Greenhouse flat field environment simulation system based on machine learning

Abstract

The present application relates to the technical field of intelligent control of greenhouse environment, and particularly relates to a greenhouse plane field environment simulation system based on machine learning, comprising: a data acquisition unit, used for acquiring greenhouse environment parameter data and control equipment operation data, wherein the control equipment operation data comprises a working state combination of multiple control equipment; a data processing cloud platform, in communication connection with the data acquisition unit, used for processing the greenhouse environment parameter data and the control equipment operation data and generating a greenhouse environment prediction model, and used for predicting the greenhouse environment by using the greenhouse environment prediction model to obtain temperature distribution data of different positions in the greenhouse at the next moment. The system realizes high-precision prediction of temperature distribution at different positions in the greenhouse, provides a scientific basis for optimization and regulation of the greenhouse environment and crop planting layout, and improves the intelligent level of greenhouse environment management.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for greenhouse environments, and in particular to a greenhouse planar field environment simulation system based on machine learning. Background Technology

[0002] Intelligent control of greenhouse environments is a crucial technological area in modern agricultural production. Environmental parameters such as temperature, humidity, light intensity, and carbon dioxide concentration within the greenhouse directly impact crop growth and yield. However, current greenhouse environment control typically relies on simple feedback control systems or experience-based adjustment methods. These methods struggle to cope with the complex changes in the internal and external greenhouse environments, especially when there are highly coupled and nonlinear relationships between environmental parameters. Traditional methods' modeling and predictive capabilities become insufficient in these situations. For example, greenhouse temperature is not only affected by external climatic conditions (such as wind speed, radiation, and rainfall) but also closely related to internal parameters like humidity, light intensity, and carbon dioxide concentration. This complex coupling makes it difficult for traditional control systems to achieve accurate prediction and efficient regulation.

[0003] Furthermore, existing greenhouse environmental control systems generally lack detailed modeling of the spatial distribution of environmental parameters within the greenhouse. They typically only predict single points or average values, failing to accurately reflect the temperature distribution at different locations within the greenhouse. This limitation leads to uneven greenhouse environmental regulation, potentially resulting in localized overheating or undercooling, thus affecting crop growth and yield. Simultaneously, existing technologies also have shortcomings in data processing. For example, sensor-collected data often contains outliers or missing values; failure to effectively address these issues significantly reduces the accuracy and reliability of the predictive models. Summary of the Invention

[0004] In view of this, the present invention proposes a greenhouse planar field environment simulation system based on machine learning to solve the problems of low prediction accuracy, insufficient spatial distribution modeling, and limited data processing capabilities of existing greenhouse environment control systems.

[0005] The technical solution of this invention is implemented as follows: This invention provides a greenhouse planar field environment simulation system based on machine learning, comprising:

[0006] The data acquisition unit is used to collect greenhouse environmental parameter data and control equipment operation data, wherein the control equipment operation data includes a combination of the working status of multiple control devices;

[0007] The data processing cloud platform is connected to the data acquisition unit and is used to process the greenhouse environmental parameter data and control equipment operation data to generate a greenhouse environment prediction model. The greenhouse environment prediction model is used to predict the greenhouse environment and obtain the temperature distribution data of different locations in the greenhouse at the next moment.

[0008] The data processing cloud platform includes:

[0009] The data processing unit is used to preprocess the collected data;

[0010] The greenhouse environment prediction model training unit is used to train and generate multiple greenhouse environment prediction sub-models using preprocessed data. Each prediction sub-model corresponds to a combination of working states, and all greenhouse environment prediction sub-models constitute the greenhouse environment prediction model.

[0011] The prediction execution unit is used to select the corresponding prediction sub-model based on the current working state of the control equipment, and input the current environmental parameters into the selected prediction sub-model to obtain the predicted temperature distribution data of different locations in the greenhouse at the next moment.

[0012] The visualization unit receives the temperature distribution prediction data output by the prediction execution unit and generates a temperature distribution heat map of the greenhouse planar field for display.

[0013] Based on the above technical solutions, preferably, the data acquisition unit includes:

[0014] Weather station sensors are used to collect data on outdoor environmental conditions;

[0015] The greenhouse environment sensors are used to collect environmental parameter data at different locations. The greenhouse environment sensors include temperature sensors, humidity sensors, light radiation sensors and carbon dioxide sensors. The greenhouse is divided into several layers along the longitudinal direction. Each layer is equipped with several temperature sensors and humidity sensors. The light radiation sensors and carbon dioxide sensors are installed on the top layer of the greenhouse.

