as a nonlinear function, and the network input value and the predicted value are independent variables and dependent variables of the function respectively. When the number of input nodes is n and the number of output nodes is m, the BP neural network expresses a function mapping relation from n independent variables to m dependent variables.
(2) Random forest principle
The random forest is an integrated algorithm composed of a plurality of decision trees, belongs to a subclass of integrated learning, can be used for classification and regression problems, and mainly samples sample units and variables so as to generate a large number of decision trees. For each sample unit, all decision trees classify the sample units in turn, and the mode in the prediction category is the category to which the sample unit predicted by the random forest belongs (when used for regression, the output is the average value of all tree prediction values).
Assuming that there are N sample cells in total and M feature attributes in a given sample set X, the random forest algorithm when used for regression problems is approximately as follows: (1) q samples are randomly extracted from a given sample set X by adopting a Bootstrap method and put back, so that a decision tree is generated; (2) randomly extracting M features (M < M) from each node, and taking the M features as candidate features for dividing the node, wherein the feature numbers at each node are consistent; (3) all decision trees are completely generated without pruning (the minimum node is 1); (4) and for the new sample point, predicting the new sample point by using all trees, and taking the average value of the predicted values of all trees as the final predicted value of the model.
When prediction is carried out, node splitting is carried out for 2 random processes by introducing random extraction subsamples and randomly selected characteristic factors, so that the correlation among regression trees is reduced, and further the generalization error of a random forest regression model is reduced. The probability of each sample not being decimated is 1-1N () N, which tends to be 0.368 when N is sufficiently large, i.e., about 37% of the samples that would not be decimated, which would constitute out-of-bag data (OOB), the model can be adjusted by observing the error of the out-of-bag data.
(3) Principle of linear regression:
since the change in socioeconomic phenomenon is often affected by a plurality of factors, multiple regression analysis is generally performed, and regression including two or more independent variables is called multiple linear regression. The basic principle and the basic calculation process of the multiple linear regression are the same as those of the single linear regression, but the calculation is quite troublesome due to the large number of independent variables, and statistical software is generally used in actual application. The multiple linear regression is similar to the unitary linear regression, the model parameters can be estimated by using a least square method, the model and the model parameters are also required to be subjected to statistical test, the selection of proper independent variables is one of the preconditions of correctly performing multiple regression prediction, and the selection of the independent variables of the multiple regression model can be solved by using a correlation matrix among the variables.
3. Inspection method
The inspection method is mainly two, namely error and error percentage, and a model with a good forecasting effect is selected for forecasting. Wherein the error refers to the use of the forecast value minus the live value as follows:
E a ＝F i -O i
where Fi is the i-th day yield forecast and Oi is the i-th day yield live.
The error percentage is the error divided by the live condition, the error of the forecast yield and the live yield is defined to be forecast accuracy within +/-20 percent, the ratio of the forecast accuracy is calculated to be the forecast accuracy, and the formula is as follows:
wherein Nr K Correct for index forecastNumber of Nf K Is the number of samples. The actual meaning of the index forecast accuracy is the percentage of the index forecast error less than or equal to 10%.
4. Yield and meteorological element change analysis
(1) Annual changes in yield and meteorological elements
From the average yield change graphs (left graph of fig. 8) of Guizhou province from 2010 to 2021, it can be seen that the yields of both the upper and lower leaves are increasing, the single leaf weight of the overall province average lower leaf is increased from 5.14g in 2010 to 6.62g in 2021, the single leaf weight of the upper leaf is increased from 9.56g in 2010 to 11.7g in 2021, the trend of the average single leaf weight increase of the upper leaf is more obvious, the single leaf weight of the middle leaf is basically the trend of keeping unchanged, and the average value is 9.34g. The variation trend of the eastern and western regions is consistent with that of the whole province, and the variation trend is shown that the single leaf weight of the middle leaf is not greatly changed, the lower leaf and the upper leaf are weight-increasing trends, and the weight increase of the upper leaf is obvious compared with that of the lower leaf (not shown).
