Astragalus sinicus cultivation management optimization system based on big data analysis

Abstract

The invention discloses an astragalus sinicus cultivation management optimization system based on big data analysis, and relates to the technical field of agricultural planting, the system comprises a field data acquisition module, a big data processing and analysis module, a cultivation scheme generation module, an intelligent regulation and control execution module, a user interaction module, a data storage and communication module and a system operation and maintenance module, and an interconnection architecture is constructed through an industrial Ethernet and LoRa; the field data acquisition module is provided with multiple types of sensors; big data analysis adopts edge and cloud collaboration, an edge end processes high-frequency data and divides regions, and the cloud end mines an association rule and identifies diseases and pests; the cultivation scheme generation module is combined with the database and the case to output an optimization scheme; the user interaction module provides a touch screen and APP visualization function; and edge caching and cloud distributed architecture are adopted for data storage. The problems that traditional cultivated varieties are blind in adaptation and disease and pest control is delayed are solved; through cooperation of multiple modules, precise seed selection, high-efficiency water and fertilizer, early prevention and control and whole-course tracing are realized, and the yield and quality are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of agricultural planting technology, and particularly relates to a cultivation management optimization system for Astragalus sinicus based on big data analysis. BACKGROUND

[0002] As high-quality green manure, forage crops and nectar plants, Astragalus sinicus plays a key role in improving soil fertility, reducing fertilizer use and developing ecological agriculture, and is widely planted in rice areas and temperate agricultural areas in southern China. The demand for large-scale planting of Astragalus sinicus is increasing, but the technical shortcomings of traditional cultivation and management modes have seriously restricted the improvement of its benefits. How to realize the precision and high-efficiency planting of Astragalus sinicus through technological innovation has become an urgent need in the industry.

[0003] The existing cultivation and management of Astragalus sinicus relies on the experience of farmers, lacks scientific data support, and leads to serious blindness in key links such as variety selection and water and fertilizer regulation. Variety selection is only based on local common types, without considering the adaptability of soil texture, climate conditions and planting purposes. For example, planting waterlogging-tolerant varieties in low-lying clay land is easy to cause seedling death, and selecting green manure varieties for forage demand leads to poor palatability. Water and fertilizer management is more extensive, irrigation is mostly based on weather and visual observation of soil moisture, drought and water shortage during seedling stage affect root development, and water accumulation during flowering stage leads to root rot; fertilization is implemented according to fixed proportion, without dynamic adjustment according to growth period demand, and problems such as phosphorus deficiency during seedling stage and potassium deficiency during pod setting stage are common, which not only causes resource waste, but also affects yield and quality.

[0004] Weak data perception and risk prevention and control capability further aggravate the cultivation difficulties. Traditional planting lacks systematic data collection means, and key data such as soil nutrients and environmental temperature and humidity are obtained late and scattered, which cannot support management decision-making. Disease and pest control relies on manual patrol, and it is difficult to detect diseases and pests such as powdery mildew and aphids at the initial stage of outbreak, and serious damage has been caused when they are found. The prevention and control scheme mostly relies on experience to select pesticides, which is easy to cause pesticide damage or exceed the standard of residue. Rotation planning lacks a long-term perspective, and continuous planting leads to soil phosphorus enrichment and pathogen accumulation, and the yield-increasing effect of subsequent crops decreases year by year. At the same time, the data of the whole cultivation process are not recorded, which cannot trace the management effect to optimize the subsequent scheme, and also cannot meet the requirement of transparency of the planting process for green agricultural certification.

[0005] These problems are superimposed on each other, leading to large fluctuations in Astragalus sinicus yield and unstable quality, and it is difficult to improve the planting efficiency. With the development of large-scale and standardized agriculture, traditional experience-based management cannot adapt to the high-efficiency planting demand of Astragalus sinicus, and it is urgent to build a cultivation and management technical system integrating precise perception, intelligent decision-making, dynamic regulation and whole-process tracing to solve the technical bottlenecks of current Astragalus sinicus planting. SUMMARY

[0006] The present invention proposes a big data analysis-based optimization system for the cultivation and management of milkvetch to solve the problems mentioned in the prior art.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: a big data analysis-based optimization system for the cultivation and management of milkvetch, comprising:

[0008] The field data acquisition module is equipped with soil parameter sensors, environmental sensors, and crop growth sensors. The soil sensors measure nitrogen, phosphorus, and potassium; the environmental sensors monitor temperature, humidity, light intensity, and precipitation; and the crop sensors detect plant height, leaf area index, and flowering rate.

[0009] The big data processing and analysis module uses edge computing nodes and cloud servers to work together to perform data filtering, normalization and anomaly removal, and divides the cultivation areas through K-means clustering; the cloud server uses a random forest regression algorithm and a CNN model to identify 12 types of diseases and pests of milk clover.

[0010] The cultivation plan generation module has a built-in database of milkvetch growth period. Combining real-time data with historical cases, it constructs a multi-objective optimization model with the goals of increasing yield, optimizing quality, and reducing costs.

[0011] The intelligent control and execution module supports Modbus-RTU and MQTT protocols, and can connect to integrated water and fertilizer equipment, intelligent seeders and plant protection drones;

[0012] The user interaction module is equipped with a 7-inch touch screen and a mobile APP, which supports cultivation data visualization, manual adjustment of plans, anomaly alarms, and query of historical plans;

[0013] The data storage and communication module adopts edge caching and cloud-distributed database, and the communication integrates 4G / 5G and LoRa. The data is encrypted using AES-128.

[0014] The system operation and maintenance module enables 24 / 7 status monitoring, tracks sensor online rate, equipment failure rate and solution execution success rate, automatically generates maintenance work orders after fault warning, and supports remote firmware upgrades and parameter backups.

[0015] Furthermore, it also includes a yield prediction submodule. This submodule constructs a three-dimensional data system encompassing historical, real-time, and predictive data, integrating nearly five years of plot yield data, hourly soil nutrient data, environmental parameters, cultivation management records, and soil microbial activity. It employs a spatiotemporal attention mechanism to optimize the prediction model. Calculate the predicted yield of milkvetch during the pod-setting stage, where Y(t) is the predicted yield at time t. For plot base yield, S(t') is the soil nutrient comprehensive value at t' time, E(t') is the environmental suitability at t' time, M(t') is the management measure in place rate at t' time, F(t') is the soil microbial activity value at t' time, a, b, c, d are the influence coefficients of soil, environment, management and microorganism respectively; the prediction model updates the parameters every 15 days, and combines with the measured biomass in the field to correct the deviation.

