CN107169610A - Aquaculture dissolved oxygen prediction method and device - Google Patents
Aquaculture dissolved oxygen prediction method and device Download PDFInfo
- Publication number
- CN107169610A CN107169610A CN201710409841.1A CN201710409841A CN107169610A CN 107169610 A CN107169610 A CN 107169610A CN 201710409841 A CN201710409841 A CN 201710409841A CN 107169610 A CN107169610 A CN 107169610A
- Authority
- CN
- China
- Prior art keywords
- data
- dissolved oxygen
- normalization
- neural network
- water quality
- 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
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 title claims abstract description 129
- 239000001301 oxygen Substances 0.000 title claims abstract description 129
- 229910052760 oxygen Inorganic materials 0.000 title claims abstract description 129
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000009360 aquaculture Methods 0.000 title claims abstract description 36
- 244000144974 aquaculture Species 0.000 title claims abstract description 36
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 94
- 238000013528 artificial neural network Methods 0.000 claims abstract description 77
- 238000010606 normalization Methods 0.000 claims abstract description 76
- 238000012545 processing Methods 0.000 claims abstract description 24
- 239000003643 water by type Substances 0.000 claims abstract description 20
- 241001269238 Data Species 0.000 claims abstract description 12
- 238000012360 testing method Methods 0.000 claims description 28
- 238000012549 training Methods 0.000 claims description 20
- XKMRRTOUMJRJIA-UHFFFAOYSA-N ammonia nh3 Chemical compound N.N XKMRRTOUMJRJIA-UHFFFAOYSA-N 0.000 claims description 14
- 230000000644 propagated effect Effects 0.000 claims description 13
- 238000012937 correction Methods 0.000 claims description 4
- 210000004218 nerve net Anatomy 0.000 claims description 3
- 210000004027 cell Anatomy 0.000 description 19
- 238000010586 diagram Methods 0.000 description 10
- 230000007935 neutral effect Effects 0.000 description 8
- 230000004913 activation Effects 0.000 description 6
- 230000006403 short-term memory Effects 0.000 description 5
- 230000005855 radiation Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- -1 pH value Substances 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 241000251468 Actinopterygii Species 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000020411 cell activation Effects 0.000 description 1
- 230000006854 communication Effects 0.000 description 1
- 210000004209 hair Anatomy 0.000 description 1
- 201000004792 malaria Diseases 0.000 description 1
- 230000015654 memory Effects 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 241000894007 species Species 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000000714 time series forecasting Methods 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
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- Life Sciences & Earth Sciences (AREA)
- Tourism & Hospitality (AREA)
- General Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- Animal Husbandry (AREA)
- Biomedical Technology (AREA)
- Operations Research (AREA)
- Agronomy & Crop Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Marine Sciences & Fisheries (AREA)
- Mining & Mineral Resources (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Primary Health Care (AREA)
- Artificial Intelligence (AREA)
- Quality & Reliability (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the present invention provides a kind of aquaculture dissolved oxygen prediction method and device.Method includes：Gather the water quality data in aquaculture waters；The water quality data collected is normalized, normalization data is obtained；The normalization data is inputted into default LSTM neural network prediction models, model prediction data is obtained；The model prediction data is subjected to renormalization processing, dissolved oxygen prediction data are obtained.The embodiment of the present invention is by inputting default LSTM neural network prediction models after the water quality data collected is normalized, obtain model prediction data, model prediction data is subjected to renormalization processing again, dissolved oxygen prediction data can be obtained, solve the problem of can not being predicted in the prior art using other water quality datas to dissolved oxygen, there is provided a kind of Forecasting Methodology of new dissolved oxygen, and improve the precision of prediction of dissolved oxygen.
Description
Technical field
The present embodiments relate to technical field of aquaculture, more particularly to a kind of aquaculture dissolved oxygen prediction method and
Device.
Background technology
China is aquaculture big country, and the guarantee of output of aquatic products and quality depends on cultivation water if appropriate for aquatic life
Thing is survived.The water quality factors of influence dissolved oxygen concentration have a lot, including pH value, water temperature, turbidity, salinity etc.；External factor includes
Solar radiation, air themperature, air pressure etc..Dissolved oxygen is aquatile existence essential condition, and dissolved oxygen concentration is less than
During 3mg/L, fish will be caused dead.Therefore, the dissolved oxygen data of acquisition Cultivated water have very heavy for aquaculture
The meaning wanted.
The Forecasting Methodology of dissolved oxygen is broadly divided into two types in the prior art：The first kind is traditional Forecasting Methodology,
Classical mathematics is predicted as theoretical foundation, including time series forecasting, regression analysis, Markov model, simulation of water quality
Method etc.；Equations of The Second Kind is the Forecasting Methodology based on artificial intelligence, including gray model, neural network prediction method, supporting vector
Machine Regression Forecast etc..These Forecasting Methodologies are predicted both for the dissolved oxygen data detected, when can not detect dissolving
Oxygen, which can only be detected, can not then know the dissolved oxygen data in water during other water quality informations, it is impossible to which the dissolved oxygen data in water are carried out
Prediction.
Dissolved oxygen data can be carried out in the case where dissolved oxygen data can not be detected there is presently no a kind of method pre-
Survey, therefore it provides a kind of is current industry skill urgently to be resolved hurrily to the method that dissolved oxygen is predicted using other water quality datas
Art problem.
The content of the invention
In order to solve problems of the prior art, the embodiment of the present invention provides a kind of aquaculture dissolved oxygen prediction side
Method and device.
