CN117495072A - Marine pasture mechanized seedling sowing scheduling system based on machine learning - Google Patents
Marine pasture mechanized seedling sowing scheduling system based on machine learning Download PDFInfo
- Publication number
- CN117495072A CN117495072A CN202410003764.XA CN202410003764A CN117495072A CN 117495072 A CN117495072 A CN 117495072A CN 202410003764 A CN202410003764 A CN 202410003764A CN 117495072 A CN117495072 A CN 117495072A
- Authority
- CN
- China
- Prior art keywords
- planned
- seeding
- seedling sowing
- sequence
- suitability
- 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
- 238000009331 sowing Methods 0.000 title claims abstract description 143
- 238000010801 machine learning Methods 0.000 title claims abstract description 20
- 238000013528 artificial neural network Methods 0.000 claims abstract description 27
- 108091026890 Coding region Proteins 0.000 claims abstract description 13
- 238000004458 analytical method Methods 0.000 claims abstract description 7
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 238000010899 nucleation Methods 0.000 claims description 113
- 239000013598 vector Substances 0.000 claims description 74
- 210000002569 neuron Anatomy 0.000 claims description 64
- 230000007613 environmental effect Effects 0.000 claims description 33
- 238000009792 diffusion process Methods 0.000 claims description 14
- 230000004913 activation Effects 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 6
- 230000008901 benefit Effects 0.000 abstract description 8
- 230000008859 change Effects 0.000 abstract description 8
- 238000000034 method Methods 0.000 description 16
- 230000008569 process Effects 0.000 description 9
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 6
- 238000001514 detection method Methods 0.000 description 4
- 239000013535 sea water Substances 0.000 description 4
- 238000009395 breeding Methods 0.000 description 3
- 230000001488 breeding effect Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 229910052760 oxygen Inorganic materials 0.000 description 2
- 239000001301 oxygen Substances 0.000 description 2
- 230000000737 periodic effect Effects 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 241000251468 Actinopterygii Species 0.000 description 1
- 241001474374 Blennius Species 0.000 description 1
- 241000282414 Homo sapiens Species 0.000 description 1
- 241001465754 Metazoa Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000002537 cosmetic Substances 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000009313 farming Methods 0.000 description 1
- 238000012067 mathematical method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 235000015097 nutrients Nutrition 0.000 description 1
- 238000009304 pastoral farming Methods 0.000 description 1
- 230000035699 permeability Effects 0.000 description 1
- 230000029553 photosynthesis Effects 0.000 description 1
- 238000010672 photosynthesis Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000029058 respiratory gaseous exchange Effects 0.000 description 1
- 235000015170 shellfish Nutrition 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 241000894007 species Species 0.000 description 1
- 230000007480 spreading Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0633—Workflow analysis
-
- A—HUMAN NECESSITIES
- A01—AGRICULTURE; FORESTRY; ANIMAL HUSBANDRY; HUNTING; TRAPPING; FISHING
- A01K—ANIMAL HUSBANDRY; AVICULTURE; APICULTURE; PISCICULTURE; FISHING; REARING OR BREEDING ANIMALS, NOT OTHERWISE PROVIDED FOR; NEW BREEDS OF ANIMALS
- A01K61/00—Culture of aquatic animals
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/02—Agriculture; Fishing; Forestry; Mining
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A40/00—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production
- Y02A40/70—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production in livestock or poultry
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Strategic Management (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Tourism & Hospitality (AREA)
- Animal Husbandry (AREA)
- Marine Sciences & Fisheries (AREA)
- Environmental Sciences (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Molecular Biology (AREA)
- General Business, Economics & Management (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Quality & Reliability (AREA)
- Computational Mathematics (AREA)
- Primary Health Care (AREA)
- Agronomy & Crop Science (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Educational Administration (AREA)
- Probability & Statistics with Applications (AREA)
- Algebra (AREA)
- Game Theory and Decision Science (AREA)
- Mathematical Analysis (AREA)
Abstract
The invention relates to a machine learning-based ocean pasture mechanized seedling sowing scheduling system, which is characterized in that coordinates, ocean current directions and environment data of a plurality of planned seedling sowing areas in an ocean pasture are obtained through a data obtaining and preprocessing module, a topological sequence is established, a two-dimensional coding sequence is established through a data coding module, a seedling sowing suitability sequence is obtained through a suitability prediction module by utilizing a preset RNN neural network prediction model, and finally a seedling sowing scheduling plan is obtained through a plan analysis module. Compared with the prior art, the invention utilizes the topological sequence to represent the dynamic change relation of the ocean, and predicts the seedling sowing suitability by presetting the RNN neural network prediction model, so that the problem that the seedling sowing suitability can still be accurately predicted under the conditions that the ocean is dynamic, the data is difficult to comprehensively detect and the accurate deduction is difficult to realize is solved, and the seedling sowing plan is scientific and reasonable, and the benefit of the ocean pasture is optimal.
Description
Technical Field
The invention relates to the technical field of ocean pastures, in particular to an ocean pasture mechanized seedling sowing scheduling system based on machine learning.
Background
The marine pasture is an innovative cultivation mode, and utilizes the characteristics of rich marine resources and wide water areas to cultivate and reproduce various aquatic animals and plants by constructing cultivation facilities in the ocean. Compared with the traditional land culture, the marine pasture can provide a larger space and a good ecological environment, and provides more growth space and resources for the culture. The marine ranches can cover various breeding objects including fishes, shellfishes, seaweeds and the like, and the breeding species not only can be eaten by human beings, but also can be used in the fields of pharmacy, cosmetics and the like.
