CN110084367A - A kind of Forecast of Soil Moisture Content method based on LSTM deep learning model - Google Patents
A kind of Forecast of Soil Moisture Content method based on LSTM deep learning model Download PDFInfo
- Publication number
- CN110084367A CN110084367A CN201910317820.6A CN201910317820A CN110084367A CN 110084367 A CN110084367 A CN 110084367A CN 201910317820 A CN201910317820 A CN 201910317820A CN 110084367 A CN110084367 A CN 110084367A
- Authority
- CN
- China
- Prior art keywords
- data
- formula
- moisture content
- deep learning
- soil moisture
- 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.)
- Granted
Links
- 239000002689 soil Substances 0.000 title claims abstract description 101
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000013136 deep learning model Methods 0.000 title claims abstract description 38
- 238000012549 training Methods 0.000 claims abstract description 18
- 238000012360 testing method Methods 0.000 claims abstract description 12
- 241000208340 Araliaceae Species 0.000 claims abstract description 8
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 claims abstract description 8
- 235000003140 Panax quinquefolius Nutrition 0.000 claims abstract description 8
- 235000008434 ginseng Nutrition 0.000 claims abstract description 8
- 230000006870 function Effects 0.000 claims description 25
- 230000015654 memory Effects 0.000 claims description 19
- 230000000694 effects Effects 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- 239000011159 matrix material Substances 0.000 claims description 3
- 230000000717 retained effect Effects 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 2
- 238000013135 deep learning Methods 0.000 abstract description 3
- 230000031016 anaphase Effects 0.000 abstract description 2
- 239000000463 material Substances 0.000 abstract description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 7
- 238000003973 irrigation Methods 0.000 description 6
- 230000002262 irrigation Effects 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 230000004913 activation Effects 0.000 description 4
- 238000012271 agricultural production Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 235000013399 edible fruits Nutrition 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000001556 precipitation Methods 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 206010053759 Growth retardation Diseases 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 231100000001 growth retardation Toxicity 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000035800 maturation Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000005080 plant death Effects 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
Classifications
-
- A—HUMAN NECESSITIES
- A01—AGRICULTURE; FORESTRY; ANIMAL HUSBANDRY; HUNTING; TRAPPING; FISHING
- A01G—HORTICULTURE; CULTIVATION OF VEGETABLES, FLOWERS, RICE, FRUIT, VINES, HOPS OR SEAWEED; FORESTRY; WATERING
- A01G25/00—Watering gardens, fields, sports grounds or the like
- A01G25/16—Control of watering
- A01G25/167—Control by humidity of the soil itself or of devices simulating soil or of the atmosphere; Soil humidity sensors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- 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
- 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)
- Life Sciences & Earth Sciences (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Economics (AREA)
- General Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Software Systems (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- Bioinformatics & Computational Biology (AREA)
- Development Economics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mining & Mineral Resources (AREA)
- Marine Sciences & Fisheries (AREA)
- Primary Health Care (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Animal Husbandry (AREA)
- Agronomy & Crop Science (AREA)
- Evolutionary Biology (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Soil Sciences (AREA)
Abstract
The invention discloses a kind of Forecast of Soil Moisture Content methods based on LSTM deep learning model, the soil physical chemistry and meteorological data first in a period of time in collection target farmland；It is divided into training sample set and test sample collection after the data being collected into are pre-processed；Construct LSTM deep learning model, LSTM deep learning model is trained by training sample set to obtain adjusting the LSTM deep learning model after ginseng, LSTM deep learning model after exchanging ginseng by test sample collection again is verified, finally using the LSTM deep learning model after verifying as Forecast of Soil Moisture Content model；The data of collection are inputted into Forecast of Soil Moisture Content model, finally by the Forecast of Soil Moisture Content result of Forecast of Soil Moisture Content model output future time instance.The present invention forecasts soil moisture content using deep learning method, has saved manpower and material resources, can really reflect influence of the Primary Stage Data to anaphase, fully demonstrate sequential character.
Description
Technical field
The present invention relates to Forecast of Soil Moisture Content method field, specifically a kind of soil moisture in the soil based on LSTM deep learning model
Feelings prediction technique.
Background technique
China is the serious country of drought and water shortage, and water resource of per capita only has the 1/4 of world average level, is the whole world
Poorest one of the country of per capita water resource.Implementing Precision Irrigation in agricultural production process can be with effectively save water resource, rush
Into crop growth.But often there is irrigating insufficient and irrigating superfluous phenomenon, irrigation is insufficient to be caused to make for field irrigation
Object growth retardation, under low output, and irrigating surplus then easily makes crop root depauperation, the phenomenon that plant death occurs, causes
Water utilization rate reduces, and is unable to reach the purpose of high yield.Therefore, Forecast of Soil Moisture Content model is established, soil moisture content is carried out
Prediction can be solved the problems, such as effectively to irrigate insufficient and irrigate surplus, be the technical way for realizing farmland Precision Irrigation.