[0016] Control equipment sensors are used to collect operating data from control equipment.

[0017] Based on the above technical solutions, preferably, the control equipment includes a fan, skylight, side window, sunshade net, and air conditioner. The combination of operating states of the control equipment is expressed by the following formula:

[0018] S={(s1,s2,s3,s4,s5)|s i ∈{0,1},i=1,2,3,4,5}

[0019] In this context, s1 represents the fan, s2 represents the skylight, s3 represents the side window, s4 represents the sunshade net, s5 represents the air conditioner, 0 represents off, and 1 represents on.

[0020] Based on the above technical solutions, the preferred method for constructing the greenhouse environment prediction sub-model includes:

[0021] S1. Acquire greenhouse environmental parameter data under the same combination of operating states of the control equipment, which are divided into training data and test data;

[0022] S2. Preprocess the greenhouse environmental parameter data;

[0023] S3. Construct a greenhouse environment prediction sub-model using a BP neural network, and train the prediction sub-model based on the training data;

[0024] S4. Verify the prediction accuracy of the prediction sub-model using test data, optimize the model parameters by calculating the error between the predicted value and the measured value, and output the greenhouse environment prediction sub-model.

[0025] Based on the above technical solutions, preferably, the greenhouse environmental parameter data includes:

[0026] Outdoor environmental parameters: wind force, wind speed, total radiation, cumulative rainfall, instantaneous rainfall, daily rainfall, air temperature, air humidity, and carbon dioxide concentration;

[0027] Indoor environmental parameters: air temperature, air humidity, and carbon dioxide concentration at different locations.

[0028] Based on the above technical solution, preferably, in step S2, the greenhouse environmental parameter data is preprocessed, including:

[0029] S21. Perform outlier detection on the acquired raw data and remove data that does not meet the preset conditions;

[0030] S22. Fill missing values ​​based on data type. Filling methods include mode filling, linear interpolation filling, and weighted average filling.

[0031] S23. Normalize the filled data to obtain preprocessed data.

[0032] Based on the above technical solutions, preferably, the wind force, wind speed, and instantaneous rainfall are calculated using the mode filling method, and the formula for the mode filling method is:

[0033] x 填充 =Mod(x) i |i indicates the day the data needs to be filled)

[0034] Where Mod represents the mode operation, x i This represents the value on the day the data needs to be filled in;

[0035] Total radiation, cumulative rainfall, daily rainfall, air temperature, air humidity, and carbon dioxide concentration were filled using a linear interpolation method. The formula for this linear interpolation method is as follows:

[0036]

[0037] Where N2 is the number of sampling points within the preset time range, x t-i and x t+i These are the data from the i-th sampling point before and after the missing point, respectively.

[0038] Based on the above technical solutions, preferably, the greenhouse temperature and humidity are filled using a weighted average, including:

[0039] (1) Determine the 8 nearest valid sampling points around the missing point, and denot them as P1, P2, ..., P8;

[0040] (2) Collect data of the same relative positions in other planar fields at different time points, calculate the standardized regression coefficients of each sampling point and the center point, normalize the standardized regression coefficients to obtain the weight coefficients, and the formula for calculating the standardized regression coefficients is as follows:

[0041]

[0042] Where, x i For the historical data sequence of the i-th sampling point, For the independent variable x i The standard deviation, y is the historical data sequence centered at point S y Let y be the standard deviation of the independent variable. The formula for calculating the standard deviation is:

[0043]

[0044] x i This represents the i-th sample point. This represents the sample mean; n represents the sample size.

[0045] (3) Calculate the weighted average filling of missing values ​​based on the weight coefficients.

[0046] Based on the above technical solutions, preferably, in step S3, the structure of the greenhouse environment prediction sub-model includes:

[0047] The input layer is used to receive greenhouse environmental parameter data and control equipment operation data; the hidden layer adopts a double hidden layer structure, with the first hidden layer having 2n+1 nodes and the second hidden layer having n+1 nodes, where n is the number of nodes in the input layer; the output layer has a number of nodes corresponding to the number of temperature points to be predicted in the greenhouse.

