CN115880433A - Crop cultivation optimization method based on digital twinning - Google Patents
Crop cultivation optimization method based on digital twinning Download PDFInfo
- Publication number
- CN115880433A CN115880433A CN202211638555.XA CN202211638555A CN115880433A CN 115880433 A CN115880433 A CN 115880433A CN 202211638555 A CN202211638555 A CN 202211638555A CN 115880433 A CN115880433 A CN 115880433A
- Authority
- CN
- China
- Prior art keywords
- crop
- data
- crops
- time
- cultivation
- 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
Images
Abstract
The invention discloses a crop cultivation optimization method based on digital twins, which comprises the following steps: collecting internal and external environmental data and appearance images of crops in real time, and storing the data and the appearance images in a crop cultivation database in an integrated manner; unifying a data time sequence and a sampling frequency by adopting a data preprocessing method; constructing a crop optimal internal and external environment data prediction network and a crop growth image prediction network, predicting the preprocessed crop internal and external environment data, and training a network model through the preprocessed data; visualizing and displaying the 3D model of the crop, the growth trend and the predicted optimal value of the internal and external environment of the crop in the twin model; the crop simulation visualization is used for optimizing the trial operation and adjustment of the digital twin model through crop real-time data; and the online deployment user makes a decision to output the digital twin model to a terminal user interface, and the direct or indirect adjustment command deployment is carried out online through the agricultural Internet of things control system, so that the user can conveniently optimize, regulate and control the crop cultivation environment in time.
Description
Technical Field
The invention relates to the technical field of digital twins, in particular to a crop cultivation optimization method based on digital twins.
Background
Under the background of application of advanced internet technologies such as big data and cloud computing, the traditional agricultural operation mode is continuously updated. The current crops grow and cultivate through intelligent irrigation and fertilization mode, saved a large amount of manpower and materials, improved the cultivation efficiency of crops. However, crops often have different life cycles, and the requirements of different species for the surrounding environment are greatly different, so that the analysis only through the traditional sensing data has certain limitations, and a more complex and more effective growth model needs to be established.
The digital twin technology makes full use of data such as physical models, sensor updating, operation history and the like to construct virtual models to accurately reflect physical objects, realizes the process of converting real into virtual and converting virtual into real in a visual platform, and integrates algorithms, data and decision analysis, thereby improving the production and management efficiency. However, the current digital twinning technology has no complete system flow in the crop growth and cultivation process. Under the current development trend of intelligent agriculture, the full-period management of the crop growth process by using a digital twin technology has important significance. Therefore, a crop cultivation optimization method based on digital twinning is proposed.
Disclosure of Invention
Aiming at the defects of the current crop growth and cultivation technology, the invention realizes data interaction between a crop entity and a virtual twin three-dimensional model by establishing a crop three-dimensional simulation model, updates and corrects internal and external data of a crop growth environment in real time by adopting a dynamic analysis algorithm and a data driving mode, provides controllable and appropriate environmental conditions such as temperature, humidity, water, illumination, chemical fertilizer and the like for crop growth and cultivation, and realizes real-time optimization of crop cultivation.
The purpose of the invention can be realized by the following technical scheme:
a crop cultivation optimization method based on digital twinning comprises the following steps:
step 1: the crop real-time data acquisition comprises the steps of acquiring internal and external environment data in the crop growth process on line through a sensor, and acquiring crop pictures through a camera, so that the dynamic real-time monitoring of the crop growth and cultivation process is realized.
And 2, step: and constructing a crop cultivation database according to the integrated storage data.
And step 3: and preprocessing the crop cultivation data.
And 4, step 4: and predicting the preprocessed data of the internal and external environments of the crops, realizing the data prediction of the growth environment of the crops on the basis of a transformer network model, and predicting the current optimal data of the internal and external environments of the crops through a training network of the excellent cultivation historical data of the crops in a database.
And 5: a crop visualization model is constructed by collecting crop pictures, 3D visualization modeling is carried out according to the collected crop pictures, and a crop virtual end twin model is established.