Forecast of Soil Moisture Content model is according to certain following a period of time of farmland meteorological data, soil physical chemistry data and past soil moisture content prediction
The soil moisture content at quarter the number of duty is determined with this, to have the function that stable high yield.There is presently no maturations
Soil moisture content is effectively predicted in technical method, establishes the Forecast of Soil Moisture Content model that generalization ability is strong, forecast accuracy is high,
The Accurate Prediction for implementing soil moisture content is that agriculture precisely production needs one of the major issue solved.Document " is based on gas within 2016
As the Qidong City soil moisture content prediction research of the factor " (Agriculture of Anhui science, 2016,44 (34): 174-176) propose using by
Homing method is walked, analyzes Qidong City 2011-2014 years soil table moisture in the soils, soil moisture and meteorological factor (precipitation of the same period respectively
Amount, temperature, humidity, sunshine, wind) correlation, filter out influence soil moisture content Key Meteorological Factors, and combine empirical equation
Method establishes Soil moisture prediction model.The result shows that the meteorological factor for influencing Qidong City soil moisture content is mainly precipitation, sunshine
And temperature, the forecasting model thus established forecast that the average relative error of soil moisture content in future 30d within 5%, examines effect
Fruit is ideal, illustrates relatively accurately forecast soil moisture content in the following 30d using the model, and be used for guiding agricultural production.
Document in 2017 " the Forecast of Soil Moisture Content Research on Accuracy based on BP neural network --- by taking Feidong County as an example " (soil
Notification, 2017,48 (02): 292-297).It proposes using BP neural network for predicting soil moisture content, main thought
Choosing wherein influences more significant temperature on average, medial humidity, amount of radiation, rainfall as mode input sample to soil moisture
This, establishes network model.In addition, the soil moisture content at the beginning of the period has a significant impact to the soil moisture content of period Mo, so the period
First soil moisture content is also by the input sample as model.
Patent " a kind of soil moisture content forecasting procedure based on soil moisture content index " (application number: N201810457976.X), the party
Model used in method is the model of half proposed by many year field experimental study, has parameter simple and easy to get, just
In the practical application the characteristics of.This method proposes to survey soil moisture content with loss of moisture sensitive layer to calculate crop root germinal layer soil
Soil moisture content index successively carries out the calculating of 20cm and 50cm soil layer actual measurement moisture content, 50cm soil layer soil moisture content index calculates, 50cm soil layer moisture in the soil
Feelings exponential forecasting, 20cm soil layer hydrated comples ion, 50cm soil layer hydrated comples ion, irrigation period prediction, irrigating water quota prediction, it is real
Show soil moisture content monitoring, forecast, update, the standardization of irrigation period and irrigating water quota Predicting Technique, constructs soil moisture content
Monitoring and prediction information system facilitate soil moisture information inquiry, drought assessment and risk management and control, suitable for vast plains region agriculture
The forecast of field soil moisture content.
These above-mentioned methods realize the soil moisture content situation of change according to the past period to predict subsequent time
Soil moisture content change curve.But they it is common the problem is that: do not fully taken into account when building model data when
Sequence feature, generalization ability and forecast accuracy are all to be improved.
Summary of the invention
The object of the present invention is to provide a kind of Forecast of Soil Moisture Content methods based on LSTM deep learning model, existing to solve
There is the problem of technology Forecast of Soil Moisture Content method does not account for data time sequence feature.
In order to achieve the above object, the technical scheme adopted by the invention is as follows:
A kind of Forecast of Soil Moisture Content method based on LSTM deep learning model, it is characterised in that: the following steps are included:
(1), the soil physical chemistry and meteorological data in a period of time in target farmland are collected；
(2), soil physical chemistry and meteorological data that step (1) is collected into are pre-processed, it then again will be pretreated
Data are divided into training sample set and test sample collection in proportion；
(3), LSTM deep learning model is constructed, the LSTM deep learning model has an input layer, 25
Hidden layer and an output layer are trained LSTM deep learning model by training sample set to obtain adjusting the LSTM after ginseng
Deep learning model, then by test sample collection exchange ginseng after LSTM deep learning model verified, finally with verifying after
LSTM deep learning model as Forecast of Soil Moisture Content model；
(4), the soil physical chemistry and meteorological data collected step (1) pass through as the input of Forecast of Soil Moisture Content model
Forecast of Soil Moisture Content model handles soil physical chemistry and meteorological data, when finally exporting following by Forecast of Soil Moisture Content model
The Forecast of Soil Moisture Content result at quarter.
A kind of Forecast of Soil Moisture Content method based on LSTM deep learning model, it is characterised in that: step (1)
In, the missing data of collected soil physical chemistry and meteorological data is repaired using linear interpolation method, wherein the public affairs of linear interpolation method
Formula is as follows:
In formula (1), i and j respectively indicate the i-th moment and jth moment value, it is desirable that 0 < i < j.xkAnd xk+jWhen being k respectively
Carve the soil physical chemistry acquired with the k+j moment and meteorological data, xk+iThe soil physical chemistry and meteorological data lost for the k+i moment.
A kind of Forecast of Soil Moisture Content method based on LSTM deep learning model, it is characterised in that: in step (2)
Pretreatment be normalization pretreatment, normalization pretreatment after using wherein 85% data as training sample set, 15% number
According to as test sample collection, normalizing pretreated formula is
By the way that pretreatment is normalized to collected data, make the mapping range [0,1] for the data value collected,
It normalizes in pretreated formula, x is initial data, xmax, xminMaxima and minima respectively in initial data, xnow
For the result data after normalized；
By normalization pretreatment to eliminate the dimension impact between index, the comparativity between data target is solved, it is former
For beginning data after pretreatment, each index is in the same order of magnitude, facilitates the building of model.