In order to increase the breeding efficiency and productivity of marine ranches, many marine ranches have employed automated seeding machinery. These mechanical devices enable the periodic placement of seedlings or seeds into the farming facility, enabling an automated sowing process. By using an automatic mechanical seedling sowing technology, labor cost and operation time can be reduced, and accuracy and consistency of seedling sowing can be improved.
However, time and location decisions for seeding remain a challenging problem for larger-scale marine rangelands. On the one hand, since the ocean is a dynamic environment, it is affected by factors such as ocean flow, water temperature variation, water quality conditions, etc., which may complicate and make difficult the selection of seeding time and location. On the other hand, it is difficult to obtain comprehensive detection data in the ocean, so that an accurate scientific basis is lacking for a decision maker. Currently, most seeding decisions are also based on experience, which, while providing some reference, often fail to achieve optimal results.
Disclosure of Invention
Therefore, the invention provides a mechanical seedling sowing scheduling system for a marine pasture based on machine learning, which is used for solving the problem that the conventional seedling sowing scheduling method cannot achieve optimal benefit.
The invention provides a machine learning-based marine pasture mechanized seedling sowing and scheduling system, which comprises:
the data acquisition and preprocessing module is used for acquiring coordinates, ocean current directions and environmental data of a plurality of planned seeding areas in the marine pasture, wherein the environmental data comprises flow velocity, and a topological sequence for representing ocean current diffusion sequence among the plurality of planned seeding areas is established according to the coordinates, the ocean current directions and the flow velocity of each planned seeding area;
the data coding module is used for establishing a two-dimensional coding sequence representing ocean current diffusion sequences and environmental conditions of a plurality of planned seedling sowing areas based on the topological sequence according to the environmental data of each planned seedling sowing area;
the fitness prediction module is used for inputting the two-dimensional coding sequence into a preset RNN neural network prediction model to obtain a seedling sowing fitness sequence of each planned seedling sowing region, wherein the seedling sowing fitness sequence is used for representing the seedling sowing fitness of a plurality of preset time points of the planned seedling sowing region in a future period;
the plan analysis module is used for obtaining a seedling sowing scheduling plan according to the seedling sowing suitability sequences of the plurality of planned seedling sowing areas, wherein the seedling sowing scheduling plan comprises a scheduling sequence formed by arranging the plurality of planned seedling sowing areas, and the scheduling sequence is used for representing the predicted optimal seedling sowing sequence of the plurality of planned seedling sowing areas at a plurality of preset time points in the future.
Preferably, the establishing a topological sequence for representing the ocean current diffusion sequence among the multiple planned seedling sowing areas according to the coordinates, ocean current directions and flow rates of each planned seedling sowing area comprises the following steps:
obtaining ocean current diffusion time between two planned seedling sowing areas according to the coordinates, the ocean current direction and the flow velocity;
taking the planned seedling sowing areas as nodes, taking the ocean current direction between two planned seedling sowing areas corresponding to two nodes as the direction of the edge between the nodes, and taking the ocean current diffusion time between the two planned seedling sowing areas corresponding to the two nodes as the weight of the edge, so as to establish a directed acyclic graph;
and performing topological sorting on the plurality of planned seedling sowing areas based on the directed acyclic graph to obtain a topological sequence of the plurality of planned seedling sowing areas.
Preferably, the two-dimensional coding sequence comprises a plurality of coding vectors, each coding vector corresponds to one planned seeding region, and the plurality of coding vectors are arranged based on the topological sequence of the planned seeding region corresponding to each coding vector; each element in the coding vector corresponds to one type of environment data respectively, and each element in the coding vector is a characteristic value of the corresponding environment data;
the preset RNN neural network prediction model comprises a first input layer, a first hidden layer, a first output layer, a second hidden layer and a second output layer which are sequentially connected, wherein:
the first hidden layer is used for calculating the initial seeding suitability of each planned seeding region and taking the initial seeding suitability as the data of the first output layer based on the topological ordering of the planned seeding regions and the environmental data of each planned seeding region;
the second hidden layer is used for predicting the initial seedling sowing suitability of each planned seedling sowing area to obtain a seedling sowing suitability sequence of each planned seedling sowing area.
Preferably, in the preset RNN neural network prediction model:
the first input layer comprises a plurality of first input units, each first input unit is used for inputting one coding vector, and the plurality of first input units are arranged according to the sequence of the corresponding coding vector in the two-dimensional coding sequence;
the first hidden layer comprises a plurality of first neurons, each first neuron corresponds to one first input unit, the plurality of first neurons are arranged along the arrangement sequence of the corresponding first input units, and the input end of each first neuron is connected with the corresponding first input unit and the output end of the other first neuron arranged at the previous position;
the first output layer comprises a plurality of first output units, each first output unit corresponds to one first neuron, the input end of each first output unit is connected with the output end of the corresponding first neuron, and each first output unit is used for outputting the initial seeding suitability of a planned seeding region corresponding to one coding vector;
the second hiding layer comprises a plurality of second hiding units, each second hiding unit corresponds to one first output unit, and the input end of each second hiding unit is connected with the output end of the corresponding first output unit;
the second output layer comprises a plurality of second output units, each second output unit corresponds to one second hidden unit, the input end of each second output unit is connected with the output end of the corresponding second hidden unit, and each second output unit is used for outputting a seedling sowing suitability sequence of a planned seedling sowing area.
Preferably, the expression adopted when the first neuron outputs the vector is:
；
wherein,is the>Position(s)>Is->An output vector of the first neuron, +.>For the first nonlinear activation function, +.>For the first adjustment coefficient based on the overall difference of the plurality of encoded vectors,/I>For the first weight, ++>Is->An output vector of the first neuron, +.>Is the>Coding vector->According to->Dividing the code vector by the +.>Second adjustment coefficients obtained by differences of other coding vectors than the individual coding vectors,/for>For the second weight, ++>Is a first bias.
