Intelligent driving device for sunshade net of greenhouse

Abstract

The invention provides an intelligent driving device for a sunshade net of a greenhouse, and relates to the technical field of intelligent driving, and the device comprises a sensor module which is used for obtaining historical data, plant data and external data through a sensor, and transmitting the data to an instruction generation module. And the instruction generation module is used for inputting associated data to the accurate prediction model established by the deep learning model optimized based on the love optimization algorithm, and generating a control signal. And the wireless communication module is used for receiving the control signal and transmitting the control signal to the control driving module. And the control driving module is used for receiving the control signal and controlling the opening and closing of the sunshade net. According to the method, the associated data is input into the accurate prediction model established by the deep learning model optimized based on the love optimization algorithm, the control signal is generated, the state of the sunshade net can be automatically adjusted according to different environmental changes and plant requirements, the opening and closing state of the sunshade net is optimized, unnecessary energy consumption and moisture loss are reduced, and the quality of the sunshade net is improved. Therefore, the resource utilization efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent drive technology, and in particular to an intelligent drive device for a greenhouse sunshade net. Background Art

[0002] Early greenhouses had simple structures and single functions, primarily using plastic film as a covering material. Their primary purpose was to extend the growing season and increase the yield of specific crops. With the increasing popularity of greenhouses, especially in developed countries, their technology and functionality have rapidly improved, with the emergence of new materials such as high-strength glass and polycarbonate panels, as well as more advanced structural designs such as multi-span greenhouses and solar greenhouses. With the application of the Internet of Things in greenhouses, new technologies such as automated control systems and climate control equipment have emerged, improving the efficiency and convenience of greenhouse use. In modern times, intelligent control technologies are widely used in greenhouses, with environmental monitoring and automatic control being key components and crucial for optimizing the crop growth environment.

[0003] like Figure 1 As shown in the figure, greenhouses are an important part of modern agricultural facilities. The regulation of internal environmental factors such as light, temperature, and humidity is crucial for the growth of crops. Especially in the summer when the sun is strong, greenhouses usually need to install sunshade nets to prevent excessive temperatures and strong light from damaging crops. However, traditional sunshade net drive devices mostly use manual or simple mechanical drive methods, which have problems such as inconvenient operation, slow response speed, and difficulty in accurately controlling the sunshade effect. With the continuous advancement of modern agricultural technology and the development trend of intelligentization, the drive method of greenhouse sunshade nets also needs to develop in the direction of intelligence and automation. Intelligent drive devices can not only achieve precise control of the sunshade net, but also automatically adjust the opening and closing degree of the sunshade net according to environmental factors such as light and temperature inside the greenhouse, thereby improving the greenhouse's environmental control capabilities and providing more suitable environmental conditions for crop growth.

[0004] Therefore, developing an intelligent driving device for greenhouse sunshade nets to realize automated and intelligent control of sunshade nets has important practical significance and application value, solving the problems of resource waste and untimely environmental adjustment caused by relying on manual management, and the traditional manual adjustment method is limited by subjective judgment and is difficult to reflect the complex relationship between plant growth needs and environmental changes in real time. Summary of the Invention

[0005] The present invention provides an intelligent driving device for a greenhouse sunshade net, which is used to solve the defects of the prior art such as waste of resources and untimely environmental adjustment caused by reliance on manual management.

[0006] The present invention provides an intelligent driving device for a greenhouse sunshade net, comprising: The sensor module is used to obtain historical data, plant data and external data through sensors, and send the plant data and external data to the instruction generation module.

[0007] The instruction generation module is used to establish a deep learning model based on the love optimization algorithm, input historical data for training, obtain an accurate prediction model, use the association analysis algorithm to obtain the correlation data between plant data and external data, input the correlation data into the accurate prediction model, and generate a control signal.

[0008] The wireless communication module is used to receive control signals and transmit them to the control drive module and send device status information to the remote terminal.

[0009] The control and drive module is used to receive control signals and control the opening and closing of the sunshade net. The control and drive module includes a microprocessor, a motor, a reducer, a drive shaft, and a drive circuit. The motor is used to drive the reducer, which controls the rotation speed of the drive shaft. The drive shaft is connected to the sunshade net, and the rotation of the drive shaft drives the opening and closing of the sunshade net. The microprocessor is used to receive detection signals from the sensor module and control signals from the wireless communication module. The drive circuit is used to drive the operation of the control and drive module.