A kind of Forecast of Soil Moisture Content method based on LSTM deep learning model, it is characterised in that: step (3)
In, the network structure of the LSTM deep learning model of building is (7,25,1), each hidden layer in LSTM deep learning model
Tool is respectively adopted there are three the LSTM unit of door, three doors of the LSTM unit are to forget door, input gate and out gate respectively, are led to
It crosses three doors and the update of complete pair state and exports target value jointly, the data handling procedure of hidden layer is as follows:
Forget door and determines the degree for forgeing information, first reading h(t-1)And x(t)To carry out Screening Treatment, wherein h to data(t-1)That indicate is the output of a upper memory cell, x(t)What is indicated is when precellular input, as shown in formula (2):
f(t)=σ (Wf·[h(t-1),x(t)]+bf) (2),
In formula (2), WfIt is weight term, bfIt is bias term, f(t)It is the forgetting degree of information, σ is sigmoid function, is taken
Value is between [0,1].Sigmoid function exports the numerical value between 0 to 1 and is used for cell state C(t)Update, wherein 1
It is expressed as that information is fully retained, 0 indicates to give up this node data completely；
Input gate determines that new information is added in concealed nodes, wherein C(t-1)It is the cell state of last moment, defines i(t)To determine the information updated, completes information addition and need to include two steps: firstly, passing through the sigmoid of an input gate
Function determines which information needs to update；Secondly, by one vector of a tanh layer generation, that is, it is alternative for updating
Content a(t), this two parts is joined together, a update is carried out to the state of cell, as shown in formula (3), (4):
i(t)=σ (Wi·[h(t-1),x(t)]+bi) (3),
a(t)=tanh (Wc·[h(t-1),x(t)]+ba) (4),
In formula 3, σ is sigmoid function, WiFor weight term, h(t-1)It is the part of last moment final output, x(t)For
When precellular input.biIt is bias term.
In formula 4, tanh is tanh function, WcFor weight term, h(t-1)It is the part of last moment final output, x(t)For
When precellular input.baIt is bias term.
In formula (4) when new and old cell state, by the content a of last moment update(t-1)It is updated to update this moment interior
Hold a(t).The cell state C of last moment(t-1)With the f in forgetting door(t)It is multiplied, and adds i(t)*a(t), reach more neoblast
The effect of state, as shown in formula (5):
C(t)=f(t)*C(t-1)+i(t)*a(t)(5),
In formula (5) formula, * indicates Hadamard product, the i.e. product of representing matrix corresponding position.C(t-1)For last moment
Cell state, a(t)For new content.C(t)For new memory state.
Out gate determines output item, is primarily based on memory cell state, runs a sigmoid function to determine that memory is thin
Which information of born of the same parents will export；Secondly, memory cell state is handled by tanh, one is obtained between -1 to 1
Value, and the value is multiplied with the output of out gate, as shown in formula (6) and formula (7):
In formula (6), o(t)To export which information, h(t-1),x(t)It is expressed as the output of last moment and inputs this moment.Wo
For weight term, boFor bias term.
H in formula (7)(t)For the part of final output, by o obtained in formula (6)(t)Multiplied by current new memory state
By the value of tanh function, achieve the effect that the information for remembeing that sequence relies on for a long time.
The advantages of the present invention over the prior art are that: the present invention utilizes deep learning algorithm, using one kind based on length
The Forecast of Soil Moisture Content method of short-term memory model.It is compared with the traditional method, soil moisture content is carried out using deep learning method
Forecast, it is not necessary to be measured in real time using manual method, save manpower and material resources.In addition, can really be reflected using LSTM unit
Influence of the Primary Stage Data to anaphase, has fully demonstrated sequential character, has improved forecasting efficiency and accuracy, has higher
Generalization ability.Forecast of Soil Moisture Content method based on LSTM deep learning model has good application value.
Detailed description of the invention
Fig. 1 is implementation flow chart of the present invention.
Fig. 2 is the LSTM cell schematics that the present invention uses.
Fig. 3 is activation primitive schematic diagram.
Fig. 4 is activation primitive schematic diagram.
Fig. 5 is the prediction graph of test set (soil moisture content).
Specific embodiment
Present invention will be further explained below with reference to the attached drawings and examples.
As shown in Figure 1, the present invention the specific implementation process is as follows:
1, the meteorological data and soil physical chemistry data in farmland are monitored.The data of acquisition in every 30 minutes.Acquisition about 3 months
Data volume amounts to more than 4000 datas record, for the data of part missing, is repaired using the linear interpolation method,
As shown in formula (1):
In formula (1), i and j respectively indicate the i-th moment and jth moment value, it is desirable that 0 < i < j.xkAnd xk+jIt is the k moment respectively
The soil physical chemistry and meteorological data acquired with the k+j moment, xk+iFor the soil physical chemistry and meteorological data lost when the k+i moment.
