Intelligent drip irrigation control method and system based on microclimate perception and medium

Abstract

The invention discloses an intelligent drip irrigation control method based on microclimate perception, and the method comprises the following steps: calling historical data related to plants, including environmental parameter data related to microclimate and actual plant moisture demand data, and dividing the data into a training set and a verification set; establishing a regression model of intelligent drip irrigation by using a neural network structure, wherein the regression model is used for predicting a mapping relation of plant water demands based on environmental parameters; iterative optimization is carried out by utilizing multi-term feature extension to update the weight and bias of the model, and training of the regression model is realized; and the trained regression model is evaluated, and the performance of the regression model is evaluated and optimized. One of the main purposes of the invention is to provide an intelligent drip irrigation control method and system based on microclimate perception, and the method and system are used for solving the technical problem that in the background technology, existing plant irrigation cannot realize refined irrigation according to requirements and quantity.

Description

Technical Field

[0001] The present invention mainly relates to the technical field of drip irrigation intelligent control, and specifically to a drip irrigation intelligent control method, system and medium based on microclimate perception. Background Art

[0002] In my country, the gardening industry mostly uses traditional irrigation methods such as manual watering and timed sprinkler irrigation. However, traditional irrigation methods have the following disadvantages: 1. Resource waste: Manual irrigation usually cannot accurately control the amount of water and irrigation time, resulting in waste of water resources; 2. Labor intensive: Traditional manual irrigation requires a lot of manpower input, with high labor costs and low efficiency; 3. Instability: The stability of manual irrigation is affected by human factors, such as improper operation or personnel changes may lead to unstable irrigation effects; 4. Unintelligent: Traditional manual irrigation technology lacks intelligence and automation features, and cannot be adjusted and optimized according to real-time environmental conditions; 5. Soil compaction: Frequent manual irrigation may cause soil compaction, affecting soil aeration and water permeability, and thus affecting crop growth.

[0003] The disadvantages of smart sprinkler irrigation with fixed time and fixed quantity are: 1. Not adaptable to changing weather conditions: The fixed time and fixed quantity sprinkler system may not be able to flexibly adapt to changing weather conditions, such as sudden rainfall or drought. This may lead to over-irrigation or under-irrigation, affecting crop growth; 2. Waste of water resources: Although the fixed quantity sprinkler can control the amount of water for each sprinkler, if the irrigation frequency is unreasonable or there are problems such as leakage in the sprinkler system, it may still lead to waste of water resources; 3. Lack of personalized irrigation management: Different crops have different growth cycles and water requirements, and the fixed time and fixed quantity sprinkler system may not be able to perform personalized irrigation management according to the actual needs of the crops, resulting in over-irrigation or under-irrigation of some crops; 4. Increased costs: The installation and maintenance of the fixed time and fixed quantity sprinkler system requires a certain investment cost, including irrigation equipment, control system, maintenance personnel, etc., which increases the economic burden on farmers; 5. Cause soil salinization: Excessive sprinkler irrigation may cause excessive accumulation of salt in the soil, thereby affecting soil quality and even causing soil salinization problems.

[0004] In response to the above problems, technical personnel in this field are also actively seeking solutions. For example, the patent application number is CN2022112946377, and the patent name is a technical solution for a plant watering control system based on cloud-edge collaboration, including: watering unit, edge computing module, temperature and humidity sensor, cloud server, mobile terminal, etc.

[0005] The above solution has the following disadvantages:

[0006] 1) It is impossible to achieve fine-grained irrigation according to demand and quantity. Because the judgment of the amount of water to be watered each time is based on the "should be watered amount" and "actual water output" obtained by querying the database. But in fact, the evaporation rate of water in the soil and the absorption rate of water by plants are closely related to the season and climate. The actual humidity of the soil cannot be accurately judged by "querying the corresponding growth habits, historical watering records, temperature and humidity information of the plant in the database".

[0007] 2) It is impossible to continuously optimize the system. The effects of soil moisture on plant growth may be different in different growth cycles, seasons and climate conditions. A sustainable optimization and iteration model should be established for different varieties of plants to determine the effects of different soil moisture on plants.