[0010] According to the greenhouse sunshade net intelligent driving device provided by the present invention, the steps of optimizing the deep learning model include: Multiple candidate individuals are randomly generated to form a love group, the candidate individuals are randomly initialized, and the loss function is used to define the fitness function. The candidate individuals are the parameters of the deep learning model.

[0011] The fitness between each team of candidate individuals is calculated according to the fitness function, and the sentiment weight is constructed based on the fitness.

[0012] The emotional weights are used for mating to generate new individuals, and Gaussian noise is used to mutate the new individuals to obtain unknown individuals. The positions of the unknown individuals are updated in each generation to obtain position-updated individuals.

[0013] After reaching the preset number of iterations, the position update individual with the best fitness and position is selected as the neural network parameters, and the deep learning model is constructed using the neural network parameters.

[0014] According to the greenhouse sunshade net intelligent driving device provided by the present invention, the formula for updating the position of the individual is expressed as: Where, is the learning rate, is a random perturbation, An unknown individual, is a normally distributed random variable, is the current location of the unknown individual, is the updated position of the unknown individual.

[0015] According to the greenhouse sunshade net intelligent driving device provided by the present invention, the steps of obtaining an accurate prediction model include: Remove missing values and outliers from historical data to obtain optimized data, and divide the optimized data into training set and validation set.

[0016] Select the hyperparameters of the deep learning model, input the training set into the deep learning model, perform forward propagation calculations according to the hyperparameters, and obtain the prediction results.

[0017] The cross entropy loss function is used to calculate the loss value of the deep learning model based on the prediction results. The formula is expressed as: Where, is the sample size, is the number of categories, is the true label of the category to which the sample belongs, Represents a sample Belong to category The predicted probability of is the loss value.

[0018] Calculate the gradient of the cross entropy loss function for the deep learning model according to the backpropagation algorithm.

[0019] The hyperparameters are updated based on the loss value and gradient, and the validation set is used to check whether the hyperparameters have reached the preset threshold at the preset time interval. If so, the update is stopped to obtain an accurate prediction model, otherwise the hyperparameters continue to be updated.

[0020] According to the intelligent driving device for a greenhouse sunshade net provided by the present invention, the gradient formula is expressed as: Where, It is a layer in the deep learning network. The error of the layer is The layer weight matrix is Layer weight matrix The transposed matrix of is the derivative of the activation function, It is The weighted input of the layer, It means element-by-element multiplication. is the gradient.

[0021] According to the greenhouse sunshade net intelligent driving device provided by the present invention, the step of obtaining associated data includes: The plant data and external data are organized into transaction datasets, and the support threshold and confidence threshold of the transaction datasets are set.

[0022] The support of each transaction item in the transaction data set is counted to see if it is less than the support threshold. If so, the current transaction item is deleted; otherwise, a frequent transaction set is formed.

[0023] For each frequent transaction in the frequent transaction set, a non-empty true subset is generated as the antecedent, and the frequent transaction set is subtracted from the antecedent to obtain the consequent. Multiple association rules are formed based on the antecedent and the consequent.

[0024] The confidence of each association rule is calculated according to the confidence formula, and the confidences greater than the confidence threshold are selected as association data.

[0025] According to the greenhouse sunshade net intelligent driving device provided by the present invention, the formula for calculating the confidence of each association rule is expressed as follows: Where, It is the previous item. It is the latter term, It also includes and The proportion of the number of transactions to the total number of transactions, is the ratio of the number of included transactions to the total number of transactions, In the association rule roll out confidence level.

[0026] According to the greenhouse sunshade net intelligent driving device provided by the present invention, the step of generating a control signal includes: The associated data is converted into a feature matrix and normalized to obtain input data, which is then input into the precise prediction model to obtain future growth data through forward propagation calculation.

[0027] A preset growth range is set based on the associated data combined with historical data, and future growth data is compared with the preset growth range to generate a control signal.

[0028] According to the greenhouse sunshade net intelligent driving device provided by the present invention, the formula for obtaining future growth data is expressed as: Where, is the neuron of the deep learning model, It's a neuron Input, is the weight vector, is the transpose of the weight vector, is the feature matrix, is the bias, is the activation derivative.

[0029] According to the present invention, a greenhouse sunshade net intelligent driving device is provided, wherein the sensor module includes a light sensor and a temperature sensor. The light sensor is used to detect the light intensity inside and outside the greenhouse. The temperature sensor is used to detect the temperature inside and outside the greenhouse.