On the one hand, the embodiment of the present invention provides a kind of aquaculture dissolved oxygen prediction method, including：
Gather the water quality data in aquaculture waters；
The water quality data collected is normalized, normalization data is obtained；
The normalization data is inputted into default long short-term memory LSTM neural network prediction models, model prediction is obtained
Data；
The model prediction data is subjected to renormalization processing, dissolved oxygen prediction data are obtained.
On the other hand, the embodiment of the present invention provides a kind of aquaculture dissolved oxygen prediction device, including：
Collecting unit, the water quality data for gathering aquaculture waters；
Normalization unit, for the water quality data collected to be normalized, obtains normalization data；
Predicting unit, for the normalization data to be inputted into default LSTM neural network prediction models, obtains model
Prediction data；
Renormalization unit, for the model prediction data to be carried out into renormalization processing, obtains dissolved oxygen prediction number
According to.
Aquaculture dissolved oxygen prediction method and device provided in an embodiment of the present invention, by by the water quality data collected
Default LSTM neural network prediction models are inputted after being normalized, model prediction data is obtained, then by model prediction
Data carry out renormalization processing, can obtain dissolved oxygen prediction data, other water quality can not be utilized in the prior art by solving
The problem of data are predicted to dissolved oxygen improves the pre- of dissolved oxygen there is provided a kind of Forecasting Methodology of new dissolved oxygen
Survey precision.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
There is the accompanying drawing used required in technology description to be briefly described, it should be apparent that, drawings in the following description are this hairs
Some bright embodiments, for those of ordinary skill in the art, on the premise of not paying creative work, can be with root
Other accompanying drawings are obtained according to these accompanying drawings.
Fig. 1 is aquaculture dissolved oxygen prediction method flow schematic diagram provided in an embodiment of the present invention；
Fig. 2 is another aquaculture dissolved oxygen prediction method flow schematic diagram provided in an embodiment of the present invention；
Fig. 3 is LSTM neutral net basic structure schematic diagrams；
Fig. 4 is aquaculture dissolved oxygen prediction apparatus structure schematic diagram provided in an embodiment of the present invention；
Fig. 5 is another aquaculture dissolved oxygen prediction apparatus structure schematic diagram provided in an embodiment of the present invention.
Embodiment
To make the purpose, technical scheme and advantage of the embodiment of the present invention clearer, below in conjunction with the embodiment of the present invention
In accompanying drawing, the technical scheme in the embodiment of the present invention is clearly and completely described, it is clear that described embodiment is
A part of embodiment of the present invention, rather than whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art
The every other embodiment obtained under the premise of creative work is not made, belongs to the scope of protection of the invention.
Fig. 1 is aquaculture dissolved oxygen prediction method flow schematic diagram provided in an embodiment of the present invention, as shown in figure 1, side
Method includes：
Step 10, the water quality data in collection aquaculture waters；
Step 20, the water quality data collected is normalized, obtains normalization data；
Step 30, the normalization data inputted into default long short-term memory LSTM neural network prediction models, obtained
Model prediction data；
Step 40, by the model prediction data carry out renormalization processing, obtain dissolved oxygen prediction data.
Specifically, when needing the dissolved oxygen data to aquaculture waters to be predicted：
(1) water quality information in one group of aquaculture waters is gathered using experimental provision；
(2) water quality information of collection has different dimension and dimensional unit, and such situation influences whether to predict the outcome,
In order to eliminate the dimension impact between data, it is necessary to carry out data normalization processing, that is, the water quality data collected is entered
Row normalized, to solve the comparativity between data target.After data normalization is handled, each index is in same number
Magnitude, is convenient for next step data processing.
(3) normalization data obtained after normalized is inputted into default long short-term memory (Long Short-Term
Memory, abbreviation LSTM) in neural network prediction model, the LSTM neural network prediction models are exclusively for prediction dissolved oxygen
The forecast model that data are set up, it is possible to use the normalization data of input obtains model prediction data.
(4) model prediction data of LSTM neural network prediction models output is to calculate one obtained by normalization data
Data, and the standard of the dissolved oxygen data for the description that is not accordant to the old routine, therefore obtained model prediction data is subjected to renormalization
Processing, you can obtain dissolved oxygen prediction data.
Aquaculture dissolved oxygen prediction method provided in an embodiment of the present invention, by the way that the water quality data collected is returned
Default LSTM neural network prediction models are inputted after one change processing, model prediction data are obtained, then model prediction data is entered
The processing of row renormalization, can obtain dissolved oxygen prediction data, other water quality datas pair can not be utilized in the prior art by solving
The problem of dissolved oxygen is predicted improves the precision of prediction of dissolved oxygen there is provided a kind of Forecasting Methodology of new dissolved oxygen.
On the basis of above-described embodiment, further, the water quality data includes：PH value, water temperature, turbidity, ammonia nitrogen contain
One or more in amount and salinity.
The factor of influence oxygen in water concentration has a lot, including the water such as pH value, water temperature, turbidity, ammonia-nitrogen content and salinity
Quality factor, and the external factor such as air themperature, solar radiation and air pressure.The water quality data of collection of the embodiment of the present invention is pH
One or more in value, water temperature, turbidity, ammonia-nitrogen content and salinity, these water quality datas are easy to collection and dense with dissolved oxygen
The correlation of degree is higher, higher using the dissolved oxygen data precision of these data predictions, also, the data class of collection is more,
The precision of obtained dissolved oxygen prediction data is also corresponding higher.