[0008] After continuous exploration, the inventor of this application has realized the intelligent control of the drip irrigation system through algorithms, integration of multi-dimensional data of microclimate perception, and model processing and analysis. Microclimate perception refers to the monitoring and understanding of microclimate conditions in the crop growth environment, which refers to the local climate environment formed at a specific location due to factors such as terrain, vegetation cover, and building structure. It is opposite to macroclimate (such as regional climate, national climate), and refers to the climate characteristics in a specific area at a smaller scale, such as a farmland, a greenhouse, an urban block, etc.

[0009] Microclimate-related environmental parameter data, including but not limited to parameters such as temperature, humidity, light intensity, wind speed, and corresponding plant water demand data. Through microclimate perception, we can more accurately understand the environmental conditions for crop growth, thereby more effectively managing crops and improving crop yield and quality. The intelligent irrigation system in this application uses microclimate perception to monitor and predict the water demand of crops and achieve precise irrigation. Summary of the invention

[0010] One of the main purposes of the present invention is to provide a drip irrigation intelligent control method, system and medium based on microclimate perception, so as to solve the technical problems raised in the above background technology that the existing plant watering cannot achieve refined watering on demand and in quantity, and cannot continuously optimize the changes in water demand according to the different growth cycles of plants in different seasons and climatic conditions.

[0011] In order to solve the above technical problems, the present invention provides a drip irrigation intelligent control method based on microclimate perception, comprising the following steps:

[0012] S1: Data collection, calling plant-related historical data, the historical data including environmental parameter data related to microclimate and actual plant water demand data, and dividing the data into a training set and a validation set;

[0013] S2: Model establishment, inputting the environmental parameter data related to the microclimate in the training set into the neural network structure, and establishing a regression model of intelligent drip irrigation to predict the mapping relationship between plant water demand and environmental parameters;

[0014] S3: Model training, using multiple feature expansions for iterative optimization to update the model weights and biases to train the regression model;

[0015] S4: Model evaluation: Use the validation set to evaluate the trained regression model, evaluate the performance of the regression model and perform optimization.

[0016] In an achievable manner of the first aspect, the model evaluation comprises the following steps:

[0017] 101. Input the environmental parameter data related to the microclimate in the validation set into the trained regression model to calculate the predicted plant water demand data;

[0018] 102. Analyze and compare the calculated predicted plant water demand data y with the actual plant water demand data y0 in the plant historical data;

[0019] If the set threshold requirement is met within the specified data deviation range, it means that the model meets the actual needs and is used as the optimal model;

[0020] If the set threshold requirement is not met within the specified data deviation range, it means that the model needs further adjustment, and the model training steps are repeated, and the model weights and biases are updated again through iterative optimization through multiple feature expansions.

[0021] In a manner that can be implemented in the first aspect, the following steps are also included:

[0022] S5: Model application: Apply the model that has passed the evaluation to the intelligent prediction and control of the water demand of plants of the same variety. Collect the environmental parameter data related to the microclimate around the plant in real time over a period of time, and use the model to predict the water demand of the plant on the same day. Then, use the drip irrigation system to carry out quantitative drip irrigation on demand, and observe the growth of the plant at the same time.

[0023] S6: Model iteration: According to the growth of the plants, determine whether the threshold in the model evaluation needs to be adjusted. If the plants grow well, there is no need to adjust the threshold. If the plants grow poorly, the threshold needs to be increased and the model evaluation needs to be re-performed.

[0024] In the method that can be implemented in the first aspect, when the plant growth is poor, it is necessary to add the environmental parameter data related to the microclimate collected in real time as new data to the training set, and use multiple feature extensions to iteratively optimize the newly supplemented data and then update the model weights and biases.

[0025] In an achievable manner of the first aspect, the plant-related historical data or real-time collected data is preprocessed, and the data preprocessing includes data cleaning, denoising and data normalization:

[0026] The data cleaning is used to remove invalid, erroneous or incomplete data;

[0027] The denoising process is used to remove the noise in the cleaned data, and adopts the median filtering technology;

[0028] The normalization process is used to scale the denoised data to a uniform scale.