[0030] The present invention provides an intelligent driving device for a greenhouse sunshade net. By training a deep learning model optimized by a love optimization algorithm with historical data as input, an accurate prediction model is obtained. This solves the problems that traditional prediction models are prone to fall into local optimal solutions, resulting in inaccurate predictions, and traditional manual adjustment methods are limited by subjective judgment and are difficult to reflect the complex relationship between plant growth needs and environmental changes in real time. It can effectively capture and learn the complex relationship between plant growth and the external environment, predict the growth status of plants in real time, and adjust different environmental variables such as light, temperature and humidity according to data and models, thereby providing optimal growth conditions for plants and improving the efficiency and quality of plant growth.

[0031] The present invention provides an intelligent driving device for a greenhouse sunshade net. It obtains correlated data through correlation analysis based on plant data and external data, and inputs the correlated data into a precise prediction model to generate a control signal, thereby solving the problem of resource waste and untimely environmental adjustment caused by reliance on manual management. It can automatically adjust the state of the sunshade net according to different environmental changes and plant needs, and optimize the opening and closing state of the sunshade net, thereby reducing unnecessary energy consumption and water loss, and thus improving resource utilization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0033] Figure 1 This is one of the flow diagrams of an intelligent driving device for a greenhouse sunshade net provided by an embodiment of the present invention.

[0034] Figure 2 This is the second flow chart of an intelligent driving device for a greenhouse sunshade net provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0035] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0036] The following combination Figure 1 and Figure 2 The present invention describes an intelligent driving device for a greenhouse sunshade net.

[0037] Figure 1 This is one of the flow diagrams of an intelligent driving device for a greenhouse sunshade net provided by an embodiment of the present invention.

[0038] like Figure 1 As shown, an embodiment of the present invention provides an intelligent driving device for a greenhouse sunshade net, and the method mainly includes the following steps: The sensor module is used to acquire historical data, plant data, and external data through sensors and send the plant data and external data to the instruction generation module. The sensor module includes a light sensor and a temperature sensor. The light sensor is used to detect the light intensity inside and outside the greenhouse. The temperature sensor is used to detect the temperature inside and outside the greenhouse.

[0039] The instruction generation module is used to establish a deep learning model based on the love optimization algorithm, input historical data for training, obtain an accurate prediction model, use the association analysis algorithm to obtain the correlation data between plant data and external data, input the correlation data into the accurate prediction model, and generate a control signal.

[0040] Figure 2 This is the second flow chart of an intelligent driving device for a greenhouse sunshade net provided by an embodiment of the present invention.

[0041] like Figure 2 As shown, the steps to optimize a deep learning model include: Multiple candidate individuals are randomly generated to form a love group, the candidate individuals are randomly initialized, and the loss function is used to define the fitness function. The candidate individuals are the parameters of the deep learning model.

[0042] The fitness between each team of candidate individuals is calculated according to the fitness function, and the sentiment weight is constructed based on the fitness.

[0043] Use sentiment weights to mate and generate new individuals, and use Gaussian noise to mutate the new individuals to obtain unknown individuals. In each generation, the positions of the unknown individuals are updated to obtain position-updated individuals. The formula is expressed as: Where, is the learning rate, is a random perturbation, An unknown individual, is a normally distributed random variable, is the current location of the unknown individual, is the updated position of the unknown individual.

[0044] After reaching the preset number of iterations, the position update individual with the best fitness and position is selected as the neural network parameters, and the deep learning model is constructed using the neural network parameters.

[0045] The steps to obtain an accurate prediction model include: Optimized data is obtained by removing missing values and outliers from historical data. This optimized data is then divided into a training set and a validation set. For missing values, data characteristics and distribution patterns are carefully considered, and adaptable methods such as mean filling, median filling, regression-based filling, or multiple imputation are used to accurately fill missing values and ensure data integrity. For outliers, statistical analysis methods such as boxplots and the 3σ principle are used to clearly define the range of abnormal data. Outliers are then properly handled, either by correcting them to return them to a reasonable data range or removing them if they meet the criteria. This series of operations results in optimized data of higher quality and greater reliability. Next, to efficiently train the model and accurately evaluate its performance, the optimized data is further subdivided into a training set and a validation set.