[0048] Based on the above technical solutions, the preferred execution process of the predictive execution unit is as follows:

[0049] Step 1: Receive the current working status information (s1, s2, s3, s4, s5) of each control device, convert it into a standard binary state vector, and match and call the corresponding prediction sub-model from the temperature prediction model based on the state vector.

[0050] Step 2: Obtain the current greenhouse environmental parameter data and perform preprocessing;

[0051] Step 3: Input the preprocessed greenhouse environmental parameter data into the selected prediction sub-model. Based on the input data, the prediction sub-model generates standardized prediction results for the preset detection points in the greenhouse at the next moment through forward propagation calculation.

[0052] Step 4: Denormalize the standardized prediction results and convert them into actual temperature values. Based on the spatial coordinates of the preset detection points, construct complete temperature distribution prediction data.

[0053] The machine learning-based greenhouse planar field environment simulation system of the present invention has the following advantages over the prior art:

[0054] (1) This invention achieves high-precision prediction of temperature distribution at different locations in a greenhouse by combining a data acquisition unit, a data processing cloud platform and a visualization display unit. The system can dynamically select a prediction sub-model based on the combination of greenhouse environmental parameters and the operating status of control equipment, and generate greenhouse temperature distribution data for the next moment. This provides a scientific basis for optimizing and regulating the greenhouse environment and crop planting layout, and improves the level of intelligent greenhouse environmental management.

[0055] (2) The data acquisition unit of the present invention combines data acquisition of outdoor environment, multi-point environment inside the greenhouse and operation status of control equipment, and comprehensively covers the core influencing factors of greenhouse environment changes; by vertically layering sensors to collect accurate indoor temperature, humidity and radiation data, it provides high-quality data support for greenhouse environment prediction; at the same time, through data preprocessing technology, including outlier detection, missing value filling and normalization, it ensures the integrity and consistency of data, further improves the accuracy and robustness of the environmental prediction model, and enhances the reliability of the system.

[0056] (3) This invention significantly improves the adaptability and predictive ability of greenhouse environments by designing multiple greenhouse environment prediction sub-models based on combinations of control device states and training and optimizing them using a BP neural network. The prediction execution unit can quickly match and call the corresponding sub-model after acquiring the current operating status of the control device in real time. The sub-model works together with the input environmental parameters to generate temperature distribution data at different locations in the greenhouse at the next moment. This modular construction and dynamic selection mechanism effectively improves the agility and relevance of the simulation system. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 This is a schematic diagram of the structure of the greenhouse planar field environment simulation system based on machine learning according to the present invention;

[0059] Figure 2 This is a system architecture diagram of the greenhouse environment of the present invention;

[0060] Figure 3 This is a system architecture diagram of the meteorological station measurement system of the present invention;

[0061] Figure 4 This is a flowchart illustrating the construction method of the greenhouse environment prediction sub-model of the present invention. Detailed Implementation

[0062] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0063] like Figure 1 and Figure 2 As shown, the present invention provides a greenhouse planar field environment simulation system based on machine learning, comprising:

[0064] The data acquisition unit is used to collect greenhouse environmental parameter data and control equipment operation data, wherein the control equipment operation data includes a combination of the working status of multiple control devices;

[0065] The data processing cloud platform is connected to the data acquisition unit and is used to process the greenhouse environmental parameter data and control equipment operation data to generate a greenhouse environment prediction model. The greenhouse environment prediction model is used to predict the greenhouse environment and obtain the temperature distribution data of different locations in the greenhouse at the next moment.

[0066] The data acquisition unit is used to comprehensively and in real-time collect greenhouse environmental parameter data and control equipment operation data, providing high-quality basic data support for the greenhouse environment simulation system. This unit includes weather station sensors, greenhouse internal environment sensors, and control equipment sensors.

[0067] Specifically, such as Figure 3 As shown, the weather station sensors are used to collect environmental condition data outside the greenhouse. The collected data includes, but is not limited to, wind force, wind speed, total radiation, cumulative rainfall, instantaneous rainfall, daily rainfall, air temperature, air humidity, and carbon dioxide concentration. By deploying weather station sensors outside the greenhouse, changes in external weather conditions can be monitored in real time.