The average air temperature in the last 10 years is basically unchanged from the key meteorological elements (right diagram of fig. 8) of the tobacco county planted in the whole province from 2010 to 2021, such as precipitation, air temperature and sunshine hours, and is maintained at about 20 ℃, the average highest air temperature in the year is about 25 ℃, and the average lowest air temperature in the year is about 17 ℃. The precipitation amount and the sunshine duration are inversely related, 2011 is the year with the least precipitation in the last 10 years, the annual precipitation amount is only 583mm,2014 is the year with the most precipitation in the last 10 years, and the annual precipitation amount is 1178mm.
(2) Yield and meteorological element correlation analysis
By using the correlation coefficient to analyze the correlation between the yield and the single weather element every ten days (table 1), it can be seen that the correlation coefficient of the eastern part of the middle part passes the significance test, each leaf condition is equivalent, and the number of ten days passing the significance test in the western part is less, which is only 54.2%. From the correlation coefficient R, the maximum positive correlation is only 0.395, the maximum negative correlation is only-0.323, and the correlation between single weather elements and the yield of tobacco leaves is proved to be certain, but the correlation coefficient is not high. From the results of the significance test, the element with the largest positive and negative correlation with the three leaf positions and the time period (table 1 and fig. 9) are found, and it is known that the sunlight is the meteorological element with the largest positive influence on the yield of each leaf position, and mainly concentrated in the middle ten days of 6 months, the precipitation of the middle ten days of 9 months and the sunlight of the middle ten days of 4 months have the largest positive influence on the yields of the upper eastern leaf and the middle western leaf respectively; the weather elements and the period of time are relatively complex, and the accumulated temperature, sunlight and precipitation before 7 months and the air temperature after 6 months can have negative influence on the yield of tobacco leaves.
TABLE 1 analysis of correlation conditions of single leaf weight and ten-day-by-ten-day single weather element
5. Modeling result analysis
The correlation between the tobacco leaf yield and the single meteorological element is not good, but still has certain correlation, and in order to know the effect of forecasting the tobacco leaf yield by a plurality of meteorological elements, the following input historical meteorological elements are divided into two types: four elements, namely the air temperature in the mature period of tobacco leaves, the rain amount in a vigorous long-term, the number of times of the large Tian Qiri hours and the available period in the field period; multiple elements, namely, average air temperature, average minimum air temperature, average maximum air temperature, precipitation, sunshine hours and effective accumulated temperature every ten days in 3-9 months. And respectively adopting four methods of linear regression, stepwise regression, BP neural network and random forest to establish a tobacco leaf position yield prediction model, and exploring an optimal prediction model and accuracy thereof.
(1) Regression model modeling results
As can be seen by comparing the prediction accuracy of the linear regression model with the prediction accuracy of the stepwise regression model (fig. 10), the prediction accuracy of the stepwise regression model is better than that of the linear regression model, especially when the prediction is performed by adopting multiple elements in the western region, but when the prediction is performed by adopting 4 elements in the western region, the stepwise regression models of the middle and upper leaves cannot be smoothly established because the correlation between independent variables and dependent variables is poor, and a mathematical model cannot be formed; for middle eastern regions, the accuracy of multi-element forecasting is slightly higher than 4 elements as a whole, but the difference is small, while for western regions, the accuracy of multi-element forecasting is far lower than 4 elements, and is not more than 50%. The optimal linear regression model in the middle eastern region is a multi-element stepwise regression model, and the average prediction accuracy is as follows: 87.85%, the middle leaf prediction accuracy is highest, 92.37%, the lower part She Ci and the upper leaf is lowest; the optimal regression model in western regions is a 4-element linear regression model, and the average prediction accuracy is: 87.96% and the middle and upper leaves She Yubao are 92.59% each with the lowest accuracy and the lower leaves.