[0046] Hyperparameters for the deep learning model are selected, the training set is fed into the model, and forward propagation calculations are performed according to the hyperparameters to obtain predictions. Hyperparameters are selected based on the data size, the richness of the feature dimensions, the complexity of the data distribution, and the expected performance metrics of the model. For example, for historical data with large-scale samples and high-dimensional features, a larger number of neural network layers and an appropriate number of neurons per layer are preferred to fully capture the complex patterns and underlying regularities in the data. Furthermore, when selecting an optimizer, a trade-off is made between common optimization algorithms such as stochastic gradient descent (SGD) and adaptive moment estimation (Adam). Based on the characteristics of the data and the convergence of the model, relevant hyperparameters such as the learning rate and momentum are fine-tuned to avoid getting stuck in local optimal solutions.

[0047] After selecting the hyperparameters for the deep learning model, we input the training set data into the constructed deep learning model, officially starting the forward propagation computational process. During this process, data begins at the model's input layer and, according to the pre-defined hyperparameter settings, passes through each hidden layer, where information is transferred and features are transformed. Within each hidden layer, neurons perform a weighted summation operation on the input data based on preset weight parameters and bias values. This is then followed by a nonlinear transformation using a specific activation function (such as the commonly used ReLU function, Sigmoid function, and Tanh function), effectively extracting high-order features and complex relationships within the data. The data is continuously refined and abstracted, gradually forming a feature representation that reflects the data's inherent patterns. Ultimately, the model's output layer outputs predictions. These predictions encompass the model's estimated information on various aspects of plant growth, such as future growth trends, potential pest and disease risks, and estimated yield ranges.

[0048] The cross entropy loss function is used to calculate the loss value of the deep learning model based on the prediction results. The formula is expressed as: Where, is the sample size, is the number of categories, is the true label of the category to which the sample belongs, Represents a sample Belong to category The predicted probability of is the loss value.

[0049] The gradient of the cross entropy loss function to the deep learning model is calculated based on the back propagation algorithm. The gradient formula is expressed as: Where, It is a layer in the deep learning network. The error of the layer is The layer weight matrix is Layer weight matrix The transposed matrix of is the derivative of the activation function, It is The weighted input of the layer, It means element-by-element multiplication. is the gradient.

[0050] The hyperparameters are updated based on the loss value and gradient, and the validation set is used to check whether the hyperparameters have reached the preset threshold at the preset time interval. If so, the update is stopped to obtain an accurate prediction model, otherwise the hyperparameters continue to be updated.

[0051] The steps to obtain linked data include: Organize plant data and external data into transactional datasets, and set support and confidence thresholds for the transactional datasets. Plant data includes variety, growth stage, and past pest and disease occurrences. External data includes meteorological data for the planting area, such as temperature, humidity, and daylight hours, soil fertility, pH, and the surrounding geographic characteristics.

[0052] We count the support of each transaction item in the transaction dataset to see if it is less than a support threshold. If so, we delete the current transaction item; otherwise, we form a frequent transaction set. For each unique transaction item, we meticulously traverse the entire transaction dataset, accurately counting its occurrence across all transactions. We then calculate the ratio of this number to the total number of transactions in the dataset to determine its support value. We then rigorously compare the support value of each transaction item against a pre-set support threshold. If the support value of a transaction item is less than this threshold, it indicates that it occurs relatively infrequently in the dataset and may contribute little to the association rules we are looking to mine. In this case, we decisively delete the current transaction item to simplify the dataset and focus on more representative data. Conversely, for transaction items whose support values are at least the support threshold, we demonstrate a certain degree of prevalence and representativeness in the dataset. These items are retained and further combined and integrated to form a frequent transaction set.

[0053] For each frequent transaction in the frequent transaction set, a non-empty true subset is generated as the antecedent. The frequent transaction set is subtracted from the antecedent to obtain the consequent. Multiple association rules are then formed based on the antecedents and consequents. For each such frequent transaction, a subset generation algorithm is applied to generate all possible non-empty true subsets of the frequent transaction, which are then used as the antecedents in the association rule. This process requires a deep understanding and precise application of the combinatorial principles of sets to ensure that no possible antecedent combinations are missed. Subsequently, a set operation is performed to subtract the determined antecedent from the current frequent transaction to obtain the corresponding consequent. This step requires high accuracy in the set difference operation to ensure that the correspondence between the antecedents and consequents truly reflects the inherent logical connections within the transaction set. Based on these precisely generated antecedents and consequents, multiple association rules are constructed according to the definition and form of association rules. Each association rule represents a potential association pattern between the plant data and external data.

