CN114297907A - Greenhouse environment spatial distribution prediction method and device - Google Patents
Greenhouse environment spatial distribution prediction method and device Download PDFInfo
- Publication number
- CN114297907A CN114297907A CN202111361824.8A CN202111361824A CN114297907A CN 114297907 A CN114297907 A CN 114297907A CN 202111361824 A CN202111361824 A CN 202111361824A CN 114297907 A CN114297907 A CN 114297907A
- Authority
- CN
- China
- Prior art keywords
- greenhouse
- environment
- prediction
- network
- features
- 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 provides a greenhouse environment spatial distribution prediction method and a device, wherein the method comprises the following steps: determining a feature vector matrix comprising a plurality of groups of time sequence features according to a plurality of groups of environment features of the current and previous moments of the predicted point, wherein the environment features comprise greenhouse internal environment features and greenhouse external environment features; inputting the feature vector matrix into a convolutional network of a trained prediction model for feature extraction to obtain an extracted global feature vector; and inputting the global feature vector into a time series network of the prediction model, and outputting prediction results of the temperature, the humidity, the illumination intensity and the carbon dioxide content after preset time. The method can accurately mine the characteristics of large hysteresis, time sequence, nonlinearity and space distribution difference of greenhouse microclimate environment data, and combines the better mining time correlation of a convolutional network and a time sequence network, so that the space distribution trend of the multipoint environment at the future moment of the greenhouse can be more accurately predicted, and a decision basis can be provided for the overall environment regulation and control of the greenhouse.
Description
Technical Field
The invention relates to the field of agricultural information processing, in particular to a greenhouse environment spatial distribution prediction method and device.
Background
The regulation and control mode of the greenhouse fruiting room mostly depends on the current actual environment variable to make regulation and control decisions, cannot solve the problem of large hysteresis of the microclimate environment in the fruiting room, and is not beneficial to the growth of crops. The development of a high-precision prediction model taking the main environmental factors of the mushroom house as output variables can master the environmental distribution trend of the main environmental factors in a future period in advance, and is an important premise for carrying out efficient and accurate environmental early warning and pre-regulation and control of fruiting houses.
Environmental variables influencing the growth and development of edible fungi in the fruiting room of the greenhouse are dynamic variables changing along with time, and are interacted and mutually coupled, and the spatial distribution has difference and nonuniformity. At present, around the characteristics of time sequence and nonlinearity, a prediction method adopted by greenhouse fruiting room microclimates mainly comprises a time sequence analysis method, a regression prediction method, a support vector machine, an artificial neural network and the like, but has great limitation on the prediction capability of complex and variable fruiting room microclimate data, and the time and space correlation of greenhouse environment data is difficult to accurately mine, so that the prediction result is inaccurate.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a greenhouse environment spatial distribution prediction method and a greenhouse environment spatial distribution prediction device.
The invention provides a greenhouse environment spatial distribution prediction method, which comprises the following steps: determining a feature vector matrix comprising a plurality of groups of time series features according to a plurality of groups of environment features of the current and previous moments of the predicted point, wherein the environment features comprise greenhouse internal environment features and greenhouse external environment features; inputting the feature vector matrix into a convolutional network of a trained prediction model for feature extraction to obtain an extracted global feature vector; inputting the global feature vector into a time series network of the prediction model, and outputting prediction results of the temperature, humidity, illumination intensity and carbon dioxide content of a prediction point in the greenhouse after a preset time; and the prediction model is obtained by determining samples of the characteristic vector matrix according to the known humidity, the illumination intensity and the carbon dioxide content in the greenhouse after the preset time, and training the samples, and comprises a convolution network and a time sequence network.
According to the greenhouse environment spatial distribution prediction method, the environment characteristics in the greenhouse comprise substrate temperature, air humidity, illumination intensity and carbon dioxide concentration; the environmental characteristics outside the greenhouse comprise air humidity, air temperature and wind speed.
According to the greenhouse environment spatial distribution prediction method, the environment features further comprise space sparse features, and the space sparse features comprise regional features, ventilation features and humidification features of the predicted points.
According to the greenhouse environment spatial distribution prediction method of an embodiment of the present invention, before determining the feature vector matrix including the plurality of sets of time series features according to the plurality of sets of environmental features at the current and previous time of the prediction point, the method further includes: interpolating missing data by a linear interpolation method, and smoothing abnormal data by an averaging method; the interpolation and smoothing methods respectively comprise:
wherein, 0<i<j，xa+iData missing at time a + i, xaAnd xa+jRaw data for time a and a + j, xkAs exception data, xk-1、xk+1Is the adjacent valid data.
