An intelligent management method and system for agricultural planting based on the Internet of Things

Abstract

The present invention discloses an intelligent management method and system for agricultural planting based on the Internet of Things, which relates to the technical field of agricultural planting. The method includes the following steps: obtaining environmental data in the farmland, extracting features from the environmental data to obtain environmental feature data; analyzing the environmental feature data by using a factor analysis algorithm to obtain the key factors affecting the growth of crops; constructing a demand prediction model, and using the demand prediction model to predict the growth demand of crops at a future time to obtain the optimal irrigation time and fertilization amount of the crops. By using the factor analysis algorithm to analyze the environmental feature data, the present invention enables agricultural managers to carry out land management and crop planting in a targeted manner, avoiding unnecessary resource consumption. By constructing a demand prediction model, it is possible to predict the demand of crops for resources such as water and fertilizers at a future time, helping agricultural managers to reasonably arrange the irrigation and fertilization time, thereby improving the yield and quality of crops.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural planting. Specifically, it relates to an intelligent management method and system for agricultural planting based on the Internet of Things. Background Art

[0002] Agricultural planting covers the cultivation of various plants, including crops, forest trees, fruit trees, flowers, medicinal plants, and ornamental plants, etc. The growth requirements and environmental adaptabilities of different plants vary. Therefore, in the planting process, it is crucial to understand and control the environmental factors that affect plant growth. The growth of crops depends not only on soil quality but also on various environmental factors, mainly including key parameters such as soil temperature and humidity, air temperature and humidity, and light intensity. These environmental factors directly affect the growth processes of plants, such as photosynthesis, nutrient absorption, and root development. Modern agriculture can, through the introduction of Internet of Things technology, sensors, and big data analysis, monitor and collect various environmental data in farmland in real time. These real-time data include information such as soil humidity, temperature, air humidity, temperature, and light, which can provide the most suitable growth conditions for crops. Through scientific data analysis, agricultural producers can timely understand the changes in the farmland environment and take corresponding adjustment measures, such as precise irrigation, fertilization, and temperature control, to ensure that crops grow under the best conditions.

[0003] In the prior art, due to the lack of efficient and real-time means for environmental data collection and analysis, managers cannot timely obtain the information on environmental changes required for crop growth, resulting in difficulties in accurately adjusting measures such as irrigation and fertilization, being unable to provide the most suitable growth conditions for crops, and being inconvenient for identifying the key environmental factors affecting crop growth, and not being able to help agricultural managers understand which factors play a decisive role in crop growth, reducing the success rate of crop growth. At the same time, it is not convenient to predict the future demand of crops for resources such as water and fertilizers, and it is not convenient to help agricultural management reasonably arrange the irrigation and fertilization times, which not only reduces the utilization efficiency of water resources and fertilizers but also reduces the healthy growth of crops, and further reduces the yield and quality of crops.

[0004] Regarding the problems in the related art, no effective solution has been proposed yet. Summary of the Invention

[0005] In view of the deficiencies of the prior art, the present invention proposes an intelligent management method and system for agricultural planting based on the Internet of Things, which solves the problems mentioned in the above background art that it is not convenient to understand the environmental conditions required for crop growth in real time, so that the managers cannot obtain timely environmental change information, which is not convenient for providing the most suitable growth conditions for crops, and it is not convenient to identify the key environmental factors affecting crop growth, and cannot help agricultural managers understand which factors play a decisive role in crop growth, reducing the success rate of crop growth. At the same time, it is not convenient to predict the future demand of crops for resources such as water and fertilizers, and it is not convenient to help agricultural management reasonably arrange irrigation and fertilization times, which not only reduces the utilization efficiency of water resources and fertilizers, but also reduces the healthy growth of crops, and further reduces the yield and quality of crops.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions:

[0007] According to one aspect of the present invention, an intelligent management method for agricultural planting based on the Internet of Things is provided, and the method includes the following steps:

[0008] S1. Obtain the environmental data in the farmland, and perform feature extraction on the environmental data to obtain environmental feature data;

[0009] S2. Analyze the environmental feature data by using a factor analysis algorithm to obtain the key factors affecting crop growth;

[0010] S3. Construct a demand prediction model, and use the demand prediction model to predict the future crop growth demand to obtain the optimal irrigation time and fertilization amount of the crops;

[0011] S2 includes the following steps:

[0012] S21. Obtain a sample set of environmental feature data, and organize the sample set to screen out a data subset related to crop growth characteristics;

[0013] S22. Based on screening technology, screen the data subset and select the environmental data that has the greatest impact on crop growth as high-quality individuals;

[0014] S23. Optimize the high-quality individuals by using a feature optimization algorithm to obtain the key factors affecting crop growth.

[0015] Furthermore, obtaining the environmental data in the farmland and performing feature extraction on the environmental data to obtain environmental feature data includes the following steps:

[0016] S11. Obtain the original environmental data in the farmland for data decomposition, and use signal separation technology to filter out high-frequency noise to obtain denoised farmland environmental data;

[0017] S12. Use dimensionality reduction analysis method to decompose the denoised farmland environment data, extract each environmental feature component, and perform smoothing processing;

[0018] S13. Calculate the correlation index between each extracted environmental feature component and the original farmland environment data, screen the features with the highest correlation, and eliminate the environmental components with the lowest correlation;

[0019] S14. Integrate the processed environmental feature components into a complete environmental feature data set to obtain environmental feature data.

[0020] Furthermore, using the dimensionality reduction analysis method to decompose the denoised farmland environment data, extract each environmental feature component, and perform smoothing processing includes the following steps:

[0021] S121. Arrange the denoised farmland environment data in chronological order into a data matrix, and perform singular value decomposition on the data matrix;

[0022] S122. Based on the decomposed data matrix, construct a linear mapping matrix, and perform dynamic mode decomposition on it to identify each environmental feature component related to the change of the farmland environment;

[0023] S123. Use low-pass filtering technology to smooth the identified key patterns and eliminate short-term fluctuations and noise.

[0024] Furthermore, based on the screening technology, screen the data subset, and select the environmental data that has the greatest impact on crop growth as high-quality individuals, including the following steps:

[0025] S221. Initialize the screening parameters, and each parameter represents a candidate configuration of the environmental data subset;

[0026] S222. Evaluate the screening parameters of each environmental data subset, and select the environmental data subset with the greatest impact as the central environmental data subset;

[0027] S223. Adjust the screening parameters of the non-central environmental data subset, and determine the number of generated new screening parameter sets according to the impact error on crop growth;

[0028] S224. Generate new environmental data subset parameters through Gaussian mutation, and generate a new screening parameter set according to the mutation range;

[0029] S225. Select the subset with the least impact on crop growth from the current screening parameters of all environmental data subsets for iterative update, and record the screening result with the greatest impact on crop growth;

[0030] S226. Check whether the preset number of iterations is reached. If so, stop the screening and record the environmental data that has the greatest impact on crop growth as the high-quality individual.

[0031] Further, the following steps are included to optimize the high-quality individual using the feature optimization algorithm to obtain the key factors affecting crop growth:

[0032] S231. Initialize the parameters of the feature optimization algorithm and the objective function of the factors affecting crop growth, define the initial optimization model, and set its objective function.

[0033] S232. Randomly initialize the feature subset configuration within the parameter space of the factors affecting crop growth, construct the initial optimization model, and calculate the corresponding objective function.

[0034] S233. Evaluate the initial optimization model using the training data, calculate the impact index of the current feature configuration on crop growth, and record the performance results.

[0035] S234. According to the adjustment rule of the feature optimization algorithm, adjust the current feature parameter combination. If the new feature combination has the best performance, update the current configuration; otherwise, update the global optimum and adjust the optimization model.

[0036] S235. If the optimal solution is not found during the optimization process, expand the feature selection range, dynamically adjust the search strategy, and eliminate the feature configuration with the least influence to obtain the key factors that have the greatest impact on crop growth.

[0037] Further, the formula of the adjustment rule is:

[0038] ;

[0039] In the formula, represents the new feature parameter combination after adjustment; T a represents the current feature parameter combination; T * represents the global optimal feature parameter combination; α represents the coefficient controlling the local search step size; β represents the coefficient controlling the local search range; T x and T m both represent randomly selected feature parameter combinations.