Fig. 2 is another aquaculture dissolved oxygen prediction method flow schematic diagram provided in an embodiment of the present invention, such as Fig. 2 institutes
Show, methods described includes：
Step 00, set up LSTM neural network prediction models；
Step 10, the water quality data in collection aquaculture waters；
Step 20, the water quality data collected is normalized, obtains normalization data；
Step 30, the normalization data inputted into default LSTM neural network prediction models, obtain model prediction number
According to；
Step 40, by the model prediction data carry out renormalization processing, obtain dissolved oxygen prediction data.
When the dissolved oxygen data in waters are predicted, first have to set up a LSTM neural network prediction model,
The precision of LSTM neural network prediction models is higher, and the model prediction data obtained by LSTM neural network prediction models is got over
Accurately, the dissolved oxygen prediction data obtained after renormalization processing are also just closer to actual value.Therefore, the embodiment of the present invention is provided
Method by setting up LSTM neural network prediction models, the precision of dissolved oxygen prediction data can be greatly improved.
On the basis of the various embodiments described above, further, the method bag for setting up LSTM neural network prediction models
Include：
M group test datas are gathered, the test data includes water quality data and corresponding dissolved oxygen data；
The M groups test data is normalized, M groups normalization test data is obtained, wherein, the water quality number
According to normalization water quality data is referred to as after being normalized, normalization dissolved oxygen number is referred to as after the dissolved oxygen data normalization
According to；
The normalization test data is randomly divided into two parts by group, wherein quantity claims more than the data of preset percentage
For training data, weights are adjusted for training pattern；Another part data are referred to as verifying data, for verifying model；
Water quality data will be normalized as the input of LSTM neural network prediction models in the training data, according to pre-
Equipment, method carries out obtaining predicting dissolved oxygen data after propagated forward；
The prediction dissolved oxygen data are compared with corresponding normalization dissolved oxygen data, dissolved using the prediction
The error of oxygen data and normalization dissolved oxygen data carries out the weights in backpropagation, correction model, the LSTM god corrected
Through Network Prediction Model；
Water quality data will be normalized as the input of the LSTM neural network prediction models of amendment in the checking data,
If the prediction dissolved oxygen data of output are being preset in error range with the normalization dissolved oxygen data in checking data, described to repair
Positive LSTM neural network prediction models are final LSTM neural network prediction models.
Fig. 3 is LSTM neutral net basic structure schematic diagrams, is followed as shown in figure 3, LSTM neutral nets are substantially one
Ring neutral net (Recurrent neural Network, abbreviation RNN), it is hidden on the basis of RNN that LSTM neutral nets, which are,
Each node of layer has opened three fan doors.LSTM neutral net basic structures are referred to as a block.
Cell is the memory of neuron state, with parameter State (S) come recording status.
Input gate Input Gate：Indicate whether to allow information to be added in current hidden node, if Input Gate
Value be 1, then it represents that allow data input, if Input Gate value be 0, then it represents that do not allow data input, thus
Some input information useless can be abandoned；
Forget door Forget Gate：Indicate whether to retain the historical information that current hidden node is stored, if Forget
Gate value is 1, then retains the data of current hidden node storage, if Forget Gate value is 0, is emptied current hidden
The historical information that node layer is stored；
Out gate Output Gate：Indicate whether to export present node output valve to next layer, if Output
Gate value is 1, then the output valve of present node will act on next layer, if Output Gate value is 0, work as prosthomere
The output valve of point will not act on next layer.
(1) collecting test data：
When setting up LSTM neural network prediction models first with corresponding harvester, according to certain time interval
M group test datas are gathered, wherein, the water quality data of t and the dissolved oxygen data at t+1 moment are used as one group of data, water quality number
According to including：PH value, water temperature, turbidity, ammonia-nitrogen content and salinity, the t+1 moment are next collection moment of t；Collection is as far as possible
Many test datas, because substantial amounts of data are conducive to improving the precision of model.
(2) test data collected is pre-processed：
(a) test data is normalized, obtains normalization water quality data and normalize dissolved oxygen data.Collection
The test data arrived includes pH value, water temperature, turbidity, ammonia nitrogen, salinity and dissolved oxygen, and these data have different dimensions and dimension
Unit, such situation influences whether to predict the outcome, in order to eliminate the dimension impact between data, it is necessary to carry out data normalization
Processing, that is, the water quality data collected is normalized, to solve the comparativity between data target.By number
After normalized, each index is in the same order of magnitude, is convenient for next step data processing.
(b) test data after M group normalizeds is upset at random, makes to be completely independent and without any between each group
System.
(c) data after upsetting are divided into two parts, wherein quantity is referred to as training data more than the data of preset percentage,
Weights are adjusted for training pattern；Another part data are referred to as verifying data, for verifying model.For example, taking 75% data
As training data, then remaining 25% data, which are used as, verifies data.
(3) parameter of LSTM neural network prediction models is trained using training data.
The basic structure for being previously noted LSTM neutral nets is that have input gate in a block, each block, forget door
Door is fanned with out gate three.It will be introduced below with the communication process of three fan doors in a certain block：
(a) propagated forward process.
As shown in figure 3, the input of input gate includes：Input layer；Previous moment hidden layer exports cell outputs,
That is the dotted line from Cell, referred to as peephole connections；And the hidden layer state of previous moment ' input gate itself '.Therefore,
Value in input gate is：
Value after input gate is：
Wherein, l represents input gate,For the value in input gate, i is the number of input layer, and h is previous moment hidden layer
The number of output, the number that c is cell in block,The value inputted for input layer,It is defeated for previous moment hidden layer
Go out the value of input,The value inputted for the hidden layer state of previous moment ' input gate itself ', wil、whlAnd wclIt is respectively each defeated
The weights entered,For by the value after input gate, f () is the activation primitive of input gate.