According to the greenhouse environment spatial distribution prediction method, the characteristic vector matrix is input into a convolutional network of a trained prediction model for characteristic extraction, and the method comprises the following steps: inputting the feature vector matrix into the convolution network, carrying out four times of convolution through the convolution network, and outputting a global feature vector subjected to dimension reduction according to leveling operation and an activation function; wherein, after every two convolutions, one maximum pooling is performed.
According to the greenhouse environment spatial distribution prediction method provided by the embodiment of the invention, the step of inputting the global feature vector into the time series network of the prediction model and outputting prediction results of the temperature, humidity, illumination intensity and carbon dioxide content of a prediction point in the greenhouse after a preset time length comprises the following steps: inputting the global feature vector into a time series network, and sequentially performing feature processing through two Gate controlled cycle Unit (GRU) network layers of the time series network; according to a full connection layer based on a Linear rectification function (RELU) activation function, after reverse normalization, prediction results of temperature, humidity, illumination intensity and carbon dioxide content in the greenhouse after preset time are output.
According to the greenhouse environment spatial distribution prediction method of an embodiment of the present invention, before determining the feature vector matrix including the plurality of sets of time series features according to the plurality of sets of environmental features at the current and previous time of the prediction point, the method further includes: acquiring sample data of temperature and humidity, illumination intensity and carbon dioxide content in the greenhouse after a preset time is known, respectively taking the temperature and humidity, the illumination intensity and the carbon dioxide content as labels of the samples, and determining a characteristic vector matrix of the samples; and adjusting model parameters based on an adaptive moment estimation (Adam) optimization algorithm by using the sample data, and training to obtain the prediction model.
The invention also provides a greenhouse environment spatial distribution prediction device, comprising: the data acquisition module is used for determining a characteristic vector matrix comprising a plurality of groups of time series characteristics according to a plurality of groups of environmental characteristics of the current and previous moments of the predicted point, wherein the environmental characteristics comprise greenhouse internal environmental characteristics and greenhouse external environmental characteristics; the first processing module is used for inputting the characteristic vector matrix into a convolutional network of a trained prediction model for characteristic extraction to obtain an extracted global characteristic vector; the second processing module is used for inputting the global feature vector into the time series network of the prediction model and outputting prediction results of the temperature, humidity, illumination intensity and carbon dioxide content of prediction points in the greenhouse after preset time; and the prediction model is obtained by determining samples of the characteristic vector matrix according to the known humidity, the illumination intensity and the carbon dioxide content in the greenhouse after the preset time, and training the samples, and comprises a convolution network and a time sequence network.
The invention also provides an electronic device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the program to realize the steps of the greenhouse environment spatial distribution prediction method.
The present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when being executed by a processor, carries out the steps of the method for predicting the spatial distribution of a greenhouse environment as defined in any one of the above.
According to the greenhouse environment spatial distribution prediction method and device, the characteristics of large hysteresis, time sequence, nonlinearity and spatial distribution difference of greenhouse microclimate environment data can be accurately mined by extracting the multiple groups of time sequence characteristics of the characteristics in the greenhouse and the characteristics outside the greenhouse, the characteristic correlation can be better processed by combining a convolution network, and the time correlation can be better mined by combining a time sequence network, so that the prediction error and the volatility are smaller, the spatial distribution trend of the multipoint environment at the future time of the greenhouse can be more accurately predicted, and a decision basis can be provided for the overall environment regulation and control of the greenhouse.
Drawings
In order to more clearly illustrate the technical solutions of the present invention or the prior art, the drawings needed for the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
FIG. 1 is a schematic flow chart of a method for predicting spatial distribution of greenhouse environment according to the present invention;
FIG. 2 is a second schematic flow chart of the method for predicting spatial distribution of greenhouse environment according to the present invention;
FIG. 3 is a block diagram of a CNN-GRU neural network model provided by the present invention;
FIG. 4 is a diagram of the internal structural elements of a GRU provided by the present invention;
FIG. 5 is a schematic structural diagram of a greenhouse environment spatial distribution prediction device provided by the present invention;
fig. 6 is a schematic structural diagram of an electronic device provided in the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. 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.
The invention can be applied to the prediction of environmental characteristic distribution in mushroom houses, and can also be applied to the prediction of greenhouse indoor environmental characteristics of other crops, and the prediction is exemplified by the mushroom house environmental prediction.