Preferably, the first adjustment coefficient is obtained according to the following formula:
wherein,for variance operator>Is the%>The position of the individual elements is determined,/>for the total number of the planned seeding areas, +.>For the total number of categories of environmental data, +.>Is the>The%>The elements.
Preferably, the second adjustment coefficient is obtained according to the following formula:
；
wherein,is the>The%>Element(s)>Is the>The%>The elements.
Preferably, the expression adopted when the first output unit outputs the suitability of the initial seeding is:
；
wherein,is->The first output unit outputs the first seeding suitability,/-for the first seeding>Is the%>Weight of the environmental data corresponding to the individual element, < +.>Is->The first neuron is the first neuron in the output vector>The elements.
Preferably, the second hiding units include a plurality of second neurons, input ends of the second neurons in each second hiding unit are connected with the first output unit and the second output unit corresponding to the second hiding unit, the second neurons in each second hiding unit are further connected in sequence, and each second hiding unit is used for outputting seeding suitability of a preset time point.
Preferably, the expression adopted when the second hiding unit outputs data is:
；
wherein,is->Output data of the second neurons, +.>For a second nonlinear activation function, +.>According to->Third adjustment coefficient obtained by flow velocity corresponding to each planned seedling sowing area, < >>For the third weight->Is the firstOutput vector of second neuron, +.>For the fourth weight, ++>Is a second bias.
The beneficial effects of adopting the embodiment are as follows:
the invention provides a machine learning-based ocean pasture mechanized seedling sowing scheduling system, which is characterized in that coordinates, ocean current directions and environment data of a plurality of planned seedling sowing areas in an ocean pasture are obtained through a data obtaining and preprocessing module, the environment data comprise flow rates, the coordinates, the ocean current directions and the flow rates are built to build a topological sequence, then a two-dimensional coding sequence is built through a data coding module according to the environment data and the topological sequence, then a seedling sowing suitability sequence of each planned seedling sowing area is obtained through a suitability prediction module by utilizing a preset RNN neural network prediction model, and finally a seedling sowing scheduling plan is obtained through a plan analysis module according to the seedling sowing suitability sequences of the planned seedling sowing areas. Compared with the prior art, the method and the device have the advantages that the topological sequence established through coordinates, ocean current directions and flow velocity is utilized to represent the dynamic change relation of the ocean, meanwhile, the prediction model of the preset RNN neural network which can process the time relation sequence predicts the seedling sowing suitability of each planned seedling sowing area, the problem of difficult decision-making caused by the dynamic ocean environment is solved, the data in the planned seedling sowing area only need to be acquired, the learning of the preset RNN neural network prediction model is utilized to realize that the accurate prediction of the seedling sowing suitability can be still carried out under the condition that the comprehensive detection of the data is difficult and the accurate deduction is difficult to realize, so that the seedling sowing plan is scientific and reasonable, and the benefit of the ocean pasture is optimal.
Drawings
FIG. 1 is a system architecture diagram of one embodiment of a machine learning based marine ranching mechanized seeding scheduling system provided by the invention;
FIG. 2 is a schematic structural diagram of a predictive model of a preset RNN neural network according to the present invention;
fig. 3 is a schematic structural diagram of a second hidden unit in the preset RNN neural network prediction model according to the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, an embodiment of the present invention discloses a machine learning-based marine ranch mechanized seeding scheduling system, comprising:
the data acquisition and preprocessing module 110 is configured to acquire coordinates, ocean current directions and environmental data of a plurality of planned seeding areas in the marine pasture, wherein the environmental data comprises flow rates, and establish a topological sequence for representing ocean current diffusion sequences among the plurality of planned seeding areas according to the coordinates, the ocean current directions and the flow rates of each planned seeding area;
the data encoding module 120 is configured to establish a two-dimensional encoding sequence that characterizes the ocean current diffusion sequence and the environmental condition of the plurality of planned seedling sowing areas based on the topological sequence according to the environmental data of each planned seedling sowing area;
the fitness prediction module 130 is configured to input the two-dimensional code sequence into a preset RNN neural network prediction model, to obtain a seeding suitability sequence of each planned seeding region, where the seeding suitability sequence is used to characterize the seeding suitability of a plurality of preset time points in a future period of time of the planned seeding region;
the plan parsing module 140 is configured to obtain a seedling-sowing scheduling plan according to a seedling-sowing suitability sequence of a plurality of planned seedling-sowing regions, where the seedling-sowing scheduling plan includes a scheduling sequence composed of a plurality of planned seedling-sowing region arrangements, and the scheduling sequence is used to characterize a predicted optimal seedling-sowing sequence of the plurality of planned seedling-sowing regions at a plurality of preset time points in the future.
Compared with the prior art, the method and the device have the advantages that the topological sequence established through coordinates, ocean current directions and flow velocity is utilized to represent the dynamic change relation of the ocean, meanwhile, the prediction model of the preset RNN neural network which can process the time relation sequence predicts the seedling sowing suitability of each planned seedling sowing area, the problem of difficult decision-making caused by the dynamic ocean environment is solved, the data in the planned seedling sowing area only need to be acquired, the learning of the preset RNN neural network prediction model is utilized to realize that the accurate prediction of the seedling sowing suitability can be still carried out under the condition that the comprehensive detection of the data is difficult and the accurate deduction is difficult to realize, so that the seedling sowing plan is scientific and reasonable, and the benefit of the ocean pasture is optimal.