[0054] The confidence of each association rule is calculated according to the confidence formula, which is expressed as: Where, It is the previous item. It is the latter term, It also includes and The proportion of the number of transactions to the total number of transactions, is the ratio of the number of included transactions to the total number of transactions, In the association rule roll out confidence level.

[0055] Filter the data with a confidence greater than the confidence threshold as related data.

[0056] The steps for generating the control signal include: The associated data is converted into a feature matrix and normalized to obtain input data. The input data is input into the precise prediction model and the future growth data is obtained through forward propagation calculation. The formula is expressed as: Where, is the neuron of the deep learning model, It's a neuron Input, is the weight vector, is the transpose of the weight vector, is the feature matrix, is the bias, is the activation derivative.

[0057] A preset growth range is set based on the associated data combined with historical data, and future growth data is compared with the preset growth range to generate a control signal. For the suitable temperature range for plant growth, an upper and lower temperature threshold is set; for soil moisture, a corresponding humidity range threshold is also set. These thresholds and ranges can be determined based on long-term accumulated planting experimental data, agricultural expert experience, etc., and may need to be continuously adjusted and optimized according to actual application conditions. If the predicted soil moisture value is lower than the set lower threshold, it means that the soil may be short of water, and a control signal needs to be generated to start the irrigation system; if the predicted temperature is higher than the upper threshold, control signals corresponding to related operations such as shading and ventilation may need to be generated.

[0058] The wireless communication module is used to receive control signals and transmit them to the control drive module and send device status information to the remote terminal.

[0059] The control and drive module is used to receive control signals and control the opening and closing of the sunshade net. The control and drive module includes a microprocessor, a motor, a reducer, a drive shaft, and a drive circuit. The motor is used to drive the reducer, which controls the rotation speed of the drive shaft. The drive shaft is connected to the sunshade net, and the rotation of the drive shaft drives the opening and closing of the sunshade net. The microprocessor is used to receive detection signals from the sensor module and control signals from the wireless communication module. The drive circuit is used to drive the operation of the control and drive module.

[0060] The power module provides power to the sensor module, command generation module, control drive module and wireless communication module.

[0061] The present invention obtains an accurate prediction model by inputting historical data into the deep learning model optimized by the love optimization algorithm for training, which solves the defects that traditional prediction models are prone to fall into local optimal solutions, resulting in inaccurate predictions, and traditional manual adjustment methods are limited by subjective judgment and are difficult to reflect the complex relationship between plant growth needs and environmental changes in real time. It can effectively capture and learn the complex relationship between plant growth and the external environment, predict the growth status of plants in real time, and adjust different environmental variables such as light, temperature and humidity according to data and models, so as to provide optimal growth conditions for plants and improve the efficiency and quality of plant growth.

[0062] The present invention obtains correlation data based on the correlation analysis of plant data and external data, and inputs the correlation data into a precise prediction model to generate a control signal, thereby solving the problem of resource waste and untimely environmental adjustment caused by reliance on manual management. It can automatically adjust the state of the sunshade net according to different environmental changes and plant needs, and optimize the opening and closing state of the sunshade net, reducing unnecessary energy consumption and water loss, thereby improving resource utilization efficiency.

[0063] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0064] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. An intelligent driving device for a greenhouse sunshade net, characterized in that: include: a sensor module, configured to acquire historical data, plant data, and external data through sensors, and send the plant data and external data to the instruction generation module; An instruction generation module is used to establish a deep learning model optimized based on the love optimization algorithm, input the historical data for training, obtain an accurate prediction model, use the association analysis algorithm to obtain the correlation data between the plant data and the external data, input the correlation data into the accurate prediction model, and generate a control signal; A wireless communication module is used to receive control signals and transmit them to the control drive module and send device status information to a remote terminal; A control drive module is used to receive the control signal and control the opening and closing of the sunshade net; the control drive module includes a microprocessor, a motor, a reducer, a transmission shaft and a drive circuit, the motor is used to drive the reducer to operate, the reducer is used to control the rotation speed of the transmission shaft, the transmission shaft is connected to the sunshade net, and the rotation of the transmission shaft drives the opening and closing of the sunshade net, the microprocessor is used to receive the detection signal of the sensor module and the control signal of the wireless communication module, and the drive circuit is used to control the operation of the control drive module.