Likewise, forgeing the input of door includes：Input layer；Previous moment hidden layer output cell outputs；With
And the hidden layer state of previous moment ' forgeing door itself '.Therefore, the value in forgetting door is：
Through forgetting door after value be：
Wherein,Represent to forget door,To forget the value in door, i is the number of input layer, and h is that previous moment is hidden
The number of layer output, the number that c is cell in block,The value inputted for input layer,For previous moment hidden layer
The value of input is exported,The value inputted for the hidden layer state of previous moment ' forgeing door itself ',WithPoint
Not Wei each input weights,For the value after forgetting door, f () is the activation primitive for forgeing door.
Cell is first calculated before Cell is calculated, and cell shows as nethermost circle in figure 3, rather than middle
Cell.Cell source has two：The historical information that input layer and previous moment ' oneself ' are transmitted to.
Therefore cell value is：
Then Cell calculation formula are：
Wherein, c represents cell,For the value in cell, i is the number of input layer, and h is that previous moment hidden layer is defeated
The number gone out,The value inputted for input layer,The value of input is exported for previous moment hidden layer, is previous moment ' oneself
Oneself ' hidden layer state input value, wicAnd whcThe weights of respectively each input,For Cell value,For by input gate it
Value afterwards,For the value after forgetting door, g () is Cell activation primitive.
The historical information that Cell state values=forgetting door * last moments ' oneself ' is preserved+ input gate * passes through g ()
Value after activation primitive activation.
Out gate is identical with the principle of input gate, and the value in out gate is：
Value after out gate is：
Wherein, ω represents out gate,To forget the value in door, i is the number of input layer, and h is that previous moment is hidden
The number of layer output, the number that c is cell in block,The value inputted for input layer,For previous moment hidden layer
The value of input is exported,The value inputted for the hidden layer state of previous moment ' oneself ', wiω、whωAnd wcωRespectively each input
Weights,For by the value after out gate, f () is the activation primitive of out gate.
Input gate, the activation primitive for forgeing door, out gate and Cell, can use S sigmoid growth curve sigmod functions.
(b) back-propagating process.
According to predicted value and the error of actual value, network backpropagation is carried out, local derviation is asked to the parameter in each door, by ladder
Spend adjustment in direction weights, it is established that the LSTM neural network prediction models of best initial weights, i.e., the LSTM neutral nets of described amendment
Forecast model.Method during back-propagating is same as the prior art, and here is omitted.
For example, pH value, water temperature, turbidity, ammonia-nitrogen content and salinity are respectively x1, x2, x3, x4 and x5 in a certain group of data,
Dissolved oxygen is y.Using x1, x2, x3, x4 and x5 numerical value as LSTM neural network prediction models input, according to current LSTM
Weights in neural network prediction model are calculated, and obtain exporting y1.Then according to y1 and y error, then reversely input
Each weights, i.e., above-mentioned w in LSTM neural network prediction models, regulation modelil、whl、wcl、 wic、
whc、wiω、whωAnd wcω.Then the next group of above-mentioned action of Data duplication is recycled, until y1 and the target y values error of output expire
Foot is required.
(4) by the LSTM neural network prediction models for the normalization water quality data Introduced Malaria verified in data.If output
Value the LSTM neural network prediction moulds are proved in error range, then with the value of normalization dissolved oxygen data in checking data
Type meets forecast demand, and LSTM neural network prediction models are successfully established.
Method provided in an embodiment of the present invention, collection mass data is by propagated forward and backpropagation to LSTM nerve nets
Network forecast model is trained, and using verifying that data verify to LSTM neural network prediction models, so as to obtain precision
Higher LSTM neural network prediction models, are conducive to improving the precision of prediction of dissolved oxygen.
Fig. 4 is aquaculture dissolved oxygen prediction apparatus structure schematic diagram provided in an embodiment of the present invention, as shown in figure 4, dress
Put including：Collecting unit 41, normalization unit 42, predicting unit 43 and renormalization unit 44；Wherein, collecting unit 41 is used for
Gather the water quality data in aquaculture waters；Normalization unit 42 is used to the water quality data collected being normalized,
Obtain normalization data；Predicting unit 43 is used to the normalization data inputting default LSTM neural network prediction models,
Obtain model prediction data；Renormalization unit 44 is used to the model prediction data carrying out renormalization processing, obtains molten
Solve oxygen prediction data.
Specifically, when needing the dissolved oxygen data to aquaculture waters to be predicted：
(1) collecting unit 41 gathers the water quality information in one group of aquaculture waters using experimental provision；
(2) water quality information of collection has different dimension and dimensional unit, and such situation influences whether to predict the outcome,
In order to eliminate the dimension impact between data, it is necessary to carry out data normalization processing, that is, the water quality data collected is entered
Row normalized, to solve the comparativity between data target.After normalization unit 42 is to data normalized, respectively
Index is in the same order of magnitude, is convenient for next step data processing.
(3) the normalization data input prediction unit 43 obtained after normalization unit 42 is normalized is default
In long short-term memory LSTM neural network prediction models, the LSTM neural network prediction models are exclusively for prediction dissolved oxygen number
According to the forecast model of foundation, predicting unit 43 can obtain model prediction data using the normalization data of input.