[0016] Further, it further includes a variety adaptation submodule, which constructs a feature database covering 20 mainstream varieties, and the subdivision parameters include waterlogging tolerance, cold tolerance, branching ability, fresh grass yield, growth period, stress resistance and utilization value; combined with the measured data of the plot: soil texture, organic matter content of 0-20 cm soil layer, annual mean temperature, annual precipitation, planting purpose and rotation mode, a three-level evaluation index system is constructed by using the analytic hierarchy process, the variety adaptation index is calculated, and the matching planting parameters: suitable sowing period, sowing density, sowing method and base fertilizer amount are generated simultaneously; mixed sowing suggestions are provided for complex plots.

[0017] Further, it further includes an irrigation optimization submodule, which dynamically decides based on the four-dimensional data of soil moisture, growth period, texture and weather; combined with the growth period characteristics of Chinese milk vetch, the irrigation strategy is refined: when the soil moisture is less than 60% during the seedling period, start light sprinkling irrigation; when the soil moisture is less than 70% during the branching period, start moderate irrigation; when the soil moisture is less than 75% during the flowering period, start precise drip irrigation; when the soil moisture is less than 65% during the podding period, start intermittent irrigation; combine with the future 72-hour weather forecast data, if the predicted precipitation is ≥10mm, delay irrigation for 24-48 hours, ≥20mm, cancel this irrigation, if accompanied by strong wind, advance irrigation and reduce irrigation amount by 10%; optimize the irrigation time window.

[0018] Further, it further includes a nutrient demand calculation submodule, which is based on the nutrient absorption law of Chinese milk vetch and the soil fertilizer supply characteristics, and refines the demand focus according to the growth period: seedling period focuses on phosphorus and molybdenum fertilizer; branching period considers nitrogen, phosphorus and potassium; flowering period needs balanced nitrogen, phosphorus and potassium; podding period strengthens potassium and phosphorus; according to the Calculate the nutrient demand amount of each growth period, wherein is the nutrient demand amount of the i-th growth period, is the nutrient demand coefficient of the i-th growth period, is the target biomass of the i-th growth period, is the nutrient loss rate of the i-th growth period, is the corresponding nutrient available content of the soil, is the nutrient utilization efficiency; generate the formula fertilizer scheme by adapting the fertilizer type: seedling period organic fertilizer accounts for 60%+ chemical fertilizer 40%, flowering period organic fertilizer accounts for 40%+ chemical fertilizer 60%; combined with the fertilization method optimization, phosphorus and potassium fertilizer is mainly base application, nitrogen fertilizer is mainly topdressing, and trace elements are foliar sprayed.

[0019] Further, it also includes an intelligent pest warning sub-module, which constructs a three-in-one identification system of image, spectrum and environment. The crop sensor collects leaf image and spectrum data every 2 hours, and the CNN model fuses image texture features and spectrum features. The risk correlation model is established by fusing hourly environmental data: the risk of powdery mildew increases when the temperature is 20-28℃, the humidity is above 75% and it rains continuously for 3 days; aphids are prone to outbreak when the temperature is 15-25℃ and the soil humidity is ≥80% after precipitation; the risk of rust disease increases when the light is insufficient for 5 days; four risk levels are set: low risk when the disease leaf rate is <5%, medium risk when the disease leaf rate is 5%-15%, high risk when the disease leaf rate is 15%-30%, and emergency risk when the disease leaf rate is >30%; a feedback closed loop of control effect is established, and the incidence of pests and diseases is reviewed 7 days after pesticide application. If the incidence of pests and diseases decreases by <50%, the type and dosage of pesticide are adjusted.

[0020] Further, it also includes a crop rotation planning sub-module, which combines the growth period of Astragalus sinicus and the annual variation of soil nutrients in the field. If the previous crop is indica rice, the recommended Astragalus sinicus variety is Yi Jiang Zi; if the previous crop is rapeseed, the recommended variety is Ping Ning 3. The following crops are optimized according to their needs: for japonica rice, 5 kg / acre of silicon fertilizer is applied after the Astragalus sinicus is turned over; for wheat, the turning over period is delayed to the early pod stage; an optimization strategy for crop rotation period is added. After 3 consecutive years of planting Astragalus sinicus, it is recommended to rotate with rapeseed for 1 year; the comprehensive benefits of crop rotation are calculated, including the increase in soil organic matter, the yield increase rate of the following crops, the reduction rate of chemical fertilizers, and the carbon sink amount.

[0021] Further, it also includes a cultivation risk assessment sub-module, which constructs a three-dimensional risk assessment system of weather, biology and soil. According to the risk assessment system, the risk assessment sub-module calculates the risk value of each risk factor and the comprehensive risk value of the whole growth period. The comprehensive risk value is calculated, where R is the comprehensive risk value, W(t') is the weather disaster risk at t', P(t') is the pest and disease risk at t', D(t') is the soil degradation risk at t', and h, e, f are the weight coefficients of weather, pests and diseases, respectively. The risk factors are calculated in detail: the low-temperature frost risk is combined with the minimum temperature in the next 48 hours, the rainstorm waterlogging risk is calculated according to the slope and precipitation of the field, and the drought risk is calculated according to the duration of soil moisture content <50%; targeted emergency solutions are pushed: cover with straw to keep warm before low temperature, dig temporary drainage ditches to prevent waterlogging before heavy rain, and start emergency drip irrigation during drought; increase pest and disease patrol frequency and spray leaf fertilizer to enhance resistance during medium risk; maintain routine management and monitor risk changes during low risk; generate an emergency material list simultaneously, and mark the procurement channel and usage method.

[0022] Further, it further includes an intelligent farming time reminding sub-module, which pushes a hierarchical reminder through a mobile APP based on the milk vetch growth period accumulated temperature model and real-time growth data in advance by 3 days: the basic reminder contains node name and optimal operation period; the detailed reminder contains operation points, required equipment and materials; the farming time node can be dynamically adjusted, if extreme weather is encountered, the seedling period is automatically extended to 37 days, the fertilizer period is synchronously delayed and the phosphorus fertilizer usage is increased by 10%, the corresponding adjustment is made for the rolling period; an operation quality monitoring mechanism is added, the sowing uniformity is detected through unmanned aerial vehicle aerial photography and ground sampling after sowing, and the soil nutrient change is detected after fertilization.