Step 6: the crop simulation visualization is used for carrying out trial operation and adjustment optimization on the digital twin model through crop real-time data.
And 7: and (3) online deployment user decision making, outputting the digital twin model to a terminal user interface, and performing direct or indirect adjustment command deployment on line through the agricultural Internet of things control system.
Furthermore, the data of the internal and external environments of the crops comprise external data such as ambient temperature, air humidity and illumination intensity, and soil pH value and soil humidity.
Further, the crop cultivation data is preprocessed in the step 3, firstly, the unification of the data time sequence and the sampling frequency is realized through a linear mean value method, and the formula is as follows:
where m, j denotes the sample number, T 1 And T 2 Representing two different sampling frequencies, X 1 Representation based on frequency T 1 Sampling the resulting data, X 1j The representation is based on X l Generating and aligning a sampling frequency T 2 New data dimension of (2), utilizing X' 1j Reference value X of the preceding and following fields 1m And X 1(m+1)l And their distance ratio on the time axis, weighted calculation X' 1j An approximation of (d).
Then, the data magnitude is unified through data normalization, and the calculation formula is as follows:
wherein u is X′1 Is X' 1 Data mean of σ X′1 Represents X' 1 Standard deviation of the data.
Further, in the step 4, correlation analysis is performed on the preprocessed time sequence data of the internal and external environments of the crop history and the historical yield data, and a calculation formula is as follows:
where x, y represent a one-dimensional vector of two variables, ρ x，y Representing the correlation coefficient of two variables, p x，y Has a value of [ -1,1]P is x，y =1 representing a completely linear positive correlation of two variables, p x，y =0 representing two variables wireless correlationSystem, rho x，y =1 represents a completely linear negative correlation of the two variables.
And training an optimal internal and external environment data prediction network of the crops through the preprocessed historical data of the crops, and predicting an optimal value of the internal and external environments for cultivating the crops. The optimal internal and external environment data prediction network structure of crops adopts a transformer network for prediction, an encoder extracts key features through distillation operation, an input sequence is mapped to three different sequence vectors which are Q (query), K (key) and V (value) through linear transformation, a sparse self-attention mechanism training weight matrix is adopted, and a sparse self-attention mechanism calculation function is as follows:
Q＝XW Q
K＝YW K
V＝YW V
wherein, the first and the second end of the pipe are connected with each other,
are all linear matrices, d = d k ，d k Is the dimension of Q and K, d v Is the dimension of V, is>
The vector Q is used to average each component, and only comprises the input sequence under sparse evaluation, K T Representing the transpose of the vector K, Q is a sparse matrix with the same dimensions as the input sequence, and it contains only the input sequence under sparse evaluation.
Further, the crop growth image prediction network in the step 5 adopts a convolution LSTM network model and an input gate i of the network t Forgetting door f t And an output gate o t The operation function is as follows:
wherein, C t Is the output of each layer, represents the operation of convolution operation,
represents the Hadamard product, C t And hidden layer state H t The operation function is as follows:
training a convolution LSTM through an appearance image sequence of crops, which is stored in a database and changes along with cultivation time, inputting a network into an appearance image of the crops in the past t time, outputting a predicted image after the t time, setting a parameter t for a user, and cultivating the crops into a three-dimensional model after the t time through a 3D modeling method
And the type is visually displayed to a user, so that the user can observe whether the growth condition of the crops meets the expectation.
Further, in the step 6, the crop simulation visualization is to establish a 3D twin model of the current crop and a 3D model after the cultivation time by using a 3D modeling method through the crop appearance image and the predicted image after the cultivation time t collected in real time, so that a user can observe the growth trend of the current crop, and the optimal value of the internal and external environments of the crop predicted by the crop is visually displayed at the twin model.
And (4) trial running the digital twin model through real-time data of crops, and adjusting and optimizing the model according to a test result.