[0040] Further, construct a demand prediction model and use the demand prediction model to predict the future crop growth demand to obtain the best irrigation time and fertilization amount of the crops, including the following steps:

[0041] S31. Obtain the key factors affecting the growth requirements of crops and divide them into a training set and a test set;

[0042] S32. Initialize the parameters of the demand prediction model and determine the key parameters for constructing the demand prediction model;

[0043] S33. Input the training set and the parameters of the initial demand prediction model into the demand prediction model, and perform iterative optimization based on the model optimization algorithm, and output the best demand prediction model parameters;

[0044] S34. Based on the best demand prediction model parameters, construct the final demand prediction model, input the test set into the final demand prediction model for verification, and use the final demand prediction model to predict the growth requirements of crops at future times to obtain the best irrigation time and fertilization amount of the crops.

[0045] Furthermore, inputting the training set and the parameters of the initial demand prediction model into the demand prediction model, and performing iterative optimization based on the model optimization algorithm, and outputting the best demand prediction model parameters includes the following steps:

[0046] S331. Initialize the parameters of the demand prediction model and set the maximum number of iterations of the model optimization algorithm;

[0047] S332. Input the training set and the parameters of the initial demand prediction model into the model, calculate the error of the prediction result as the fitness value;

[0048] S333. Calculate the number of parameter configurations of the current demand prediction model according to the rules of the model optimization algorithm and adjust the parameter configuration of the demand prediction model;

[0049] S334. Perform quick sorting on the current demand prediction model according to the fitness value, and classify the demand prediction model into the optimal configuration and the auxiliary configuration;

[0050] S335. Mutate and update the demand prediction model according to the fitness value to generate a new demand prediction model, determine whether the model optimization algorithm has reached the maximum number of iterations. If it has reached, stop and output the best demand prediction model parameters. Otherwise, return to step S343 to continue optimization until the maximum number of iterations is reached.

[0051] Furthermore, the rule formula of the model optimization algorithm is:

[0052] ;

[0053] In the formula, P i represents the number of parameter configurations of the demand prediction model at the i th iteration; Pmax and P min respectively represent the upper and lower limits of the parameter configuration quantity; z represents the total cycle delay; y represents the delay imbalance degree; represents the maximum fitness of the total cycle delay; represents the minimum fitness of the total cycle delay; represents the average fitness of the total cycle delay; δ represents the threshold parameter; represents the maximum fitness of the delay imbalance degree; represents the minimum fitness of the delay imbalance degree; represents the average fitness of the delay imbalance degree.

[0054] According to another aspect of the present invention, there is also provided an intelligent management system for agricultural planting based on the Internet of Things, and the system includes:

[0055] A data acquisition module, configured to acquire environmental data in the farmland, extract features from the environmental data, and obtain environmental feature data;

[0056] A data analysis module, configured to analyze the environmental feature data by using a factor analysis algorithm to obtain key factors affecting the growth of crops;

[0057] A model prediction module, configured to build a demand prediction model and use the demand prediction model to predict the growth demand of crops at a future moment, and obtain the optimal irrigation time and fertilization amount of the crops;

[0058] Among them, the data acquisition module is connected through the data analysis module and the model prediction module.

[0059] The beneficial effects of the present invention are as follows:

[0060] 1. By acquiring environmental data in the farmland and extracting key features, agricultural managers can understand in real time the environmental conditions required for crop growth, enabling agricultural managers to obtain timely information on environmental changes, providing the most suitable growth conditions for crops, thus effectively avoiding over-irrigation and over-fertilization, reducing resource waste. By using a factor analysis algorithm to analyze environmental feature data, agricultural managers can manage the land and plant crops in a targeted manner, avoid unnecessary resource consumption, and improve the success rate of crop growth. By building an accurate demand prediction model, the demand for resources such as water and fertilizers by crops at a future moment can be predicted, helping agricultural managers reasonably arrange irrigation and fertilization times, and thus improving the yield and quality of crops.

[0061] 2. By using the factor analysis algorithm to analyze the environmental characteristic data, the present invention can identify the key environmental factors affecting the growth of crops, helping agricultural managers understand which factors play a decisive role in crop growth. Combining these key factors, agricultural managers can make scientific decisions, conduct targeted land management and crop planting, avoid unnecessary resource consumption, and increase the success rate of crop growth.

[0062] 3. By constructing an accurate demand prediction model, the present invention can predict the future demand of crops for resources such as water and fertilizers, helping agricultural managers reasonably arrange the irrigation and fertilization time. This precise regulation can effectively avoid over-irrigation and over-fertilization, not only improving the utilization efficiency of water resources and fertilizers, but also promoting the healthy growth of crops, thereby increasing the yield and quality of crops. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0064] Figure 1 is a flowchart of an intelligent management method for agricultural planting based on the Internet of Things according to an embodiment of the present invention;

[0065] Figure 2 is a schematic block diagram of an intelligent management system for agricultural planting based on the Internet of Things according to an embodiment of the present invention.

[0066] In the figure:

[0067] 1. Data acquisition module; 2. Data analysis module; 3. Model prediction module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0068] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments.

[0069] In the description of the present invention, unless otherwise specified, "a plurality of" means two or more. In addition, the terms "first", "second", "third", etc. are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0070] According to an embodiment of the present invention, an intelligent management method and system for agricultural planting based on the Internet of Things are provided.

[0071] The present invention will be further described in conjunction with the accompanying drawings and specific embodiments. As Figure 1 shown, according to the intelligent management method for agricultural planting based on the Internet of Things according to an embodiment of the present invention, the method includes the following steps:

[0072] S1. Obtain the environmental data in the farmland, and perform feature extraction on the environmental data to obtain environmental feature data;

[0073] In this alternative embodiment, obtaining the environmental data in the farmland and performing feature extraction on the environmental data to obtain environmental feature data includes the following steps:

[0074] S11. Obtain the original environmental data in the farmland for data decomposition, and filter high-frequency noise using signal separation technology to obtain the denoised farmland environmental data;

[0075] S12. Decompose the denoised farmland environmental data using a dimensionality reduction analysis method, extract each environmental feature component, and perform smoothing processing;

[0076] S13. Calculate the correlation index between each extracted environmental feature component and the original farmland environmental data, select the feature with the highest correlation, and eliminate the environmental component with the lowest correlation;

[0077] S14. Integrate the processed environmental feature components into a complete environmental feature data set to obtain environmental feature data.

[0078] It should be noted that first, environmental data is collected in real time in farmland by deploying an Internet of Things sensor network (such as soil moisture sensors, temperature and humidity sensors, light sensors, etc.). Suppose the collected data includes: soil temperature, soil moisture, air temperature, air humidity, light intensity, wind speed, etc.; there will be certain high-frequency noise in the obtained original data (such as sensor errors, weather interference, etc.), and these noises may affect the analysis results. Therefore, signal separation techniques, such as wavelet transform, Fourier transform, etc., are used to decompose the original data into multiple frequency levels, remove the high-frequency noise, and only retain the low-frequency signals related to crop growth; by denoising the data, the most representative farmland environmental data is retained. At this time, the obtained is the denoised farmland environmental data, removing the short-term fluctuations caused by factors such as sensor vibration and electromagnetic interference. Through dimensionality reduction analysis (i.e., dynamic mode decomposition), the multi-dimensional original data is compressed into fewer characteristic components, and these components can capture the main change trends of the data. For example, key information such as the water retention capacity of the soil and the impact of temperature changes on crop growth is extracted from the changes in soil moisture and temperature. The correlation between each environmental characteristic component and the original environmental data is calculated through the Pearson correlation coefficient, and the characteristics most relevant to crop growth are selected. For example, if soil moisture and soil temperature have a strong impact on the water demand and growth status of crops, then these two characteristic components will be selected as important characteristics; according to the correlation index, the environmental characteristic components with low correlation with the original data are removed, and these characteristics may no longer play a significant role in subsequent analysis. The environmental characteristic components after screening and optimization are integrated into a complete environmental characteristic data set. These data sets contain the most representative and highly correlated environmental parameters in the farmland; for example, after screening, the final data set may include environmental characteristics such as soil temperature, soil moisture, air humidity, light intensity, etc. These characteristics can accurately describe the farmland environment and provide data support for the subsequent crop growth demand prediction model; after completing the data integration, the finally obtained environmental characteristic data set will be used as the model input for subsequent agricultural production decision-making analysis, such as irrigation, fertilization, and crop growth prediction.