[0023] Further, it further includes a cultivation data traceability sub-module, which constructs a three-dimensional traceability system of land, time and operation, automatically collects and correlates full growth period data: land basic information, sowing information, field management records, environmental data, growth monitoring data and harvesting data; the data are all attached with unique land identification and time stamp, the core data are stored by using a block chain technology; multi-channel traceability query is supported; a visual traceability atlas is generated, the key nodes in the growth period are displayed in the form of a time axis, the corresponding data, images and operation records can be viewed by clicking the nodes; meanwhile, the national agricultural product quality and safety traceability management information platform is interfaced, and authoritative data support is provided for milk vetch green fertilizer certification and organic agricultural subsidy application.

[0024] Compared with the existing technology, the beneficial effects of the present application are:

[0025] Through multi-module collaborative innovation and technical scheme optimization, the present application solves the core pain points of traditional milk vetch cultivation variety adaptation blindness, water and fertilizer regulation roughness, risk prevention and control lag and management traceability, constructs a full-cycle precision cultivation management system, and significantly improves the scientificity and benefit of milk vetch planting.

[0026] The upgrading of data perception and analysis capability lays a solid foundation for accurate decision-making. The field data acquisition module realizes real-time and accurate acquisition of multi-dimensional data of soil, environment and crops, combines fault self-diagnosis and supplement transmission mechanism, and ensures data integrity and accuracy. The edge and cloud collaborative architecture realizes efficient data processing, the edge quickly filters out abnormalities and divides cultivation zones, and the cloud deeply mines the correlation rules of soil, environment and yield, and provides data support for management scheme optimization. The yield prediction sub-module integrates multi-dimensional data and integral algorithm, and predicts the yield trend in advance, so that the cultivation strategy can be adjusted in time to avoid the loss of benefit caused by lagging management, and solve the problem of traditional planting relying on experience and lacking data from the source.

[0027] The precision and dynamic of the cultivation scheme realize efficient use of resources and yield and quality improvement. The variety adaptation sub-module constructs a multi-variety characteristic database, calculates an adaptation index in combination with field conditions and planting purposes, accurately recommends varieties and supporting parameters, solves the problem of blindness in variety selection, and significantly improves resistance and yield potential. The irrigation optimization sub-module dynamically decides based on soil moisture, growth period and meteorological data, refines irrigation strategies and time windows at each stage, reduces water waste, and at the same time guarantees crop water demand. The nutrient demand calculation sub-module refines nutrient priorities according to the growth period, generates a formula fertilizer scheme in combination with soil fertility characteristics, adapts to fertilizer types and application methods, avoids excess or insufficient nutrients, improves fertilizer utilization rate, and prevents soil degradation.

[0028] The strengthening of risk prevention and control and long-term management capability reduces planting losses and improves sustainability. The disease and pest intelligent early warning sub-module fuses image, spectrum and environmental data, realizes early identification and risk grading of diseases and pests, recommends biological or chemical control schemes, significantly improves the timeliness and accuracy of prevention and control, and reduces yield loss and pesticide residues. The cultivation risk assessment sub-module integrates meteorological, biological and soil risks to generate emergency schemes and material lists, which can respond to disasters such as low temperature and heavy rain in advance and reduce the degree of loss. The rotation planning sub-module generates a whole cycle scheme in combination with the needs of the previous and next crops, optimizes the rotation period and management measures, avoids soil degradation, improves long-term planting efficiency, and adapts to the needs of ecological agricultural development.

[0029] Whole-process traceability and agricultural time reminding function further perfect the cultivation management closed loop. The agricultural time intelligent reminding sub-module accurately pushes key node operation reminders based on the growth period model and real-time growth data, supports dynamic adjustment and quality monitoring, and ensures timely and standardized agricultural time operation. The cultivation data traceability sub-module records whole growth period data and stores them using blockchain, supports multi-channel query and visual display, provides basis for subsequent cultivation optimization, and interfaces with the quality and safety traceability platform to improve product market acceptance.

[0030] Overall, the present application constructs a precise perception, intelligent decision-making, dynamic regulation and whole-process traceability Chinese milk vetch cultivation management technology system, effectively solves the problem of extensive traditional planting, improves yield and quality and resource utilization efficiency, enhances risk resistance and planting sustainability, and provides strong technical support for large-scale and green Chinese milk vetch planting. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 A schematic block diagram of the Chinese milk vetch cultivation management optimization system based on big data analysis proposed by the present application;

[0032] Figure 2 A comparison bar chart of adaptation indexes for different varieties of Chinese milk vetch;

[0033] Figure 3A broken line graph for the change trend of Astragalus sinicus yield prediction error;

[0034] Figure 4 A combination chart for water use efficiency comparison of different irrigation strategies;

[0035] Figure 5 A multi-dimensional radar chart for disease and pest control effect. DETAILED DESCRIPTION

[0036] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0037] In the description of the present application, it should be understood that the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise" and the like indicate the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application.

[0038] In addition, the terms "first", "second" are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the technical features indicated. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "multiple" is two or more, unless otherwise explicitly and specifically limited. In addition, the terms "mounting", "connecting", "connection" should be broadly understood, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium, or it can be the communication between two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances, and the present application will be further described in detail with reference to the drawings.

[0039] Reference Figures 1 to 5 : An Astragalus sinicus cultivation management optimization system based on big data analysis, comprising a field data acquisition module, a big data processing and analysis module, a cultivation scheme generation module, an intelligent control execution module, a user interaction module, a data storage and communication module, and a system operation and maintenance module;

[0040] The field data acquisition module deploys soil parameter sensors, environmental sensors, and crop growth sensors. The soil sensors measure nitrogen 0-500 mg / kg, phosphorus 0-200 mg / kg, and potassium 0-600 mg / kg, with an accuracy of ±5 mg / kg. The soil moisture content measurement range is 0-100%, with an accuracy of ±2%. The environmental sensors monitor temperature -10-50°C (accuracy ±0.3°C), humidity 20-90% RH (accuracy ±3% RH), light 0-200000 lux (accuracy ±5%), and precipitation 0-50 mm / h (accuracy ±0.1 mm). The crop sensors detect plant height 0-150 cm (accuracy ±0.5 cm), leaf area index 0-8 (accuracy ±0.1), and flowering rate 0-100% (accuracy ±1%). The sampling frequency is adjustable from 1-30 Hz, and supports fault self-diagnosis and data transmission.