Further, the step 7 of deploying user decisions online, synchronizing the calculation result of the digital twin model to a user terminal interface online, and selecting the digital twin model to directly output a regulation command or indirectly output the regulation command through an operator by the user according to the specific conditions of the crops.
Further, the crop visualization comprises the current 3D model simulation visualization of crops, the 3D model visualization after the crop cultivation time t and the optimal value visualization of the internal and external environments of the crops.
Further, the optimal internal and external environment data of the crops are predicted in a network, and the network input is past T 1 Time sequence data of internal and external environments of crops in time, and network output is future T 2 Time series data of internal and external environments of crops in time, wherein T 1 And T 2 Can be finely adjusted according to specific crops.
Further, the crop growth image prediction network has the network input of past t 1 The appearance image sequence of crops in time and the network output is cultivation t 2 Appearance image of the crop after time, wherein t 1 And t 2 Can be finely adjusted according to specific crops.
The invention has the beneficial effects that:
1. the crop cultivation optimization method is based on a digital twin model, adopts a neural network to predict the current optimal internal and external environment data of crops, provides an optimal cultivation scheme visually through the twin model, and realizes the simulation visualization of a 3D model and a growth trend of the crops, so that a user can observe the growth state of the crops in real time and deploy control commands online;
2. the crop cultivation optimization method facilitates users to optimize and regulate crop cultivation environment in time, effectively saves manpower and material resources, and realizes more refined and intelligent management of crops.
Drawings
The invention will be further described with reference to the accompanying drawings.
FIG. 1 is a block diagram of a method of optimizing crop cultivation according to the present invention;
FIG. 2 is a diagram of a network for predicting optimum internal and external environmental data of crops in the optimization method of the present invention;
FIG. 3 is a network structure diagram of the prediction of the appearance image of crops in the optimization method of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1, a crop cultivation optimization method based on digital twinning establishes a crop digital twinning model, the digital twinning model comprises a crop cultivation database, a crop optimal internal and external environment data prediction network, a crop growth image prediction network, a visual display module and an online deployment module, and the crop cultivation optimization method comprises the following steps:
step 1: crop real-time data acquisition
The method comprises the steps of obtaining internal and external environment data of crops through a temperature sensor, a humidity sensor and a light intensity sensor which are arranged in the air and a soil sensor which is connected to a seedling raising plate, wherein the internal and external environment data of the crops comprise external data such as environment temperature, air humidity and illumination intensity and internal data such as soil pH value and soil humidity, the internal and external environment data are collected in real time through the sensors, a high-definition image collecting device is adopted to collect multiple groups of external images of the crops at every interval unit time t, and the collected internal and external environment data and the collected external images of the crops are integrally stored in a database according to specific time points.
Step 2: building crop cultivation database according to integrated storage data
Crop real-time and historical data required by the digital twin model are stored in a centralized manner, and the crop real-time and historical data mainly comprise internal and external environment time sequence data of successful crop cultivation cases, crop yield data and appearance images which change along with time in the crop cultivation process, and can be inquired and filed in real time by a system.
And 3, step 3: preprocessing crop cultivation data
Preprocessing input data, unifying time sequence and sampling frequency of the input data by a linear mean value method, and setting the sampling interval time of air humidity and ambient temperature as T 1 ，T 2 ，T 1 And T 2 The time sequence of original data representing two different sampling frequencies and air humidity sampling is X l The time sequence of the environmental temperature sampling is X 2 ，X 1j Representation is based on X 1 Generating and aligning a sample interval time T of an ambient temperature 2 According to X, the new data dimension of 1 Generating time sequence data X 'unified with the sampling frequency of the environment temperature' 1 The calculation formula of (c) is as follows:
wherein, m is more than j and less than m +1. Then, the data magnitude is unified through data normalization, m and j represent the serial number of sampling, T 1 And T 2 Representing two different sampling frequencies, X 1 Representation based on frequency T 1 Sample the resulting data, X' 1j Representation is based on X l Generating and aligning a sampling frequency T 2 New data dimension of (2), utilizing X' 1j Reference value of the preceding and following fields, X 1m And X l(m+1) l and their distance ratio on the time axis, weighted calculation X' 1j The approximate calculation formula of (c) is as follows:
wherein u is X′1 Is X' 1 Data mean of σ X′1 Is X' 1 The standard deviation of the data of (2) and the processing of other environment data are the same. The processed data can be subjected to standard normal distribution, and the subsequent calculation amount is reduced.