[0079] In this alternative embodiment, the denoised farmland environmental data is decomposed by using the dimensionality reduction analysis method, each environmental characteristic component is extracted, and the smoothing process includes the following steps:

[0080] S121. Arrange the denoised farmland environmental data in chronological order into a data matrix, and perform singular value decomposition on the data matrix;

[0081] S122. Based on the decomposed data matrix, construct a linear mapping matrix, and perform dynamic mode decomposition on it to identify each environmental characteristic component related to the changes in the farmland environment;

[0082] S123. Smooth the identified key patterns using low - pass filtering technology to eliminate short - term fluctuations and noise.

[0083] Specifically, the environmental data includes:

[0084] 1) Soil data:

[0085] Soil temperature: Affects the growth and metabolism of plant roots.

[0086] Soil humidity: Is directly related to the water supply of plants and is a key factor determining irrigation.

[0087] Soil pH value: Affects the absorption of nutrients by plants and determines whether plants can grow healthily in this environment.

[0088] Soil nutrient content: Includes macro - elements such as nitrogen, phosphorus, and potassium, as well as trace elements such as iron, manganese, and copper, which directly affect plant growth.

[0089] 2) Climate data:

[0090] Air temperature: Affects physiological processes such as photosynthesis and respiration of plants.

[0091] Air humidity: Affects the transpiration of plants, and too high or too low humidity may be unfavorable for plant growth.

[0092] Light intensity: The photosynthesis of plants depends on light intensity, so it is an important factor affecting crop growth.

[0093] Wind speed and direction: Excessive wind speed may cause too fast water evaporation, affecting the water supply of crops.

[0094] 3) Other data:

[0095] Precipitation: Affects the water retention of the soil and thus affects the irrigation requirements of crops.

[0096] CO2 concentration: Carbon dioxide is the raw material for plant photosynthesis, and changes in CO2 concentration will affect the growth rate of crops.

[0097] Specifically, the environmental characteristic data includes:

[0098] Climate index: Such as the daily average value and weekly average value of air temperature, etc., which can reflect the trend of climate change.

[0099] Soil moisture status: Based on soil humidity data, calculate indicators such as soil water retention capacity and drainage performance.

[0100] Growth environment index: Integrate data such as air temperature, humidity, and light to evaluate whether the current growth environment of crops is suitable.

[0101] Light utilization efficiency: Based on the relationship between light intensity and plant growth rate, the impact of light on crops is calculated.

[0102] Precipitation utilization rate: The relationship between precipitation and soil moisture change reflects the impact of water resources on crop growth.

[0103] It should be noted that Internet of Things (IoT) technology is used to obtain farmland environmental data. Specifically, it includes:

[0104] 1) Sensor deployment:

[0105] Temperature and humidity sensor: Used to measure the temperature and humidity in the air. These are key factors affecting crop growth.

[0106] Soil moisture sensor: Used to detect the water content in the soil to help determine the optimal irrigation timing.

[0107] Light sensor: Monitors light intensity to help evaluate whether crops receive sufficient sunlight.

[0108] Gas sensor: Monitors the gas composition in the farmland, such as carbon dioxide (CO2), oxygen (O2), ammonia (NH3), etc. These gases affect crop growth.

[0109] Soil pH sensor: Used to monitor the acidity and alkalinity of the soil for soil improvement.

[0110] Weather station: Integrates multiple sensors, such as temperature, humidity, precipitation, wind speed, air pressure, etc., to obtain meteorological data of the farmland.

[0111] 2) Data collection:

[0112] Wireless communication: Utilizes wireless technologies such as Wi-Fi, Bluetooth, LoRa, ZigBee, etc. to transmit the data collected by sensors to the centralized data processing system. This method has the advantages of simple wiring and flexible deployment.

[0113] Wired communication: Transmits data through a wired network (such as Ethernet), which is suitable for environments that require stability and high bandwidth.

[0114] Satellite remote sensing: Uses remote sensing technologies (such as satellite images, sensors carried by drones, etc.) to monitor the farmland environment over a large area, especially to obtain geographical information of the farmland, vegetation index, etc.

[0115] 3) Data transmission:

[0116] Wi-Fi: Suitable for short-distance, high-data-traffic applications and can upload data to the cloud through the Internet.

[0117] Bluetooth and BLE (Bluetooth Low Energy): Suitable for low-power devices, usually used for short-distance transmission, and suitable for small-scale agricultural applications.

[0118] LoRa (Long Range): Suitable for long-distance, low-power communication, capable of achieving long-distance data transmission, especially suitable for large-scale farmland or remote areas.

[0119] NB-IoT (NarrowBand Internet of Things): Suitable for low-power, large-scale connection environments, suitable for farmland monitoring in rural or remote areas.

[0120] ZigBee: Used for short-distance, low-power communication between devices, suitable for sensor networks in farmland.

[0121] S2. Use the factor analysis algorithm to analyze the environmental characteristic data to obtain the key factors affecting crop growth;

[0122] Specifically, the key factors affecting crop growth include:

[0123] 1) Soil factors:

[0124] Soil temperature: Affects the growth and metabolism of plant roots and is one of the important environmental conditions for crop growth.

[0125] Soil humidity: Directly related to the water supply of plants and is a key factor determining irrigation requirements.

[0126] Soil pH value: Affects the absorption of nutrients by plants, and different crops have different adaptation ranges to soil pH values.

[0127] Soil nutrient content: Includes macronutrients such as nitrogen (N), phosphorus (P), and potassium (K), as well as micronutrients such as iron (Fe), manganese (Mn), and copper (Cu). These nutrients directly affect the growth and development of plants.

[0128] Soil structure: The physical structure of the soil, such as particle size and porosity, affects the retention capacity of water and nutrients.

[0129] 2) Climate factors:

[0130] Air temperature: Affects physiological processes such as photosynthesis and respiration of plants, and different crops have different temperature requirements.

[0131] Air humidity: Affects the transpiration of plants, and too high or too low humidity may be unfavorable for plant growth.

[0132] Light intensity: The photosynthesis of plants depends on light intensity and is an important factor affecting crop growth.

[0133] Light duration: Different crops have different requirements for light duration, which affects the flowering period and fruit ripening.

[0134] Precipitation: Affects the water retention of the soil, and thus affects the irrigation requirements of crops.

[0135] Wind speed and direction: Excessive wind speed may cause too rapid water evaporation, affecting the water supply of crops, and the wind direction may affect the spread of pests and diseases.

[0136] 3) Biological factors:

[0137] Pests and diseases: The occurrence of pests and diseases will affect the growth and yield of crops, and timely monitoring and control are required.

[0138] Microbial activity: Microbial activity in the soil affects the soil fertility and the root health of crops.

[0139] 4) Human factors:

[0140] Irrigation management: Reasonable irrigation management can ensure that crops obtain sufficient water and avoid over-irrigation or water shortage.

[0141] Fertilization management: Scientific fertilization can provide the nutrients required by crops and avoid nutrient deficiency or excess.

[0142] Tillage methods: Different tillage methods (such as crop rotation, intercropping, cover crops, etc.) have important impacts on soil quality and crop growth.

[0143] 5) Other factors:

[0144] CO2 concentration: Carbon dioxide is the raw material for plant photosynthesis, and changes in CO2 concentration will affect the growth rate of crops.

[0145] Soil salinity: Excessive soil salinity will affect the growth of plants, especially for salt-sensitive crops.

[0146] Soil organic matter content: Soils with high organic matter content usually have better water and fertilizer retention capabilities, which are beneficial to crop growth.

[0147] S3. Build a demand prediction model and use the demand prediction model to predict the growth requirements of crops at future times to obtain the optimal irrigation time and fertilization amount of crops;

[0148] S2 includes the following steps:

[0149] S21. Obtain a sample set of environmental characteristic data, organize the sample set, and screen out the data subset related to the growth characteristics of crops;

[0150] It should be noted that not all of the environmental feature data samples are necessarily related to the growth characteristics of crops. The environmental feature data samples may contain various different environmental information, and this information does not necessarily directly affect the growth of crops. For example, environmental data may include meteorological data (such as wind speed, air pressure, etc.), soil properties (such as soil type, chemical composition, etc.), and some external factors unrelated to crop growth (such as pollution source data in the surrounding area, or other meteorological characteristics unrelated to agricultural activities). For example:

[0151] The meteorological data of farmland may contain information such as air pressure and wind speed. Although these data may have an impact on crop growth in some cases, they do not directly affect the growth process of crops in all cases. Perhaps the impact of wind speed on some crops is relatively small, while it may be more important for other crops.