2, data prediction: before model training, need to collected farmland meteorological data and soil physical chemistry data into
Row normalized.So-called normalized exactly maps the data into [0,1] or [- 1,1] section.Guarantee different data model
The input data enclosed plays identical effect.The normalized processing formula used in the present invention forIts
Middle x is the initial data to normalized, xmax, xminMaxima and minima respectively in initial data, xnowFor normalizing
Data after change processing.
Data after normalized are divided into two parts of training sample set and test sample collection data.Data institute accounting
Example is respectively 85%, 15%, training and test for LSTM deep learning model.
3, model structure: as shown in Fig. 2, the present invention is used with 7 input layers, 25 hidden layers and 1 output layer
LSTM deep learning model, the hiding layer unit of LSTM deep learning model is all made of LSTM (long short-term memory) unit, this
The problem of subsequent node of structure very good solution declines the timing node sensing capability of front, overcomes Conventional Time sequence
The shortcoming of model structure can be good at handling timing data.LSTM is a kind of unit for being called memory cell, the list
Member has 3 doors.
Forget door and determines the degree for forgeing information.Read h(t-1)And x(t), h(t-1)What is indicated is a upper memory cell
The output of cell, x(t)What is indicated is to complete the Screening Treatment of data when precellular input, as shown in formula (2):
f(t)=σ (Wf·[h(t-1),x(t)]+bf) (2),
In formula (2), WfIt is weight term, bfIt is bias term, f(t)It is the forgetting degree of information, σ is sigmoid function, is taken
Value is between [0,1].Sigmoid function exports the numerical value between 0 to 1 and is used for cell state C(t)Update, wherein 1
It is expressed as that information is fully retained, 0 indicates to give up this node data completely；
Input gate determines that new information is added in concealed nodes.C(t-1)It is the cell state of last moment, i(t)It is determining
The part of information update.It completes information addition to need to include two steps: firstly, passing through the sigmoid function of an input gate
Determine which information needs to update；Secondly, one vector of a tanh layers of generation, that is, the alternative content a for updating(t).This two parts is joined together, and is updated to cell state, as shown in formula (3) and formula (4):
i(t)=σ (Wi·[h(t-1),x(t)]+bi) (3),
a(t)=tanh (Wc·[h(t-1),x(t)]+ba) (4),
In formula 3, σ is sigmoid function, WiFor weight term, h(t-1)It is the part of last moment final output, x(t)For
When precellular input.biIt is bias term.
In formula 4, tanh is tanh function, WcFor weight term, h(t-1)It is the part of last moment final output, x(t)For
When precellular input.baIt is bias term.
In formula (4) when new and old cell state, by the content a of last moment update(t-1)It is updated to update this moment interior
Hold a(t).The cell state C of last moment(t-1)With the f in forgetting door(t)It is multiplied, and adds i(t)*a(t), reach more neoblast
The effect of state, as shown in formula (5):
C(t)=f(t)*C(t-1)+i(t)*a(t)(5),
In formula (5) formula, * indicates Hadamard product, the i.e. product of representing matrix corresponding position.C(t-1)For last moment
Cell state, a(t)For new content.C(t)For new memory state.
Out gate determines output item.Firstly, being based on memory cell state, one sigmoid layers are run to determine that memory is thin
Which information of born of the same parents will export；Secondly, memory cell state is handled by tanh (one is obtained between -1 to 1
Value) and it is multiplied with the output of out gate.Output item is calculated according to (6) formula and (7) formula.
In formula (6), o(t)To export which information, h(t-1),x(t)It is expressed as the output of last moment and inputs this moment.Wo
For weight term, boFor bias term.
H in formula (7)(t)For the part of final output, by o obtained in formula (6)(t)Multiplied by current new memory state
By the value of tanh function, achieve the effect that the information for remembeing that sequence relies on for a long time.
Door uses sigmoid activation primitive (as shown in Figure 3):
In above formula (8), x is as input data, by sigmoid function data vector value ' compression ' between [0,1],
It is 1 if the value of input is that positive number is very big if the value of input is for negative and very big, numerical value 0.
When the state to cell is updated, tanh activation primitive has been used (as shown in Figure 4.):
In above formula (9), x is mapped between [- 1,1] as input data by function f (x).
In network training, using the final layer state of hiding of current batch processing (batch) as subsequent initial hidden
(traversing entire training set in order).The size that batch is arranged is 72.What the LSTM deep learning model that the present invention uses used
Network structure is (7,25,1)., learning rate (learning rate) is set as 0.01.In the training process, according to averagely absolutely
Error is calculated error (MeanAbsoluteError), and according to back-propagation algorithm for updating weight.
In above formula (11), m is the total number of training data, x(i)Indicate data input values, k (x(i)) it is prediction output valve, y(i)It is real output value.The error amount obtained according to above formula.It inputs the sequence in training sample data as training, constantly instructs
Practice network model and adjust ginseng, terminates when the number of iterations reaches 300 times.It obtains a stable prediction model, which is used
Make the prediction model of soil moisture content.
There are two hidden state h for LSTM model(t), C(t), Model Parameter is more.