[0068] Indoor environmental data sensors are used to collect environmental parameter data inside the greenhouse, specifically including air temperature, air humidity, light radiation intensity, and carbon dioxide concentration. To achieve three-dimensional sensing of the greenhouse environment, the sensors are arranged in layers according to the greenhouse's vertical height. Each layer has several temperature and humidity sensors evenly distributed, with the top layer serving as the crop growth layer, where light radiation and carbon dioxide sensors are installed. This layered arrangement comprehensively covers environmental information at different heights within the greenhouse, ensuring the spatial resolution and accuracy of the collected data. More preferably, the greenhouse environment is divided into five layers, with 12 temperature and humidity sensors installed on each layer.

[0069] Indoor control equipment sensors are used to collect operational status data of the control equipment within the greenhouse, including fans, skylights, side windows, shading nets, and air conditioners. Each control device has two operating states: on and off. Sensors monitor the operational status of each device in real time and convert it into a standardized binary state vector representation. For example, the on / off state of a fan can be represented as "1" or "0". The combined states of all control devices form a multi-dimensional vector describing the current control conditions of the greenhouse.

[0070] Specifically, the combination of operating states of the control equipment is represented by the following formula:

[0071] S={(s1,s2,s3,s4,s5)|s i ∈{0,1},i=1,2,3,4,5}

[0072] In this context, s1 represents the fan, s2 represents the skylight, s3 represents the side window, s4 represents the sunshade net, s5 represents the air conditioner, 0 represents off, and 1 represents on.

[0073] The data acquisition unit, in conjunction with various sensors via a real-time acquisition module, collects outdoor meteorological data, indoor environmental data, and control equipment status data at preset time intervals (e.g., every 10 minutes). The collected data is transmitted wirelessly or via wired network to a data processing cloud platform and stored in a database. Through the collaborative work of multiple sensor types, the data acquisition unit provides high-precision, timely environmental parameter data and control equipment status data, fully considering the complexity and dynamism of the greenhouse environment, enabling comprehensive perception of both the internal and external environments. Furthermore, the data acquisition unit supports real-time data display and export functions, facilitating user monitoring and analysis of the greenhouse environment.

[0074] The data processing cloud platform includes a data processing unit, a greenhouse environment prediction model training unit, a prediction execution unit, and a visualization unit. The data processing unit preprocesses the collected data. The greenhouse environment prediction model training unit uses the preprocessed data to train and generate multiple greenhouse environment prediction sub-models. Each sub-model corresponds to a specific combination of operating states, and all sub-models together form the greenhouse environment prediction model. The prediction execution unit selects the appropriate sub-model based on the current operating state combination of the control equipment and inputs the current environmental parameters into the selected sub-model to obtain predicted temperature distribution data for different locations within the greenhouse at the next moment.

[0075] The visualization unit receives temperature distribution prediction data output by the prediction execution unit and generates a temperature distribution heatmap of the greenhouse planar field for display. Through a communication interface with the prediction execution unit, the visualization unit receives real-time temperature distribution prediction data from different locations within the greenhouse. This data is calculated based on the greenhouse environment prediction model and includes temperature values ​​and their corresponding spatial coordinates at multiple preset monitoring points within the greenhouse. The visualization unit processes this data to generate a heatmap reflecting the temperature distribution within the greenhouse, which is then visually displayed through display devices (such as monitoring screens, mobile terminals, or remote monitoring platforms).

[0076] Specifically, such as Figure 4 As shown, the construction method of the greenhouse environment prediction sub-model includes:

[0077] S1. Obtain greenhouse environmental parameter data under the same combination of operating states of the control equipment, which is divided into training data and test data.

[0078] Specifically, a fixed control device is used to collect outdoor and indoor environmental parameter data of the greenhouse under certain conditions. The collected data is then divided into training data and test data. The training data is used to train the model, and the test data is used to validate the model. The division of training and test data usually follows a certain ratio (e.g., 8:2 or 7:3) to ensure that the model can fully learn the data features during training.

[0079] S2. Preprocess the greenhouse environmental parameter data, including data cleaning, data filling and data normalization.