[0041] The big data processing and analysis module uses edge computing nodes (computing power ≥15 TOPS) and cloud servers to work together. The edge nodes process high-frequency sampling data (≥10 Hz) in real time, perform data filtering, normalization, and abnormality removal, and divide the cultivation zones through K-means clustering. The cloud server uses random forest regression algorithm to mine the soil-environment-yield correlation rules, uses CNN model to identify 12 types of milk vetch diseases and pests (identification accuracy ≥96%), and generates cultivation status analysis report daily.

[0042] The cultivation scheme generation module has a milk vetch growth period database (covering seedling stage, branching stage, flowering stage, and pod setting stage parameters), combines real-time data and historical cases (1000+ plots of data), and constructs a multi-objective optimization model with yield improvement, quality optimization, and cost reduction as the target. It outputs seeding amount 2-3 kg / acre, fertilizer ratio (phosphorus and potassium as the main), irrigation amount 20-30 / acre, and disease and pest control scheme, with single scheme generation time ≤200 ms.

[0043] The intelligent control execution module supports Modbus-RTU and MQTT protocols, connects water and fertilizer integrated equipment (flow 0-50 / h, accuracy ±1 / h), intelligent seeding machine (seeding depth 1-3 cm adjustable), and plant protection unmanned aerial vehicle (load 10 kg, endurance 30 min), and can accurately execute irrigation, fertilization, and pesticide application operations according to the zoning, with control deviation ≤5%.

[0044] The user interaction module is equipped with a 7-inch touch screen (resolution 1280×720) and a mobile APP, which supports cultivation data visualization (plot heat map, growth curve), scheme manual adjustment, abnormal alarm (response time ≤2 s), and historical scheme query (supports data tracing for nearly 3 years).

[0045] The data storage and communication module uses edge caching (local storage of 15 days of data) and a cloud-based distributed database (capacity ≥30TB, read / write speed ≥150MB / s). Communication integrates 4G / 5G (upload speed ≥5Mbps) and LoRa (transmission distance ≥3km). Data is encrypted using AES-128, and the packet loss rate is ≤0.2%.

[0046] The system operation and maintenance module enables 24 / 7 status monitoring, tracking sensor online rate (≥98%), equipment failure rate (threshold ≤2%), and solution execution success rate (≥99%). After a fault warning, it automatically generates a maintenance work order (dispatch response ≤15 minutes) and supports remote firmware upgrades and parameter backups.

[0047] This invention also includes a yield prediction submodule. This submodule constructs a three-dimensional data system encompassing historical, real-time, and predictive data, integrating plot yield data from the past five years, hourly soil nutrient data (nitrogen, phosphorus, potassium, and organic matter), environmental parameters (temperature, humidity, light, accumulated temperature), cultivation management records (sowing amount, water and fertilizer application, number of pest control treatments), and soil microbial activity (number of actinomycetes and nitrogen-fixing bacteria). It employs a spatiotemporal attention mechanism to optimize the prediction model. Calculate the predicted yield of milkvetch during the pod-setting stage, where Y(t) is the predicted yield at time t (in kg / mu). The base yield of the plot is 90% of the average yield of the past 3 years, in kg / mu. S(t') is the comprehensive value of soil nutrients at time t' (normalized 0-1, nitrogen, phosphorus and potassium weights 0.4, 0.3 and 0.3, respectively). E(t') is the environmental suitability at time t' (normalized 0-1, accumulated temperature ratio 0.5, humidity 0.3, light 0.2). M(t') is the implementation rate of management measures at time t' (normalized 0-1, calculated according to the deviation of the plan implementation). F(t') is the soil microbial activity value at time t' (normalized 0-1, nitrogen-fixing bacteria ratio 0.6). a, b, c and d are the influence coefficients of soil, environment, management and microorganisms, respectively (the sum is 1, seedling stage a=0.4, b=0.3, c=0.2, d=0.1, flowering stage a=0.3, b=0.4, c=0.2, d=0.1). The prediction model updates its parameters every 15 days and corrects deviations by combining field-measured biomass (plant height, fresh weight). When the error between the predicted and measured values ​​exceeds 8%, it automatically incorporates similar data from neighboring plots to optimize the calculation, predicting yield 45 days in advance with a prediction accuracy of ≥92%. This provides a precise basis for dynamic adjustments to cultivation plans (such as additional fertilization or irrigation adjustments).

[0048] In the application, a variety adapter module is also included, which builds a characteristic database covering 20 mainstream varieties such as Minzi No. 6, Yujiang Daye, Yijiang Seed, and Pinning No. 3. The subdivided parameters include waterlogging tolerance (1-5 levels), cold tolerance (-5℃ / -8℃ / -10℃ tolerance levels), branching ability (5-15 branches per plant), fresh grass yield (1500-3000 kg / mu), growth period (80-120 days), stress resistance (powdery mildew / aphid resistance levels), and utilization value (green manure type / forage type / honey source type / multi-purpose type). Combined with the actual measurement data of the land: soil texture (sandy loam, loam, clay, sandy clay, and silt loam), 0-20 cm soil organic matter content (10-40 g / kg), annual mean temperature (10-25℃), annual precipitation (800-1800 mm), planting purpose (green manure for soil improvement, forage for livestock, and honey production), and rotation mode (rice-vetch, oil-vetch, etc.), a three-level evaluation index system (target layer: adaptation index; criterion layer: environmental adaptation, use adaptation, and cost adaptation; index layer: 12 specific parameters) is constructed by using the analytic hierarchy process, and the variety adaptation index (0-100) is calculated. Varieties with an index of 80 or more are included in the recommended list, and the supporting planting parameters are generated simultaneously: suitable sowing period (late September to early November, determined by daily mean temperature of 15-20℃), sowing density (15-20 thousand plants per mu, clay should be sparse, sandy loam should be dense), sowing method (broadcasting / strip sowing, strip sowing spacing 20-30 cm), and base fertilizer amount (15-20 kg of superphosphate per mu). For complex land, mixed sowing is recommended, such as Minzi No. 6 (waterlogging tolerance) and Yujiang Daye (high yield) mixed at a ratio of 3:2 in low-lying clay land to improve stress resistance and yield, solving the problem of relying on experience and poor adaptability in traditional variety selection.