[0029] In an achievable manner of the first aspect, the regression model formula is:

[0030] y=b0+b1·d1+b2·d2+...+b n ·d n +e (1)

[0031] Where: y is the dependent variable, i.e. the predicted plant water requirement;

[0032] b0 is the bias, which represents the expected value of the dependent variable when all independent variables are 0;

[0033] b1, b2...b n is the regression coefficient of each independent variable, also called weight, which indicates the influence of each independent variable on the dependent variable;

[0034] x1, x2...x n are the independent variables, i.e., environmental parameters related to the microclimate;

[0035] e is the error term, which is the parameter that needs to be adjusted in the model and represents the difference between the model prediction value and the actual demand value.

[0036] In a manner that can be implemented in the first aspect, polynomial feature expansion is performed using PolynomialFeatures, and the specific steps are as follows:

[0037] 101. Define an original feature matrix X;

[0038] 102. Create a PolynomialFeatures object and specify the polynomial degree;

[0039] 103. Use the fit_transform method to fit and transform the original feature matrix to generate a polynomial feature expansion matrix containing the original features;

[0040] And / or, the polynomial feature expansion is to perform a polynomial transformation on the original feature to generate a new feature, which includes high-order terms of the original feature and cross terms between different original features.

[0041] The second aspect of the present invention provides a drip irrigation intelligent control system based on microclimate perception, which is used to implement the above-mentioned drip irrigation intelligent control method. The system includes:

[0042] A data acquisition module, which uses sensors to collect environmental parameter data related to the microclimate and transmits it to the data processing module;

[0043] A data processing module is used to preprocess the collected real-time data or the called historical data. The data preprocessing includes data cleaning, denoising and data normalization, and the preprocessed historical data is divided into a training set and a validation set;

[0044] The model building module inputs the environmental parameter data related to the microclimate in the training set into the neural network structure to establish a regression model for intelligent drip irrigation, which is used to predict the mapping relationship between plant water demand and environmental parameters;

[0045] The model training optimization module uses multiple feature extensions to perform iterative optimization to update the model's weights and biases to achieve training of the regression model.

[0046] In a manner that can be implemented in the second aspect, a model evaluation module is also included, which uses a validation set to evaluate the trained regression model, evaluate the performance of the regression model and perform optimization;

[0047] The model application module applies the approved model to the intelligent prediction and control of the water demand of plants of the same variety, collects environmental parameter data related to the microclimate in real time, and predicts the water demand of the plant on the day through the model, and then realizes quantitative drip irrigation on demand through the drip irrigation system;

[0048] The model iteration module iteratively optimizes the model by adjusting the threshold in the model evaluation.

[0049] A third aspect of the present invention provides a computer-readable storage medium storing a program, which implements the above-mentioned drip irrigation intelligent control method when executed by a processor.

[0050] A fourth aspect of the present invention provides a computing device, including a processor and a memory for storing a program executable by the processor, and when the processor executes the program stored in the memory, the above-mentioned drip irrigation intelligent control method is implemented.

[0051] Beneficial effects of the invention:

[0052] (1) Significant water-saving effect: Microclimate sensing technology can adjust the operation of the drip irrigation system based on real-time meteorological data, soil moisture and other factors, so that it can more accurately meet the water needs of plants, avoiding the water waste in traditional sprinkler irrigation systems, thereby achieving water-saving effects.

[0053] (2) Precision water application: By sensing microclimate changes, the drip irrigation system can accurately apply water based on factors such as plant needs, soil moisture and meteorological conditions, ensuring that plants receive the right amount of water, thereby improving the efficiency and quality of plant growth.

[0054] (3) Reduce soil erosion and water pollution: The drip irrigation system can deliver water directly to the roots of plants, reducing the loss of water on the soil surface, thereby reducing the risk of soil erosion. At the same time, it reduces the loss of nutrients and chemicals in the soil into surface water, which is beneficial to protecting water quality.

[0055] The present invention will be explained in detail below with reference to the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 The present invention is a flowchart of a drip irrigation intelligent control method based on microclimate perception according to an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] Example

[0059] In order to solve the problems that the existing drip irrigation cannot achieve refined irrigation on demand and in quantity, and cannot continuously optimize the changes in water demand according to the different growth cycles of plants in different seasons and climatic conditions, the present invention combines the multi-dimensional data of microclimate perception and uses artificial neural networks to establish a mapping relationship between environmental parameters and plant water requirements, and realizes intelligent control of the drip irrigation system through learning and optimization of the neural network.