2. The intelligent driving device for greenhouse sunshade net according to claim 1, characterized in that: The steps of optimizing the deep learning model include: Randomly generate multiple candidate individuals to form a love group, randomly initialize the candidate individuals, and use a loss function to define a fitness function, wherein the candidate individuals are parameters of the deep learning model; Calculating the fitness between the candidate individuals in each team according to the fitness function, and constructing the sentiment weight according to the fitness; Using the sentiment weights to perform mating to generate new individuals, using Gaussian noise to mutate the new individuals to obtain unknown individuals, and updating the positions of the unknown individuals in each generation to obtain position-updated individuals; After reaching the preset number of iterations, the position update individual with the best fitness and position is selected as the neural network parameters, and the deep learning model is constructed using the neural network parameters.

3. The intelligent driving device for greenhouse sunshade net according to claim 2, characterized in that: The formula for updating the location of an individual is expressed as: Where, is the learning rate, is a random perturbation, An unknown individual, is a normally distributed random variable, is the current location of the unknown individual, is the updated position of the unknown individual.

4. The intelligent driving device for greenhouse sunshade net according to claim 1, characterized in that: The steps of obtaining the accurate prediction model include: Removing missing values and outliers from the historical data to obtain optimized data, and dividing the optimized data into a training set and a validation set; Selecting hyperparameters of the deep learning model, inputting the training set into the deep learning model, performing forward propagation calculation according to the hyperparameters, and obtaining a prediction result; The cross entropy loss function is used to calculate the loss value of the deep learning model according to the prediction results. The formula is expressed as: Where, is the sample size, is the number of categories, is the true label of the category to which the sample belongs, Represents a sample Belong to category The predicted probability of is the loss value; Calculate the gradient of the cross entropy loss function with respect to the deep learning model according to the back propagation algorithm; The hyperparameters are updated according to the loss value and the gradient, and the validation set is used to determine whether the hyperparameters reach a preset threshold at a preset time interval. If so, the update is stopped to obtain an accurate prediction model; otherwise, the hyperparameters continue to be updated.

5. The intelligent driving device for greenhouse sunshade net according to claim 4, characterized in that: The gradient is expressed as: Where, It is a layer in the deep learning network. The error of the layer is The layer weight matrix is Layer weight matrix The transposed matrix of is the derivative of the activation function, It is The weighted input of the layer, It means element-by-element multiplication. is the gradient.

6. The intelligent driving device for greenhouse sunshade net according to claim 1, characterized in that: The steps of obtaining the associated data include: Arrange the plant data and external data into a transaction data set, and set a support threshold and a confidence threshold for the transaction data set; Counting whether the support of each transaction item in the transaction data set is less than the support threshold, if so, deleting the current transaction item, otherwise forming a frequent transaction set; For each frequent transaction in the frequent transaction set, a non-empty proper subset is generated as a preceding term, the preceding term is subtracted from the frequent transaction set to obtain a subsequent term, and a plurality of association rules are formed based on the preceding term and the subsequent term; The confidence of each association rule is calculated according to a confidence formula, and the confidences greater than the confidence threshold are selected as association data.

7. The intelligent driving device for greenhouse sunshade net according to claim 6, characterized in that: The formula for calculating the confidence of each association rule is expressed as: Where, It is the previous item. It is the latter term, It also includes and The proportion of the number of transactions to the total number of transactions, is the ratio of the number of included transactions to the total number of transactions, In the association rule roll out confidence level.

8. The intelligent driving device for greenhouse sunshade net according to claim 1, characterized in that: The step of generating the control signal comprises: Converting the associated data into a feature matrix and performing normalization processing to obtain input data, inputting the input data into the precise prediction model to obtain future growth data through forward propagation calculation; A preset growth range is set according to the associated data in combination with the historical data, and the future growth data is compared with the preset growth range to generate a control signal.

9. The intelligent driving device for greenhouse sunshade net according to claim 8, characterized in that: The formula for obtaining the future growth data is expressed as: Where, is the neuron of the deep learning model, It's a neuron Input, is the weight vector, is the transpose of the weight vector, is the feature matrix, is the bias, is the activation derivative.

10. The intelligent driving device for greenhouse sunshade net according to claim 1, characterized in that: The sensor module includes a light sensor and a temperature sensor. The light sensor is used to detect the light intensity inside and outside the greenhouse; the temperature sensor is used to detect the temperature inside and outside the greenhouse.

Images