And 4, step 4: predicting the preprocessed internal and external environment data of the crops, realizing the prediction of the growth environment data of the crops on the basis of a transformer network model, and predicting the current optimal data of the internal and external environment of the crops through a training network of excellent cultivation historical data of the crops in a database
Carrying out correlation analysis on the preprocessed time sequence data of the historical internal and external environments of the crops and the historical yield data, wherein the calculation formula is as follows:
where x, y represent a one-dimensional vector of two variables, p x， y denotes the correlation coefficient of two variables, ρ x，y Has a value of [ -1,1]P is x，y =1 represents a completely linear positive correlation of the two variables, p x，y =0 represents the wireless correlation of two variables, ρ x，y =1 represents a completely linear negative correlation of the two variables; ρ is a unit of a gradient x，y The value of (2) can visually reflect the correlation between the two variables, so that the feature selection in the subsequent steps is facilitated, a variable-yield correlation matrix diagram is constructed through the correlation coefficient, the dimension with the correlation coefficient of the crop yield being larger than 0.2 is screened, and the repeated dimension with the highly consistent features in the internal and external environment data is reduced.
Training an optimal internal and external environment data prediction network of the crops through the preprocessed historical data of the crops, and predicting optimal values of internal and external environments for crop cultivation; the optimal internal and external environment data prediction network structure of crops is shown in figure 2, and the input of the optimal internal and external environment data prediction network of crops is past T 1 Time sequence data of internal and external environments of crops in time, and network output is future T 2 Time sequence data of internal and external environments of crops in time. Wherein, T 1 And T 2 Can be finely adjusted according to specific crops.
Considering that the crop cultivation period is long, the problems that gradient disappearance, explosion and the like easily occur in the conventional RNN for long-term time series data prediction, therefore, the method adopts a transformer network for prediction, an encoder extracts key features through distillation operation, an input sequence is mapped to three different sequence vectors which are Q (query), K (key) and V (value) through linear transformation, a sparse self-attention mechanism training weight matrix is adopted, and a sparse self-attention mechanism calculation function is as follows:
Q＝XW Q
K＝YW K
V＝YW V
wherein the content of the first and second substances,
are all linear matrices, d = d k ，d k Is the dimension of Q and K, d v Is the dimension of a V, <' >>
The vector Q is used to average each component, and only comprises the input sequence under sparse evaluation, K T Representing the transpose of the vector K, Q is a sparse matrix with the same dimensions as the input sequence and it contains only the input sequence under sparse evaluation, so the sparse self-attention mechanism only needs to compute O (lnL) for each input sequence q ) The dot product operation effectively reduces the computational complexity and memory consumption of the network model; d represents the dimensions of Q and K, the attention layer will @>
And the corresponding K are aggregated, and the obtained result is aggregated with V again to generate a final output vector.
Pre-processed internal and external environment data
Dividing the model into a training set and a verification set, and carrying out normalization or standardization processing, wherein the input of the network model is D t-N+1 ，...，D t-M The output is D t-M+1 ，...，D t-M+2 ，D t Training the model by using a training set, verifying a set tuning model, selecting the optimal model as a prediction model and storing the optimal model.
And 5: the visual display module builds a crop visual model by collecting crop pictures, carries out 3D visual modeling according to the collected crop pictures, builds a crop virtual end twin model, predicts an appearance image of the crops after t unit time of cultivation through a neural network, and builds a visual model of the crops after t unit time of growth through 3D modeling, so that a user can observe the visual model conveniently.