[0152] Therefore, there may be features in the environmental feature data sample set that have a weak relationship with crop growth or have an indirect impact, rather than all data being directly related to the growth characteristics of crops.

[0153] It should be noted that the screening process is to improve the relevance and quality of the data, making the model or analysis results more accurate and efficient. The selected data subset only includes those features that have a significant impact on crop growth, which helps to:

[0154] Reduce noise data: The environmental data sample set may contain irrelevant information (such as some secondary features in meteorological factors). If not screened, this information may interfere with the analysis results and affect the prediction accuracy.

[0155] Improve the accuracy of the model: Selecting the feature data closely related to crop growth can significantly improve the efficiency and accuracy of modeling. For example, by screening the key factors affecting crop growth (such as soil humidity, temperature, light intensity, etc.), the model will focus more on the key environmental parameters, thus providing more accurate predictions.

[0156] Optimize computing resources: After screening, the data dimension is reduced and the computational complexity is lowered. This is especially important for processing large amounts of data (such as the large amount of sensor data generated by Internet of Things devices).

[0157] It should be noted that the environmental feature data sample set does not necessarily only contain features related to growth. Environmental feature data often includes various factors, and some may only have an indirect impact or be less related to agricultural growth. For example, certain weather factors (such as air pressure, wind speed) may have an indirect impact on crop growth, but this is not applicable in all cases. If not screened, this information may increase redundant data and affect the efficiency and accuracy of subsequent analysis. For example:

[0158] Soil type: Although the particle size, texture, etc. of the soil indirectly affect crop growth, they are not environmental factors themselves and may not be included during screening.

[0159] Air pressure, wind speed: Although they may affect crop growth under certain specific circumstances (such as strong winds may affect the photosynthesis of crops), the impact of these factors on most crops is relatively small, so they may be considered as non-critical data to be screened out. The basis for screening the subset of environmental data is usually data analysis. For example, through statistical methods (such as correlation analysis, principal component analysis, etc.), those variables highly correlated with crop growth are identified. The screening process generally consists of the following steps:

[0160] Correlation analysis: Evaluate the correlation between each environmental characteristic and crop growth characteristics, and screen out significantly correlated variables.

[0161] Domain knowledge: Combine domain knowledge such as agronomy and crop growth models to select environmental parameters known to affect crop growth.

[0162] Data modeling: By constructing a preliminary model, evaluate the contribution of each environmental factor to crop growth prediction, and screen out effective features.

[0163] S22. Based on the screening technology, screen the data subset, and select the environmental data that has the greatest impact on crop growth as high-quality individuals;

[0164] S23. Use the feature optimization algorithm to optimize the high-quality individuals to obtain the key factors affecting crop growth.

[0165] It should be noted that first, environmental characteristic data are collected from multiple sources, including meteorological data, soil data, light intensity, etc. These data may come from sensors, weather forecasts, soil test reports, etc.

[0166] Collect the sample set: Collect environmental data related to crop growth, such as soil moisture, air temperature, light intensity, air humidity, precipitation, soil temperature, carbon dioxide concentration, wind speed, etc. These data can be obtained through automated sensors, weather stations, or remote sensing data.

[0167] Organize the sample set: Organize the collected data, remove missing values or outliers, and ensure the accuracy of the data. For example, check whether there are missing timestamps or incorrect temperature values.

[0168] Based on existing agricultural research and literature, initially determine which environmental factors are most relevant to the growth characteristics of crops. For example, soil moisture, air temperature, and light intensity are considered key factors affecting crop growth. Therefore, screen out these data from the sample set and remove irrelevant factors (such as wind speed, carbon dioxide concentration, etc.) to obtain a more accurate data subset.

[0169] Apply screening techniques to further screen the data subset and select the environmental data that has the greatest impact on crop growth as high-quality individuals. Screening techniques can include correlation analysis, feature selection algorithms (such as chi-square test, information gain, fireworks algorithm, etc.). Through the screening technique, select the environmental data characteristics that have the greatest impact on crop growth, such as soil moisture and air temperature. These high-quality individuals will serve as the basis for subsequent feature optimization. Use feature optimization algorithms to optimize the selected high-quality individuals to finally determine the key factors affecting crop growth. Common feature optimization algorithms include genetic algorithms, particle swarm optimization algorithms, lightning search algorithms, etc. During the optimization process, use the high-quality features screened out in step S22 as the initial individuals (such as soil moisture and air temperature); define a fitness function to measure the quality of each feature combination. The fitness function is usually calculated through training data and can use the mean squared error (MSE) or other appropriate metrics. This fitness function measures the accuracy of the feature combination in predicting crop growth; use feature optimization algorithms (such as genetic algorithms, particle swarm optimization, lightning search algorithms, etc.) to adjust the feature parameters, and the optimization goal is to make the fitness value optimal. For example, through operations such as crossover and mutation of the genetic algorithm, continuously optimize the feature combination. After multiple iterations of optimization, finally obtain a set of optimized feature combinations, and these feature combinations represent the key factors affecting crop growth. The optimized features include:

[0170] Soil moisture: May have an important impact on the water requirements of crops.

[0171] Air temperature: Directly affects the growth rate and health of crops.

[0172] Light intensity: Plays a key role in the photosynthesis and yield of crops.

[0173] In this alternative embodiment, based on the screening technique, screening the data subset and selecting the environmental data that has the greatest impact on crop growth as high-quality individuals includes the following steps:

[0174] S221. Initialize the screening parameters, and each parameter represents a candidate configuration of the environmental data subset;

[0175] S222. Evaluate the screening parameters of each environmental data subset and select the environmental data subset with the greatest impact as the central environmental data subset;

[0176] S223. Adjust the screening parameters for the non - central environmental data subset, and determine the number of new screening parameter sets generated according to the impact error on crop growth.

[0177] S224. Generate new environmental data subset parameters through Gaussian mutation, and generate a new screening parameter set according to the mutation range.

[0178] It should be noted that generating new environmental data subset parameters through Gaussian mutation and generating a new screening parameter set according to the mutation range are for the screening parameters, rather than directly acting on the original time - series data. The screening parameters refer to the combination of each parameter in the feature subset (such as the weight configuration of features like temperature, humidity, soil humidity, etc.). Gaussian mutation is used to explore the optimal configuration of these parameter combinations to help screen out the most influential environmental features. This operation does not involve modifying or adding noise to the original data itself (such as temperature and humidity data in the time series). In the present invention, the time - series structure of the original time - series data will be completely retained in the subsequent modeling stage. The purpose of Gaussian mutation is to optimize the feature selection process and will not directly destroy the time - series nature of the data, because in the subsequent analysis and prediction processes, the time - series data will be used as input variables and processed through methods such as regression analysis and time - series modeling. Through this method, the model can understand and predict the time - series dynamic process of crop growth. The present invention obtains and analyzes environmental data in the farmland through Internet of Things technology to predict and optimize the irrigation and fertilization strategies of crops. Although growth characteristics of crops such as leaf area index and chlorophyll content are very important, the extraction and analysis of these characteristics and the dynamic prediction of the growth process generally belong to a part of the subsequent modeling. In the present invention, the screening and optimization of environmental data are to improve the indirect prediction accuracy of crop growth status. Crop growth characteristics (such as leaf area index and chlorophyll content) can be used as input labels in the subsequent modeling and evaluation stages, and thus assist in formulating more accurate irrigation and fertilization strategies. It should be noted that crop growth characteristics (such as leaf area index and chlorophyll content) can be used as input labels in the subsequent modeling and evaluation stages, and thus assist in formulating more accurate irrigation and fertilization strategies, including:

[0179] 1) Objectives:

[0180] Use the environmental data collected by sensors and crop growth characteristics for modeling to improve the accuracy of predicting irrigation and fertilization amounts.

[0181] Establish a multivariate regression model or a machine - learning model, taking environmental data and crop growth characteristics as inputs, and outputting irrigation and fertilization requirements.

[0182] 2) Data collection:

[0183] Environmental data: The environmental data in the farmland is collected in real time through Internet of Things devices (such as temperature and humidity sensors, soil moisture sensors, light sensors, etc.).

[0184] Crop growth data: Growth characteristics such as the leaf area index (LAI) and chlorophyll content (Chl) of the crops are obtained through means such as image recognition technology or leaf sensors.