The following describes the greenhouse environment spatial distribution prediction method and apparatus of the present invention with reference to fig. 1-6. Fig. 1 is a schematic flow chart of a greenhouse environment spatial distribution prediction method provided by the present invention, and as shown in fig. 1, the present invention provides a greenhouse environment spatial distribution prediction method, which includes:
101. determining a feature vector matrix comprising a plurality of groups of time series features according to a plurality of groups of environment features of the current and previous moments of the predicted point, wherein the environment features comprise greenhouse internal environment features and greenhouse external environment features;
at present, the foundation of the whole model construction needs massive indoor environment data and outdoor meteorological data of greenhouse mushroom houses for training. For example, three data collectors are uniformly arranged in two mushroom houses to collect environmental data in the greenhouse, the instrument transmits the data to a server by using a GPRS network module every twenty minutes for storage, the downloaded data is stored in an Excel file format, the environmental characteristics are coded by using one-hot codes, and the environmental characteristics in the greenhouse are obtained after the coding. The outdoor meteorological data are acquired by an outdoor meteorological station at a position of 20m of the greenhouse mushroom house, 90% of data of each data acquisition point can be used as a training set, 10% of data of each data acquisition point can be used as a test set, and the characteristics of the environment outside the greenhouse can be obtained after the data are processed. And the plurality of sets of environmental characteristics can be a plurality of sets of data collected at fixed time steps. For example, environmental characteristics are collected once in 20 minutes.
During prediction, the collection and training of multiple groups of environmental features are the same, and finally, a feature vector matrix is generated. For example, the environment characteristics inside the greenhouse, including the substrate temperature, the air humidity, the illumination intensity and the carbon dioxide concentration, and the environment characteristics outside the greenhouse, including the air humidity, the air temperature and the wind speed. Finally, a two-dimensional matrix of the time step size multiplied by the eigenvector is obtained.
The invention can include various data preprocessing modes, such as normalization processing:
and normalizing each parameter and target variable. The calculation formula comprises:
in the formula, xmaxIs the maximum value; x is the number ofminIs the minimum value; x is the number of*Is a normalized value.
102. And inputting the characteristic vector matrix into a convolution network of a trained prediction model for characteristic extraction to obtain an extracted global characteristic vector.
The convolution network mainly captures deep feature rules in an input historical sequence, reduces dimensions of extracted high-dimensional features, compresses data, accelerates operation efficiency, and finally obtains a global feature vector after dimension reduction through convolution network processing. For the specific construction of the input layer, reference may be made to the convolutional network used for the image recognition process.
103. And inputting the global feature vector into a time series network of the prediction model, and outputting prediction results of the temperature, humidity, illumination intensity and carbon dioxide content of the prediction point in the greenhouse after a preset time length. And the prediction model is obtained by determining samples of the characteristic vector matrix according to the known humidity, the illumination intensity and the carbon dioxide content in the greenhouse after the preset time, and training the samples, and comprises a convolution network and a time sequence network.
The time series network learns the global feature vector extracted from the convolutional network, and may include GRU and LSTM networks.
The prediction model is obtained by training samples of temperature and humidity, illumination intensity and carbon dioxide content in the greenhouse after the preset time is known, and comprises a convolution network and a time sequence network two-layer network structure. Wherein, the parameters of the sample input model are the characteristic vector matrix extracted by the same method in 101.
According to the greenhouse environment spatial distribution prediction method, through the extraction of multiple groups of time sequence characteristics of the characteristics in the greenhouse and the characteristics outside the greenhouse, the characteristics of large hysteresis, time sequence, nonlinearity and spatial distribution difference of greenhouse microclimate environment data can be accurately mined, the characteristic correlation can be better processed by combining a convolution network, and the time correlation can be better mined by combining a time sequence network, so that the prediction error and the volatility are smaller, the spatial distribution trend of the multipoint environment at the future time of the greenhouse can be more accurately predicted, and a decision basis can be provided for the overall environmental regulation and control of the greenhouse.
The greenhouse mushroom house prediction early warning system and the deep learning are applied to greenhouse edible mushroom production, and the intelligent perception of the edible mushroom environment, growth and information is realized. The method takes the predicted environmental change as the basis for changing the growth environment, changes the traditional environmental control and decision management mode, reduces human resources, improves the management efficiency and realizes the intellectualization and automation of greenhouse crop production.
In one embodiment, the greenhouse internal environment characteristics include substrate temperature, air humidity, light intensity, and carbon dioxide concentration; the environmental characteristics outside the greenhouse, including air humidity, air temperature and wind speed, are as exemplified above.
In one embodiment, the environmental features further include spatially sparse features including regional features, ventilation features, and humidification features of the predicted point.
In the moment of window opening ventilation and humidifier opening and closing, and the phenomenon of short sudden change of outdoor weather, the indoor temperature and humidity value can fluctuate greatly, and certain influence can be caused on prediction precision. In order to obtain a more accurate prediction result, the embodiment of the invention takes the spatial sparsity feature into account.
The regional characteristics can include the distance between the predicted point and the ventilation system and the distance between the predicted point and the humidification system, the ventilation characteristics can include whether ventilation is performed or not, the air volume, and the humidification characteristics can include whether humidification is performed or not and the humidification degree. And quantizing the information to obtain corresponding space sparse characteristics.