Training a crop growth image prediction network through a crop historical appearance image sequence, predicting an appearance image after t time of crop cultivation, and predicting a network structure of the crop growth image, wherein as shown in figure 3, the input of the crop growth image prediction network is t in the past 1 The appearance image sequence of crops in time and the network output is cultivation t 2 Appearance image of the crop after time. Wherein, t 1 And t 2 Can be finely adjusted according to specific crops.
Training convolution LSTM through appearance images of crop cultivation processes in a database, and capturing space-time sequence characteristics through an LSTM network model; the input and output of the convolution LSTM are three-dimensional tensors that hold all spatial information, and are composed of two networks, a coding network and a prediction network, the initial state and output of the prediction network are copied from the final state of the coding network, the input gates i of the network t Door f for forgetting to leave t And an output gate o t Each layer outputting C t And hidden layer state H t The operation function of (1) is as follows:
wherein, C t Is the output of each layer, represents the operation of convolution operation,
representing the Hadamard product，C t And hidden layer state H t The operation function is as follows:
the convolution LSTM is trained through an appearance image sequence of crops, which is stored in a database and changes along with the cultivation time, the appearance images of the crops in the past t time are input into a network, the predicted images in the past t time are output, the parameter t can be set by a user, and a 3D twin model of the current crops and a 3D model after the cultivation t time are established by a 3D modeling method for the user to observe the growth trend of the current crops. And the predicted optimal values of the internal and external environments of the crops are visually displayed at the twin model, so that the optimal cultivation suggestions are visually displayed.
Step 6: commissioning and adjusting optimization of digital twin model through real-time crop data
The trained model is tested in a real-time crop cultivation environment using the most recent data to verify its accuracy and validity. Since the training set used by the network model is mainly historical data, it is necessary to re-verify the validity and make necessary updates in a real-time crop cultivation environment. The commissioning of step 6 must be performed before the digital twin model is actually deployed online to the crop cultivation environment.
After the trial operation, the model needs to be further optimized according to the test result and the feedback of the first-line crop cultivation personnel, and then the model can be put into crop cultivation practice.
And 7: the online deployment module outputs the digital twin model to a terminal user interface for a user to observe the crop cultivation condition in real time, and direct or indirect adjustment command deployment can be carried out online through the agricultural Internet of things control system
The digital twin model establishes connection with other related systems (such as a LIMS system and a real-time database system) through the agricultural Internet of things to obtain real-time internal and external environment data and an external appearance image of the crops as an input data source of the model.
The digital twin model can directly convert the predicted current optimal internal and external environment data into a control command through an online deployment decision of an agricultural Internet of things and output the control command to an automatic control system, or integrate cultivation optimization suggestion information into a control management system through a visual user terminal interface, show the visual growth trend of crops to cultivation personnel and output the control command online through the cultivation personnel; aiming at crops with higher cultivation cost, an indirect control mode can be adopted; aiming at crops with lower cultivation cost and higher real-time control requirement, a direct control mode can be adopted.
The method is based on a digital twin model, adopts a neural network to predict the current optimal internal and external environmental data of the crops, provides an optimal cultivation scheme visually through the twin model, and realizes the simulation visualization of a 3D model and a growth trend of the crops, so that a user can observe the growth state of the crops in real time and deploy control commands online. The invention is convenient for users to optimize and regulate the crop cultivation environment in time, effectively saves manpower and material resources, and realizes more refined and intelligent management of crops.
In the description herein, reference to a term means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. While the present disclosure has been described in considerable detail and with particular reference to a few illustrative embodiments thereof, it is not intended to be limited to any such details or embodiments or any particular embodiments, but it is to be construed as effectively covering the intended scope of the disclosure by providing a broad, potential interpretation of such claims in view of the prior art with reference to the appended claims. Furthermore, the foregoing description of the present disclosure has been presented in terms of embodiments foreseen by the inventors for purposes of providing a useful description, and enabling one of ordinary skill in the art to devise equivalent variations of the present disclosure that are not presently foreseen.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and so forth) having computer-usable program code embodied therein. The scheme in the embodiment of the application can be implemented by adopting various computer languages, such as object-oriented programming language Java and transliterated scripting language JavaScript.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While the preferred embodiments of the present application have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including the preferred embodiment and all changes and modifications that fall within the scope of the present application.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present application without departing from the spirit and scope of the application. Thus, if such modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is intended to include such modifications and variations as well.