[0080] The preprocessing specifically includes the following steps:

[0081] S21. Perform outlier detection on the acquired raw data and remove data that does not meet preset conditions. Outlier detection identifies and removes abnormal data points in the raw data caused by sensor malfunctions, sudden environmental changes, or other uncontrollable factors. Outliers are usually extreme values ​​or unreasonable numerical changes that deviate significantly from the normal data distribution, such as a sudden jump in greenhouse temperature to abnormally high or low values, or obvious discontinuous fluctuations in humidity sensors. Detected outliers will be removed or marked as missing values ​​and proceed to the next step of data filling.

[0082] S22. Fill missing values ​​based on data type. Filling methods include mode filling, linear interpolation filling, and weighted average filling.

[0083] Specifically, wind force, wind speed, and instantaneous rainfall change significantly over a short period of time, and are therefore discrete variables. The mode filling method is used, and the formula for the mode filling method is:

[0084] x 填充 =Mod(x) i |i indicates the day the data needs to be filled)

[0085] Where Mod represents the mode operation, x i This represents the value on the day the data needs to be filled.

[0086] Total radiation, cumulative rainfall, daily rainfall, air temperature, air humidity, and carbon dioxide concentration are continuous variables. A linear interpolation method is used to replace them, calculating the average of the sampling data within 10 minutes before and after the current data point. The formula for the linear interpolation method is:

[0087]

[0088] Where N2 is the number of sampling points within the preset time range, x t-i and x t+i These are the data points of the i-th sampling point before and after the missing point, respectively. The weighting coefficients are calculated on an average basis, ensuring that the filled value is the average of the previous and subsequent sampling points.

[0089] Data such as greenhouse air temperature and humidity are significantly influenced by the surrounding environment. This type of data is not only time-dependent but also spatially related, therefore the influence of surrounding spatial sampling points needs to be considered. Thus, the estimation of greenhouse air temperature and humidity data is based on surrounding sensors and their historical time series, achieved by calculating correlations and applying a weighted average. The specific weighted average filling method includes:

[0090] (1) In the greenhouse, the air temperature and humidity at the missing point can be estimated using data from multiple sensors around it. First, determine the eight nearest valid sampling points around the missing point (denoted as P1, P2, ... P8). These points are provided by environmental sensors installed in the greenhouse. At the same time, outliers or invalid data should be removed to ensure the reliability of the sampling points used for estimation.

[0091] (2) To comprehensively consider the correlation and influence of surrounding sensors and missing points in spatial distribution, it is necessary to calculate the correlation coefficient between each surrounding sampling point and the missing point. Data from other planar fields at different time points with the same relative position are collected, and the standardized regression coefficients between each sampling point and the center point are calculated. These standardized regression coefficients are then normalized to obtain the weighting coefficients. The formula for calculating the standardized regression coefficients is as follows:

[0092]

[0093] Where, x i For the historical data sequence of the i-th sampling point, For the independent variable x i The standard deviation, y is the historical data sequence centered at point S y Let y be the standard deviation of the independent variable. The formula for calculating the standard deviation is:

[0094]

[0095] x i This represents the i-th sample point. represents the mean of the sample; n represents the number of samples.

[0096] The calculated correlation coefficient represents the similarity between a sampled data point and a missing data point; a higher correlation coefficient indicates a greater influence of the sampled data point on the missing data point. Then, the standardized regression coefficients are normalized to obtain weight coefficients, ensuring that the sum of all weight coefficients equals 1. The formula for calculating the weight coefficients is:

[0097]

[0098] wi r is the weight coefficient for the i-th sampling point, representing the contribution of that sampling point to the missing point imputation value. i Let be the standardized regression coefficient of the i-th sampling point; The sum of the standardized regression coefficients for all 8 sampling points is used to normalize the weights.

[0099] (3) Based on the calculated weighting coefficients, the temperature or humidity data from the surrounding sensors are weighted and averaged to estimate the values ​​of missing points. The formula for weighted average filling is:

[0100]

[0101] Among them, w i P represents the weighting coefficient for the i-th sampling point. i This represents the actual data value of the i-th sampling point.

[0102] For filling in temperature and humidity data within the greenhouse, the dense sensor network within the greenhouse is fully utilized, and missing values ​​are effectively filled in using compensatory data from surrounding sensors. Compared to simple linear interpolation or fixed mean filling, weighted average extrapolation fully considers the spatial correlation and historical behavior patterns of the greenhouse environment. This not only improves the accuracy of the predicted values ​​but also preserves the spatial distribution characteristics of temperature and humidity within the greenhouse, thus providing higher-quality input data for subsequent greenhouse environment predictions.