[0060] Specifically, a first aspect of an embodiment of the present invention provides a drip irrigation intelligent control method based on microclimate perception, comprising the following steps:

[0061] S1: Data collection, calling plant-related historical data, the historical data including environmental parameter data related to microclimate and actual plant water demand data, and dividing the data into a training set and a validation set;

[0062] S2: Model establishment: input the environmental parameter data related to the microclimate in the training set into the neural network structure to establish a regression model for intelligent drip irrigation, which is used to predict the mapping relationship between plant water demand and environmental parameters. The regression model formula is:

[0063] y=b0+b1·d1+b2·d2+...+b n ·d n +e(1)

[0064] Where: y is the dependent variable, i.e. the predicted plant water requirement;

[0065] b0 is the bias, which represents the expected value of the dependent variable when all independent variables are 0;

[0066] b1, b2...b n is the regression coefficient of each independent variable, also called weight, which indicates the influence of each independent variable on the dependent variable;

[0067] x1, x2...x n are the independent variables, i.e., environmental parameters related to the microclimate;

[0068] e is the error term, which is the parameter that needs to be adjusted in the model and represents the difference between the model prediction value and the actual demand value;

[0069] S3: Model training: Use the training set to train the regression model, and iteratively optimize through multiple feature expansions to update the model's weights and biases, so that the model can accurately learn the mapping relationship between environmental parameters and plant water requirements;

[0070] S4: Model evaluation: Use the validation set to evaluate the trained regression model, evaluate the performance of the regression model and perform optimization.

[0071] In this embodiment, the model evaluation includes the following steps:

[0072] 101. Input the environmental parameter data related to the microclimate in the validation set into the trained regression model to calculate the predicted plant water demand data;

[0073] 102. Analyze and compare the calculated predicted plant water demand data y with the actual plant water demand data y0 in the plant historical data;

[0074] If the set threshold requirement is met within the specified data deviation range, it means that the model meets the actual needs and is used as the optimal model;

[0075] If the set threshold requirement is not met within the specified data deviation range, it means that the model needs further adjustment, and the model training steps are repeated, and the model weights and biases are updated again through iterative optimization through multiple feature expansions.

[0076] For example, among all the predicted plant water demand data, if the number of data whose deviation value a (a=y / y0) meets the requirement of ±5% reaches 80% (that is, the threshold requirement) of the total number, it means that the model meets the actual needs and is used as the optimal model.

[0077] If the requirement of 80% of the total number cannot be met, it means that the model needs further adjustment, and the model training steps must be repeated and the model must be retrained until the threshold requirement is met during model evaluation. It can then be used as the optimal model and deployed in the actual microclimate perception system to achieve intelligent prediction and control of the water demand of the same variety of plants.

[0078] S5: Model application: Apply the model that has passed the evaluation to the intelligent prediction and control of the water demand of plants of the same variety. Collect the environmental parameter data related to the microclimate around the plant in real time over a period of time, and use the model to predict the water demand of the plant on the same day. Then, use the drip irrigation system to carry out quantitative drip irrigation on demand, and observe the growth of the plant at the same time.

[0079] S6: Model iteration: According to the growth of the plants, determine whether the threshold in the model evaluation needs to be adjusted. If the plants grow well, there is no need to adjust the threshold. If the plants grow poorly, the threshold needs to be increased and the model evaluation needs to be re-performed.

[0080] When the plants are growing well, the collected plant-related data (including environmental parameter data related to the microclimate and the actual amount of plant watering) only needs to be preprocessed and stored in the database to provide new data for the database and prepare for subsequent model iteration optimization.

[0081] When plants grow poorly, the real-time collected environmental parameter data related to the microclimate needs to be added to the training set as new data supplements, and the newly supplemented data needs to be iteratively optimized using multiple feature extensions to update the model's weights and biases to achieve iterative optimization of the model.