Claims (10)
1. A crop cultivation optimization method based on digital twinning is characterized by comprising the following steps:
step 1: the method comprises the following steps of (1) crop real-time data acquisition, namely acquiring internal and external environment data in the crop growth process on line through a sensor, and acquiring crop pictures through a camera, so as to realize dynamic real-time monitoring on the crop growth and cultivation process;
step 2: constructing a crop cultivation database according to the integrated storage data;
and step 3: preprocessing crop cultivation data;
and 4, step 4: predicting the preprocessed internal and external environment data of the crops, realizing the prediction of the growth environment data of the crops based on a transformer network model, and predicting the current optimal data of the internal and external environment of the crops through a training network of excellent cultivation historical data of the crops in a database;
and 5: establishing a crop visual model by collecting crop pictures, performing 3D visual modeling according to the collected crop pictures, and establishing a crop virtual end twin model;
step 6: the crop simulation visualization is used for carrying out trial operation and adjustment optimization on the digital twin model through crop real-time data;
and 7: and (3) deploying user decisions online, outputting the digital twin model to a terminal user interface, and performing direct or indirect adjustment command deployment online through an agricultural Internet of things control system.
2. The crop cultivation optimizing method based on the digital twinning as claimed in claim 1, wherein the data of the internal and external environment of the crop includes external data such as ambient temperature, air humidity, illumination intensity, soil pH value and soil humidity.
3. The method as claimed in claim 1, wherein the crop cultivation data in step 3 is preprocessed, and the unification of the data time sequence and the sampling frequency is firstly realized by a linear mean value method, wherein the formula is as follows:
where m, j denotes the sample number, T 1 And T 2 Representing two different sampling frequencies, X 1 Representation based on frequency T 1 Sample the resulting data, X' 1j Representation is based on X 1 Generating and aligning a sampling frequency T 2 X 'is used as the new data dimension of (1)' 1j Reference value X of the preceding and following fields 1m And X 1(m+1) And distance ratio on their time axis, weight calculation X' 1j An approximation of (d);
then, the data magnitude is unified through data normalization, and the calculation formula is as follows:
wherein, the first and the second end of the pipe are connected with each other,
is X' 1 Data of (2)Mean value,. Or>
Is X' 1 Standard deviation of the data of (1).
4. The digital twinning-based crop cultivation optimization method as claimed in claim 1, wherein the correlation analysis is performed in step 4 on the preprocessed time sequence data of the historical internal and external environments of the crop and the historical yield data, and the calculation formula is as follows:
where x, y represent a one-dimensional vector of two variables, ρ x，y Representing the correlation coefficient, p, of two variables x，y Has a value of [ -1,1]P is x,y =1 represents a completely linear positive correlation of the two variables, p x,y =0 represents the wireless correlation of two variables, ρ x,y =1 represents a completely linear negative correlation of the two variables;
training an optimal internal and external environment data prediction network of the crops through the preprocessed historical data of the crops, and predicting optimal values of internal and external environments for crop cultivation; the optimal internal and external environment data prediction network structure of crops adopts a transformer network for prediction, an encoder extracts key features through distillation operation, an input sequence is mapped to three different sequence vectors which are Q (query), K (key) and V (value) through linear transformation, a sparse self-attention mechanism training weight matrix is adopted, and a sparse self-attention mechanism calculation function is as follows:
Q＝XW Q
K＝YW K
V＝YW V
wherein, the first and the second end of the pipe are connected with each other,
are all linear matrices, d = d k ,d k Is the dimension of Q and K, d v Is the dimension of a V, <' >>
The vector Q is used for averaging each component, and only comprises an input sequence under sparse evaluation, K T Representing the transpose of the vector K, Q is a sparse matrix with the same dimensions as the input sequence, and it contains only the input sequence under sparse evaluation.