In the above process, the planned seeding area is an area in the ocean pasture where seeding is planned, and may be a plurality of set cultivation fences in the ocean pasture, or a certain area in the ocean pasture divided by people. When the environment data are collected, only one point in the planned seedling sowing area is required to be collected, and sea areas outside the planned seedling sowing area are not required to be collected at the same time. Specifically, the environmental data in the present invention are data in the ocean that may affect seeding, for example:
ocean temperature: the temperature change of seawater has an important influence on the growth and reproduction of aquatic organisms.
Salinity: salinity in seawater can affect aquatic organisms, ocean currents and the like, and is an important factor of the ocean environment.
Ocean flow rate: the water flow rate can affect the stability of sowing and nutrient delivery, and the adaptability of sowing equipment needs to be considered.
Tide: tide is a periodic variation of the marine environment, with an impact on sowing timing and depth.
Ocean turbidity: the turbidity of the water area can influence the light permeability, and plays an important role in the generated photosynthesis and the water area ecological system.
Dissolved oxygen: the dissolved oxygen content of seawater has a critical impact on the respiration and growth of organisms.
It is understood that in practice, any suitable environmental data may be selected according to specific situations, such as sea conditions, cultivation requirements, and seed types, and only it is required to ensure that the data can be represented by a feature value.
Further, in a preferred embodiment, in the above process data acquiring and preprocessing module, the establishing a topological sequence for characterizing a current spreading sequence between the plurality of planned seeding regions according to coordinates, current directions and flow rates of each planned seeding region specifically includes:
obtaining ocean current diffusion time between two planned seedling sowing areas according to the coordinates, the ocean current direction and the flow velocity;
taking the planned seedling sowing areas as nodes, taking the ocean current direction between two planned seedling sowing areas corresponding to two nodes as the direction of the edge between the nodes, and taking the ocean current diffusion time between the two planned seedling sowing areas corresponding to the two nodes as the weight of the edge, so as to establish a directed acyclic graph;
and performing topological sorting on the plurality of planned seedling sowing areas based on the directed acyclic graph to obtain a topological sequence of the plurality of planned seedling sowing areas.
Obviously, most of the environmental data in the ocean change along with the flowing of the seawater, and the topological sequence established in the process represents the sequence of ocean currents passing through a plurality of planned seedling sowing areas, so that the ocean dynamic situation related to the planned seedling sowing areas in the ocean pasture is further represented, and the ocean dynamic situation can be regarded as the sequence of environmental influence among the plurality of planned seedling sowing areas, so that a theoretical basis is provided for the follow-up prediction of the seedling sowing suitability under the dynamic situation.
Further, in a preferred embodiment, the two-dimensional code sequence established by the data coding module specifically includes a plurality of code vectors, each code vector corresponds to a planned seeding region, and the plurality of code vectors are arranged based on a topological sequence of the planned seeding region corresponding to each code vector; each element in the coding vector corresponds to one type of environment data respectively, and each element in the coding vector is a characteristic value of the corresponding environment data.
In this embodiment, the plurality of encoding vectors are arranged based on the topological sequence, so that the two-dimensional encoding sequence used as the model input data not only includes the characteristics of the environmental data, but also carries the environmental dynamic change relations of the plurality of planned seedling sowing areas, so that the dynamic change relations can be taken into consideration in the subsequent analysis and prediction, and a more accurate prediction result is achieved.
It is understood that the characteristic value is a value capable of representing an environmental data, for example, the salinity data may be normalized to obtain 0-1 data as the characteristic value corresponding to the salinity data in the salinity code vector, for example, the ocean temperature may be classified, and different grades may be represented by specific values as the characteristic value, and so on. It is understood that the eigenvalues are mainly used for encoding and inputting into a preset RNN neural network prediction model, and that one data is represented by the eigenvalues as the prior art that can be understood by those skilled in the art, so that the description thereof will not be repeated herein.
Further, in a preferred embodiment, in the fitness prediction module, the preset RNN neural network prediction model is a model built based on an RNN model, where RNN (Recurrent Neural Networks, RNN) is a cyclic neural network, and is a neural network model applicable to sequence data, which can capture dynamic characteristics of time sequence data. At present, RNN is generally used for natural language processing, but in this embodiment, the sequence of the topological sequence (i.e. the influencing sequence of a plurality of planned seeding regions) is regarded as the time sequence in the RNN, so that the RNN can be converted into a method for analyzing the dependency relationship of a plurality of planned seeding regions in a dynamic environment, so as to accurately predict the seeding suitability, and further obtain the future optimal seeding time and place. In practice, for prediction of ocean seeding time and place, RNN models that may be used include any existing RNN model such as Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU, gated loop unit).
Further, in a preferred embodiment, the preset RNN neural network prediction model includes a first input layer, a first hidden layer, a first output layer, a second hidden layer, and a second output layer that are sequentially connected, where:
the first hidden layer is used for calculating the initial seeding suitability of each planned seeding region and taking the initial seeding suitability as the data of the first output layer based on the topological ordering of the planned seeding regions and the environmental data of each planned seeding region;
the second hidden layer is used for predicting the initial seedling sowing suitability of each planned seedling sowing area to obtain a seedling sowing suitability sequence of each planned seedling sowing area.
The structure is actually that two continuous RNN models are utilized to predict the seeding suitability sequence, and the first RNN model is embodied in the first hidden layer and is mainly used for combining dynamic influence relations of a plurality of planned seeding areas to obtain a reasonable initial seeding suitability. And the second RNN model is mainly used for continuously predicting the seeding suitability of a plurality of preset time points in the future based on the initial seeding suitability. While ensuring reasonable and accurate output results, the design ensures that each RNN model is not too complex, and can reach satisfactory speed no matter training or running.