Specifically, according to the embodiment of the invention, the air temperature, the air relative humidity and the carbon dioxide concentration at a certain moment in the mushroom house, the outdoor air temperature, the outdoor air humidity, the outdoor wind speed, the indoor matrix temperature, the indoor illumination intensity, the environment distribution characteristic, the ventilation characteristic and the humidification characteristic are connected in series to form a brand new time series characteristic vector, the outdoor meteorological data, the indoor microclimate data and the current outdoor meteorological data of the mushroom house and the indoor microclimate data of the mushroom house are expressed to be a two-dimensional matrix with time step length multiplied by the characteristic vector, and the two-dimensional matrix is preprocessed and then input into the prediction model.
According to the method, the regional characteristics, the ventilation characteristics, the humidification characteristics and the like are added into the time sequence characteristics, and on the basis of time relevance, the spatial relevance among the characteristics can be further mined, so that the prediction precision is improved.
In one embodiment, before determining the feature vector matrix of the plurality of time series according to the in-greenhouse environment features and the out-greenhouse environment features at different time moments before the current time moment, the method further includes: and interpolating the missing data by a linear interpolation method, and smoothing the abnormal data by an averaging method.
The interpolation and smoothing methods respectively comprise:
wherein x isa+iData missing at time a + i, xaAnd xa+jRaw data for time a and a + j, xkAs exception data, xk-1、xk+1Is the adjacent valid data.
As the humidity value in the isothermal chamber of the mushroom house is larger, the precision of the sensor which is placed in the mushroom house for monitoring for a long time can be influenced to a certain extent. Meanwhile, the problems of network transmission quality, equipment faults, interference of human factors and the like cause the conditions of data abnormity, data loss and the like to occur in the data acquisition process of the sensor. Therefore, the linear interpolation method of the formula (2) is adopted to interpolate the short-time data, and if more lost data or larger time interval exists, the data at the same time in the adjacent days with the same weather type are adopted to fill up the missing data. And for abnormal data processing, performing smooth restoration on data by adopting an averaging method of a formula (3).
In one embodiment, the inputting the feature vector matrix into a convolutional network layer of a trained prediction model for feature extraction includes: inputting the feature vector matrix into the convolution network, carrying out four times of convolution through the convolution network, and outputting a global feature vector subjected to dimension reduction according to leveling operation and an activation function; wherein, after every two convolutions, one maximum pooling is performed.
When the greenhouse is a mushroom house, the characteristics of complex and variable mushroom house microclimate environment input data are considered, the invention is particularly designed with four layers of convolution layers (Conv2D), the number of the convolution layers is 16, 16, 32 and 32 in sequence, and a RELU activation function is selected for activation. Maximum pooling was performed every two consecutive convolutions (Maxplating 2D). In order to fully utilize the existing mushroom house environment distribution data, the size of a convolution kernel is set to be 3 multiplied by 3, the size of a pool is 2, and finally, through leveling operation, namely, Flatten operation, a multidimensional feature vector can be changed into a one-dimensional global feature vector after the Flatten operation, so that dimension reduction is carried out, and the one-dimensional global feature vector is converted into a global feature vector to be used as the input of a GRU layer.
According to the greenhouse environment spatial distribution prediction method provided by the embodiment of the invention, through the characteristics of CNN local connection and weight sharing, the dimension can be reduced by the characteristics of a pooling layer while high-level characteristics are captured, and parameters are reduced.
In one embodiment, the inputting the global feature vector into the time series network of the prediction model, and outputting the prediction results of the temperature, humidity, illumination intensity and carbon dioxide content of the prediction point in the greenhouse after a preset time period includes: inputting the global feature vector into a time series network, and sequentially performing feature processing through two GRU network layers of the time series network; according to the full-connection layer based on the RELU activation function, after reverse normalization, prediction results of the temperature, humidity, illumination intensity and carbon dioxide content in the greenhouse after preset time are output.
Fig. 2 is a second schematic flow chart of the greenhouse environment spatial distribution prediction method provided by the present invention, and as shown in the figure, it is found through the continuous improvement of the experiment of the present invention that the best prediction effect is achieved by constructing a two-layer GRU structure, the activation function is a RELU activation function, and the number of neurons is 64 and 128, respectively. And finally, outputting preset time duration, such as temperature, humidity and carbon dioxide vectors after 20 minutes, after the reverse normalization of the full connection layer (Dense), wherein fig. 3 is a structure diagram of the CNN-GRU neural network model provided by the invention, and can be seen in fig. 3. Fig. 4 is a diagram of an internal structural unit of a GRU provided in the present invention, and as shown in fig. 4, a calculation formula of a gated cyclic unit neural network (GRU) is as follows:
in the formula, rtAnd ztRespectively reset gate and refresh gate, WrTo reset the weight matrix of the gate, WzThe weight matrix of the gate is updated. x is the number oftFor input, htIn order to hide the output of the layer,
is to input xtAnd past hidden state ht-1σ (·) is the activation function sigmoid; tanh (-) is the activation function tanh.