The model is built by adopting linear regression on 9 states in Guizhou province to obtain the highest and lowest prediction accuracy of each state (figure 11), and it is known that the models of the other 8 state 4 elements except for the Anshun city are obviously superior to the model with multiple elements in both the highest prediction accuracy and the lowest prediction accuracy, but the model with multiple elements in the Anshun city can obtain higher prediction accuracy.
(2) BP neural network yield prediction model establishment and inspection
In BP neural network modeling, the inventor tries that the number of hidden layers is 1 and 2 respectively, and the number of nodes of each hidden layer neuron is 1 to 10, and 110 network structures are altogether. And selecting the model with highest prediction accuracy for comparison and inspection. From the viewpoint of prediction accuracy of different regions and different leaf positions (fig. 12), the prediction accuracy of 4-element and multi-element models on the leaf yield of each part exceeds 85%; in the middle eastern region, the multi-element model is better, the average forecasting accuracy is 90.96%, wherein the forecasting accuracy of the middle leaf is 94.07%, the forecasting accuracy of the lower leaf is She Ci, and the forecasting accuracy of the upper leaf is 87.29%; in the western region, the model of 4 elements is better, the average forecasting accuracy is 92.59%, the forecasting accuracy of the upper leaf is the highest, 96.3%, the middle She Ci, and the lower leaf is the lowest, 88.89%.
The model is built by BP neural network for 9 states in Guizhou province, so that the highest and lowest prediction accuracy of each state are obtained (figure 13), and it is known that the highest prediction accuracy and the lowest prediction accuracy of the models of the multi-element in the rest 7 states are obviously better than those of the model of the 4 elements except for the graduation city and the six-disc water city, but the model of the 4 elements is adopted in the Anshun city, so that the higher prediction accuracy can be obtained, and the prediction accuracy of the 4 elements and the multi-element in the six-disc water city is equal to 100%.
(3) Random forest modeling results
From the viewpoint of the forecasting accuracy of each leaf position yield by adopting random forest modeling (figure 14), the forecasting accuracy of each leaf yield by the 4-element and multi-element models is more than 62%; the forecasting accuracy of the middle eastern area is slightly better than that of the western area on the whole; in the middle eastern region, the multi-element model is more excellent, the average forecasting accuracy is 86.44%, wherein the forecasting accuracy of the middle leaves is the highest and is 90.68%, the lower leaves are She Ci, and the upper leaves are the lowest and are 81.36%; in the western region, the difference of the leaf position prediction accuracy of different element models is larger, the yield prediction accuracy of the lower leaf is better than that of the 4 element model, which is 85.19%, and the middle leaf and the upper leaf are better than that of the multi-element model, which is 88.89% and 92.59%, respectively.
The highest and lowest prediction accuracy of each city state are obtained by respectively adopting BP neural network to establish models for 9 city states in Guizhou province (figure 15), and it is known that the models of the multi-element models of the rest 6 city states except for Anshun city, liujinshui city and Qian southeast state are obviously superior to the model of the 4 elements in both the highest prediction accuracy and the lowest prediction accuracy, but the model of the 4 elements in Liujinshui city and Qian southeast state can obtain higher prediction accuracy, and the prediction accuracy of the 4 elements and the multi-element models of Anshun city are equivalent.
(4) Multiple mode contrast analysis
By comparing the prediction models established by the four methods of linear regression, stepwise regression, BP neural network and random forest, the prediction model of the BP neural network is far superior to the other three methods, and the prediction accuracy of the 4-element linear regression model in the western region is consistent with the prediction accuracy of the multi-element stepwise regression model in the western region in the upper leaf and the prediction accuracy of the BP neural network of the corresponding element category. Therefore, the BP neural network model with multiple elements is better in the middle eastern region of Guizhou province, the yield prediction accuracy of each leaf position is more than 87%, the average value is 90.96%, the BP neural network model with 4 elements is better in the western region of Guizhou province, the yield prediction accuracy of each leaf position is more than 88%, and the average value is 92.59%.