[0049] In the application, an irrigation optimization submodule is also included, which dynamically decides based on soil moisture, growth period, texture, and meteorological four-dimensional data, and real-time collects soil moisture data of 5m*5m grid. The land soil moisture thermal map is generated by Kriging interpolation to avoid single point sensor data deviation. Combined with the growth characteristics of vetch, the irrigation strategy is refined: when the soil moisture is less than 60% during the seedling stage (1-30 days after emergence), light sprinkling irrigation is started (20 / mu, water spraying intensity 10 mm / h, avoiding washing out seedlings); when the soil moisture is less than 70% during the branching period (31-60 days), moderate irrigation is started (25 / mu for sandy loam, 28 / mu for clay); when the soil moisture is less than 75% during the flowering period (61-90 days), precise drip irrigation is started (30 / mu, drip head flow 2 L / h, wetting depth 15-20 cm); when the soil moisture is less than 65% during the podding period (91-120 days), intermittent irrigation is started (22-25 / acre, divided into 2 irrigations, 4 hours apart. Integrating 72-hour weather forecast data, if predicted rainfall is ≥10mm, irrigation is delayed by 24-48 hours; if ≥20mm, irrigation is cancelled. If accompanied by strong winds, irrigation is advanced and the irrigation volume is reduced by 10%. The irrigation time window is optimized, choosing 5-7 am or 5-7 pm to reduce water evaporation loss (evaporation rate reduced by more than 30%). Irrigation water use efficiency (IWUE = crop water consumption or total irrigation volume) is calculated in real time. When IWUE is below 0.7, the cause is analyzed (e.g., dripper blockage, uneven soil moisture) and optimization suggestions are pushed, such as cleaning drippers, adjusting irrigation zones, and continuously optimizing the irrigation cycle (adjustable every 5-10 days) to achieve precise and efficient water resource utilization.

[0050] This invention also includes a nutrient requirement calculation submodule. This submodule, based on the nutrient absorption patterns of milkvetch and soil fertility characteristics, constructs a dynamic nutrient requirement model, refining the key requirements according to the growth stage: During the seedling stage, the focus is on phosphorus (to promote root development, phosphorus accounting for 40%) and molybdenum fertilizer (nitrogenase activity, 0.1 kg / mu); during the branching stage, nitrogen, phosphorus, and potassium are considered (nitrogen 25%, phosphorus 35%, potassium 40%), with supplemental boron fertilizer (0.05 kg / mu, to prevent flowering without fruiting); during the flowering stage, a balanced nitrogen, phosphorus, and potassium ratio is required (approximately 30% each); during the pod-setting stage, potassium (45%) and phosphorus (30%) are strengthened to improve pod setting rate and grain plumpness. Calculate the nutrient requirements for each growth stage, among which This represents the nutrient requirement for the i-th growth stage (unit: kg / mu). The nutrient requirement coefficient for the i-th growth stage is 0.8 for seedling stage, 1.2 for branching stage, 1.5 for flowering stage, and 1.3 for pod-setting stage. The target biomass for the i-th growth stage is (95% of the highest historical biomass for the same growth stage, adjusted based on the environmental forecast for the year, in kg / mu). Nutrient loss rate during the i-th reproductive period (before rainy days) =0.3, sunny day =0.1), The available nutrient content in the soil (unit: mg / kg, converted to kg / acre). Nutrient utilization efficiency is set at (phosphorus 60%, potassium 75%, nitrogen 55%, molybdenum 90%). When generating the formulated fertilizer, the appropriate fertilizer types are: 60% organic fertilizer (well-rotted farmyard manure) + 40% chemical fertilizer during the seedling stage; 40% organic fertilizer + 60% chemical fertilizer during the flowering stage, avoiding excessive chemical fertilizer leading to soil compaction. Combined with optimized fertilization methods, phosphorus and potassium fertilizers are primarily applied as basal fertilizer (70%), nitrogen fertilizer is primarily applied as top dressing (divided into two applications, 15 days apart), and micronutrients are applied via foliar spraying (concentration 0.1-0.2%) to ensure sufficient nutrient absorption and avoid abnormal growth caused by excess or deficiency.

[0051] The present application also includes a pest intelligent early warning sub-module, which constructs a three-in-one identification system of image, spectrum and environment. The crop sensor collects leaf images (resolution 20 million pixels) and spectrum data (400-1000 nm waveband) every 2 hours. The CNN model fuses image texture features (white powder layer of powdery mildew, body morphology of aphids) and spectrum features (15% decrease in reflectivity of powdery mildew leaves at 550 nm waveband, 20% increase in reflectivity of aphids at 700 nm waveband), and identifies 15 types of diseases and pests such as powdery mildew, rust, aphids and thrips, with an identification accuracy of ≥97% and an identification time of ≤5s. The risk correlation model is established by fusing hourly environmental data: the risk of powdery mildew increases when the temperature is 20-28℃ and the humidity is above 75% for 3 consecutive rainy days; aphids are prone to outbreak when the temperature is 15-25℃ and the soil humidity is ≥80% after precipitation; the risk of rust increases when the light is insufficient (daily average <4h) for 5 consecutive days. Four risk levels are set: disease leaf rate <5% is low risk, daily patrol and leaf cleaning suggestions are pushed; 5%-15% is medium risk, biological control (ladybugs control aphids, trichoderma control powdery mildew) is recommended, and the amount of natural enemy release (1000 heads / acre of ladybugs) and the use time (evening without strong wind) are specified; 15%-30% is high risk, a chemical control scheme is generated (fungicide type: kresoxim-methyl for powdery mildew, imidacloprid for aphids; dosage: 100ml / acre; application time: morning or evening, avoiding the honey source period); >30% is an emergency risk, and the plant protection unmanned aerial vehicle is preferentially applied, and the artificial auxiliary control suggestion is pushed simultaneously. A feedback loop of control effect is established, and the incidence of diseases and pests is reviewed 7 days after application. If the decrease is <50%, the type and dosage of pesticide are adjusted to ensure that the control accuracy is ≥95%.