[0082] In some feasible embodiments, PolynomialFeatures may be used to perform polynomial feature expansion, and the specific steps are as follows:

[0083] 201. Define an original feature matrix X;

[0084] 202. Create a PolynomialFeatures object, specify the polynomial degree;

[0085] 203. Use the fit_transform method to fit and transform the original feature matrix to generate a polynomial feature expansion matrix containing the original features. The polynomial feature expansion is to perform a polynomial transformation on the original features to generate new features, which include the higher-order terms of the original features and the cross terms between different original features.

[0086] The role and purpose of polynomial feature expansion are:

[0087] 1. Capturing nonlinear relationships: There may be nonlinear relationships between the original features and the target variable. Through polynomial feature expansion, these nonlinear relationships can be captured, thereby improving the model's fitting ability and prediction accuracy. For example, the relationship between temperature and humidity may not be a simple linear relationship, but a more complex polynomial relationship.

[0088] 2. Improve the flexibility of the model: Polynomial feature expansion increases the degree of freedom of the model, enabling the model to fit more complex patterns. Especially when the amount of data is large and the relationship between features is complex, adding polynomial features can significantly improve the performance of the model.

[0089] 3. Enhance feature interactions: Interactions between features can have a significant impact on the prediction results. By generating cross terms of the features, these interactions can be explicitly introduced into the model. For example, the interaction between temperature and humidity may have a significant impact on wind speed, and generating features such as TH can help the model better capture this relationship.

[0090] For example, suppose there are three original features: temperature (T), humidity (H), and wind speed (W). The polynomial feature expansion process is as follows:

[0091] Original features: T, H, W;

[0092] Secondary features: T2, H2, W2, TH, TW, HW;

[0093] Tertiary features: T3, H3, W3, T2 H, T2W, H2 T, H2 W, W2 T, W2 H, THW;

[0094] Creating a new feature has the following effects:

[0095] 1. Improve model performance: Polynomial feature expansion increases the dimensionality of the feature space, allowing the model to better fit complex patterns and relationships, thereby improving model performance.

[0096] 2. Improve the explanatory power of the model: By introducing more features, especially interaction terms, it can help understand the relationship and influence between different sensor data features. For example, by analyzing the coefficients of the interaction terms of temperature and humidity, we can understand their joint influence on the target variable.

[0097] 3. Dealing with the limitations of linear models: Linear models assume that there is a linear relationship between features and target variables, but the actual situation is often more complicated. Polynomial feature expansion enables linear models to handle nonlinear relationships and improve the applicability of the model.

[0098] In some feasible embodiments, the generated new features are further standardized so that the generated new features are on the same scale. The standardization step is:

[0099] 301. Calculate the statistics of the new features. For each feature, calculate its mean and standard deviation.

[0100] 302. Apply the normalization formula and normalize each feature using the following formula:

[0101]

[0102] Where X is the original eigenvalue, μ is the mean of the feature, and σ is the standard deviation of the feature;

[0103] 303. Generate a standardized feature matrix, apply the above formula to each feature, and generate a new feature matrix in which all features are standardized.

[0104] If standardization is not performed, the following problems may occur after the original features are expanded using the fit_transform method for polynomial features:

[0105] 1. Inconsistent feature scales: Polynomial feature expansion generates powers and combinations of the original features, which may result in new features with very different ranges and dimensions from the original features. For example, the square or cube of a feature may be much larger than the original feature, which makes comparisons between features meaningless.

[0106] 2. Impact on model performance: Many machine learning algorithms (such as gradient descent, support vector machine, k-nearest neighbor, etc.) are sensitive to feature scale. If the feature scale is inconsistent, these algorithms may not be able to learn effectively or converge to the optimal solution.

[0107] 3. Increased difficulty in model training: Large numerical differences between features may lead to numerical stability issues during model training. For example, in gradient descent, feature scale differences may cause the gradient direction to be biased towards features with larger values, thereby affecting the weight update of other features.

[0108] 4. Overfitting risk: Features with large values ​​may dominate the model, causing the model to focus too much on these features and ignore other features that may be equally important. This may cause the model to perform well on the training data but poorly on unseen data, which is called overfitting.