(4) model prediction data of LSTM neural network prediction models output is to calculate one obtained by normalization data
Data, and the standard of the dissolved oxygen data for the description that is not accordant to the old routine, therefore by renormalization unit 44 by obtained model prediction
Data carry out renormalization processing, you can obtain dissolved oxygen prediction data.
Aquaculture dissolved oxygen prediction device provided in an embodiment of the present invention, by the way that the water quality data collected is returned
Default LSTM neural network prediction models are inputted after one change processing, model prediction data are obtained, then model prediction data is entered
The processing of row renormalization, can obtain dissolved oxygen prediction data, other water quality datas pair can not be utilized in the prior art by solving
The problem of dissolved oxygen is predicted improves the precision of prediction of dissolved oxygen there is provided a kind of Forecasting Methodology of new dissolved oxygen.
Fig. 5 is another aquaculture dissolved oxygen prediction apparatus structure schematic diagram provided in an embodiment of the present invention, such as Fig. 5 institutes
Show, device includes：Collecting unit 41, normalization unit 42, predicting unit 43 and collecting unit bag described in renormalization unit 44
Include：PH value acquisition module 411, water temperature acquisition module 412, turbidity acquisition module 413, ammonia-nitrogen content acquisition module 414 and salinity
Acquisition module 415.
Specifically, when collecting unit 41 carries out data acquisition, the pH value in waters is gathered by pH value acquisition module 411, by water
Warm acquisition module 412 gathers the water temperature in waters, and the turbidity in waters is gathered by turbidity acquisition module 413, mould is gathered by ammonia-nitrogen content
Block 414 gathers the ammonia-nitrogen content in waters, and the salinity in waters is gathered by salinity acquisition module 415.
The factor of influence oxygen in water concentration has a lot, including the water such as pH value, water temperature, turbidity, ammonia-nitrogen content and salinity
Quality factor, and the external factor such as air themperature, solar radiation and air pressure.The embodiment of the present invention using H values acquisition module 411,
The water that water temperature acquisition module 412, turbidity acquisition module 413, ammonia-nitrogen content acquisition module 414 and salinity acquisition module 415 are gathered
Prime number according to be pH value, water temperature, turbidity, ammonia-nitrogen content and salinity in one or more, these water quality datas be easy to collection and
Correlation with dissolved oxygen concentration is higher, higher using the dissolved oxygen data precision of these data predictions, also, the data of collection
Species is more, and the precision of obtained dissolved oxygen prediction data is also corresponding higher.
On the basis of above-described embodiment, further, described device also includes LSTM neural network prediction models and set up
Unit, for setting up LSTM neural network prediction models.
When the dissolved oxygen data in waters are predicted, first have to set up a LSTM neural network prediction model,
The precision of LSTM neural network prediction models is higher, and the model prediction data obtained by LSTM neural network prediction models is got over
Accurately, the dissolved oxygen prediction data obtained after renormalization processing are also just closer to actual value.Therefore, the embodiment of the present invention is provided
Device set up unit using LSTM neural network prediction models and set up LSTM neural network prediction models, can greatly improve molten
Solve the precision of oxygen prediction data.
On the basis of the various embodiments described above, further, the LSTM neural network prediction models of setting up set up unit
Including：Acquisition module, normalization module, data division module, propagated forward module, backpropagation module and authentication module, its
In, acquisition module is used to gather M group test datas, and the test data includes water quality data and corresponding dissolved oxygen number
According to；Normalization module is used to the M groups test data being normalized, and obtains M groups normalization test data, wherein,
The water quality data is referred to as being referred to as normalizing after normalization water quality data, the dissolved oxygen data normalization after being normalized
Dissolved oxygen data；Data division module is used to the normalization test data by group being randomly divided into two parts, wherein group number compared with
Many data are referred to as training data, for training pattern adjusting parameter；Group number less data is referred to as verifying data, for verifying
Model；Propagated forward module is used to regard the normalization water quality data in the training data as LSTM neural network prediction models
Input, according to presetting method carry out propagated forward after obtain predict dissolved oxygen data；Backpropagation module is used for will be described pre-
Dissolved oxygen data are surveyed to be compared with corresponding normalization dissolved oxygen data, it is molten with normalizing using the prediction dissolved oxygen data
Solve the weights in the error progress backpropagation of oxygen data, correction model, the LSTM neural network prediction models corrected；Test
Card module is used to regard the normalization water quality data in the checking data as the defeated of the LSTM neural network prediction models of amendment
Enter, if the prediction dissolved oxygen data of output are being preset in error range with the normalization dissolved oxygen data in checking data, institute
The LSTM neural network prediction models for stating amendment are final LSTM neural network prediction models.
Specifically, acquisition module gathers M groups water quality data and dissolved oxygen data, every group of data according to certain time interval
Middle water quality data is the data of t, and dissolved oxygen data are the data of the subsequent time of t+1 moment, i.e. water quality data.
This M group data collected are normalized normalization module, to carry out subsequent operation.
Data division module upsets this M groups data at random, is separated into two parts, regard a fairly large number of part as training
Data, using the part of negligible amounts as checking data, for example, taking 75% data as training data, remaining 25% makees
For checking data.
Using the normalization water quality data in training data as the input of propagated forward module, the defeated of propagated forward is obtained
Go out.
The error of normalization dissolved oxygen data in output and training data that propagated forward module is obtained is as reverse
The input of propagation module, and the weights in LSTM neural network prediction models are corrected during carrying out backpropagation, are obtained
The LSTM neural network prediction models of amendment.