Table 2 comparison of various modes
Defining the predicted single leaf weight minus the actual single leaf weight as an error, dividing the error by the actual single leaf weight as an error percentage, taking the multi-element in the western region as an independent variable, and basically setting the error of the rest prediction model between +/-2 except that the error of the linear regression prediction result is larger, wherein the prediction error of the multi-element is lower than that of the four-element. As shown in fig. 16, from the error percentage, the prediction conclusion of the four models in the middle eastern region is relatively large, and the abnormal values are basically larger by more than 50%. The outliers in the western region are mostly 100% greater.
6. Analysis of results
(1) The variation trend of the eastern and western regions is consistent with that of the whole province, the variation of the single leaf weight of the middle leaf is not great, the weight of the lower leaf and the weight of the upper leaf are increased, and the weight of the upper leaf is obviously increased compared with that of the lower leaf.
(2) A linear and nonlinear multiple method is used for establishing a forecasting model of single-leaf weight and meteorological elements (four key growth period meteorological elements and ten-day meteorological elements) in two areas (middle east and west) of Guizhou province, the forecasting model of the BP neural network is superior to the models of other three methods, the accuracy rate of establishing the forecasting model by multiple elements is generally higher than that of using four elements in the middle east area, and the forecasting effect of four elements is generally higher in the west area than that of multiple elements. Therefore, when detailed weather elements cannot be acquired, a prediction model can be established by using 4 elements, and the predicted weather elements are brought into the prediction to predict the yield of the current year.
(3) As can be seen from the error and error percentage distribution of different prediction methods, the artificial intelligence prediction mode is significantly better than the linear prediction mode, and the model has a larger trend for the prediction of the yield.
The background management module comprises a user management module, a basic information management module and a configuration management module. Wherein, the user management module: adding and deleting users, modifying user passwords, distributing user rights and the like; and the basic information management module: tobacco field GIS information, site information and the like; and (3) a configuration management module: and configuring the respective information independently according to the special conditions of each tobacco field.
The foregoing examples illustrate only a few embodiments of the invention and are described in detail herein without thereby limiting the scope of the invention. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the invention, which are all within the scope of the invention. Accordingly, the scope of protection of the present invention is to be determined by the appended claims.
Claims (10)
1. The intelligent digital tobacco field agriculture system is characterized by comprising a flue-cured tobacco planting information module, a meteorological big data module and a production suggestion module, wherein,
The flue-cured tobacco planting information module is used for collecting flue-cured tobacco plots, flue-cured tobacco plants, house owner information, curing barn information and seedling shed information, and simultaneously collecting soil fertility data of each flue-cured tobacco plot according to grids, and superposing the soil fertility data on a grid digital base map of flue-cured tobacco planting layout;
the weather big data module comprises a tobacco weather live sub-module, a real-time tobacco weather early-warning sub-module and a tobacco weather forecast sub-module, wherein the tobacco weather live sub-module is used for displaying tobacco weather live data of each grid point on the gridding digital base map, the tobacco weather early-warning sub-module is used for carrying out weather disaster early warning on each flue-cured tobacco plot according to the tobacco weather live data, and the tobacco weather forecast sub-module is used for displaying future weather data of each grid point on the gridding digital base map;
the production suggestion module comprises a disaster response sub-module, a production technology scheme sub-module, a farming reminding sub-module and a tobacco yield forecasting sub-module, wherein the disaster response sub-module is used for providing corresponding treatment measures according to weather disaster early warning, the production technology scheme sub-module is used for providing production suggestions according to soil fertility data and tobacco weather live data, the farming reminding sub-module is used for displaying the current key farming stage, and the tobacco yield forecasting sub-module is used for forecasting the yield of a flue-cured tobacco land according to weather data.
2. The digital tobacco field intelligent agriculture system according to claim 1, wherein the grid digital base map is divided according to a map of a flue-cured tobacco planting area of 1 x 1, and the soil fertility data comprises organic matter, alkaline hydrolysis nitrogen, quick-acting phosphorus, quick-acting potassium and soil pH information.