[0109] 5. Computational efficiency issues: During the numerical optimization process, feature scale differences may require more iterations to converge, or require more sophisticated numerical stability control, which will reduce the training efficiency of the model.

[0110] Therefore, after polynomial feature expansion, the features need to be standardized or normalized to ensure that all features are on the same scale, thereby improving the training efficiency and prediction performance of the model.

[0111] In this embodiment, Apache Flink is used to build a real-time data processing flow to process the environmental parameter data related to the microclimate collected by the sensor in real time, and the median filtering algorithm is implemented in Flink to perform real-time denoising on the data. The processed data is then output to the Kafka message queue or storage system for subsequent processing and analysis.

[0112] Data preprocessing includes data cleaning, denoising and data normalization (also called standardization). Data cleaning is used to remove invalid, erroneous or incomplete data; normalization is used to scale the denoised data to a uniform scale. You can refer to the above standardization process. Denoising is used to remove noise from the cleaned data and uses median filtering technology. Specifically, the steps of median filtering are as follows:

[0113] (1) Window selection: Determine the window size of the filter, which is usually an odd-sized square or rectangle (such as 3x3, 5x5, etc.). The window size determines the smoothness of the filter.

[0114] (2) Window movement: Slide the window over the data, which can be the pixels of an image or the data points of a signal. The window moves one pixel or data point at a time.

[0115] (3) Sorting: In each window, the grayscale values ​​(or signal intensities) of all pixels (or data points) are sorted.

[0116] (4) Median selection: Select the middle value from the sorted data as the new value of the center pixel (or data point) of the window. If the window size is NxN, the median is the (N^2+1) / 2th value.

[0117] (5) Assignment: Assign the calculated median value to the pixel at the center of the window.

[0118] (6) Iterative process: Repeat steps (2) to (5) until the entire image or signal has been processed.

[0119] Median filtering is used to remove outliers or abnormal values ​​in data. For soil sensor data, median filtering can effectively handle some abnormal values, such as abnormal readings caused by stray effects or sensor errors. Its main functions are:

[0120] Removing impulse noise: For data containing impulse noise, median filtering is very effective because impulse noise is usually located at both ends of the sequence after sorting and will not affect the median.

[0121] Signal fidelity: While removing noise, median filtering can better maintain the original characteristics of the signal and will not blur the edges of the signal like mean filtering.

[0122] Real-time processing: In hardware implementation, a high level of real-time processing effect can be achieved, which is suitable for sensor data processing with high real-time requirements.

[0123] After the model passes the evaluation, the water demand can be predicted for the system. In this implementation, the optimal regression model for intelligent drip irrigation of different varieties of plants can be established in the above manner, and a model library can be established. After the crop variety is entered, the corresponding regression model can be called from the model library, and the microclimate parameters related to the crop (temperature, humidity, wind speed, light intensity, etc.) collected can be input. The model can be used to predict the water demand of the crop on the day, and then the system can perform quantitative drip irrigation on demand.

[0124] Specifically, Figure 1 As shown in the figure, in practice, the irrigation amount and frequency of the drip irrigation system can be adjusted according to the measured water demand. The water demand of the plant can be met by adjusting the parameters such as the irrigation time, water flow and irrigation interval of the drip irrigation system. During the drip irrigation process, the temperature and humidity of the soil and the meteorological data are monitored in real time, and the drip irrigation parameters of the drip irrigation system are adjusted dynamically. After drip irrigation, the soil humidity around the crop is monitored regularly to calculate whether the humidity range is in line with the growth habits of the crop. If the humidity is lower than the normal range, supplementary irrigation is performed.

[0125] The drip irrigation intelligent control method implemented in this paper can solve the shortcomings of water resource waste and the problem that the existing sprinkler irrigation cannot flexibly perceive the changes in the crop growth environment and the needs of crops at different growth stages, resulting in excessive or insufficient sprinkler irrigation. In addition, it can also achieve cost savings and increase efficiency.