[0185] 3) Data processing:

[0186] Time series processing: The environmental data and crop growth characteristic data are processed in a time series to ensure the temporal continuity of the data and avoid data loss or noise.

[0187] Feature extraction: Features are extracted from the original environmental data and feature fusion is performed in combination with crop growth data (such as LAI, Chl).

[0188] 4) Model construction:

[0189] Regression analysis model: Methods such as multiple linear regression and support vector regression (SVR) are used to construct a prediction model. The inputs of the model are environmental data and crop growth characteristics, and the outputs are the predicted values of irrigation and fertilization amounts.

[0190] Machine learning model: Machine learning models such as random forest regression, gradient boosting machine (GBM), and deep neural network (DNN) can be used to perform more complex modeling in combination with environmental data and crop growth characteristics.

[0191] 5) Model training and validation:

[0192] Training dataset: Farmland data at different growth stages are selected for training to ensure coverage of different seasons, weather conditions, and crop states.

[0193] Validation and testing: Methods such as the Hold-Out method and cross-validation are used to validate the model to ensure the generalization ability and prediction accuracy of the model.

[0194] 6) Model optimization:

[0195] Feature screening: Feature selection techniques (such as LASSO regression, principal component analysis (PCA), etc.) are used to screen the input data, and environmental features and crop growth features that are closely related to irrigation and fertilization requirements are retained.

[0196] Model tuning: The hyperparameters of the model are optimized through methods such as grid search and random search to improve the prediction accuracy.

[0197] 7) Prediction results:

[0198] Output: Based on the environmental data and crop growth characteristics, the model outputs the irrigation amount and fertilization amount at different growth stages. Specifically, the model predicts the water and nutrients required by the crops according to the real-time data and provides decision-making support for farmland management.

[0199] Feedback and adjustment: If the prediction error in a certain stage is large, the parameters of the model can be feedback-adjusted or the model can be retrained.

[0200] It should be noted that the present invention obtains the environmental data in the farmland in real time through the Internet of Things technology and conducts time series prediction based on these data. In different growth stages of crops, the influence of environmental factors (such as temperature, humidity, soil moisture, etc.) on crop requirements is different. By constructing a prediction model based on environmental data (such as regression model, neural network, machine learning model, etc.), the irrigation and fertilization requirements of crops can be dynamically predicted in different growth stages. This process can predict the most suitable irrigation and fertilization strategies through environmental parameters without directly measuring the crop growth characteristics by means of multi-dimensional analysis of environmental data, combined with the growth laws and requirements of crops. As the crops grow, the environmental conditions change continuously, and the prediction model will dynamically adjust its parameters to optimize the irrigation and fertilization strategies in real time. This not only depends on a single data point, but comprehensively considers the long-term environmental data trends to ensure that the irrigation and fertilization amounts are highly accurate and adaptable in different stages.

[0201] S225. From the currently selected parameters of all environmental data subsets, select the subset with the least impact on crop growth for iterative update, and record the screening result with the greatest impact on crop growth;

[0202] S226. Check whether the preset number of iterations is reached. If so, stop the screening and record the environmental data with the greatest impact on crop growth as the high-quality individual.

[0203] Specifically, the screening technology is the improved fireworks algorithm. In the basic fireworks algorithm, the number of explosion sparks generated by an individual and the size of the explosion radius are completely determined by the difference in the fitness values between this individual and other individuals. In the later stage of algorithm optimization, when the population approaches the optimal value, the numerator term of the explosion operator approaches 0, that is, the explosion radius tends to 0, and the subsequent optimization process is basically ineffective, and its local optimization ability has certain limitations. Therefore, some improvements have been made to the basic fireworks algorithm, that is, the generation of the explosion radius of the optimal individual in the fireworks population is dynamically changed, rather than simply changed according to the fitness value. The improved fireworks algorithm has greater local exploration ability compared with the basic fireworks algorithm. Because its optimization radius is dynamically changed and gradually reduced, the improved algorithm does not have the problem of ineffective optimization, and the accuracy of its local optimization is greatly improved, so as to achieve precise classification of environmental data in the behavioral feature space and improve the classification accuracy.

[0204] In this alternative embodiment, the key factors affecting crop growth are optimized using a feature optimization algorithm, which includes the following steps:

[0205] S231. Initialize the parameters of the feature optimization algorithm and the objective function of the factors affecting crop growth, define the initial optimization model, and set its objective function;

[0206] S232. Randomly initialize the feature subset configuration within the parameter space of the factors affecting crop growth, construct the initial optimization model, and calculate the corresponding objective function;

[0207] S233. Evaluate the initial optimization model using the training data, calculate the influence index of the current feature configuration on crop growth, and record the performance results;

[0208] S234. According to the adjustment rules of the feature optimization algorithm, adjust the current feature parameter combination. If the new feature combination has the best performance, update the current configuration; otherwise, update the global optimum and adjust the optimization model;

[0209] S235. If no optimal solution is found during the optimization process, expand the feature selection range, dynamically adjust the search strategy, eliminate the feature configuration with the least influence, and obtain the key factors that have the greatest influence on crop growth.

[0210] Specifically, the feature optimization algorithm is an improved lightning search algorithm, which is mainly implemented through a transitional discharge body, a spatial discharge body, and a leader discharge body. The present invention improves the lightning search algorithm by updating the position of the leader discharge body and changing the discharge body iteration method.

[0211] To facilitate the understanding of the above technical solution of the present invention, the following will detail how the present invention uses the feature optimization algorithm to optimize high-quality individuals and obtain the key factors affecting crop growth during the actual process.

[0212] Step 1. Initialize the parameters of the optimization algorithm and define the objective function:

[0213] 1) Initialize the parameters of the feature optimization algorithm, including the initial value of the feature combination, the setting of the objective function, and the coefficient for controlling the step size ( α and β ).

[0214] Suppose the key factors selected to affect crop growth include soil humidity ( g 1), air humidity ( g 2), and light intensity ( g 3), and the objective function is:

[0215] Y = ( g 1,g 2, g 3) = w 1 × g 1 + w 2 × g 2 + w 3 × g 3.

[0216] Among them, w 1, w 2, w 3 are the weights of each feature, and all weights are set to 1 during initialization.

[0217] For example:

[0218] T a = [0.45, 0.60, 300] represents the current feature parameter combination, which includes soil humidity, air humidity, and light intensity.

[0219] T * = [0.50, 0.65, 350] is the global optimal feature combination (assuming the global optimal configuration is these values).

[0220] T x = [0.48, 0.62, 320] and T m = [0.47, 0.61, 310] are randomly selected feature parameter combinations.

[0221] The coefficient for controlling the local search step size α = 0.8, and the coefficient for controlling the local search range β = 0.6.

[0222] Step 2: Randomly initialize the feature subset configuration, construct the initial optimization model, and calculate the corresponding objective function:

[0223] Randomly initialize the feature subset configuration and calculate the objective function. Select soil humidity ( g 1) and light intensity ( g 3) as the initial feature subset and calculate the objective function.

[0224] According to the formula of the adjustment rule, substitute the parameters:

[0225] = [0.45, 0.60, 300] + 0.8 × ([0.50, 0.65, 350] - [0.45, 0.60, 300]) + 0.6 × ([0.48, 0.62, 320] - [0.47, 0.61, 310]).

[0226] Calculate item by item:

[0227] =[0.45, 0.60, 300] + 0.8×[0.05, 0.05, 50] + 0.6×[0.01, 0.01, 10].

[0228] =[0.45, 0.60, 300] + [0.04, 0.04, 40] + [0.006, 0.006, 6].

[0229] =[0.496, 0.646, 346].

[0230] Therefore, the new feature combination is =[0.496, 0.646, 346].

[0231] Step 3: Evaluate the initial optimization model using the training data and calculate the impact index of the current feature configuration on crop growth:

[0232] Evaluate the impact of the current feature configuration on crop growth. Assume that the training data shows that the impacts of soil humidity, air humidity, and light intensity are respectively:

[0233] The impact of soil humidity on growth is 0.7.

[0234] The impact of air humidity on growth is 0.5.

[0235] The impact of light intensity on growth is 0.9.

[0236] Therefore, the current impact index is calculated as:

[0237] Impact index = 0.7×0.496 + 0.5×0.646 + 0.9×346.

[0238] The calculation result is:

[0239] Impact index = 0.346 + 0.323 + 311.4 = 311.676.