According to the greenhouse environment spatial distribution prediction method, by combining the CNN-GRU, compared with the LSTM neural network and the GRU neural network, the training time is obviously accelerated, the loss value of the model is lower, and the prediction effect is better. The result shows that the CNN-GRU has good adaptability to complex and variable microclimate data of the mushroom house, and the operation efficiency of the model is improved.
In one embodiment, before determining the feature vector matrix including the plurality of sets of time-series features according to the plurality of sets of environmental features at the current and previous time of the predicted point, the method further includes: acquiring sample data of temperature and humidity, illumination intensity and carbon dioxide content in the greenhouse after a preset time is known, respectively taking the temperature and humidity, the illumination intensity and the carbon dioxide content as labels of the samples, and determining a characteristic vector matrix of the samples; and adjusting model parameters based on an Adam optimization algorithm by using the sample data, and training to obtain the prediction model.
The time step of the model is set to be 9 minutes, and parameters of the model are continuously adjusted by using an Adam optimization algorithm. The Adam optimization algorithm is an expansion type of a random gradient descent algorithm, and meanwhile, the advantages of the Adagrad and the RMSProp algorithms are obtained, so that the Adam optimization algorithm is suitable for large-scale data, and the performance of the model can be optimized more efficiently. In order to avoid that the model falls into the local minimum value, through multiple experiments and analysis, the initial learning rate is set to be 0.001, and the learning rate is adjusted to be 10% of the original learning rate every iteration for 100 times until the effect of the model reaches the best when the loss value tends to be stable.
Examples of the invention:
taking the application of greenhouse mushroom houses as an example, the greenhouse prediction early warning system can evaluate the overall environmental performance of the mushroom houses through the greenhouse mushroom house spatial distribution prediction result, and accurately early warn and provide regulation and control decision information.
Data acquisition and preprocessing: the indoor environment data is monitored by six data collectors (three for each greenhouse mushroom house) distributed in two adjacent greenhouse mushroom houses, and data points are coded by One-hot codes. The monitoring equipment adopts greenhouse cloud environment data collectors, and each data collector is installed on a tripod. Each data collector can measure air temperature, air relative humidity, soil temperature and CO2And the illumination intensity is sent to a remote cloud server for storage at set time intervals (20 min). Outdoor air temperature, air relative humidity, wind speed and other meteorological data are automatically acquired by an HOBO meteorological station about 20m away from the outside, the greenhouse prediction and early warning system can perform real-time online crawling on indoor environment data and outdoor meteorological data through a crawler technology, and a manager can also manually call downloaded data to read the data. Before the greenhouse prediction early warning system carries out prediction early warning, managers need to process abnormal data through a data preprocessing link.
Constructing a greenhouse prediction early warning system: the prediction early warning system is built in an environment of Pycharm, development is completed through PyQt5 in cooperation with Qt Designer, the prediction model is realized based on a Keras deep learning tool, a Tensorflows deep learning framework is used as a rear-end support, and the programming language is Python. The prediction early warning system mainly comprises three modules, namely a data acquisition module, a data prediction module and a data early warning module, and encapsulates the suitable growth environmental conditions of different edible fungi, can early warn each area and provide regulation and control decision information by combining the spatial distribution prediction result with the encapsulated suitable environmental conditions, and the decision information is transmitted to an application layer, so that the greenhouse mushroom house production related personnel can clearly master the growth state and the environmental information of the greenhouse edible fungi.
The following describes the greenhouse environment spatial distribution prediction apparatus provided by the present invention, and the greenhouse environment spatial distribution prediction apparatus described below and the greenhouse environment spatial distribution prediction method described above may be referred to with each other.
Fig. 5 is a schematic structural diagram of a greenhouse environment spatial distribution prediction apparatus according to the present invention, and as shown in fig. 5, the greenhouse environment spatial distribution prediction apparatus includes: a data acquisition module 501, a first processing module 502 and a second processing module 503. The data acquisition module 501 is configured to determine a feature vector matrix including a plurality of sets of time series features according to a plurality of sets of environmental features of a predicted point at a current time and a previous time, where the environmental features include an environment feature in a greenhouse and an environment feature outside the greenhouse; the first processing module 502 is configured to input the feature vector matrix into a convolutional network of a trained prediction model to perform feature extraction, so as to obtain an extracted global feature vector; the second processing module 503 is configured to input the global feature vector into the time series network of the prediction model, and output prediction results of the temperature, humidity, illumination intensity, and carbon dioxide content of a prediction point in the greenhouse after a preset time; and the prediction model is obtained by determining samples of the characteristic vector matrix according to the known humidity, the illumination intensity and the carbon dioxide content in the greenhouse after the preset time, and training the samples, and comprises a convolution network and a time sequence network.