Specifically, in a preferred embodiment, the preset RNN neural network prediction model is that:
the first input layer comprises a plurality of first input units, each first input unit is used for inputting one coding vector, and the plurality of first input units are arranged according to the sequence of the corresponding coding vector in the two-dimensional coding sequence;
the first hidden layer comprises a plurality of first neurons, each first neuron corresponds to one first input unit, the plurality of first neurons are arranged along the arrangement sequence of the corresponding first input units, and the input end of each first neuron is connected with the corresponding first input unit and the output end of the other first neuron arranged at the previous position;
the first output layer comprises a plurality of first output units, each first output unit corresponds to one first neuron, the input end of each first output unit is connected with the output end of the corresponding first neuron, and each first output unit is used for outputting the initial seeding suitability of a planned seeding region corresponding to one coding vector;
the second hiding layer comprises a plurality of second hiding units, each second hiding unit corresponds to one first output unit, and the input end of each second hiding unit is connected with the output end of the corresponding first output unit;
the second output layer comprises a plurality of second output units, each second output unit corresponds to one second hidden unit, the input end of each second output unit is connected with the output end of the corresponding second hidden unit, and each second output unit is used for outputting a seedling sowing suitability sequence of a planned seedling sowing area.
In a preferred embodiment of the preset RNN neural network prediction model, the content input by the first input layer is a two-dimensional coding sequence composed of a plurality of coding vectors, and the content output by the second output layer is a plurality of vectors. The preset RNN neural network prediction model in the embodiment has a simple and understandable structure, and is convenient for engineers to build and train. The specific form of the above structure is shown in FIG. 2, in whichFor a preset input value for calculating the first +.>，/>Is->Second hidden units corresponding to the first output units, ">Is->A seedling sowing suitability sequence corresponding to each planned seedling sowing area,is->In the seedling-sowing suitability sequence corresponding to the planned seedling-sowing region, the +.>The seedling-sowing degree of each preset time point is proper, < >>K is the total number of the preset time points. It is understood that the preset time point is set manually according to the specific situation. It will be appreciated that for different letter designations in the drawings, those skilled in the art will understand the meaning based on the description herein, and thus will not be described in detail herein.
Further, to improve accuracy again, in a preferred embodiment, the expression used when the first neuron outputs a vector is:
；
wherein,is the>Position(s)>Is->An output vector of the first neuron, +.>For the first nonlinear activation function, +.>For the first adjustment coefficient based on the overall difference of the plurality of encoded vectors,/I>For the first weight, ++>Is->An output vector of the first neuron, +.>Is the>Coding vector->According to->Dividing the code vector by the +.>Second adjustment coefficients obtained by differences of other coding vectors than the individual coding vectors,/for>For the second weight, ++>Is a first bias.
The expression is an improvement on the expression of the neuron in the traditional RNN model, and the common RNN model can continuously update the first weight, the second weight and the first bias to improve the model accuracy during training iteration. On the basis of the method, two parameters of a first adjustment coefficient and a second adjustment coefficient are further added, wherein the first adjustment parameter can further correct the influence of the output vector of the previous first neuron on the first neuron to be calculated currently according to the overall difference of the plurality of coding vectors, for example, when the overall difference of the plurality of coding vectors is large, the environmental data difference representing the plurality of planned seeding regions is also large, and then the mutual influence among the planned seeding regions is also large, and at the moment, the influence of the output vector of the previous first neuron on the first neuron to be calculated currently can be further increased through the first adjustment coefficient so as to achieve a more accurate result.
Similarly, the second adjustment parameter is obtained according to the difference between the coding vector corresponding to the current first neuron and other coding vectors, and is used for further correcting the influence of the current coding vector on the first neuron to be calculated currently. For example, if the difference between the current coding vector and other coding vectors is large, it is indicated that the environments of the planned seeding region corresponding to the coding vector and other regions are slightly different, and at this time, the influence of the planned seeding region on other regions is also large, and at this time, the influence of the current coding vector on the first neuron to be calculated currently can be further increased by means of the second adjustment coefficient, so as to achieve a more accurate result.
Further, in a preferred embodiment, the first adjustment factor is obtained according to the following formula:
；
wherein,for variance operator>Is the%>Element position->For the total number of the planned seeding areas, +.>For the total number of categories of environmental data, +.>Is the>The%>The elements.
The above formula adopts a way of calculating variance to measure the overall difference of a plurality of coding vectors, and the meaning is that the variance of the same type of environmental data between different planned seeding areas is calculated, and then the average value of the variances of all environmental data is counted to be used as a first adjustment coefficient. The formula is easy to understand and convenient for code realization, and has a faster running speed.
Further, in a preferred embodiment, the second adjustment factor is obtained according to the following formula:
；
wherein,is the>The%>Element(s)>Is the>The%>The elements.
The meaning of the formula is to calculate the environmental data difference between the planned seeding region corresponding to the current coding vector and the two adjacent planned seeding regions in the topological sequence to judge whether the planned seeding region is worth adjusting the input influence duty ratio in the current first neuron. In the above process, only two adjacent planned seeding areas are compared, and compared with the comparison with all planned seeding areas, the second adjustment coefficient calculation mode of the embodiment analyzes the influence of the second adjustment coefficient to a certain extent, and simultaneously avoids the complex redundant calculation process with lower meaning, so that the whole model has both accuracy and operation efficiency.
Further, in a preferred embodiment, the expression adopted when the first output unit outputs the suitability of the initial seeding is:
；
wherein,is->The initial seeding output by the first output units is suitableDegree (f)>Is the%>Weight of the environmental data corresponding to the individual element, < +.>Is->The first neuron is the first neuron in the output vector>The elements.