[0240] Step 4: According to the adjustment rules of the feature optimization algorithm, adjust the current feature parameter combination. If the new feature combination has the best performance, update the current configuration; otherwise, update the global optimum and adjust the optimization model:

[0241] If the new feature combination is better than the previous one in terms of performance, update the current configuration. If there is no improvement, update the global optimum configuration. Assume that during the adjustment process, it is found that the new feature combination brings a higher impact index (such as 311.676 is higher than the previous index), so update the current configuration.

[0242] Step 5: If the optimal solution is not found during the optimization process, expand the feature selection range, dynamically adjust the search strategy, eliminate the feature configuration with the least influence, and obtain the key factors that have the greatest impact on crop growth:

[0243] If the optimal solution is still not found during the optimization process, expand the feature selection range. Assume that the air temperature ( g 4) is introduced as a new feature, and the objective function is recalculated.

[0244] Assume that the new configuration after calculation is:

[0245] =[0.45, 0.60, 300, 25] + 0.8×([0.05, 0.65, 350, 27] - [0.45, 0.60, 300, 25]) + 0.6×([0.48, 0.62, 320, 26] - [0.47, 0.61, 310, 24]).

[0246] Finally, after adjustment, the new optimal feature configurations can be obtained, and these configurations have the greatest impact on crop growth.

[0247] It should be noted that when the optimization algorithm of the present invention expands the feature selection range, it not only uses the condition of "updating the configuration if the performance is better", but also sets clear evaluation indicators and convergence criteria. Specifically, the evaluation indicators may include prediction accuracy, stability of feature selection, and performance changes after each iteration. The algorithm will track these indicators. When the performance improvement is lower than a certain threshold, it is considered that the algorithm is close to the optimal solution, and the iteration is stopped to avoid entering an infinite loop. During the optimization process, the algorithm will perform local optimization through multiple iterations, and at the same time, the screening parameters will be adjusted in each iteration to find a better feature combination. When a new feature combination is selected each time, a performance evaluation will be carried out. If the performance of the new combination is better than the current combination, it will be updated to the new configuration. By continuously optimizing within a limited number of steps, the performance improvement each time ensures that the algorithm moves towards the optimal solution and completes the search within a fixed number of steps. The optimization algorithm of the present invention introduces a Gaussian mutation mechanism. In some iteration steps, if the algorithm cannot further find a better solution, or the performance improvement of the current feature combination is small, feature expansion will be performed according to the mutation range. This mutation not only changes the current feature combination, but may also explore new combination methods, thus jumping out of the local optimal solution and exploring a wider feature space to find a possible global optimal solution.

[0248] In this optional embodiment, the formula of the adjustment rule is:

[0249] ;

[0250] In the formula, represents the adjusted new combination of feature parameters; T a represents the current combination of feature parameters; T * represents the globally optimal combination of feature parameters; α represents the coefficient for controlling the local search step size; β represents the coefficient for controlling the local search range; T x and T m both represent randomly selected combinations of feature parameters.

[0251] In this alternative embodiment, a demand prediction model is constructed, and the demand prediction model is used to predict the future crop growth demand to obtain the optimal irrigation time and fertilization amount of the crops, including the following steps:

[0252] S31. Obtain the key factors affecting crop growth demand and divide them into a training set and a test set;

[0253] S32. Initialize the parameters of the demand prediction model and determine the key parameters for constructing the demand prediction model;

[0254] S33. Input the training set and the parameters of the initial demand prediction model into the demand prediction model, and perform iterative optimization based on the model optimization algorithm to output the optimal demand prediction model parameters;

[0255] S34. Based on the optimal demand prediction model parameters, construct the final demand prediction model, input the test set into the final demand prediction model for verification, and use the final demand prediction model to predict the future crop growth demand to obtain the optimal irrigation time and fertilization amount of the crops.

[0256] It should be explained that, first, it is necessary to obtain the key factors affecting crop growth, which usually include soil humidity, air temperature, light intensity, precipitation, etc. Then, historical data is collected, including the values of these key factors at different time points and their corresponding crop growth demand data.

[0257] Key factors: Assume that soil humidity, air temperature, light intensity, and precipitation are selected as the influencing factors.

[0258] Data division: The collected data is divided into a training set and a test set. Usually, the training set accounts for 70%-80%, and the test set accounts for 20%-30%. The training set is used for model training, and the test set is used for model verification and effect evaluation.

[0259] Select a suitable demand forecasting model and initialize the model's parameters. Common forecasting models include regression models, support vector machines (SVMs), neural networks, etc. Assume that a regression model is used as the demand forecasting model.

[0260] Initialize the model: Set the initial parameters of the regression model, such as initial weights, bias terms, etc.

[0261] Determine the key parameters: Determine the key parameters of the model, such as the learning rate, regularization coefficient, number of training iterations, etc. These parameters will have a significant impact on the training effect and prediction ability of the model.

[0262] Use a randomly generated set of parameters to initialize multiple different model scenarios. Each scenario corresponds to a set of parameters for the demand forecasting model, and these initial scenarios are used for subsequent model training and optimization.

[0263] Generate initial scenarios: For example, randomly initialize parameters such as weights, bias terms, learning rates, etc. to generate multiple candidate scenarios.

[0264] Scenario evaluation: Evaluate each initial scenario using the training set data, and calculate the prediction error (such as mean squared error MSE) or other evaluation metrics for each model.

[0265] Next, input the training set data into the demand forecasting model and iteratively optimize the initial parameters according to the optimization algorithm. Common optimization algorithms include gradient descent, genetic algorithms, particle swarm optimization, dandelion algorithm, etc.

[0266] Iterative optimization: Adjust the model's parameters (such as weights, biases, etc.) through the model optimization algorithm to gradually reduce the prediction error of the model. After each iteration, evaluate the performance of the current model and compare it with the historical best result.

[0267] Output the best parameters: Through multiple rounds of iteration, finally obtain a set of best demand forecasting model parameters. These parameters can minimize the prediction error and improve the accuracy of the model.

[0268] Finally, use the best parameters obtained through the optimization algorithm to construct the final demand forecasting model. Then, input the test set data into the final model for verification and evaluate its prediction performance.

[0269] Model verification: Verify the generalization ability of the model by calculating the prediction error of the model on the test set. If the model performs well on the test set, it indicates that it can effectively predict the demand for agricultural crops.

[0270] Predicting demand: Use the final demand prediction model to predict future moments and obtain the optimal irrigation time and fertilization amount for crops. For example, if the model predicts low soil moisture and high temperature in the future, it may recommend irrigation and increased fertilization at that moment.

[0271] It should be noted that in the present invention, the prediction demand model is actually a numerical prediction model based on environmental data and crop growth characteristics, and does not directly generate "natural language suggestions". The core function of this model is to predict the numerical values of irrigation and fertilization (such as irrigation amount, fertilization amount, time point, etc.) based on the input environmental data (such as soil moisture, temperature, etc.), rather than generating natural language suggestions. The following is a specific explanation:

[0272] 1) Difference between numerical prediction and suggestion generation

[0273] Numerical prediction: The model generates precise irrigation and fertilization amounts based on the input environmental data, such as soil moisture, temperature, light intensity, etc. For example: "At a future moment, it is recommended to irrigate 20 mm of water and fertilize 10 kg". This prediction is based on data analysis and the model results of environmental conditions.

[0274] Suggestion generation: If it is necessary to convert these prediction results into specific "suggestion" language expressions (such as: "According to the prediction, irrigation should be increased currently and more fertilizer should be applied at this time"), although a language model (such as LLM) can be used to generate natural language suggestions, this is not the core function or necessary requirement of this application.

[0275] 2) Core objective:

[0276] The core objective of the present invention is to accurately predict the irrigation and fertilization amounts required for crops to ensure the optimal allocation of farmland resources, and the generation of natural language expressions of suggestions is not an essential part of this model.

[0277] Model output: The output is specific numerical values of irrigation and fertilization amounts (such as: irrigate 10 L / ㎡, fertilize 5 kg / ㎡), and the farmland management system can automatically perform irrigation and fertilization operations based on these numerical values.

[0278] Automated execution: Without manual intervention or additional language processing models, the prediction results can be directly executed through a control system, such as an automated irrigation system or a fertilizer applicator.