The device embodiment provided in the embodiments of the present invention is for implementing the above method embodiments, and for details of the process and the details, reference is made to the above method embodiments, which are not described herein again.
According to the greenhouse environment spatial distribution prediction device provided by the embodiment of the invention, through the extraction of a plurality of groups of time sequence features of the features in the greenhouse and the features outside the greenhouse, the features of large hysteresis, time sequence, nonlinearity and spatial distribution difference of greenhouse microclimate environment data can be accurately mined, the feature correlation can be better processed by combining a convolution network, and the time correlation can be better mined by combining a time sequence network, so that the prediction error and the volatility are smaller, the spatial distribution trend of the multipoint environment at the future moment of the greenhouse can be more accurately predicted, and a decision basis can be provided for the overall environmental regulation and control of the greenhouse.
Fig. 6 is a schematic structural diagram of an electronic device provided in the present invention, and as shown in fig. 6, the electronic device may include: a processor (processor)601, a communication Interface (Communications Interface)602, a memory (memory)603 and a communication bus 604, wherein the processor 601, the communication Interface 602 and the memory 603 complete communication with each other through the communication bus 604. The processor 601 may invoke logic instructions in the memory 603 to perform a greenhouse environment spatial distribution prediction method, the method comprising: determining a feature vector matrix comprising a plurality of groups of time series features according to a plurality of groups of environment features of the current and previous moments of the predicted point, wherein the environment features comprise greenhouse internal environment features and greenhouse external environment features; inputting the feature vector matrix into a convolutional network of a trained prediction model for feature extraction to obtain an extracted global feature vector; inputting the global feature vector into a time series network of the prediction model, and outputting prediction results of the temperature, humidity, illumination intensity and carbon dioxide content of a prediction point in the greenhouse after a preset time; and the prediction model is obtained by determining samples of the characteristic vector matrix according to the known humidity, the illumination intensity and the carbon dioxide content in the greenhouse after the preset time, and training the samples, and comprises a convolution network and a time sequence network.
In addition, the logic instructions in the memory 603 may be implemented in the form of software functional units and stored in a computer readable storage medium when the logic instructions are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
In another aspect, the present invention also provides a computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, which when executed by a computer, enable the computer to perform the greenhouse environment spatial distribution prediction method provided by the above methods, the method comprising: determining a feature vector matrix comprising a plurality of groups of time series features according to a plurality of groups of environment features of the current and previous moments of the predicted point, wherein the environment features comprise greenhouse internal environment features and greenhouse external environment features; inputting the feature vector matrix into a convolutional network of a trained prediction model for feature extraction to obtain an extracted global feature vector; inputting the global feature vector into a time series network of the prediction model, and outputting prediction results of the temperature, humidity, illumination intensity and carbon dioxide content of a prediction point in the greenhouse after a preset time; and the prediction model is obtained by determining samples of the characteristic vector matrix according to the known humidity, the illumination intensity and the carbon dioxide content in the greenhouse after the preset time, and training the samples, and comprises a convolution network and a time sequence network.
In yet another aspect, the present invention further provides a non-transitory computer-readable storage medium, on which a computer program is stored, the computer program being implemented by a processor to perform the greenhouse environment spatial distribution prediction method provided in the above embodiments, the method including: determining a feature vector matrix comprising a plurality of groups of time series features according to a plurality of groups of environment features of the current and previous moments of the predicted point, wherein the environment features comprise greenhouse internal environment features and greenhouse external environment features; inputting the feature vector matrix into a convolutional network of a trained prediction model for feature extraction to obtain an extracted global feature vector; inputting the global feature vector into a time series network of the prediction model, and outputting prediction results of the temperature, humidity, illumination intensity and carbon dioxide content of a prediction point in the greenhouse after a preset time; and the prediction model is obtained by determining samples of the characteristic vector matrix according to the known humidity, the illumination intensity and the carbon dioxide content in the greenhouse after the preset time, and training the samples, and comprises a convolution network and a time sequence network.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A greenhouse environment spatial distribution prediction method is characterized by comprising the following steps:
determining a feature vector matrix comprising a plurality of groups of time series features according to a plurality of groups of environment features of the current and previous moments of the predicted point, wherein the environment features comprise greenhouse internal environment features and greenhouse external environment features;
inputting the feature vector matrix into a convolutional network of a trained prediction model for feature extraction to obtain an extracted global feature vector;
inputting the global feature vector into a time series network of the prediction model, and outputting prediction results of the temperature, humidity, illumination intensity and carbon dioxide content of a prediction point in the greenhouse after a preset time;
and the prediction model is obtained by determining samples of the characteristic vector matrix according to the known humidity, the illumination intensity and the carbon dioxide content in the greenhouse after the preset time, and training the samples, and comprises a convolution network and a time sequence network.