The input data of the first neuron obtained in the above manner is also in the form of a vector, that isEach element in the vector is a semantic value obtained preliminarily, and each element corresponds to each environmental data, so that the suitability of the initial seeding is calculated by adopting a weighted summation mode in the embodiment.
Further, in an preferred embodiment, the second hiding unit includes a plurality of second neurons, input ends of the plurality of second neurons in each second hiding unit are connected to a first output unit and a second output unit corresponding to the second hiding unit, and the plurality of second neurons in each second hiding unit are further connected in sequence, and each second hiding unit is used for outputting a seeding suitability of a preset time point. The RNN model with the structure of 1vN is used for obtaining a sequence with a time sequence relationship according to fewer inputs, and is very suitable for the scene in the embodiment. The structure of the second hidden unit in this embodiment is shown in FIG. 3, and in FIG. 3Is another preset input value.
Further, based on the same idea, in order to further improve the prediction accuracy, in a preferred embodiment, the expression adopted when the second concealment unit outputs the data is:
；
wherein,is->Output data of the second neurons, +.>For a second nonlinear activation function, +.>According to->Third adjustment coefficient obtained by flow velocity corresponding to each planned seedling sowing area, < >>For the third weight->Is the firstOutput vector of second neuron, +.>For the fourth weight, ++>Is a second bias.
The above procedure is also an improvement based on the expression of neurons in the existing RNN model, i.e. a third adjustment factor based on the flow rate per planned seeding area is introduced in addition to the third weight, the fourth weight and the second bias obtained by training iterations. For example, when the ocean current flow velocity in the planned seeding region is large, it is obvious that the planned seeding region affects other regions more rapidly, and then the influence of the output result of the previous neuron should be increased when calculating the output of the next neuron, which can be achieved by the third adjustment coefficient. It can be understood that in practice, the calculation method of the third adjustment coefficient according to the flow rate may be flexibly set according to the specific situation.
After the seeding suitability sequence of each planned seeding region is obtained, a seeding scheduling plan can be obtained through a plan analysis module according to the seeding suitability sequences of a plurality of planned seeding regions. The seedling scheduling plan comprises a scheduling sequence formed by arranging a plurality of planned seedling areas, wherein the scheduling sequence is used for representing the predicted optimal seedling sequence of the planned seedling areas at a plurality of preset time points in the future.
For example, three elements are provided in each seeding suitability sequence, and the three elements represent the seeding suitability degree on the following first day, second day and third day according to the sequence, and the higher the seeding suitability value is, the more suitable the seeding is. At present, three planned seedling sowing areas are respectively an area A, an area B and an area C, and the seedling sowing suitability sequences of the three areas are respectively as follows:
planned seedling sowing area A: 1. 2, 3;
planned seedling sowing area B: 3. 2, 1;
planned seedling sowing area C: 1. 3, 2;
then the optimal seedling order should be B, C, A.
It will be appreciated that the actual situation is far more complex than the above example, and the optimal seeding order cannot be intuitively obtained, and when the more complex seeding suitability sequence is faced in practice, any existing mathematical method may be adopted to solve, so that no excessive description is made herein.
The invention provides a machine learning-based ocean pasture mechanized seedling sowing scheduling system, which is characterized in that coordinates, ocean current directions and environment data of a plurality of planned seedling sowing areas in an ocean pasture are obtained through a data obtaining and preprocessing module, the environment data comprise flow rates, the coordinates, the ocean current directions and the flow rates are built to build a topological sequence, then a two-dimensional coding sequence is built through a data coding module according to the environment data and the topological sequence, then a seedling sowing suitability sequence of each planned seedling sowing area is obtained through a suitability prediction module by utilizing a preset RNN neural network prediction model, and finally a seedling sowing scheduling plan is obtained through a plan analysis module according to the seedling sowing suitability sequences of the planned seedling sowing areas. Compared with the prior art, the method and the device have the advantages that the topological sequence established through coordinates, ocean current directions and flow velocity is utilized to represent the dynamic change relation of the ocean, meanwhile, the prediction model of the preset RNN neural network which can process the time relation sequence predicts the seedling sowing suitability of each planned seedling sowing area, the problem of difficult decision-making caused by the dynamic ocean environment is solved, the data in the planned seedling sowing area only need to be acquired, the learning of the preset RNN neural network prediction model is utilized to realize that the accurate prediction of the seedling sowing suitability can be still carried out under the condition that the comprehensive detection of the data is difficult and the accurate deduction is difficult to realize, so that the seedling sowing plan is scientific and reasonable, and the benefit of the ocean pasture is optimal.
It should be noted that, in the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described as different from other embodiments, and identical and similar parts between the embodiments are all enough to be referred to each other.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. The utility model provides a marine ranch mechanized seeding scheduling system based on machine learning which characterized in that includes:
the data acquisition and preprocessing module is used for acquiring coordinates, ocean current directions and environmental data of a plurality of planned seeding areas in the marine pasture, wherein the environmental data comprises flow velocity, and a topological sequence for representing ocean current diffusion sequence among the plurality of planned seeding areas is established according to the coordinates, the ocean current directions and the flow velocity of each planned seeding area;
the data coding module is used for establishing a two-dimensional coding sequence representing ocean current diffusion sequences and environmental conditions of a plurality of planned seedling sowing areas based on the topological sequence according to the environmental data of each planned seedling sowing area;
the fitness prediction module is used for inputting the two-dimensional coding sequence into a preset RNN neural network prediction model to obtain a seedling sowing fitness sequence of each planned seedling sowing region, wherein the seedling sowing fitness sequence is used for representing the seedling sowing fitness of a plurality of preset time points of the planned seedling sowing region in a future period;
the plan analysis module is used for obtaining a seedling sowing scheduling plan according to the seedling sowing suitability sequences of the plurality of planned seedling sowing areas, wherein the seedling sowing scheduling plan comprises a scheduling sequence formed by arranging the plurality of planned seedling sowing areas, and the scheduling sequence is used for representing the predicted optimal seedling sowing sequence of the plurality of planned seedling sowing areas at a plurality of preset time points in the future.