[0103] S23. Normalize the filled data to obtain preprocessed data.

[0104] The normalization formula is:

[0105]

[0106] Where is the normalized data value, ranging from [0,1], and x is the original data value. min x is the minimum value in the dataset. max This represents the maximum value in the dataset.

[0107] In the preprocessing stage, by employing a classification-based imputation method for greenhouse environmental parameter data according to their different characteristics, the accuracy and completeness of the data were significantly improved, enhancing the prediction accuracy and stability of the greenhouse environmental simulation system. For data with small fluctuations and discrete characteristics, such as wind force, wind speed, and instantaneous rainfall, the mode imputation method can quickly capture the dominant feature values ​​within a time period, compensating for the impact of data gaps. For data with continuous variation characteristics, such as total radiation, cumulative rainfall, air temperature, air humidity, and carbon dioxide concentration, the linear interpolation imputation method can better preserve the trend and progression of the data, avoiding the accumulation of model errors caused by discontinuities. For data such as greenhouse temperature and humidity, which are greatly affected by spatial distribution, the weighted average imputation method is used, dynamically weighting the data based on the correlation of sampling points around the missing points, thereby more accurately restoring the missing values ​​and preserving the spatial distribution characteristics and physical consistency of the data.

[0108] S3. Construct a greenhouse environment prediction sub-model using a BP neural network, and train the prediction sub-model based on the training data.

[0109] Specifically, the structure of the greenhouse environment prediction sub-model based on a BP neural network includes an input layer, an output layer, and hidden layers. The input layer receives data on external environmental conditions (such as wind speed and air temperature), internal environmental conditions (such as current indoor temperature and humidity), and the operating status data of various control devices (such as fans, skylights, and side windows). The number of input nodes is n, representing the total dimension of environmental parameters and control status. The hidden layer adopts a double-hidden-layer structure: the first hidden layer has 2n+1 nodes, and the second hidden layer has n+1 nodes, where n is the number of nodes in the input layer. The number of nodes in the output layer corresponds to the number of temperature points to be predicted within the greenhouse; each node outputs a predicted temperature value for one greenhouse monitoring point.

[0110] The entire network structure forms a complete mapping relationship through a hierarchical connection of input layer, hidden layer, and output layer: input data is connected to hidden layer neurons through fully connected layers, and each node performs weighted summation and nonlinear transformation; the hidden layer uses activation functions (such as Sigmoid or ReLU) to nonlinearly activate the data, thereby enhancing the expressive power of the model.

[0111] During network training, the input layer receives processed training data, including various environmental parameters (outdoor meteorological data and greenhouse sensor data) and combinations of control equipment operating states. In each training iteration, the network performs calculations through forward propagation. After forward propagation, the network output value is compared with the actual values ​​in the training data to calculate the prediction error. Using the error backpropagation algorithm, gradient descent is applied to adjust the weights and biases of each layer in the network based on the error. The training process is repeated multiple times until the network's error function E converges or the preset number of training iterations is reached. After training, the optimized network model parameters are saved, forming a greenhouse environment prediction sub-model under the current control equipment state combination.

[0112] S4. Verify the prediction accuracy of the prediction sub-model using test data, optimize the model parameters by calculating the error between the predicted value and the measured value, and output the greenhouse environment prediction sub-model.

[0113] Specifically, the execution process of the predictive execution unit is as follows:

[0114] Step 1: Receive the current operating status information (s1, s2, s3, s4, s5) of each control device, convert it into a standard binary state vector, and use encoding to reflect the on / off state of each control device. Based on this state vector, the system matches the corresponding prediction sub-model in the temperature prediction model library and calls that model for subsequent prediction calculations.

[0115] Step 2: Obtain the greenhouse environmental parameter data at the current moment and perform preprocessing, including outlier detection, missing value imputation and normalization, in order to remove invalid data, fill in missing data and standardize the scale of input variables;

[0116] Step 3: Input the preprocessed greenhouse environmental parameter data into the selected prediction sub-model. Based on the input data, the prediction sub-model generates standardized prediction results for the preset detection points in the greenhouse at the next moment through forward propagation calculation.