[0126] A second aspect of the embodiments of the present invention provides a drip irrigation intelligent control system based on microclimate perception, which is used to implement the above-mentioned drip irrigation intelligent control method. The system includes:

[0127] A data acquisition module, which uses sensors to collect environmental parameter data related to the microclimate and transmits it to the data processing module;

[0128] A data processing module is used to preprocess the collected real-time data or the called historical data. The data preprocessing includes data cleaning, denoising and data normalization, and the preprocessed historical data is divided into a training set and a validation set;

[0129] The model building module inputs the environmental parameter data related to the microclimate in the training set into the neural network structure to establish a regression model for intelligent drip irrigation, which is used to predict the mapping relationship between plant water demand and environmental parameters;

[0130] The model training optimization module uses multiple feature extensions to perform iterative optimization to update the model's weights and biases to achieve training of the regression model.

[0131] The system of this embodiment also includes a model evaluation module, which uses a validation set to evaluate the trained regression model, evaluate the performance of the regression model and perform optimization;

[0132] The model application module applies the approved model to the intelligent prediction and control of the water demand of plants of the same variety, collects environmental parameter data related to the microclimate in real time, and predicts the water demand of the plant for the day through the model, and then implements on-demand quantitative drip irrigation through the drip irrigation system; the model iteration module iteratively optimizes the model by adjusting the threshold in the model evaluation.

[0133] A third aspect of the embodiments of the present invention provides a computer-readable storage medium storing a program, which implements the above-mentioned drip irrigation intelligent control method when executed by a processor.

[0134] A fourth aspect of an embodiment of the present invention provides a computing device, including a processor and a memory for storing a program executable by the processor, and when the processor executes the program stored in the memory, the above-mentioned drip irrigation intelligent control method is implemented.

[0135] The specific implementation of each module in this embodiment can refer to the above-mentioned method embodiment, which will not be described one by one here; it should be noted that the system provided in this embodiment is only illustrated by the division of the above-mentioned functional modules. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure can be divided into different functional modules to complete all or part of the functions described above.

[0136] A computer-readable storage medium of this embodiment may be a storage medium such as a ROM, RAM, a disk, or an optical disk, which stores one or more programs. When the program is executed by a processor, the above-mentioned small sample financial text classification method is implemented.

[0137] A computing device of this embodiment may be a desktop computer, a laptop computer, a smart phone, a PDA handheld terminal, a tablet computer or other terminal device with a display function, including a processor and a memory for storing programs executable by the processor, the memory stores one or more programs, and when the processor executes the program stored in the memory, the above-mentioned small sample financial text classification method is implemented.

[0138] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0139] It should be noted that, in the description of this application, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of this application, unless otherwise specified, the meaning of "plurality" refers to at least two.

[0140] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0141] It should be understood that the various parts of the present application can be implemented with hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented with software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented with hardware, as in another embodiment, it can be implemented with any one of the following techniques well known in the art or a combination thereof.

[0142] A person skilled in the art may understand that all or part of the steps of the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0143] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0144] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0145] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A drip irrigation intelligent control method based on microclimate perception, characterized in that: The following steps are involved: S1: Data collection, calling plant-related historical data, the historical data including environmental parameter data related to microclimate and actual plant water demand data, and dividing the data into a training set and a validation set; S2: Model establishment, inputting the environmental parameter data related to the microclimate in the training set into the neural network structure, and establishing a regression model of intelligent drip irrigation to predict the mapping relationship between plant water demand and environmental parameters; S3: Model training, using multiple feature expansions for iterative optimization to update the model weights and biases to train the regression model; S4: Model evaluation: Use the validation set to evaluate the trained regression model, evaluate the performance of the regression model and perform optimization.

2. The drip irrigation intelligent control method based on microclimate perception according to claim 1 is characterized in that: The model evaluation includes the following steps:

101. Input the environmental parameter data related to the microclimate in the validation set into the trained regression model to calculate the predicted plant water demand data; 102. Analyze and compare the calculated predicted plant water demand data y with the actual plant water demand data y0 in the plant historical data; If the set threshold requirement is met within the specified data deviation range, it means that the model meets the actual needs and is used as the optimal model; If the set threshold requirement is not met within the specified data deviation range, it means that the model needs further adjustment, and the model training steps are repeated, and the model weights and biases are updated again through iterative optimization through multiple feature expansions.