2. The machine learning based marine ranch mechanized seeding scheduling system of claim 1 wherein the establishing a topological sequence for characterizing the ocean current diffusion order between the plurality of planned seeding regions based on the coordinates, ocean current direction and flow rate of each planned seeding region comprises:
obtaining ocean current diffusion time between two planned seedling sowing areas according to the coordinates, the ocean current direction and the flow velocity;
taking the planned seedling sowing areas as nodes, taking the ocean current direction between two planned seedling sowing areas corresponding to two nodes as the direction of the edge between the nodes, and taking the ocean current diffusion time between the two planned seedling sowing areas corresponding to the two nodes as the weight of the edge, so as to establish a directed acyclic graph;
and performing topological sorting on the plurality of planned seedling sowing areas based on the directed acyclic graph to obtain a topological sequence of the plurality of planned seedling sowing areas.
3. The machine learning based marine ranch mechanized seeding scheduling system of claim 1 wherein the two-dimensional coding sequence comprises a plurality of coding vectors, each coding vector corresponding to a planned seeding region, the plurality of coding vectors arranged based on a topological sequence of the planned seeding region to which each coding vector corresponds; each element in the coding vector corresponds to one type of environment data respectively, and each element in the coding vector is a characteristic value of the corresponding environment data;
the preset RNN neural network prediction model comprises a first input layer, a first hidden layer, a first output layer, a second hidden layer and a second output layer which are sequentially connected, wherein:
the first hidden layer is used for calculating the initial seeding suitability of each planned seeding region and taking the initial seeding suitability as the data of the first output layer based on the topological ordering of the planned seeding regions and the environmental data of each planned seeding region;
the second hidden layer is used for predicting the initial seedling sowing suitability of each planned seedling sowing area to obtain a seedling sowing suitability sequence of each planned seedling sowing area.
4. The machine learning based marine ranch mechanized seeding scheduling system of claim 3 wherein, in the pre-set RNN neural network prediction model:
the first input layer comprises a plurality of first input units, each first input unit is used for inputting one coding vector, and the plurality of first input units are arranged according to the sequence of the corresponding coding vector in the two-dimensional coding sequence;
the first hidden layer comprises a plurality of first neurons, each first neuron corresponds to one first input unit, the plurality of first neurons are arranged along the arrangement sequence of the corresponding first input units, and the input end of each first neuron is connected with the corresponding first input unit and the output end of the other first neuron arranged at the previous position;
the first output layer comprises a plurality of first output units, each first output unit corresponds to one first neuron, the input end of each first output unit is connected with the output end of the corresponding first neuron, and each first output unit is used for outputting the initial seeding suitability of a planned seeding region corresponding to one coding vector;
the second hiding layer comprises a plurality of second hiding units, each second hiding unit corresponds to one first output unit, and the input end of each second hiding unit is connected with the output end of the corresponding first output unit;
the second output layer comprises a plurality of second output units, each second output unit corresponds to one second hidden unit, the input end of each second output unit is connected with the output end of the corresponding second hidden unit, and each second output unit is used for outputting a seedling sowing suitability sequence of a planned seedling sowing area.
5. The machine learning based marine ranch mechanized seeding scheduling system of claim 4 wherein the first neuron outputs a vector using the expression:
；
wherein,is the>Position(s)>Is->An output vector of the first neuron, +.>For the first nonlinear activation function, +.>For the first adjustment coefficient based on the overall difference of the plurality of encoded vectors,/I>For the first weight, ++>Is->An output vector of the first neuron, +.>Is the>The number of the encoded vectors is the number,according to->Dividing the code vector by the +.>Second adjustment coefficients obtained by differences of other coding vectors than the individual coding vectors,/for>For the second weight, ++>Is a first bias.
6. The machine learning based marine ranch mechanized seeding scheduling system of claim 5 wherein the first adjustment factor is derived according to:
；
wherein,for variance operator>Is the%>Element position->For the total number of the planned seeding areas, +.>For the total number of categories of environmental data, +.>Is the>The%>The elements.
7. The machine learning based marine ranch mechanized seeding scheduling system of claim 6 wherein the second adjustment factor is derived according to:
；
wherein,is the>The%>Element(s)>Is the>The%>The elements.
8. The machine learning based marine ranch mechanized seeding scheduling system of claim 7 wherein the expression employed when the first output unit outputs the suitability of the initial seeding is:
；
wherein,is->The first output unit outputs the first seeding suitability,/-for the first seeding>Is the%>Weight of the environmental data corresponding to the individual element, < +.>Is->The first neuron is the first neuron in the output vector>The elements.
9. The machine learning based marine ranch mechanized seeding scheduling system of claim 8, wherein the second hidden units include a plurality of second neurons, the input ends of the plurality of second neurons in each second hidden unit are connected with the first output unit and the second output unit corresponding to the second hidden unit, the plurality of second neurons in each second hidden unit are further connected in sequence, and each second hidden unit is used for outputting the seeding suitability of a preset time point.