[0117] Step 4: Denormalize the standardized prediction results to restore the standardized values ​​to the actual temperature values. Based on the spatial coordinate information of the preset detection points (such as the geometric position and height level of the detection points), use interpolation techniques or spatial mapping algorithms to generate complete greenhouse internal temperature distribution prediction data.

[0118] It should be noted that the greenhouse environment prediction simulation constructed by this invention can start with temperature modeling at a single point within the greenhouse crop layer, and simulate and analyze the general impact of greenhouse environmental conditions and control conditions on a single-point sensor. It can then be developed into temperature modeling of planar fields at different heights within the greenhouse, and simulate and analyze the specific impact of different greenhouse environmental conditions and control conditions on temperatures at different heights within the greenhouse. This allows for the rational planning of the spatial layout within the greenhouse and the improvement of crop yield.

[0119] In both single-point greenhouse environment simulation and planar greenhouse environment simulation, the input data comes from greenhouse environmental parameter data. However, due to the different objectives of the two simulations, there may be differences in the data preprocessing stage. In single-point greenhouse environment simulation, the input data is mainly used to predict the temperature of a single sensor, directly using the temperature parameters related to the target sensor; while in planar greenhouse environment simulation, the input data needs to be used to predict the temperature distribution of the entire planar field, which may require incorporating the spatial location information of the sensors or other features to reflect the overall characteristics of the planar field.

[0120] In one embodiment, two days of data under stable control conditions in the greenhouse are selected as the training data and test data of the model, respectively. The average value of the sampled data every 10 minutes within a day is selected as the feature value of that time period. The input data is a 144*n-dimensional vector matrix generated from n items of greenhouse environmental parameter data (including wind force, wind speed, total radiation, cumulative rainfall, instantaneous rainfall, daily rainfall, air temperature, air humidity, carbon dioxide, etc.). The output data is a 144*1-dimensional vector matrix generated from the air temperature of a single sensor in the greenhouse.

[0121] In another embodiment, two days of data under stable control conditions within the greenhouse are selected as the training and testing data for the model, respectively. The data acquisition module selects the average value of the data sampled every 10 minutes within a day as the feature value for that time period. The input data is a 144*n dimensional vector matrix generated from n items of greenhouse environmental parameters (including wind force, wind speed, total radiation, cumulative rainfall, instantaneous rainfall, daily rainfall, air temperature, air humidity, carbon dioxide, etc.), and the output data is the air temperature of m sensors in a single layer of the greenhouse, i.e., a 144*m dimensional vector matrix.