3. The drip irrigation intelligent control method based on microclimate perception according to claim 2 is characterized in that: Also includes the following steps: S5: Model application: Apply the model that has passed the evaluation to the intelligent prediction and control of the water demand of plants of the same variety. Collect the environmental parameter data related to the microclimate around the plant in real time over a period of time, and use the model to predict the water demand of the plant on the same day. Then, use the drip irrigation system to carry out quantitative drip irrigation on demand, and observe the growth of the plant at the same time. S6: Model iteration: According to the growth of the plants, determine whether the threshold in the model evaluation needs to be adjusted. If the plants grow well, there is no need to adjust the threshold. If the plants grow poorly, the threshold needs to be increased and the model evaluation needs to be re-performed.

4. The drip irrigation intelligent control method based on microclimate perception according to claim 3 is characterized in that: When plants grow poorly, the environmental parameter data related to the microclimate collected in real time needs to be added to the training set as new data supplements, and the newly supplemented data needs to be iteratively optimized using multiple feature extensions to update the model weights and biases.

5. The drip irrigation intelligent control method based on microclimate perception according to claim 4 is characterized in that: Preprocess the plant-related historical data or real-time collected data, and the data preprocessing includes data cleaning, denoising and data normalization: The data cleaning is used to remove invalid, erroneous or incomplete data; The denoising process is used to remove the noise in the cleaned data, and adopts the median filtering technology; The normalization process is used to scale the denoised data to a uniform scale.

6. The drip irrigation intelligent control method based on microclimate perception according to claim 1 is characterized in that: The regression model formula is: y=b0+b1·d1+b2·d2+...+b n ·d n +e (1) Where: y is the dependent variable, i.e. the predicted plant water requirement; b0 is the bias, which represents the expected value of the dependent variable when all independent variables are 0; b1, b2...b n is the regression coefficient of each independent variable, also called weight, which indicates the influence of each independent variable on the dependent variable; x1, x2...x n are the independent variables, i.e., environmental parameters related to the microclimate; e is the error term, which is the parameter that needs to be adjusted in the model and represents the difference between the model prediction value and the actual demand value.

7. The drip irrigation intelligent control method based on microclimate perception according to claim 6 is characterized in that: Use PolynomialFeatures to expand polynomial features. The specific steps are as follows:

201. Define an original feature matrix X; 202. Create a PolynomialFeatures object, specify the polynomial degree; 203. Use the fit_transform method to fit and transform the original feature matrix to generate a polynomial feature expansion matrix containing the original features; And / or, the polynomial feature expansion is to perform a polynomial transformation on the original feature to generate a new feature, which includes high-order terms of the original feature and cross terms between different original features.

8. A drip irrigation intelligent control system based on microclimate perception, used to implement any one of the drip irrigation intelligent control methods according to claims 1-7, characterized in that: The system comprises: A data acquisition module, which uses sensors to collect environmental parameter data related to the microclimate and transmits it to the data processing module; A data processing module is used to preprocess the collected real-time data or the called historical data. The data preprocessing includes data cleaning, denoising and data normalization, and the preprocessed historical data is divided into a training set and a validation set; The model building module inputs the environmental parameter data related to the microclimate in the training set into the neural network structure to establish a regression model for intelligent drip irrigation, which is used to predict the mapping relationship between plant water demand and environmental parameters; The model training optimization module uses multiple feature extensions to perform iterative optimization to update the model's weights and biases to achieve training of the regression model.

9. The drip irrigation intelligent control system based on microclimate perception according to claim 8 is characterized in that: It also includes a model evaluation module, which uses the validation set to evaluate the trained regression model, evaluate the performance of the regression model and perform tuning; The model application module applies the approved model to the intelligent prediction and control of the water demand of plants of the same variety, collects environmental parameter data related to the microclimate in real time, and predicts the water demand of the plant on the day through the model, and then realizes quantitative drip irrigation on demand through the drip irrigation system; The model iteration module iteratively optimizes the model by adjusting the threshold in the model evaluation.

10. A computer-readable storage medium storing a program, characterized in that: When the program is executed by a processor, the drip irrigation intelligent control method as described in any one of claims 1 to 9 is implemented.

Images