(1) process of the propagated forward process in each sequence index position are as follows:
Door output is forgotten 1. updating
f(t)=σ (Wf·[h(t-1),x(t)]+bf),
2. updating two parts output of input gate
i(t)=σ (Wi·[h(t-1),x(t)]+bi),
a(t)=tanh (Wc·[h(t-1),x(t)]+ba),
3. updating cell state
C(t)=f(t)*C(t-1)+i(t)*a(t),
4. updating the output of out gate
5. updating current sequence index prediction output
Y (t)=σ (Wyh(t)+c)
(2) back-propagation algorithm: for reverse propagated error, pass through hidden state h(t)Gradient δ(t)Gradually to forward pass
It broadcasts.The backpropagation of LSTM, there are two hidden state h(t)And C(t).Define two δ, it may be assumed thatWith
For the ease of deriving, loss function is divided into two pieces, one piece be moment t position loss l(t), another piece is the moment
L is lost after t(t+1), it may be assumed that
And in last sequence index position τ,WithIt is respectively as follows:
Then byReverse-direction derivation Gradient by this layer of t moment output gradient error and
Error two parts greater than t moment determine, it may be assumed that
The key of LSTM backpropagation isCalculating.Due to h(t)=o(t)*tanh(C(t)) in first item o(t)In
Recurrence relation comprising a h, so that Section 2 tanh (C(t)) become more sophisticated, tanh function variable can be expressed as:
C(t)=C(t-1)*f(t)+i(t)*a(t),
AndReversed gradient error consist of two parts, i.e. preceding layerGradient error and this layer slave h(t)It passes
Gradient error back.
It is knownWithIt can obtain WfGradient calculating process, other parameters are same as above.
4, network test (adjusting ginseng and optimization).The LSTM depth constructed will be input to by pretreated training set data
Learning model continues to optimize parameter, steps up precision of prediction, in order to prevent over-fitting, needs to carry out regularization.Finally obtain
Model, output is the Forecast of Soil Moisture Content at the following a certain moment as a result, the average relative error < 0.25% predicted, prediction knot
Fruit is as shown in figure 5, prediction result is preferable.
The method applied in the present invention largely using, incorporate relationship between historical data, fully take into account
The timing feature of data, builds reasonable time series models, to the predictive ability improved to soil moisture content, enhances model
Generalization ability have good application value.
Claims (4)
1. a kind of Forecast of Soil Moisture Content method based on LSTM deep learning model, it is characterised in that: the following steps are included:
(1), the soil physical chemistry and meteorological data in a period of time in target farmland are collected；
(2), soil physical chemistry and meteorological data that step (1) is collected into are pre-processed, then again by pretreated data
It is divided into training sample set and test sample collection in proportion；
(3), LSTM deep learning model is constructed, the LSTM deep learning model is hidden with an input layer, 25
Layer and an output layer are trained LSTM deep learning model by training sample set to obtain adjusting the LSTM depth after ginseng
Learning model, then by test sample collection exchange ginseng after LSTM deep learning model verified, finally with verifying after
LSTM deep learning model is as Forecast of Soil Moisture Content model；
(4), the soil physical chemistry and meteorological data collected step (1) pass through soil as the input of Forecast of Soil Moisture Content model
Soil moisture content prediction model handles soil physical chemistry and meteorological data, finally by Forecast of Soil Moisture Content model output future time instance
Forecast of Soil Moisture Content result.
2. a kind of Forecast of Soil Moisture Content method based on LSTM deep learning model according to claim 1, feature exist
In: in step (1), the missing data of collected soil physical chemistry and meteorological data is repaired using linear interpolation method, wherein linearly
The formula of interpolation method is as follows:
In formula (1), i and j respectively indicate the i-th moment and jth moment value, it is desirable that 0 < i < j；xkAnd xk+jIt is k moment and k+j respectively
The soil physical chemistry and meteorological data of moment acquisition, xk+iFor the soil physical chemistry and meteorological data lost when the k+i moment.
3. a kind of Forecast of Soil Moisture Content method based on LSTM deep learning model according to claim 1, feature exist
In: the pretreatment in step (2) is normalization pretreatment, using wherein 85% data as training sample after normalization pretreatment
Collection, as test sample collection, normalize pretreated formula is 15% data
By the way that pretreatment is normalized to collected data, make the mapping range [0,1] for the data value collected, normalizing
Change in pretreated formula, x is initial data, xmax, xminMaxima and minima respectively in initial data, xnowTo return
Result data after one change processing；
By normalization pretreatment to eliminate the dimension impact between index, the comparativity between data target, original number are solved
According to after pretreatment, each index is in the same order of magnitude, facilitates the building of model.