2. The greenhouse environment spatial distribution prediction method according to claim 1, wherein the greenhouse internal environment characteristics include substrate temperature, air humidity, illumination intensity and carbon dioxide concentration;
the environmental characteristics outside the greenhouse comprise air humidity, air temperature and wind speed.
3. The greenhouse environment spatial distribution prediction method of claim 1, wherein the environmental features further comprise spatial sparsity features comprising regional features, ventilation features, and humidification features of predicted points.
4. The method of predicting spatial distribution of greenhouse environment according to claim 1, wherein before determining the eigenvector matrix including the plurality of sets of time-series characteristics from the plurality of sets of environmental characteristics at the current and previous times of the predicted point, the method further comprises:
interpolating missing data by a linear interpolation method, and smoothing abnormal data by an averaging method;
the interpolation and smoothing methods respectively comprise:
wherein, 0<i<j，xa+iData missing at time a + i, xaAnd xa+jRaw data for time a and a + j, xkAs exception data, xk-1、xk+1Is xkAdjacent valid data.
5. The method for predicting the spatial distribution of the greenhouse environment according to claim 1, wherein the step of inputting the feature vector matrix into a convolutional network of a trained prediction model for feature extraction comprises the following steps:
inputting the feature vector matrix into the convolution network, carrying out four times of convolution through the convolution network, and outputting a global feature vector subjected to dimension reduction according to leveling operation and an activation function;
wherein, after every two convolutions, one maximum pooling is performed.
6. The greenhouse environment spatial distribution prediction method of claim 5, wherein the step of inputting the global feature vector into the time series network of the prediction model and outputting prediction results of temperature, humidity, illumination intensity and carbon dioxide content of predicted points in the greenhouse after a preset time period comprises:
inputting the global feature vector into a time series network, and sequentially performing feature processing through two gate control circulation unit network layers of the time series network;
according to the full-connection layer based on the linear rectification activation function, after reverse normalization, prediction results of the temperature, humidity, illumination intensity and carbon dioxide content in the greenhouse after preset time are output.
7. The method of predicting spatial distribution of greenhouse environment according to claim 1, wherein before determining the eigenvector matrix including the plurality of sets of time-series characteristics from the plurality of sets of environmental characteristics at the current and previous times of the predicted point, the method further comprises:
acquiring sample data of temperature and humidity, illumination intensity and carbon dioxide content in the greenhouse after a preset time is known, respectively taking the temperature and humidity, the illumination intensity and the carbon dioxide content as labels of the samples, and determining a characteristic vector matrix of the samples;
and adjusting model parameters based on an adaptive moment estimation optimization algorithm by using the sample data, and training to obtain the prediction model.
8. A greenhouse environment spatial distribution prediction apparatus, comprising:
the data acquisition module is used for determining a characteristic vector matrix comprising a plurality of groups of time series characteristics according to a plurality of groups of environmental characteristics of the current and previous moments of the predicted point, wherein the environmental characteristics comprise greenhouse internal environmental characteristics and greenhouse external environmental characteristics;
the first processing module is used for inputting the characteristic vector matrix into a convolutional network of a trained prediction model for characteristic extraction to obtain an extracted global characteristic vector;
the second processing module is used for inputting the global feature vector into the time series network of the prediction model and outputting prediction results of the temperature, humidity, illumination intensity and carbon dioxide content of prediction points in the greenhouse after preset time;
and the prediction model is obtained by determining samples of the characteristic vector matrix according to the known humidity, the illumination intensity and the carbon dioxide content in the greenhouse after the preset time, and training the samples, and comprises a convolution network and a time sequence network.
9. An electronic device comprising a memory, a processor and a computer program stored on said memory and executable on said processor, characterized in that said processor, when executing said program, carries out the steps of the greenhouse environment spatial distribution prediction method according to any one of claims 1 to 7.