[0279] 3) Optional function: Natural language generation (the role of LLM):

[0280] If it is necessary to provide natural language suggestions in the system (such as converting numerical results into easily understandable text information), a large language model (LLM) or a similar natural language processing module can be integrated. This part of the function belongs to an optional extension and is not an essential core function.

[0281] For example, on the user interface of the farmland management system, the system can generate natural language prompts based on the results predicted by the model, such as: "According to the current environmental prediction, the soil moisture is low and the temperature is high. It is recommended to immediately carry out irrigation and increase the fertilization amount." The implementation of this function can be completed by calling the LLM without affecting the implementation of the core prediction function.

[0282] It should be explained that the prediction model of the present invention processes and analyzes environmental data and crop growth characteristics, and generates irrigation amounts and fertilization amounts for specific moments based on certain model algorithms (such as regression models, machine learning models, etc.). This process includes:

[0283] 1) Input data:

[0284] Environmental data: such as soil moisture, temperature, light intensity, precipitation, etc.;

[0285] Crop growth data: such as the growth stage of the crop, leaf area index (LAI), chlorophyll content, etc.

[0286] 2) Prediction process:

[0287] The model learns from historical data to identify the relationship between environmental conditions, crop growth status, and irrigation and fertilization amounts, so as to predict the most appropriate irrigation and fertilization amounts given future environmental data. For example:

[0288] If it is predicted that the soil moisture is low, the model may recommend increasing irrigation;

[0289] If the temperature is high and the soil moisture is moderate, the model may recommend increasing the fertilization amount or adjusting the irrigation strategy.

[0290] 3) Output result:

[0291] The output will be the exact irrigation and fertilization amounts, which can directly control automated irrigation systems, fertilization systems, etc.

[0292] Final output: The model will provide irrigation and fertilization suggestions for a period of time in the future to ensure that crops obtain sufficient water and nutrients under the best growth conditions.

[0293] In this optional embodiment, inputting the parameters of the training set and the initial demand prediction model into the demand prediction model and performing iterative optimization based on the model optimization algorithm, and outputting the best demand prediction model parameters includes the following steps:

[0294] S331. Initialize the parameters of the demand prediction model and set the maximum number of iterations of the model optimization algorithm;

[0295] S332. Input the parameters of the training set and the initial demand prediction model into the model, calculate the error of the prediction result, and use it as the fitness value.

[0296] S333. Calculate the number of parameter configurations of the current demand prediction model according to the rules of the model optimization algorithm, and adjust the parameter configuration of the demand prediction model.

[0297] S334. Perform a quick sort on the current demand prediction model according to the fitness value, and classify the demand prediction models into the optimal configuration and the auxiliary configuration.

[0298] S335. Mutate and update the demand prediction model according to the fitness value to generate a new demand prediction model. Determine whether the model optimization algorithm has reached the maximum number of iterations. If so, stop and output the parameters of the best demand prediction model. Otherwise, return to step S343 to continue optimization until the maximum number of iterations is reached.

[0299] It should be noted that, first, initialize the relevant parameters of the demand prediction model and set the maximum number of iterations of the model optimization algorithm. Assume that the selected demand prediction model is a regression model. It is necessary to initialize the parameters of the model (such as weights and bias terms) and set the maximum number of iterations.

[0300] Initialize model parameters: Assume that a regression model is selected, then initialize the weights of the model (for example, the weights of soil humidity, temperature, and light intensity), as well as other model parameters (such as learning rate, bias, etc.).

[0301] Set the maximum number of iterations: Set the maximum number of iterations, such as 1000 times. This means that if the parameters of the model do not improve further after 1000 iterations, the optimization will stop.

[0302] Input the training set data and the parameters of the initial demand prediction model into the model, calculate the prediction result of the model, and calculate the error based on the prediction result. The error is usually measured by the mean square error (MSE) or other appropriate metrics.

[0303] Input the training set and initial parameters: Input the training set data (including soil humidity, temperature, light intensity, etc.) into the model and make predictions using the initialized model parameters.

[0304] Calculate the prediction error: Calculate the error (such as MSE) based on the difference between the predicted value and the actual value. The smaller the error, the closer the prediction result of the model is to the real data.

[0305] Fitness value: The error value will be used as the fitness value for the subsequent optimization process. The fitness value is used to measure the quality of the current parameter configuration of the model. The smaller the error, the higher the fitness value, and vice versa.

[0306] According to the rules of the model optimization algorithm, calculate the number of parameter configurations of the current demand prediction model, and adjust the model parameters according to the current fitness value.

[0307] Calculate the number of parameter configurations: In the optimization algorithm, the number of parameter configurations determines how many different model versions there are. Assuming the use of genetic algorithms or particle swarm optimization algorithms, this rule determines the number of models in each generation of iteration.

[0308] Adjust the model parameters: Adjust the parameters according to the current fitness value, usually through operations such as crossover and mutation in the algorithm (such as changing weight values, adjusting the learning rate, etc.). This process aims to optimize the performance of the model so that it can obtain better prediction results in the next iteration.

[0309] Sort all demand prediction models according to the fitness value. After sorting, the models are divided into optimal configurations and auxiliary configurations.

[0310] Quick sort: Perform a quick sort according to the fitness value of each model, and select the model with the highest fitness value as the optimal configuration.

[0311] Optimal configuration and auxiliary configuration classification: Classify the model with the highest fitness value as the "optimal configuration", while the models with lower fitness values are used as "auxiliary configurations". The optimal configuration will be key adjusted in the next step of optimization, while the auxiliary configuration is used as a backup plan.

[0312] Use the optimal configuration and auxiliary configuration to mutate and update the model to generate a new demand prediction model. Then, judge whether the maximum number of iterations has been reached. If so, stop the optimization and output the final best demand prediction model parameters.

[0313] Mutation and update: According to the optimal configuration, perform mutation operations (such as randomly adjusting some parameters of the model) and update operations (such as updating the weights of the model according to the fitness value). This process aims to further optimize the model.

[0314] Judge whether the maximum number of iterations has been reached: Check whether the current number of iterations has reached the preset maximum number. If the maximum number is reached, stop the optimization and output the current optimal model as the final demand prediction model.

[0315] Continue optimization: If the maximum number of iterations has not been reached, return to step S343 and continue to further optimize according to the current model parameters until the termination condition is met (i.e., the maximum number of iterations is reached).

[0316] Specifically, the model optimization algorithm is the dandelion optimization algorithm. The principle of the dandelion optimization algorithm is as follows: mimicking the seeding and reproduction characteristics of dandelions, it uses a diffuse parallel search for the optimal solution. The algorithm has high calculation accuracy and good robustness. In the dandelion optimization algorithm, the population consists of a group of dandelions, and each dandelion represents a feasible solution. These dandelions are divided into core dandelions and auxiliary dandelions according to a certain rule, and different types of dandelions adopt different seeding schemes. After seeding, both the parent and offspring dandelions are used as candidate solutions. By continuously seeding and selecting high-quality dandelions with high fitness values, when the convergence condition is finally met, the dandelion with the highest fitness in the dandelion population is regarded as the optimal solution.

[0317] In this alternative embodiment, the rule formula of the model optimization algorithm is:

[0318] ;

[0319] In the formula, P i represents the number of parameter configurations of the demand forecasting model at the i th iteration; P max and P min respectively represent the upper and lower limits of the number of parameter configurations; z represents the total cycle delay; y represents the delay imbalance degree; represents the maximum fitness of the total cycle delay; represents the minimum fitness of the total cycle delay; represents the average fitness of the total cycle delay; δ represents the threshold parameter; represents the maximum fitness of the delay imbalance degree; represents the minimum fitness of the delay imbalance degree; represents the average fitness of the delay imbalance degree.

[0320] According to another embodiment of the present invention, as Figure 2 shown, there is also provided an intelligent management system for agricultural planting based on the Internet of Things. The system includes:

[0321] A data acquisition module 1, which is used to acquire environmental data in the farmland and extract features from the environmental data to obtain environmental feature data;

[0322] A data analysis module 2, which is used to analyze the environmental feature data by using the factor analysis algorithm to obtain the key factors affecting the growth of crops;

[0323] The model prediction module 3 is used to build a demand prediction model and use the demand prediction model to predict the future crop growth demand, so as to obtain the optimal irrigation time and fertilization amount of the crops.

[0324] Among them, the data acquisition module 1 is connected to the model prediction module 3 through the data analysis module 2.