3. The digital tobacco field smart agriculture system of claim 2 wherein the method of displaying the tobacco weather live data and future weather data for each grid point on the grid digital base map is:
acquiring original meteorological element data to form a meteorological element product with 5km resolution;
interpolation is carried out on weather element products with 5km resolution to grids with 1km resolution by using an inverse distance weight method, and then regression models are utilized to correct the forecast values after interpolation, so that weather elements with 1km resolution are obtained;
and (3) interpolating the grid data with the resolution of 1km to the geometric center position of each grid on the grid digital base map by using a bilinear interpolation method, and accurately matching the meteorological elements to each grid.
4. The digital tobacco field intelligent agricultural system according to claim 1, wherein the disaster handling submodule comprises a disaster grade determining submodule and a disaster matching measure submodule, wherein the disaster grade determining submodule is used for reading meteorological elements according to each flue-cured tobacco plot and determining whether a tobacco field is in the above disasters and the grades of the disasters according to the related standards for drought, hail, wind disasters and waterlogging in the QX/T363-2016 flue-cured tobacco meteorological disaster grade standard; the disaster matching measure submodule is used for matching disaster countermeasures and remedial measures after the flue-cured tobacco land disaster and the grade are determined.
5. The digital tobacco field intelligent agriculture system of claim 1, wherein the production technology scheme submodule comprises a fertilization recommendation submodule, a transplanting period recommendation submodule, a topdressing time recommendation submodule, a worthless foot leaf processing time submodule, a topping time recommendation submodule and a picking and baking period recommendation submodule.
6. The digital tobacco field intelligent agriculture system of claim 1, wherein the processing steps of the agriculture reminder sub-module are: and reading the current time, and comparing the current time with the time period in the production overall technical scheme to display the current key agronomic stage.
7. The digital tobacco field intelligent agriculture system of claim 1, wherein the processing steps of the tobacco yield forecasting sub-module are:
calling a corresponding flue-cured tobacco yield prediction model according to the position information of the flue-cured tobacco land block;
and acquiring meteorological data corresponding to the flue-cured tobacco land, inputting the meteorological data into the flue-cured tobacco yield prediction model, and calculating the tobacco yield corresponding to the flue-cured tobacco land.
8. The digital tobacco field intelligent agriculture system of claim 7, wherein the flue-cured tobacco yield prediction model comprises a first BP neural network prediction model and a second BP neural network prediction model, the first BP neural network prediction model is invoked as a flue-cured tobacco yield prediction model when the flue-cured tobacco plot is in a first predetermined region, and the second BP neural network prediction model is invoked as a flue-cured tobacco yield prediction model when the flue-cured tobacco plot is in a second predetermined region.
9. The digital tobacco field intelligent agriculture system of claim 8, wherein the first BP neural network prediction model is constructed by:
collecting the historical tobacco leaf yield of the flue-cured tobacco plots as a first dependent variable;
collecting first historical meteorological data of a flue-cured tobacco land as a first independent variable, wherein the first historical meteorological data comprises average air temperature, average highest air temperature, average lowest air temperature, sunshine hours and precipitation amount from 3 to 9 months to ten days;
and training the BP neural network by taking the first independent variable as a network input value and taking the first independent variable as a network predicted value to obtain a first BP neural network predicted model.