10. A non-transitory computer readable storage medium, having a computer program stored thereon, wherein the computer program, when being executed by a processor, implements the steps of the greenhouse environment spatial distribution prediction method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111361824.8A CN114297907A (en) | 2021-11-17 | 2021-11-17 | Greenhouse environment spatial distribution prediction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111361824.8A CN114297907A (en) | 2021-11-17 | 2021-11-17 | Greenhouse environment spatial distribution prediction method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114297907A true CN114297907A (en) | 2022-04-08 |
Family
ID=80965417
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111361824.8A Pending CN114297907A (en) | 2021-11-17 | 2021-11-17 | Greenhouse environment spatial distribution prediction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114297907A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113467529A (en) * | 2021-05-25 | 2021-10-01 | 北京农业信息技术研究中心 | Greenhouse ozone accurate control method and device based on multi-model fusion |
| CN114778774A (en) * | 2022-04-21 | 2022-07-22 | 平安国际智慧城市科技股份有限公司 | Greenhouse gas monitoring method based on artificial intelligence and related equipment |
| CN114879786A (en) * | 2022-05-23 | 2022-08-09 | 连云港银丰食用菌科技有限公司 | Method, system, device and medium for acquiring edible fungus decision scheme |
| CN116678086A (en) * | 2023-04-27 | 2023-09-01 | 深圳市众信海科技有限公司 | Indoor temperature control method and system based on convolutional neural network |
| CN117469774A (en) * | 2023-12-28 | 2024-01-30 | 北京市农林科学院智能装备技术研究中心 | Air conditioning system regulation and control method and device, electronic equipment and storage medium |
-
2021
- 2021-11-17 CN CN202111361824.8A patent/CN114297907A/en active Pending
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113467529A (en) * | 2021-05-25 | 2021-10-01 | 北京农业信息技术研究中心 | Greenhouse ozone accurate control method and device based on multi-model fusion |
| CN114778774A (en) * | 2022-04-21 | 2022-07-22 | 平安国际智慧城市科技股份有限公司 | Greenhouse gas monitoring method based on artificial intelligence and related equipment |
| CN114879786A (en) * | 2022-05-23 | 2022-08-09 | 连云港银丰食用菌科技有限公司 | Method, system, device and medium for acquiring edible fungus decision scheme |
| CN114879786B (en) * | 2022-05-23 | 2023-09-01 | 连云港银丰食用菌科技有限公司 | Method, system, device and medium for obtaining edible fungus decision scheme |
| CN116678086A (en) * | 2023-04-27 | 2023-09-01 | 深圳市众信海科技有限公司 | Indoor temperature control method and system based on convolutional neural network |
| CN116678086B (en) * | 2023-04-27 | 2024-01-30 | 深圳市众信海科技有限公司 | Indoor temperature control method and system based on convolutional neural network |
| CN117469774A (en) * | 2023-12-28 | 2024-01-30 | 北京市农林科学院智能装备技术研究中心 | Air conditioning system regulation and control method and device, electronic equipment and storage medium |
| CN117469774B (en) * | 2023-12-28 | 2024-04-02 | 北京市农林科学院智能装备技术研究中心 | Air conditioning system regulation and control method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114297907A (en) | Greenhouse environment spatial distribution prediction method and device | |
| CN110705684A (en) | Environment self-adaptive learning method and system based on end cloud cooperation | |
| CN110685868A (en) | Wind turbine generator fault detection method and device based on improved gradient elevator | |
| CN115018021B (en) | Machine room abnormity detection method and device based on graph structure and abnormity attention mechanism | |
| CN110594107B (en) | Wind turbine generator fault detection method and device based on rapid gradient elevator | |
| CN107622236B (en) | Crop disease diagnosis and early warning method based on swarm and gradient lifting decision tree algorithm | |
| CN113391607A (en) | Hydropower station gate control method and system based on deep learning | |
| CN114240000A (en) | Air quality prediction method based on space-time graph convolution network | |
| CN111553806B (en) | Self-adaptive crop management system and method based on low-power-consumption sensor and Boost model | |
| CN115880433A (en) | Crop cultivation optimization method based on digital twinning | |
| CN115115830A (en) | Improved Transformer-based livestock image instance segmentation method | |
| CN112784920A (en) | Cloud-side-end-coordinated dual-anti-domain self-adaptive fault diagnosis method for rotating part | |
| CN114444561A (en) | PM2.5 prediction method based on CNNs-GRU fusion deep learning model | |
| CN117114913A (en) | Intelligent agricultural data acquisition system based on big data | |
| CN115128978A (en) | Internet of things environment big data detection and intelligent monitoring system | |
| CN117290800B (en) | Timing sequence anomaly detection method and system based on hypergraph attention network | |
| CN215813842U (en) | Hydropower station gate control system based on deep learning | |
| CN111027436A (en) | Northeast black fungus disease and pest image recognition system based on deep learning | |
| 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 | |
| CN115713044B (en) | Method and device for analyzing residual life of electromechanical equipment under multi-condition switching | |
| CN116706992A (en) | Self-adaptive power prediction method, device and equipment for distributed photovoltaic cluster | |
| CN117033923A (en) | Method and system for predicting crime quantity based on interpretable machine learning | |
| CN114489200A (en) | Warmhouse booth environmental control system | |
| CN113469013A (en) | Motor fault prediction method and system based on transfer learning and time sequence | |
| CN112906856A (en) | Livestock and poultry house ambient temperature detection system based on cloud platform |
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 |