[0122] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A greenhouse planar field environment simulation system based on machine learning, characterized in that, include: The data acquisition unit is used to collect greenhouse environmental parameter data and control equipment operation data, wherein the control equipment operation data includes a combination of the working status of multiple control devices; The data acquisition unit includes weather station sensors, greenhouse environment sensors, and control equipment sensors. Weather station sensors are used to collect data on outdoor environmental conditions; The greenhouse environment sensors are used to collect environmental parameter data at different locations. The greenhouse environment sensors include temperature sensors, humidity sensors, light radiation sensors and carbon dioxide sensors. The greenhouse is divided into several layers along the longitudinal direction. Each layer is equipped with several temperature sensors and humidity sensors. The light radiation sensors and carbon dioxide sensors are installed on the top layer of the greenhouse. Control equipment sensors are used to collect operating data from the control equipment. The control equipment includes a fan, skylight, side windows, sunshade net, and air conditioner. The combination of operating states of the control equipment is expressed by the following formula: ， Where s1 represents the fan, s2 represents the skylight, s3 represents the side window, s4 represents the sunshade net, s5 represents the air conditioner, 0 represents off, and 1 represents on; The data processing cloud platform is connected to the data acquisition unit and is used to process the greenhouse environmental parameter data and control equipment operation data to generate a greenhouse environment prediction model. The greenhouse environment prediction model is used to predict the greenhouse environment and obtain the temperature distribution data of different locations in the greenhouse at the next moment. The data processing cloud platform includes: The data processing unit is used to preprocess the collected data; The greenhouse environment prediction model training unit is used to train and generate multiple greenhouse environment prediction sub-models using preprocessed data. Each prediction sub-model corresponds to a combination of working states, and all greenhouse environment prediction sub-models constitute the greenhouse environment prediction model. The methods for constructing greenhouse environment prediction sub-models include: S1. Acquire greenhouse environmental parameter data under the same combination of operating states of the control equipment, which are divided into training data and test data; S2. Preprocess the greenhouse environmental parameter data, including: S21. Perform outlier detection on the acquired raw data and remove data that does not meet the preset conditions; S22. Fill missing values ​​based on data type: Wind force, wind speed and instantaneous rainfall are filled using the mode method; total radiation, cumulative rainfall, daily rainfall, air temperature, air humidity and carbon dioxide concentration are filled using the linear interpolation method; greenhouse temperature and greenhouse humidity are filled using the weighted average method. Weighted average filling includes: (1) Determine the 8 nearest valid sampling points around the missing point, denoted as P1, P2, ..., P8; (2) Collect data of the same relative positions in other planar fields at different time points, calculate the standardized regression coefficients of each sampling point and the center point, normalize the standardized regression coefficients to obtain the weight coefficients, and the formula for calculating the standardized regression coefficients is: ， Where, x i For the historical data sequence of the i-th sampling point, For the independent variable x i The standard deviation of y is the historical data sequence centered at y. Let y be the standard deviation of the independent variable. The formula for calculating the standard deviation is: ， x i This represents the i-th sample point. This represents the sample mean; n represents the sample size. (3) Calculate the weighted average imputation of missing values ​​based on the weighting coefficients; The formula for calculating the weighting coefficient is: , Let be the weight coefficient for the i-th sampling point. Let be the standardized regression coefficient of the i-th sampling point; This is the sum of the standardized regression coefficients for all 8 sampling points; The formula for weighted average filling is: ,in, Let Pi be the weight coefficient for the i-th sampling point, and Pi be the actual data value of the i-th sampling point. S23. Normalize the filled data to obtain preprocessed data; S3. Construct a greenhouse environment prediction sub-model using a BP neural network, and train the prediction sub-model based on the training data; S4. Verify the prediction accuracy of the prediction sub-model using test data, optimize the model parameters by calculating the error between the predicted value and the measured value, and output the greenhouse environment prediction sub-model. The prediction execution unit is used to select the corresponding prediction sub-model based on the current operating state of the control equipment, and input the current environmental parameters into the selected prediction sub-model to obtain the predicted temperature distribution data at different locations within the greenhouse at the next moment; the execution process of the prediction execution unit is as follows: Step 1: Receive the current operating status information of each control device. This is converted into a standard binary state vector, and the corresponding prediction sub-model is matched and called from the temperature prediction model based on the state vector. Step 2: Obtain the current greenhouse environmental parameter data and perform preprocessing; Step 3: Input the preprocessed greenhouse environmental parameter data into the selected prediction sub-model. Based on the input data, the prediction sub-model generates standardized prediction results for the preset detection points in the greenhouse at the next moment through forward propagation calculation. Step 4: Denormalize the standardized prediction results and convert them into actual temperature values. Based on the spatial coordinate information of the preset detection points, construct complete temperature distribution prediction data. The visualization unit receives the temperature distribution prediction data output by the prediction execution unit and generates a temperature distribution heat map of the greenhouse planar field for display.

2. The greenhouse planar field environment simulation system based on machine learning as described in claim 1, characterized in that, The greenhouse environmental parameter data includes: Outdoor environmental parameters: wind force, wind speed, total radiation, cumulative rainfall, instantaneous rainfall, daily rainfall, air temperature, air humidity, and carbon dioxide concentration; Indoor environmental parameters: air temperature, air humidity, and carbon dioxide concentration at different locations.

3. The greenhouse planar field environment simulation system based on machine learning as described in claim 1, characterized in that, The formula for the mode filling method is: ， Here, Mod represents the mode operation. This represents the value on the day the data needs to be filled in; The formula for the linear interpolation filling method is: ， in, The number of sampling points within a preset time range. and These are the data from the i-th sampling point before and after the missing point, respectively.

4. The greenhouse planar field environment simulation system based on machine learning as described in claim 1, characterized in that, In step S3, the structure of the greenhouse environment prediction sub-model includes: The input layer is used to receive greenhouse environmental parameter data and control equipment operation data; the hidden layer adopts a double hidden layer structure, with the first hidden layer having 2n+1 nodes and the second hidden layer having n+1 nodes, where n is the number of nodes in the input layer; the output layer has a number of nodes corresponding to the number of temperature points to be predicted in the greenhouse.

Images