4. a kind of Forecast of Soil Moisture Content method based on LSTM deep learning model according to claim 1, feature exist
In: in step (3), the network structure of the LSTM deep learning model of building is (7,25,1), in LSTM deep learning model
Tool is respectively adopted there are three the LSTM unit of door in each hidden layer, and three doors of the LSTM unit are to forget door, input gate respectively
And out gate, the update of complete pair state and target value is exported jointly by three doors, the data processing of hidden layer
Journey is as follows:
Forget door and determines the degree for forgeing information, first reading h(t-1)And x(t)To carry out Screening Treatment, wherein h to data(t-1)Table
That show is the output of a upper memory cell, x(t)What is indicated is when precellular input, as shown in formula (2):
f(t)=σ (Wf·[h(t-1),x(t)]+bf) (2),
In formula (2), WfIt is weight term, bfIt is bias term, f(t)It is the forgetting degree of information, σ is sigmoid function, and value is
Between [0,1], sigmoid function exports the numerical value between 0 to 1 and is used for cell state C(t)Update, wherein 1 indicate
It indicates to give up this node data completely for information is fully retained, 0；
Input gate determines that new information is added in concealed nodes, wherein C(t-1)It is the cell state of last moment, defines i(t)For
It determines the information updated, completes information addition and need to include two steps: firstly, passing through the sigmoid function of an input gate
Determine which information needs to update；Secondly, by one vector of a tanh layer generation, that is, it is alternative interior for updating
Hold a(t), this two parts is joined together, a update is carried out to the state of cell, as shown in formula (3), (4):
i(t)=σ (Wi·[h(t-1),x(t)]+bi) (3),
a(t)=tanh (Wc·[h(t-1),x(t)]+ba) (4),
In formula 3, σ is sigmoid function, WiFor weight term, h(t-1)It is the part of last moment final output, x(t)It is current
The input of cell, biIt is bias term；
In formula 4, tanh is tanh function, WcFor weight term, h(t-1)It is the part of last moment final output, x(t)It is current
The input of cell, baIt is bias term；
In formula (4) when new and old cell state, by the content a of last moment update(t-1)It is updated to the content a updated this moment(t), the cell state C of last moment(t-1)With the f in forgetting door(t)It is multiplied, and adds i(t)*a(t), reach update cell state
Effect, as shown in formula (5):
C(t)=f(t)*C(t-1)+i(t)*a(t)(5),
In formula (5) formula, * indicates Hadamard product, the i.e. product of representing matrix corresponding position, C(t-1)For the cell of last moment
State, a(t)For new content, C(t)For new memory state；
Out gate determines output item, is primarily based on memory cell state, runs a sigmoid function to determine memory cell
Which information will export；Secondly, memory cell state is handled by tanh, a value between -1 to 1 is obtained,
And the value is multiplied with the output of out gate, as shown in formula (6) and formula (7):
In formula (6), o(t)To export which information, h(t-1),x(t)It is expressed as the output of last moment and inputs this moment, WoFor power
Weight item, boFor bias term；
H in formula (7)(t)For the part of final output, by o obtained in formula (6)(t)Pass through multiplied by current new memory state
The value of tanh function achievees the effect that the information for remembeing that sequence relies on for a long time.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910317820.6A CN110084367B (en) | 2019-04-19 | 2019-04-19 | Soil moisture content prediction method based on LSTM deep learning model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910317820.6A CN110084367B (en) | 2019-04-19 | 2019-04-19 | Soil moisture content prediction method based on LSTM deep learning model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110084367A true CN110084367A (en) | 2019-08-02 |
| CN110084367B CN110084367B (en) | 2022-10-25 |
Family
ID=67415755
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910317820.6A Active CN110084367B (en) | 2019-04-19 | 2019-04-19 | Soil moisture content prediction method based on LSTM deep learning model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110084367B (en) |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN110567534A (en) * | 2019-09-10 | 2019-12-13 | 广东工业大学 | Method for predicting flow of combustion air outlet in glass melting furnace and related device |
| CN111561972A (en) * | 2020-06-18 | 2020-08-21 | 华南农业大学 | Soil water content prediction system and method based on time sequence |
| CN111610406A (en) * | 2020-04-24 | 2020-09-01 | 国网河北省电力有限公司电力科学研究院 | Grounding grid corrosion prediction method based on deep learning |
| CN111859814A (en) * | 2020-07-30 | 2020-10-30 | 中国电建集团昆明勘测设计研究院有限公司 | Rock aging deformation prediction method and system based on LSTM deep learning |
| CN112115984A (en) * | 2020-08-28 | 2020-12-22 | 安徽农业大学 | Tea garden abnormal data correction method and system based on deep learning and storage medium |
| CN112255095A (en) * | 2020-09-25 | 2021-01-22 | 汕头大学 | Soil stress-strain relation determining method |
| CN112287294A (en) * | 2020-09-10 | 2021-01-29 | 河海大学 | Time-space bidirectional soil water content interpolation method based on deep learning |