5. The method as claimed in claim 1, wherein the prediction network of the crop growth image in step 5 adopts a convolution LSTM network model, and the input gate i of the network t Forgetting door f t And an output gate o t The operation function is as follows:
wherein, C t Is the output of each layer, represents the operation of convolution operation,
representing the Hadamard product, C t And hidden layer state H t The operation function is as follows:
the convolution LSTM is trained through an appearance image sequence of crops, which is stored in a database and changes along with cultivation time, the appearance images of the crops in the past t time are input into a network, predicted images after t time are output, the parameter t can be set by a user, and then a three-dimensional model of the crops after t time cultivation is displayed to the user in a visualized mode through a 3D modeling method for the user to observe whether the growth conditions of the crops meet expectations or not.
6. The crop cultivation optimization method based on the digital twin according to claim 1, wherein in the step 6, crop simulation visualization is performed by acquiring a crop appearance image and a predicted image of the crops after the cultivation time t in real time, a 3D twin model of the current crops and a 3D model of the crops after the cultivation time t are established by using a 3D modeling method, so that a user can observe the growth trend of the current crops, and the optimal values of the internal and external environments of the crops, which are predicted by the crops, are visually displayed at the twin model;
and (4) trial running the digital twin model through real-time data of crops, and adjusting and optimizing the model according to a test result.
7. The crop cultivation optimization method based on digital twins as claimed in claim 1, wherein the step 7 deploys user decisions online, synchronizes the calculation results of the digital twins model to the user terminal interface online, and the user selects the digital twins model to directly output the regulation and control command or indirectly output the regulation and control command through an operator according to the specific conditions of the crops.
8. The digital twinning-based crop cultivation optimization method as claimed in claim 6, wherein the crop visualization includes a current 3D model simulation visualization of the crop, a 3D model visualization after t time of crop cultivation, and an optimal value visualization of the internal and external environment of the crop.
9. The method as claimed in claim 4, wherein the optimal internal and external environmental data of the crop is predicted network, and the network input is past T 1 Time sequence data of internal and external environments of crops in time, and network output is future T 2 Time series data of internal and external environments of crops in time, wherein T 1 And T 2 Can be finely adjusted according to specific crops.
10. A method as claimed in claim 5 wherein said crop growth image predicts a network where the network input is past t 1 The appearance image sequence of crops in time and the network output is cultivation t 2 Appearance image of the crop after time, wherein t 1 And t 2 Can be finely adjusted according to specific crops.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211638555.XA CN115880433A (en) | 2022-12-20 | 2022-12-20 | Crop cultivation optimization method based on digital twinning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211638555.XA CN115880433A (en) | 2022-12-20 | 2022-12-20 | Crop cultivation optimization method based on digital twinning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115880433A true CN115880433A (en) | 2023-03-31 |
Family
ID=85754046
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211638555.XA Pending CN115880433A (en) | 2022-12-20 | 2022-12-20 | Crop cultivation optimization method based on digital twinning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115880433A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116664332A (en) * | 2023-08-01 | 2023-08-29 | 安徽金海迪尔信息技术有限责任公司 | Agricultural production monitoring system based on digital twinning |
| CN116738766A (en) * | 2023-08-11 | 2023-09-12 | 安徽金海迪尔信息技术有限责任公司 | Intelligent agriculture online industrialization service system based on digital twinning |