[0325] In summary, by means of the above technical solutions of the present invention, the present invention can identify the key environmental factors affecting crop growth by using the factor analysis algorithm to analyze the environmental characteristic data, helping agricultural managers understand which factors play a decisive role in crop growth. Combining these key factors, agricultural managers can make scientific decisions, carry out land management and crop planting in a targeted manner, avoid unnecessary resource consumption, and improve the success rate of crop growth. By building an accurate demand prediction model, the present invention can predict the future demand of crops for resources such as water and fertilizers, helping agricultural managers reasonably arrange irrigation and fertilization times. This precise regulation can effectively avoid over-irrigation and over-fertilization, not only improving the utilization efficiency of water resources and fertilizers, but also promoting the healthy growth of crops, thereby increasing the yield and quality of crops.

[0326] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. An intelligent management method for agricultural planting based on the Internet of Things, characterized in that, include: S1. Obtain environmental data in the farmland, and extract features from the environmental data to obtain environmental feature data; S2. Analyze the environmental characteristic data using factor analysis algorithm to obtain the key factors affecting crop growth; S3. Build a demand forecasting model, and use it to predict the growth demand of crops in the future, and obtain the optimal irrigation time and fertilizer amount for crops; The S2 includes: S21, obtaining a sample set of environmental characteristic data, and sorting the sample set to select a data subset related to crop growth characteristics; S22. Based on the screening technology, the data subsets are screened and the environmental data with the greatest impact on crop growth are selected as high-quality individuals; The S22 includes: S221, initializing screening parameters, each parameter represents a candidate environment data subset configuration; S222, evaluating the screening parameters of each environmental data subset, and selecting the environmental data subset with the greatest impact as the central environmental data subset; S223, adjusting the screening parameters of the non-central environmental data subset, and determining the number of new screening parameter sets to be generated according to the error of their impact on crop growth; S224, generating new environmental data subset parameters through Gaussian mutation, and generating a new screening parameter set according to the range of the mutation; S225. From all current environmental data subset screening parameters, select the subset with the least impact on crop growth for iterative update, and record the screening result with the greatest impact on crop growth; S226, checking whether the preset number of iterations has been reached, if so, stopping the screening, and recording the environmental data that has the greatest impact on the growth of crops as high-quality individuals; The S22 also includes: S23, using feature optimization algorithm to optimize high-quality individuals and obtain key factors affecting crop growth; The S23 includes: adjusting the current feature parameter combination according to the adjustment rules of the feature optimization algorithm, and if the new feature combination has the best performance, updating the current configuration; otherwise, updating the global optimum and adjusting the optimization model; if the optimal solution is not found during the optimization process, expanding the feature selection range, dynamically adjusting the search strategy, eliminating the feature configuration with the least influence, and obtaining the key factors that have the greatest impact on crop growth; The formula of the adjustment rule is: ； In the formula, represents the adjusted new feature parameter combination; T a represents the current feature parameter combination; T * represents the globally optimal feature parameter combination; α represents the coefficient for controlling the local search step size; β represents the coefficient for controlling the local search range; T x and T m both represent randomly selected feature parameter combinations; The model optimization algorithm is the dandelion optimization algorithm, and the rule formula of the model optimization algorithm is: ； Wherein, P i represents the number of parameter configurations of the demand prediction model at the i -th iteration; P max and P min respectively represent the upper and lower limits of the number of parameter configurations; z represents the total cycle delay; y represents the delay imbalance degree; represents the maximum fitness of the total cycle delay; represents the minimum fitness of the total cycle delay; represents the average fitness of the total cycle delay; δ represents the threshold parameter; represents the maximum fitness of the delay imbalance degree; represents the minimum fitness of the delay imbalance degree; represents the average fitness of the delay imbalance degree.

2. The intelligent management method for agricultural planting based on the Internet of Things according to claim 1, characterized in that, The step of obtaining environmental data in the farmland and extracting features from the environmental data to obtain environmental feature data comprises the following steps: S11, obtaining the original environmental data in the farmland for data decomposition, and using signal separation technology to filter high-frequency noise to obtain denoised farmland environmental data; S12, using a dimensionality reduction analysis method to decompose the denoised farmland environmental data, extract each environmental characteristic component, and perform smoothing processing; S13, calculating the correlation index between each extracted environmental characteristic component and the original farmland environmental data, selecting the characteristics with the highest correlation, and eliminating the environmental components with the lowest correlation; S14, integrating the processed environmental characteristic components into a complete environmental characteristic data set to obtain environmental characteristic data.

3. The intelligent management method for agricultural planting based on the Internet of Things according to claim 2, characterized in that, The denoised farmland environment data is decomposed using the dimensionality reduction analysis method, each environmental feature component is extracted, and the smoothing process includes the following steps: S121. Arrange the denoised farmland environment data in chronological order into a data matrix, and perform singular value decomposition on the data matrix; S122. Based on the decomposed data matrix, construct a linear mapping matrix, and perform dynamic mode decomposition on it to identify each environmental feature component related to the changes in the farmland environment; S123. Use low-pass filtering technology to smooth the identified key patterns to eliminate short-term fluctuations and noise.

4. The intelligent management method for agricultural planting based on the Internet of Things according to claim 1, wherein, Before adjusting the current feature parameter combination according to the adjustment rules of the feature optimization algorithm, if the new feature combination has the best performance, update the current configuration; otherwise, update the global optimum and adjust the optimization model, and it also includes: S231. Initialize the parameters of the feature optimization algorithm and the objective function of the influencing factors of crop growth, define the initial optimization model, and set its objective function; S232. Randomly initialize the feature subset configuration within the parameter space of the influencing factors of crop growth, construct the initial optimization model, and calculate the corresponding objective function; S233. Use the training data to evaluate the initial optimization model, calculate the influence index of the current feature configuration on crop growth, and record the performance results.

5. A smart management method for agricultural planting based on the Internet of Things according to claim 1, characterized in that, The steps of constructing a demand prediction model and using the demand prediction model to predict the crop growth demand at future times to obtain the optimal irrigation time and fertilization amount of the crops include the following steps: S31. Obtain the key factors affecting the crop growth demand and divide them into a training set and a test set; S32. Initialize the parameters of the demand prediction model and determine the key parameters for constructing the demand prediction model; S33. Input the training set and the parameters of the initial demand prediction model into the demand prediction model, and perform iterative optimization based on the model optimization algorithm to output the optimal parameters of the demand prediction model; S34. Based on the optimal parameters of the demand prediction model, construct the final demand prediction model, input the test set into the final demand prediction model for verification, and use the final demand prediction model to predict the crop growth demand at future times to obtain the optimal irrigation time and fertilization amount of the crops.

6. The intelligent management method for agricultural planting based on the Internet of Things according to claim 5, characterized in that, The steps of inputting the training set and the parameters of the initial demand prediction model into the demand prediction model and performing iterative optimization based on the model optimization algorithm to output the optimal parameters of the demand prediction model include the following steps: S331. Initialize the parameters of the demand prediction model and set the maximum number of iterations of the model optimization algorithm; S332. Input the training set and the parameters of the initial demand prediction model into the model, calculate the error of the prediction result as the fitness value; S333. Calculate the number of parameter configurations of the current demand prediction model according to the rules of the model optimization algorithm and adjust the parameter configuration of the demand prediction model; S334. Perform quick sorting on the current demand prediction model according to the fitness value and classify the demand prediction model into an optimal configuration and an auxiliary configuration; S335. Mutate and update the demand prediction model according to the fitness value to generate a new demand prediction model. Determine whether the model optimization algorithm has reached the maximum number of iterations. If so, stop and output the parameters of the optimal demand prediction model. Otherwise, return to step S343 to continue optimization until the maximum number of iterations is reached.

7. An intelligent management system for agricultural planting based on the Internet of Things, which is used to implement the intelligent management method for agricultural planting based on the Internet of Things described in any one of claims 1-6, and is characterized in that, The system includes: A data acquisition module, configured to acquire environmental data in the farmland and perform feature extraction on the environmental data to obtain environmental feature data; A data analysis module, configured to analyze the environmental feature data by using a factor analysis algorithm to obtain the key factors affecting the growth of crops; A model prediction module, configured to construct a demand prediction model and use the demand prediction model to predict the growth demand of crops at future times to obtain the optimal irrigation time and fertilization amount of the crops; Among them, the data acquisition module is connected through the data analysis module and the model prediction module.

Images