| CN112784331A (en) * | 2020-09-25 | 2021-05-11 | 汕头大学 | Soil stress-strain relation determination method based on improved LSTM deep learning method |
| CN113257008A (en) * | 2021-05-12 | 2021-08-13 | 兰州交通大学 | Pedestrian flow dynamic control system and method based on deep learning |
| CN113323676A (en) * | 2021-06-03 | 2021-08-31 | 上海市隧道工程轨道交通设计研究院 | Method for determining cutter head torque of shield tunneling machine by using principal component analysis-length memory model |
| CN113435707A (en) * | 2021-06-03 | 2021-09-24 | 大连钜智信息科技有限公司 | Soil testing and formulated fertilization method based on deep learning and weighted multi-factor evaluation |
| CN113435649A (en) * | 2021-06-29 | 2021-09-24 | 布瑞克农业大数据科技集团有限公司 | Global agricultural data sorting method, system, device and medium |
| CN113468810A (en) * | 2021-07-01 | 2021-10-01 | 天行智控(成都)科技有限公司 | Intelligent floor sensing indoor tumble prediction model and establishment method thereof |
| CN113796228A (en) * | 2021-09-26 | 2021-12-17 | 南京邮电大学 | Plant cultivation system and method based on digital twinning |
| CN114080953A (en) * | 2021-11-05 | 2022-02-25 | 山东省农业机械科学研究院 | Illumination management method and system for mushroom house |
| CN114451257A (en) * | 2021-12-23 | 2022-05-10 | 珠海格力电器股份有限公司 | Irrigation method and device based on neural network, storage medium and electronic equipment |
| WO2022245312A1 (en) * | 2021-05-18 | 2022-11-24 | Hi̇t Bi̇li̇şi̇m Danişmanlik Turi̇zm Hi̇zmetleri̇ Sanayi̇ Ve Ti̇caret Li̇mi̇ted Şi̇rketi̇ | A method for estimating soil moisture and a system operating according to said method |
| CN117272813A (en) * | 2023-09-26 | 2023-12-22 | 长春师范大学 | Soil humidity prediction method based on water balance constraint deep learning |
| CN117494902A (en) * | 2023-11-22 | 2024-02-02 | 山东大学 | Soil moisture content prediction method and system based on soil moisture correlation analysis |
| CN117933946A (en) * | 2024-03-20 | 2024-04-26 | 华兴国创(北京)科技有限公司 | Rural business management method based on big data |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018120016A1 (en) * | 2016-12-30 | 2018-07-05 | 上海寒武纪信息科技有限公司 | Apparatus for executing lstm neural network operation, and operational method |
| CN108738444A (en) * | 2018-03-01 | 2018-11-06 | 洛阳中科龙网创新科技有限公司 | A kind of tractor methods of cultivation based on deep learning system |
-
2019
- 2019-04-19 CN CN201910317820.6A patent/CN110084367B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018120016A1 (en) * | 2016-12-30 | 2018-07-05 | 上海寒武纪信息科技有限公司 | Apparatus for executing lstm neural network operation, and operational method |
| CN108738444A (en) * | 2018-03-01 | 2018-11-06 | 洛阳中科龙网创新科技有限公司 | A kind of tractor methods of cultivation based on deep learning system |
Non-Patent Citations (2)
| Title |
|---|
| 张学艺等: "几种干旱监测模型在宁夏的对比应用", 《农业工程学报》 * |
| 聂红梅等: "基于PCA-SVR的冬小麦土壤水分预测", 《土壤》 * |
Cited By (32)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN110567534B (en) * | 2019-09-10 | 2021-08-13 | 广东工业大学 | Method for predicting flow of combustion air outlet in glass melting furnace and related device |
| CN110567534A (en) * | 2019-09-10 | 2019-12-13 | 广东工业大学 | Method for predicting flow of combustion air outlet in glass melting furnace and related device |
| CN111610406A (en) * | 2020-04-24 | 2020-09-01 | 国网河北省电力有限公司电力科学研究院 | Grounding grid corrosion prediction method based on deep learning |
| CN111561972A (en) * | 2020-06-18 | 2020-08-21 | 华南农业大学 | Soil water content prediction system and method based on time sequence |
| CN111859814A (en) * | 2020-07-30 | 2020-10-30 | 中国电建集团昆明勘测设计研究院有限公司 | Rock aging deformation prediction method and system based on LSTM deep learning |
| CN111859814B (en) * | 2020-07-30 | 2023-07-28 | 中国电建集团昆明勘测设计研究院有限公司 | Rock aging deformation prediction method and system based on LSTM deep learning |
| CN112115984A (en) * | 2020-08-28 | 2020-12-22 | 安徽农业大学 | Tea garden abnormal data correction method and system based on deep learning and storage medium |
| CN112287294A (en) * | 2020-09-10 | 2021-01-29 | 河海大学 | Time-space bidirectional soil water content interpolation method based on deep learning |
| CN112287294B (en) * | 2020-09-10 | 2024-02-27 | 河海大学 | Space-time bidirectional soil water content interpolation method based on deep learning |
| CN112255095A (en) * | 2020-09-25 | 2021-01-22 | 汕头大学 | Soil stress-strain relation determining method |
| CN112255095B (en) * | 2020-09-25 | 2023-12-01 | 汕头大学 | Soil stress-strain relation determination method |
| CN112784331A (en) * | 2020-09-25 | 2021-05-11 | 汕头大学 | Soil stress-strain relation determination method based on improved LSTM deep learning method |
| CN113257008A (en) * | 2021-05-12 | 2021-08-13 | 兰州交通大学 | Pedestrian flow dynamic control system and method based on deep learning |
| WO2022245312A1 (en) * | 2021-05-18 | 2022-11-24 | Hi̇t Bi̇li̇şi̇m Danişmanlik Turi̇zm Hi̇zmetleri̇ Sanayi̇ Ve Ti̇caret Li̇mi̇ted Şi̇rketi̇ | A method for estimating soil moisture and a system operating according to said method |
| CN113435707A (en) * | 2021-06-03 | 2021-09-24 | 大连钜智信息科技有限公司 | Soil testing and formulated fertilization method based on deep learning and weighted multi-factor evaluation |