Authentication module will verify that the normalization water quality data in data is used as the LSTM neural network prediction models of amendment
Input, if obtained predicted value is with verifying that the normalization dissolved oxygen data in data in default error range, prove this
The LSTM neural network prediction models of amendment meet prediction and required, and using this LSTM neural network prediction model as final
LSTM neural network prediction models
Device provided in an embodiment of the present invention, the acquisition module set up by LSTM neural network prediction models in unit is adopted
Collect mass data, LSTM neural network prediction models are trained by propagated forward module and backpropagation module, and profit
The LSTM neural network prediction models obtained with authentication module to training are verified, have obtained the higher LSTM nerve nets of precision
Network forecast model, is conducive to improving the precision of prediction of dissolved oxygen.
Device embodiment described above is only schematical, wherein the unit illustrated as separating component can
To be or may not be physically separate, the part shown as unit can be or may not be physics list
Member, you can with positioned at a place, or can also be distributed on multiple NEs.It can be selected according to the actual needs
In some or all of module realize the purpose of this embodiment scheme.Those of ordinary skill in the art are not paying creativeness
Work in the case of, you can to understand and implement.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can
Realized by the mode of software plus required general hardware platform, naturally it is also possible to which some functions are realized by hardware processor
Module.Understood based on such, the part that above-mentioned technical proposal substantially contributes to prior art in other words can be with soft
The form of part product is embodied, and the computer software product can be stored in a computer-readable storage medium, such as ROM/
RAM, magnetic disc, CD etc., including some instructions to cause a computer equipment (can be personal computer, server, or
Person's network equipment etc.) perform method described in some parts of each embodiment or embodiment.
Finally it should be noted that：The above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although
The present invention is described in detail with reference to the foregoing embodiments, it will be understood by those within the art that：It still may be used
To be modified to the technical scheme described in foregoing embodiments, or equivalent substitution is carried out to which part technical characteristic；
And these modification or replace, do not make appropriate technical solution essence depart from various embodiments of the present invention technical scheme spirit and
Scope.
Claims (8)
1. a kind of aquaculture dissolved oxygen prediction method, it is characterised in that including：
Gather the water quality data in aquaculture waters；
The water quality data collected is normalized, normalization data is obtained；
The normalization data is inputted into default LSTM neural network prediction models, model prediction data is obtained；
The model prediction data is subjected to renormalization processing, dissolved oxygen prediction data are obtained.
2. according to the method described in claim 1, it is characterised in that the water quality data includes：PH value, water temperature, turbidity, ammonia nitrogen
One or more in content and salinity.
3. method according to claim 1 or 2, it is characterised in that water quality of the methods described in collection aquaculture waters
Before data, in addition to：Set up LSTM neural network prediction models.
4. method according to claim 3, it is characterised in that the method bag for setting up LSTM neural network prediction models
Include：
M group test datas are gathered, the test data includes water quality data and corresponding dissolved oxygen data；
The M groups test data is normalized, M groups normalization test data is obtained, wherein, the water quality data enters
It is referred to as being referred to as normalization dissolved oxygen data after normalization water quality data, the dissolved oxygen data normalization after row normalization；
The normalization test data is randomly divided into two parts by group, wherein quantity is referred to as instruction more than the data of preset percentage
Practice data, weights are adjusted for training pattern；Another part data are referred to as verifying data, for verifying model；
Water quality data will be normalized as the input of LSTM neural network prediction models in the training data, according to default side
Method carries out obtaining predicting dissolved oxygen data after propagated forward；
The prediction dissolved oxygen data are compared with corresponding normalization dissolved oxygen data, the prediction dissolved oxygen number is utilized
The weights in backpropagation, correction model, the LSTM nerve nets corrected are carried out according to the error with normalization dissolved oxygen data
Network forecast model；
Water quality data will be normalized as the input of the LSTM neural network prediction models of amendment in the checking data, if defeated
The prediction dissolved oxygen data gone out with the normalization dissolved oxygen data in checking data in default error range, then the amendment
LSTM neural network prediction models are final LSTM neural network prediction models.
5. a kind of aquaculture dissolved oxygen prediction device, it is characterised in that including：
Collecting unit, the water quality data for gathering aquaculture waters；
Normalization unit, for the water quality data collected to be normalized, obtains normalization data；
Predicting unit, for the normalization data to be inputted into default LSTM neural network prediction models, obtains model prediction
Data；
Renormalization unit, for the model prediction data to be carried out into renormalization processing, obtains dissolved oxygen prediction data.
6. device according to claim 5, it is characterised in that the collecting unit includes：PH value acquisition module, water temperature are adopted
Collect module, turbidity acquisition module, ammonia-nitrogen content acquisition module and salinity acquisition module.
7. the device according to claim 5 or 6, it is characterised in that described device also includes LSTM neural network prediction moulds
Type sets up unit, for setting up LSTM neural network prediction models.