10. The digital tobacco field intelligent agriculture system of claim 8, wherein the second BP neural network prediction model is constructed by:
collecting the historical tobacco leaf yield of the flue-cured tobacco plots as a second dependent variable;
collecting second historical meteorological data of a flue-cured tobacco plot as a second independent variable, wherein the second historical meteorological data comprises air temperature in a tobacco maturity stage, rainfall capacity in a vigorous and long-term period, a time of Tian Qiri hours and available time in a field growth period;
and training the BP neural network by taking the second independent variable as a network input value and taking the second independent variable as a network predicted value to obtain a second BP neural network predicted model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310415073.6A CN116523673A (en) | 2023-04-18 | 2023-04-18 | Intelligent agricultural system for digital tobacco field |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310415073.6A CN116523673A (en) | 2023-04-18 | 2023-04-18 | Intelligent agricultural system for digital tobacco field |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116523673A true CN116523673A (en) | 2023-08-01 |
Family
ID=87405764
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310415073.6A Pending CN116523673A (en) | 2023-04-18 | 2023-04-18 | Intelligent agricultural system for digital tobacco field |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116523673A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117313993A (en) * | 2023-09-15 | 2023-12-29 | 湖南省烟草公司湘西自治州公司 | Flue-cured tobacco cultivation management method based on temperature and altitude |
| CN118120428A (en) * | 2024-05-07 | 2024-06-04 | 中国农业科学院农业环境与可持续发展研究所 | Intelligent fertilizer preparation decision-making method and system based on meteorological data |
-
2023
- 2023-04-18 CN CN202310415073.6A patent/CN116523673A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117313993A (en) * | 2023-09-15 | 2023-12-29 | 湖南省烟草公司湘西自治州公司 | Flue-cured tobacco cultivation management method based on temperature and altitude |
| CN118120428A (en) * | 2024-05-07 | 2024-06-04 | 中国农业科学院农业环境与可持续发展研究所 | Intelligent fertilizer preparation decision-making method and system based on meteorological data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11893648B2 (en) | Methods and systems for recommending agricultural activities | |
| US11785879B2 (en) | Methods and systems for managing agricultural activities | |
| US11847708B2 (en) | Methods and systems for determining agricultural revenue | |
| US20210383290A1 (en) | Methods and systems for recommending agricultural activities | |
| CN112602563B (en) | Water-saving irrigation system and accurate irrigation method | |
| Chen et al. | A neural network integrated approach for rice crop monitoring | |
| US20160309646A1 (en) | Agronomic systems, methods and apparatuses | |
| CN116523673A (en) | Intelligent agricultural system for digital tobacco field | |
| CN106951979A (en) | The crop maturity phase Forecasting Methodology that remote sensing, crop modeling are merged with weather forecast | |
| Guo et al. | Optimizing irrigation schedule in a large agricultural region under different hydrologic scenarios | |
| CN110909973A (en) | Comprehensive drought monitoring and evaluating method considering underlying surface condition | |
| Swain et al. | A framework for managing irrigation water requirements under climatic uncertainties over Beed district, Maharashtra, India | |
| CN110516943B (en) | Surface temperature-based dynamic monitoring and remote sensing method for irrigation area in spring irrigation period | |
| Potopová et al. | Application of hydroclimatic drought indicators in the transboundary Prut River basin | |
| CN114897423B (en) | Farm irrigation probability index calculation method based on crop moisture indication line | |
| CN115205695A (en) | Method and system for determining planting strategy according to planting data | |
| Reddy et al. | A novel approach to analysis district level long scale seasonal forecasting of monsoon rainfall in Andhra Pradesh and Telangana | |
| CN113570273A (en) | Spatial method and system for irrigation farmland statistical data | |
| WO2023245399A1 (en) | Rice production potential simulation method based on land system and climate change coupling | |
| Łabędzki et al. | Indicator-based monitoring and forecasting water deficit and surplus in agriculture in Poland | |
| CN116644837A (en) | Method for predicting tobacco leaf yield based on meshed regional meteorological conditions | |
| CN117436710B (en) | Drought assessment method based on crop model | |
| Millogo | Application of remote sensing for irrigation performance assessment: preliminary evaluation at di perimeters in Sourou valley, Burkina Faso | |
| Kakoti | AGRO-CLIMATIC CHARACTERIZATION FOR PLANNING SMART AGRICULTURAL PRACTICES IN ASSAM | |
| Blagojević et al. | Multi-Criteria Decision-Making for Irrigation Management in Serbia |
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 |