[0052] In the application, a crop rotation planning submodule is also included, which collects cultivation data (growth period, fertilizer consumption characteristics, and residual stubble nutrients) of 10 types of previous crops such as rice (indica rice / japonica rice), rapeseed, wheat, and corn, combines the growth period (80-120 days) of Astragalus sinicus with the annual variation law (annual increase / decrease range of organic matter, nitrogen, phosphorus, and potassium) of soil nutrients in the plot, and generates a whole-cycle rotation scheme of previous crops, Astragalus sinicus, and subsequent crops. If the previous crop is indica rice (high nitrogen consumption and low nitrogen content in residual stubble), the recommended Astragalus sinicus variety is Yi Jiang Zi (moisture-tolerant and strong nitrogen-fixing ability), the seeding rate is 3 kg / mu, superphosphoric acid calcium 8 kg / mu is added, the plowing period (flowering period, flowering rate 70%) is 15-20 days apart from rice transplanting, the plowing depth is 15 cm, and the soil nitrogen content is increased by 10-15%; if the previous crop is rapeseed (high potassium consumption and low potassium content in residual stubble), the recommended variety is Pingning No. 3 (fertilizer-tolerant and high branching), the seeding rate is 2.5 kg / mu, potassium chloride 7 kg / mu is added, the plowing period is advanced by 5-7 days (flowering rate 60%), and after plowing, lime is applied to adjust the pH value (suitable for clay soil). Optimization according to the needs of subsequent crops: for japonica rice, 5 kg / mu of silicon fertilizer is added after Astragalus sinicus is plowed; for wheat, the plowing period is delayed to the early pod stage to increase soil organic matter content. Add rotation period optimization strategy: after planting Astragalus sinicus for 3 consecutive years, it is recommended to rotate with rapeseed for 1 year to avoid soil phosphorus enrichment and pathogen accumulation. Calculate the comprehensive benefits of rotation: soil organic matter increase (0.5-1 g / kg per year), subsequent crop yield increase rate (8-15%), fertilizer reduction rate (15-20%), and carbon sink amount (200-300 ), and generate a visual rotation benefit report to provide data support for long-term cultivation planning and green agriculture certification.

[0053] In the application, a cultivation risk assessment submodule is also included, which constructs a three-dimensional risk assessment system of "meteorology-biology-soil", and calculates the risk value of each factor according to the weight of each factor The comprehensive risk value is calculated throughout the whole growth period, wherein R is the comprehensive risk value (0-10, >6 is high risk, 4-6 is medium risk, and <4 is low risk), W(t') is the meteorological disaster risk at t' (normalized 0-1, 0.4 for low-temperature freezing damage, 0.3 for rainstorm waterlogging, and 0.3 for drought), P(t') is the disease and pest risk at t' (normalized 0-1, calculated according to the occurrence rate and damage level of diseases and pests), D(t') is the soil degradation risk at t' (normalized 0-1, 0.5 for the decrease rate of organic matter, 0.3 for the increase of soil bulk density, and 0.2 for acidification), and h, e and f are the weight coefficients of meteorology, diseases and pests and soil respectively (the sum is 1, h=0.4, e=0.3 and f=0.3 in the seedling stage, and h=0.3, e=0.4 and f=0.3 in the flowering stage). The calculation of each risk factor is refined: the low-temperature freezing risk is combined with the minimum temperature in the next 48 hours (≤0℃ for 2 hours, the risk is increased to 0.8), the rainstorm waterlogging risk is calculated according to the land slope (<5°, the risk is 0.7, and >15°, the risk is 0.3) and the precipitation (24 hours >50mm, the risk is 0.9), and the drought risk is calculated according to the duration of the soil moisture content <50% (≥7 days, the risk is 0.8). The risk spatial distribution map is generated, the key areas such as low-lying areas (high rainstorm risk) and heavy clay soil areas (high degradation risk) are marked, and the emergency plan is pushed in a targeted manner: in the high risk period, cover straw (thickness 3-5cm) to keep warm before low temperature, dig temporary drainage ditch (depth 30cm, spacing 5m) to prevent waterlogging before rainstorm, and start emergency drip irrigation (preferably in the flowering period) in drought; in the medium risk period, increase the disease and pest patrol frequency (once a day), and spray leaf fertilizer to enhance the stress resistance; in the low risk period, maintain the routine management and monitor the risk change. The emergency material list is generated synchronously, the required straw, drainage pump, antifreeze, fungicide and the like are listed in the high risk period, the procurement channel and use method are marked, and the disaster loss rate is reduced (decrease ≥40%).

[0054] In the present application, an intelligent agricultural time reminding sub-module is further included, which is based on the Astragalus sinicus growth period accumulated temperature model (seedling stage accumulated temperature 200-300 , branching stage 300-400 , flowering stage 400-500 , and pod setting stage 300-400 ) and real-time growth data (plant height, leaf age), accurately calculate 8 key agricultural time nodes such as sowing period, emergence period, branching period, fertilization period, irrigation period, prevention period and turning period, and push hierarchical reminders through mobile APP 3 days in advance: basic reminders include node name, best operation period (such as sowing period on cloudy days or evening); detailed reminders include operation points (sowing depth 1-3cm, covering 0.5-1cm), required equipment (intelligent sowing machine, harrowing machine) and materials (seed, base fertilizer, pesticide); risk reminders include operation taboos (rainy days do not sow, windy days do not spray). The agricultural time node can be dynamically adjusted. If extreme weather (such as low temperature during seedling period leading to growth delay for 7 days) is encountered, the seedling period is automatically extended to 37 days, the fertilization period is synchronously delayed and the phosphorus fertilizer usage is increased by 10%, and the turning period is adjusted accordingly. Join the operation quality monitoring mechanism, after sowing, the sowing uniformity is detected by unmanned aerial vehicle aerial photography and ground sampling (deviation > 10% to push the reseeding suggestion), and after fertilization, the soil nutrient change is detected (not reaching the target value to push the supplementary fertilization scheme). Support multi-person collaboration reminder, divide large plots into responsibility fields according to regions, push exclusive agricultural time tasks to corresponding farmers, and synchronously display completion progress (completed, to be completed, overdue), to ensure accurate, timely and standardized agricultural operation.