| CN116760908A (en) * | 2023-08-18 | 2023-09-15 | 浙江大学山东(临沂)现代农业研究院 | Agricultural information optimization management method and system based on digital twin |
| CN117291444A (en) * | 2023-11-27 | 2023-12-26 | 华兴国创(北京)科技有限公司 | Digital rural business management method and system |
-
2022
- 2022-12-20 CN CN202211638555.XA patent/CN115880433A/en active Pending
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116664332A (en) * | 2023-08-01 | 2023-08-29 | 安徽金海迪尔信息技术有限责任公司 | Agricultural production monitoring system based on digital twinning |
| CN116664332B (en) * | 2023-08-01 | 2023-10-13 | 安徽金海迪尔信息技术有限责任公司 | Agricultural production monitoring system based on digital twinning |
| CN116738766A (en) * | 2023-08-11 | 2023-09-12 | 安徽金海迪尔信息技术有限责任公司 | Intelligent agriculture online industrialization service system based on digital twinning |
| CN116738766B (en) * | 2023-08-11 | 2023-10-13 | 安徽金海迪尔信息技术有限责任公司 | Intelligent agriculture online industrialization service system based on digital twinning |
| CN116760908A (en) * | 2023-08-18 | 2023-09-15 | 浙江大学山东(临沂)现代农业研究院 | Agricultural information optimization management method and system based on digital twin |
| CN116760908B (en) * | 2023-08-18 | 2023-11-10 | 浙江大学山东(临沂)现代农业研究院 | Agricultural information optimization management method and system based on digital twin |
| CN117291444A (en) * | 2023-11-27 | 2023-12-26 | 华兴国创(北京)科技有限公司 | Digital rural business management method and system |
| CN117291444B (en) * | 2023-11-27 | 2024-01-19 | 华兴国创(北京)科技有限公司 | Digital rural business management method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115880433A (en) | Crop cultivation optimization method based on digital twinning | |
| WO2022253057A1 (en) | Intelligent moisture precise irrigation control system and method for fruit and vegetable cultivation in solar greenhouse | |
| CN111008733B (en) | Crop growth control method and system | |
| CN106650212A (en) | Intelligent plant breeding method and system based on data analysis | |
| Li et al. | Prediction of plant transpiration from environmental parameters and relative leaf area index using the random forest regression algorithm | |
| CN108633697A (en) | A kind of foster culture method of the intelligent plant based on the daily data analysis of plant and cloud | |
| CN114297907A (en) | Greenhouse environment spatial distribution prediction method and device | |
| CN110532936A (en) | A kind of method and system identifying field crop growing way monitoring image Green plant | |
| CN109344738A (en) | The recognition methods of crop diseases and pest crop smothering and device | |
| CN117036088A (en) | Data acquisition and analysis method for identifying growth situation of greening plants by AI | |
| CN114529097B (en) | Multi-scale crop phenological period remote sensing dimensionality reduction prediction method | |
| CN116661530B (en) | Intelligent control system and method in edible fungus industrial cultivation | |
| CN115294518B (en) | Intelligent monitoring method and system for precise greenhouse cultivation of horticultural plants | |
| CN106989778A (en) | A kind of Tailings Dam on-line monitoring system | |
| Shadrin et al. | Pervasive agriculture: Measuring and predicting plant growth using statistics and 2D/3D imaging | |
| KR102387765B1 (en) | Method and apparatus for estimating crop growth quantity | |
| KR20230061034A (en) | Method of training machine learning model for estimating plant chlorophyll contents, method of estimating plant growth quantity and plant growth system | |
| WO2020228568A1 (en) | Method for training power generation amount prediction model of photovoltaic power station, power generation amount prediction method and device of photovoltaic power station, training system, prediction system and storage medium | |
| CN113796228B (en) | Plant cultivation system and method based on digital twinning | |
| CN108053418A (en) | A kind of animal background modeling method and device | |
| CN116579509A (en) | Photovoltaic power generation prediction method based on virtual reality | |
| CN109828623B (en) | Production management method and device for greenhouse crop context awareness | |
| Tidake et al. | Design efficient model to increase crop yield using deep learning | |
| Lu et al. | Image classification and identification for rice leaf diseases based on improved WOACW_SimpleNet | |
| CN112784477A (en) | WRF-LES and BP-PSO-Bagging combined wind power prediction method |
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 |