8. device according to claim 7, it is characterised in that the LSTM neural network prediction models of setting up set up unit
Including：
Acquisition module, for gathering M group test datas, the test data includes water quality data and corresponding dissolved oxygen number
According to；
Module is normalized, for the M groups test data to be normalized, M groups normalization test data is obtained, its
In, the water quality data is referred to as being referred to as returning after normalization water quality data, the dissolved oxygen data normalization after being normalized
One changes dissolved oxygen data；
Data division module, for the normalization test data to be randomly divided into two parts by group, wherein the more number of group number
It is stated to be training data, for training pattern adjusting parameter；Group number less data is referred to as verifying data, for verifying model；
Propagated forward module, for regarding the normalization water quality data in the training data as LSTM neural network prediction models
Input, according to presetting method carry out propagated forward after obtain predict dissolved oxygen data；
Backpropagation module, for the prediction dissolved oxygen data to be compared with corresponding normalization dissolved oxygen data, profit
The weights in backpropagation, correction model are carried out with the error of the prediction dissolved oxygen data and normalization dissolved oxygen data, are obtained
To the LSTM neural network prediction models of amendment；
Authentication module, for water quality data will to be normalized as the LSTM neural network prediction moulds of amendment in the checking data
The input of type, if the prediction dissolved oxygen data of output are with verifying the normalization dissolved oxygen data in data in default error range
Interior, then the LSTM neural network prediction models of the amendment are final LSTM neural network prediction models.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710409841.1A CN107169610A (en) | 2017-06-02 | 2017-06-02 | Aquaculture dissolved oxygen prediction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710409841.1A CN107169610A (en) | 2017-06-02 | 2017-06-02 | Aquaculture dissolved oxygen prediction method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107169610A true CN107169610A (en) | 2017-09-15 |
Family
ID=59824346
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710409841.1A Pending CN107169610A (en) | 2017-06-02 | 2017-06-02 | Aquaculture dissolved oxygen prediction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107169610A (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108133297A (en) * | 2018-01-26 | 2018-06-08 | 广州大学 | Dissolved oxygen prediction method and system based on shot and long term memory network |
| CN108460492A (en) * | 2018-03-19 | 2018-08-28 | 国通广达（北京）技术有限公司 | A kind of pipe gallery internal oxygen concentration LSTM prediction techniques and system |
| CN108647831A (en) * | 2018-05-16 | 2018-10-12 | 华南农业大学 | Ammonia nitrogen/content of nitrite forecasting system and method in a kind of culture environment of aquatic products of Multi-source Information Fusion |
| CN108665106A (en) * | 2018-05-15 | 2018-10-16 | 中国农业大学 | A kind of aquaculture dissolved oxygen prediction method and device |
| CN108717586A (en) * | 2018-05-10 | 2018-10-30 | 江南大学 | A kind of culture environment of aquatic products dissolved oxygen prediction method based on long memory network in short-term |
| CN108894308A (en) * | 2018-07-04 | 2018-11-27 | 余蝶双 | A kind of reservoir and its water quality monitoring method |
| CN109165247A (en) * | 2018-09-30 | 2019-01-08 | 中冶华天工程技术有限公司 | Sewage measurement data intelligence preprocess method |
| CN109359624A (en) * | 2018-11-14 | 2019-02-19 | 浙江农林大学 | The prediction technique and prediction meanss of laying hen foraging behaviour neural network based |
| CN110308254A (en) * | 2019-04-22 | 2019-10-08 | 江南大学 | The accurate monitoring method of low dissolved oxygen |
| CN110334845A (en) * | 2019-04-30 | 2019-10-15 | 江南大学 | One kind being based on GRU dissolved oxygen long-time prediction technique |
| CN110598895A (en) * | 2019-07-26 | 2019-12-20 | 天津科技大学 | Aquaculture dissolved oxygen prediction method |
| CN110824914A (en) * | 2019-09-30 | 2020-02-21 | 华南师范大学 | Intelligent wastewater treatment monitoring method based on PCA-LSTM network |
| CN112529234A (en) * | 2019-09-18 | 2021-03-19 | 上海交通大学 | Surface water quality prediction method based on deep learning |
| CN113504348A (en) * | 2021-05-08 | 2021-10-15 | 中国农业大学 | Method, device, equipment and storage medium for estimating dissolved oxygen |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102183621A (en) * | 2011-02-28 | 2011-09-14 | 中国农业大学 | Aquaculture dissolved oxygen concentration online forecasting method and system |
| KR20160044790A (en) * | 2014-10-16 | 2016-04-26 | 한국수자원공사 | Group management method for hydraulic and water quality model, and recording medium storing program for executing the same, and recording medium storing program for executing the same |
| CN105956702A (en) * | 2016-04-29 | 2016-09-21 | 中国农业大学 | Aquaculture water quality dissolved oxygen prediction method based on data fusion |
| CN106198909A (en) * | 2016-06-30 | 2016-12-07 | 中南大学 | A kind of aquaculture water quality Forecasting Methodology based on degree of depth study |
| CN106599520A (en) * | 2016-12-31 | 2017-04-26 | 中国科学技术大学 | LSTM-RNN model-based air pollutant concentration forecast method |
-
2017
- 2017-06-02 CN CN201710409841.1A patent/CN107169610A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102183621A (en) * | 2011-02-28 | 2011-09-14 | 中国农业大学 | Aquaculture dissolved oxygen concentration online forecasting method and system |
| KR20160044790A (en) * | 2014-10-16 | 2016-04-26 | 한국수자원공사 | Group management method for hydraulic and water quality model, and recording medium storing program for executing the same, and recording medium storing program for executing the same |