[0055] In the present application, a cultivation data traceability sub-module is also included, which constructs a three-dimensional traceability system of plot, time and operation, automatically collects and correlates full growth period data: plot basic information (soil type, area, latitude and longitude, altitude, previous crop), sowing information (variety, seed source, sowing time, amount, depth, method, base fertilizer amount), field management records (irrigation time, amount, method and fertilization time, type, dose, pesticide application time and pesticide, dose, executor and cultivation and weeding records), environmental data (hourly temperature and humidity, precipitation, light, accumulated temperature), growth monitoring data (plant height, leaf area index, flowering rate, weekly measured values of biomass and unmanned aerial vehicle aerial image) and harvesting data (yield, fresh-dry ratio, nutrient content, turning time and depth). Data are all attached with unique plot identifier (two-dimensional code) and time stamp, and core data (sowing, fertilization and pesticide application records) are stored using blockchain technology to ensure non-tamperability. Support multi-channel traceability query: farmers can view plot full cycle data through APP code scanning, agricultural technicians can export traceability reports through Web, and regulatory departments can retrieve authentication data through special interface. Generate visual traceability atlas to display key nodes during growth period in the form of time axis, and click the node to view the corresponding data, image and operation record. Data is saved for ≥5 years, and is connected to the national agricultural product quality and safety traceability management information platform, to provide authoritative data support for purple vetch green manure authentication and organic agricultural subsidy application, and to improve product market recognition.

[0056] The specific implementation of the system is further illustrated by two embodiments as follows:

[0057] Example 1: Rice-vetch rotation green manure type cultivation in southern rice area (Hengyang, Hunan, previous crop indica rice)

[0058] This example is aimed at a 200 mu rice-vetch rotation plot (soil type clay, 0-20 cm organic matter 25 g / kg, annual mean temperature 18℃, annual precipitation 1400 mm) in an ecological farm in Hengyang, Hunan, the previous crop was indica rice (high nitrogen consumption), the purpose of planting was to improve the soil with green manure, and the traditional experience was poor in variety adaptation and blind irrigation, with fresh grass yield of only 1600 kg / mu, and soil organic matter increased by less than 0.3 g / kg per year. The system of the present application is used for precise cultivation, and the specific implementation is as follows.

[0059] 1. System deployment and parameter configuration

[0060] The field data acquisition module deploys sensors in a 5m x 5m grid: soil sensors (TDR-300) are buried 15 cm deep to monitor nitrogen, phosphorus, potassium, and soil moisture; environmental sensors (SHT35) are installed on 1.2m high poles, one group per 10 mu; crop sensors (MS-120) are fixed on the edges of the plot, one group per 20 mu, with a default sampling frequency of 2Hz, which increases to 30Hz when the soil moisture changes by more than 10%.

[0061] The cultivation scheme generation module loads 20 variety characteristic databases, and the intelligent control execution module connects water and fertilizer integrated equipment (ZNX-100, flow 0-50 / h), intelligent seeding machine (2BXF-10, seeding depth 1-3cm) and plant protection unmanned aerial vehicle (T20, load 10kg). The user interaction module is equipped with a 7-inch touch screen in the control room, and farmers are equipped with a mobile APP (supports Android / iOS). The data storage edge node is equipped with a 512GB SSD, and the cloud is equipped with a 30TB distributed database, with LoRa (transmission distance 3km) + 4G dual link communication.

[0062] 2. Core process implementation details

[0063] 2.1 Variety adaptation and seeding planning

[0064] The variety matching submodule inputs the following plot data: clay soil, organic matter 25g / kg, average annual temperature 18℃, previous crop indica rice, and green manure usage. The analytic hierarchy process (AHP) calculates the matching index: Minzi 6 85 points, Yujiang Daye 82 points, and Yijiangzi 78 points. Minzi 6 is recommended as the main variety, with 20% Yujiang Daye (flood-tolerant) mixed in. The sowing parameters are generated: sowing on September 25th (average daily temperature 18℃), sowing rate 3kg / mu, row spacing 25cm, and basal fertilizer superphosphate 18kg / mu. The seeder executes according to the parameters, and drone aerial photography is used to check uniformity; a deviation of 5% meets the requirements.

[0065] 2.2 Data Acquisition and Production Forecasting

[0066] Sensor data collected during the seedling stage (October 10th): Soil nitrogen 80mg / kg, phosphorus 45mg / kg, potassium 120mg / kg, soil moisture 65%, ambient temperature 22℃, humidity 70%, plant height 8cm. Edge node filtering and normalization were then uploaded to the cloud, and K-means clustering was used to divide the area into three cultivation zones (low-lying area, flat area, and high-slope area).

[0067] Yield forecasting will begin on November 15th (branching stage), utilizing data from the past 5 years, and will be based on... Calculations show that Y0 = 1500 kg / mu (90% average over the past 3 years), S(t') = 0.6, E(t') = 0.7, M(t') = 0.9, F(t') = 0.6, a = 0.4, b = 0.3, c = 0.2, d = 0.1. The integral result is 18.5, and the solution is Y(t) = 1518.5 kg / mu. Based on a plant height of 25 cm, this is corrected to 1700 kg / mu, with a prediction error of 7%.

[0068] 2.3 Irrigation and Nutrient Regulation

[0069] On October 28th, soil moisture during the seedling stage dropped to 58%. The irrigation optimization submodule generated a solution: light sprinkler irrigation 20 / acre, executed at 6 AM (low evaporation rate), predicting 8mm of rainfall in 3 days, no delay required. Branching stage on November 20th, nutrient requirement calculation submodule according to... Calculate K i =1.2 (branching stage), G i =2000kg / mu, L i =0.1 (sunny day), P i =45mg / kg, Q i =0.6, the phosphorus requirement is 74 kg / mu. Combined with the soil phosphorus supply of 45 mg / kg, the recommended top dressing is 15 kg / mu of superphosphate + 8 kg / mu of potassium chloride. The water and fertilizer equipment was used on November 22. Two days later, the soil phosphorus was tested and found to be 60 mg / kg, which meets the target.

[0070] 2.4 Pest and Disease Early Warning and Risk Response

[0071] Crop sensor collects leaf images on December 5, CNN model identifies aphids (8% of diseased leaves), environmental data temperature 18°C, humidity 75% (2 consecutive days of rain), medium risk in early warning. Push biological control scheme: release 1000 ladybugs per mu, execute in the evening, recheck the disease rate to 2% after 7 days.