| CN113323676B (en) * | 2021-06-03 | 2024-03-22 | 上海市隧道工程轨道交通设计研究院 | Method for determining cutter torque of shield machine by using principal component analysis-long and short memory model |
| CN113435707B (en) * | 2021-06-03 | 2023-11-10 | 大连钜智信息科技有限公司 | Soil testing formula fertilization method based on deep learning and weighting multi-factor evaluation |
| CN113323676A (en) * | 2021-06-03 | 2021-08-31 | 上海市隧道工程轨道交通设计研究院 | Method for determining cutter head torque of shield tunneling machine by using principal component analysis-length memory model |
| CN115511194A (en) * | 2021-06-29 | 2022-12-23 | 布瑞克农业大数据科技集团有限公司 | Agricultural data processing method, system, device and medium |
| CN113435649A (en) * | 2021-06-29 | 2021-09-24 | 布瑞克农业大数据科技集团有限公司 | Global agricultural data sorting method, system, device and medium |
| CN113468810A (en) * | 2021-07-01 | 2021-10-01 | 天行智控(成都)科技有限公司 | Intelligent floor sensing indoor tumble prediction model and establishment method thereof |
| CN113796228A (en) * | 2021-09-26 | 2021-12-17 | 南京邮电大学 | Plant cultivation system and method based on digital twinning |
| CN113796228B (en) * | 2021-09-26 | 2022-08-23 | 南京邮电大学 | Plant cultivation system and method based on digital twinning |
| CN114080953A (en) * | 2021-11-05 | 2022-02-25 | 山东省农业机械科学研究院 | Illumination management method and system for mushroom house |
| CN114451257A (en) * | 2021-12-23 | 2022-05-10 | 珠海格力电器股份有限公司 | Irrigation method and device based on neural network, storage medium and electronic equipment |
| CN117272813A (en) * | 2023-09-26 | 2023-12-22 | 长春师范大学 | Soil humidity prediction method based on water balance constraint deep learning |
| CN117272813B (en) * | 2023-09-26 | 2024-04-30 | 长春师范大学 | Soil humidity prediction method based on water balance constraint deep learning |
| CN117494902A (en) * | 2023-11-22 | 2024-02-02 | 山东大学 | Soil moisture content prediction method and system based on soil moisture correlation analysis |
| CN117494902B (en) * | 2023-11-22 | 2024-04-16 | 山东大学 | Soil moisture content prediction method and system based on soil moisture correlation analysis |
| CN117933946A (en) * | 2024-03-20 | 2024-04-26 | 华兴国创(北京)科技有限公司 | Rural business management method based on big data |
| CN117933946B (en) * | 2024-03-20 | 2024-06-14 | 华兴国创(北京)科技有限公司 | Rural business management method based on big data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110084367B (en) | 2022-10-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110084367A (en) | A kind of Forecast of Soil Moisture Content method based on LSTM deep learning model | |
| Han et al. | Crop evapotranspiration prediction by considering dynamic change of crop coefficient and the precipitation effect in back-propagation neural network model | |
| CN102550374B (en) | Crop irrigation system combined with computer vision and multi-sensor | |
| CN110084417A (en) | A kind of strawberry greenhouse environment parameter intelligent monitor system based on GRNN neural network | |
| CN113159439B (en) | Crop yield prediction method, system, storage medium and electronic equipment | |
| CN107705556A (en) | A kind of traffic flow forecasting method combined based on SVMs and BP neural network | |
| CN107466816A (en) | A kind of irrigation method based on dynamic multilayer extreme learning machine | |
| Priya et al. | An IoT based gradient descent approach for precision crop suggestion using MLP | |
| Yu et al. | A deep learning approach for multi-depth soil water content prediction in summer maize growth period | |
| CN110163254A (en) | A kind of cucumber green house yield intelligent Forecasting device based on recurrent neural network | |
| CN110119169A (en) | A kind of tomato greenhouse temperature intelligent early warning system based on minimum vector machine | |
| CN112396152A (en) | Flood forecasting method based on CS-LSTM | |
| CN110119086A (en) | A kind of tomato greenhouse environmental parameter intelligent monitoring device based on ANFIS neural network | |
| CN113554522A (en) | Vineyard accurate drip irrigation control system based on dynamic neural network | |
| CN110119767A (en) | A kind of cucumber green house temperature intelligent detection device based on LVQ neural network | |
| CN117114374B (en) | Intelligent agricultural irrigation management system based on weather prediction | |
| CN105184400A (en) | Tobacco field soil moisture prediction method | |
| CN112270124A (en) | Real-time irrigation method and system | |
| CN108984995A (en) | A kind of ecology garden landscape design method of evaluation simulation | |
| Shang et al. | Research on intelligent pest prediction of based on improved artificial neural network | |
| Zuxing et al. | A Prediction Model of Forest Preliminary Precision Fertilization Based on Improved GRA‐PSO‐BP Neural Network | |
| CN109190810A (en) | The prediction technique of grassland in northern China area NDVI based on TDNN | |
| CN114357737A (en) | Agent optimization calibration method for time-varying parameters of large-scale hydrological model | |
| CN109272144A (en) | The prediction technique of grassland in northern China area NDVI based on BPNN | |
| Lyu et al. | Design of irrigation control system for vineyard based on lora wireless communication and dynamic neural network |
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 |