| CN105956702A (en) * | 2016-04-29 | 2016-09-21 | 中国农业大学 | Aquaculture water quality dissolved oxygen prediction method based on data fusion |
| CN106198909A (en) * | 2016-06-30 | 2016-12-07 | 中南大学 | A kind of aquaculture water quality Forecasting Methodology based on degree of depth study |
| CN106599520A (en) * | 2016-12-31 | 2017-04-26 | 中国科学技术大学 | LSTM-RNN model-based air pollutant concentration forecast method |
Non-Patent Citations (1)
| Title |
|---|
| 龚怀瑾: ""智能算法的研究及其在水质预测上的应用"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108133297A (en) * | 2018-01-26 | 2018-06-08 | 广州大学 | Dissolved oxygen prediction method and system based on shot and long term memory network |
| CN108460492A (en) * | 2018-03-19 | 2018-08-28 | 国通广达（北京）技术有限公司 | A kind of pipe gallery internal oxygen concentration LSTM prediction techniques and system |
| CN108717586A (en) * | 2018-05-10 | 2018-10-30 | 江南大学 | A kind of culture environment of aquatic products dissolved oxygen prediction method based on long memory network in short-term |
| CN108665106A (en) * | 2018-05-15 | 2018-10-16 | 中国农业大学 | A kind of aquaculture dissolved oxygen prediction method and device |
| CN108647831A (en) * | 2018-05-16 | 2018-10-12 | 华南农业大学 | Ammonia nitrogen/content of nitrite forecasting system and method in a kind of culture environment of aquatic products of Multi-source Information Fusion |
| CN108894308B (en) * | 2018-07-04 | 2020-09-11 | 台州市湛迪精密模具配件有限公司 | Reservoir and water quality monitoring method thereof |
| CN108894308A (en) * | 2018-07-04 | 2018-11-27 | 余蝶双 | A kind of reservoir and its water quality monitoring method |
| CN109165247A (en) * | 2018-09-30 | 2019-01-08 | 中冶华天工程技术有限公司 | Sewage measurement data intelligence preprocess method |
| CN109165247B (en) * | 2018-09-30 | 2021-07-23 | 中冶华天工程技术有限公司 | Intelligent pretreatment method for sewage measurement data |
| CN109359624A (en) * | 2018-11-14 | 2019-02-19 | 浙江农林大学 | The prediction technique and prediction meanss of laying hen foraging behaviour neural network based |
| CN109359624B (en) * | 2018-11-14 | 2021-10-08 | 浙江农林大学 | Neural network-based prediction method and prediction device for feeding behavior of laying hens |
| CN110308254A (en) * | 2019-04-22 | 2019-10-08 | 江南大学 | The accurate monitoring method of low dissolved oxygen |
| CN110334845A (en) * | 2019-04-30 | 2019-10-15 | 江南大学 | One kind being based on GRU dissolved oxygen long-time prediction technique |
| CN110598895A (en) * | 2019-07-26 | 2019-12-20 | 天津科技大学 | Aquaculture dissolved oxygen prediction method |
| CN112529234A (en) * | 2019-09-18 | 2021-03-19 | 上海交通大学 | Surface water quality prediction method based on deep learning |
| CN110824914A (en) * | 2019-09-30 | 2020-02-21 | 华南师范大学 | Intelligent wastewater treatment monitoring method based on PCA-LSTM network |
| CN110824914B (en) * | 2019-09-30 | 2022-07-12 | 华南师范大学 | Intelligent wastewater treatment monitoring method based on PCA-LSTM network |
| CN113504348A (en) * | 2021-05-08 | 2021-10-15 | 中国农业大学 | Method, device, equipment and storage medium for estimating dissolved oxygen |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107169610A (en) | Aquaculture dissolved oxygen prediction method and device | |
| CN105488528B (en) | Neural network image classification method based on improving expert inquiry method | |
| CN102622515B (en) | A kind of weather prediction method | |
| CN109190665A (en) | A kind of general image classification method and device based on semi-supervised generation confrontation network | |
| Sha et al. | A breast cancer classifier using a neuron model with dendritic nonlinearity | |
| RU2586864C2 (en) | Method and arrangement for local rule of training competition, which leads to a sparse connectivity | |
| CN108427921A (en) | A kind of face identification method based on convolutional neural networks | |
| CN107862331A (en) | It is a kind of based on time series and CNN unsafe acts recognition methods and system | |
| CN108763857A (en) | A kind of process soft-measuring modeling method generating confrontation network based on similarity | |
| CN106982359A (en) | A kind of binocular video monitoring method, system and computer-readable recording medium | |
| CN105956702A (en) | Aquaculture water quality dissolved oxygen prediction method based on data fusion | |
| CN107590569A (en) | A kind of data predication method and device | |
| CN108647583A (en) | A kind of face recognition algorithms training method based on multiple target study | |
| CN106503802A (en) | A kind of method of utilization genetic algorithm optimization BP neural network system | |
| CN103105246A (en) | Greenhouse environment forecasting feedback method of back propagation (BP) neural network based on improvement of genetic algorithm | |
| CN109360604A (en) | A kind of oophoroma molecule parting forecasting system | |
| CN107016453A (en) | A kind of aquaculture dissolved oxygen prediction method and device | |
| CN108665109A (en) | A kind of reservoir parameter log interpretation method based on recurrence committee machine | |
| CN107506862A (en) | A kind of online real-time estimate system and method for grinding particle size based on Internet of Things | |
| CN109919229A (en) | Monitoring pernicious gas prediction technique and system based on artificial bee colony and neural network | |
| CN108416460A (en) | Cyanobacterial bloom prediction technique based on the random depth confidence network model of multifactor sequential- | |
| CN113487165A (en) | Intelligent factory production operation scheduling method and system based on deep reinforcement learning | |
| CN108027896A (en) | System and method for decoding the pulse reservoir with continuous synaptic plasticity | |
| CN105956715A (en) | Soil moisture status prediction method and device | |
| CN107729988A (en) | Blue-green alga bloom Forecasting Methodology based on dynamic depth confidence 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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170915 |
|
| RJ01 | Rejection of invention patent application after publication |