10. The machine learning based marine ranch mechanized seeding scheduling system of claim 9 wherein the second hidden unit outputs data using the expression:
；
wherein,is->Output data of the second neurons, +.>For a second nonlinear activation function, +.>According to the firstThird adjustment coefficient obtained by flow velocity corresponding to each planned seedling sowing area, < >>For the third weight->Is->Output vector of second neuron, +.>For the fourth weight, ++>Is a second bias.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410003764.XA CN117495072B (en) | 2024-01-03 | 2024-01-03 | Marine pasture mechanized seedling sowing scheduling system based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410003764.XA CN117495072B (en) | 2024-01-03 | 2024-01-03 | Marine pasture mechanized seedling sowing scheduling system based on machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117495072A true CN117495072A (en) | 2024-02-02 |
| CN117495072B CN117495072B (en) | 2024-03-12 |
Family
ID=89669483
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410003764.XA Active CN117495072B (en) | 2024-01-03 | 2024-01-03 | Marine pasture mechanized seedling sowing scheduling system based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117495072B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030094141A1 (en) * | 2000-06-28 | 2003-05-22 | Davis Russell P. | Integrated system for shellfish production: encompassing hatchery, nursery, brood-stock conditioning and market conditioning phases; also water treatment, food supplement, propulsion, anchoring, security, and devices for the integration of neighborhood values and shellfish production. |
| US20180292826A1 (en) * | 2017-04-11 | 2018-10-11 | Defelice Thomas Peter | Intelligent systems for weather modification programs |
| US20210089947A1 (en) * | 2020-10-25 | 2021-03-25 | Mote Marine Laboratory | Aquaculture Decision Optimization System Using A Learning Engine |
-
2024
- 2024-01-03 CN CN202410003764.XA patent/CN117495072B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030094141A1 (en) * | 2000-06-28 | 2003-05-22 | Davis Russell P. | Integrated system for shellfish production: encompassing hatchery, nursery, brood-stock conditioning and market conditioning phases; also water treatment, food supplement, propulsion, anchoring, security, and devices for the integration of neighborhood values and shellfish production. |
| US20180292826A1 (en) * | 2017-04-11 | 2018-10-11 | Defelice Thomas Peter | Intelligent systems for weather modification programs |
| US20210089947A1 (en) * | 2020-10-25 | 2021-03-25 | Mote Marine Laboratory | Aquaculture Decision Optimization System Using A Learning Engine |
Non-Patent Citations (3)
| Title |
|---|
| "海洋牧场海草床生境构建技术", 专题讨论, vol. 16, no. 2, 31 December 2020 (2020-12-31), pages 200 - 205 * |
| 徐龙琴 等: "基于EMD和ELM的工厂化育苗水温组合预测模型", 农业机械学报, vol. 47, no. 04, 30 April 2016 (2016-04-30), pages 265 - 308 * |
| 颜慧慧;王凤霞;: "中国海洋牧场研究文献综述", 科技广场, no. 06, 30 June 2016 (2016-06-30), pages 162 - 167 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117495072B (en) | 2024-03-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kalimuthu et al. | Crop prediction using machine learning | |
| Chang et al. | Artificial intelligence approaches to predict growth, harvest day, and quality of lettuce (Lactuca sativa L.) in a IoT-enabled greenhouse system | |
| Mulia et al. | Hybrid ANN–GA model for predicting turbidity and chlorophyll-a concentrations | |
| Rijnsdorp et al. | Modelling the spatial dynamics and fisheries of North Sea plaice (Pleuronectes platessa L.) based on tagging data | |
| Velo-Suárez et al. | Artificial neural network approaches to one-step weekly prediction of Dinophysis acuminata blooms in Huelva (Western Andalucía, Spain) | |
| CN110069032A (en) | A kind of eggplant greenhouse intelligent checking system based on wavelet neural network | |
| Dai et al. | Comparative performance of modelling approaches for predicting fish species richness in the Yangtze River Estuary | |
| Abdollahzadeh et al. | Puma optimizer (PO): A novel metaheuristic optimization algorithm and its application in machine learning | |
| Pezy et al. | The environmental impact from an offshore windfarm: Challenge and evaluation methodology based on an ecosystem approach | |
| CN117495072B (en) | Marine pasture mechanized seedling sowing scheduling system based on machine learning | |
| CN109726857A (en) | Cyanobacterial bloom prediction technique based on GA-Elman network | |
| Negrete | Artificial neural networks in mexican agriculture, A overview | |
| Wang et al. | An aquaponics system design for computational intelligence teaching | |
| Della Croce et al. | Simulating the effects of stream network topology on the spread of introgressive hybridization across fish populations | |
| Hrabik et al. | Predation risk and prey fish vertical migration in Lake Superior: insights from an individual based model of siscowet (Salvelinus namaycush) | |
| CN113065247B (en) | Novel fishing situation forecasting model and method based on high-resolution ocean forecasting system | |
| Guo et al. | Automatic and accurate acquisition of stem-related phenotypes of mature soybean based on deep learning and directed search algorithms | |
| Mendigoria et al. | In Situ Indirect Measurement of Nitrate Concentration in Outdoor Tilapia Fishpond Based on Physico-limnological Sensors | |
| CN110414146A (en) | A kind of water environment treatment Project design parameter optimization method based on deep learning | |
| Nielsen | Estimating fish movement | |
| Özden | Estimating Corn Yield Using Statistical, Machine Learning and Deep Learning Methods | |
| de Andrade et al. | A data-driven approach to simulate collective behaviors | |
| Mariani et al. | KYSTFISK III. Population connectivity of cod and plaice in Danish coastal waters | |
| Angelini | Towards an Ecosystem Approach to Fisheries for Nephrops Norvegicus and Meluccius Merluccius Inhabiting the Central Adriatic Sea | |
| Brown | The conservation biology of tropical inshore dolphins in north-western Australian waters |
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 |