[0072] On December 10, the weather forecast heavy rain (55 mm in 24 hours), the cultivation risk assessment submodule calculates , W(t') = 0.9 (heavy rain), P(t') = 0.2, D(t') = 0.4, h = 0.4, e = 0.3, f = 0.3, R = 7.2 (high risk). Push emergency scheme: dig 30 cm deep drainage ditches in low-lying areas (spacing 5 m), drain water in time after heavy rain, and no waterlogging occurs.

[0073] 2.5 Rotation planning and data traceability

[0074] The rotation planning submodule generates a scheme: turn over Astragalus sinicus in full bloom (February 20, 70% flowering rate), depth 15 cm, and interval 18 days from rice transplanting, and increase silicon fertilizer by 5 kg per mu after turning over, which is expected to increase soil nitrogen by 12% and increase rice yield by 10% in the following crop. The cultivation data traceability submodule records the whole cycle data, generates a land block two-dimensional code, and scans the code to view the records of sowing, water and fertilizer, prevention and treatment, and connects to the Hunan Province Green Agriculture Certification Platform.

[0075] 3. Running effect data representation

[0076] Table 1: Comparison of traditional planting and the system effect of rice-Astragalus sinicus rotation

[0077] Index Traditional experience planting The system of the present application Variety adaptation rate 65% 92% Fresh grass yield (kg / mu) 1600 2800 Soil organic matter annual promotion (g / kg) 0.3 0.9 Disease and pest loss rate 25% 5% Irrigation water use efficiency 0.5 0.85

[0078] Table 1 data comes from a comparison test of 200 mu of land, traditional planting relies on experience to select seeds, with an adaptation rate of only 65%, and high seedling mortality rate in clay low-lying areas due to poor waterlogging tolerance of the variety; irrigation is judged by the naked eye, water use efficiency is 0.5, and water resources are wasted seriously; disease and pest discovery is delayed, loss rate is 25%, and fresh grass yield is only 1600 kg per mu. The present application accurately recommends a mixed planting scheme through the variety adaptation submodule, with an adaptation rate of 92%; the irrigation optimization submodule regulates in combination with soil moisture and weather, and the water use efficiency is increased to 0.85; the disease and pest early warning and risk response reduce the loss rate to 5%, and the fresh grass yield is increased to 2800 kg per mu, and the soil organic matter is increased by 3 times, perfectly adapting to the green manure type cultivation demand in the southern rice area.

[0079] Example 2: Large-scale planting of forage Astragalus sinicus in the north (Zhumadian, Henan)

[0080] The embodiment is directed to a 300-mu feed use Chinese artichoke land plot (soil type sandy loam, 0-20 cm organic matter 18 g / kg, annual mean temperature 15°C, annual precipitation 900 mm) of a breeding base in Zhumadian, Henan, the previous crop is wheat (high potassium consumption), the planting purpose is silage feed, the traditional planting has poor palatability, low protein content, and the pest control relies on chemical agents, the feed safety risk is high, and the system is used to realize efficient cultivation, and the specific implementation is as follows.

[0081] 1. System deployment and parameter configuration

[0082] The field data acquisition module is deployed according to a 6m*6m grid: the soil sensor (HH2, measuring nitrogen, phosphorus, potassium and moisture) is buried at a depth of 12cm, the environmental sensor (BME280, measuring temperature, humidity and light) has one group every 15 mu, the crop sensor (CropCircleACS-430, measuring leaf area and spectrum) has one group every 25 mu, a new palatability monitoring sensor (measuring crude fiber content) is added, the sampling frequency is 3Hz by default, and the daily increase of plant height is greater than 1cm, which is increased to 20Hz.

[0083] The cultivation scheme generation module loads 20 variety databases, focusing on feed indicators (crude fiber <25%, crude protein >18%). The intelligent control execution module is connected to the drip irrigation system (Netafim, drip head flow 2L / h), precision seeder (John Deere1725) and electric sprayer (3WBD-20, special for low-residue pesticides). The user interaction module is provided with a central control touch screen and a management personnel APP, and supports palatability data visualization. The data storage edge node is equipped with a 1TB SSD, and the cloud end has a 50TB database, and the communication adopts LoRa+5G dual link.

[0084] 2. Core process implementation details

[0085] 2.1 Variety adaptation and seeding preparation

[0086] The variety adaptation sub-module inputs data: sandy loam, organic matter 18g / kg, annual mean temperature 15°C, previous crop wheat, feed use, calculates the adaptation index: Pingning No. 3 88 points (crude protein 19%), Yi Jiang Zi 83 points (crude fiber 22%), and recommends Pingning No. 3. Generate seeding parameters: seeding on October 5 (daily mean temperature 16°C), seeding amount 2.5kg / mu, broadcasting, base fertilizer superphosphate 15kg / mu+potassium chloride 7kg / mu (to supplement the potassium consumption of the previous crop). After seeding, the unmanned aerial vehicle detects uniformity, the deviation is 8%, and the reseeding suggestion is pushed (reseeding 0.2kg / mu).

[0087] 2.2 Data acquisition and yield prediction

[0088] October 20th seedling stage data: soil nitrogen 70 mg / kg, phosphorus 40 mg / kg, potassium 90 mg / kg, soil moisture 62%, temperature 18℃, light 60000 lux, plant height 10 cm. After edge node processing, K-means is divided into 2 zones (sandy area, loam area).

[0089] November 30th (early flowering stage) to start yield prediction, according to Calculation, Y0=1800 kg / acre, S(t')=0.5, E(t')=0.8, M(t')=0.85, F(t')=0.7, a=0.3, b=0.4, c=0.2, d=0.1, integral result 20.3, Y(t)=1820.3 kg / acre, corrected to 2200 kg / acre combined with leaf area index 3.2, prediction error 6.5%.

[0090] 2.3 Irrigation and nutrient regulation

[0091] November 5th branch period soil moisture dropped to 68%, irrigation optimization sub-module scheme: sandy area sprinkler 25 / acre, loam area drip irrigation 22 / acre, executed at 17 o'clock in the evening, no precipitation is predicted. December 10th flowering period, nutrient demand calculation sub-module according to Calculation, K i =1.5 (flowering period), G i =2500 kg / acre, L i =0.2 (forecast light rain), P i =70 mg / kg, Q i =0.55, the solution is nitrogen demand 98 kg / acre, recommended topdressing urea 10 kg / acre + boron fertilizer 0.05 kg / acre (foliar spraying), executed on December 12th, 5 days later, crude protein content was detected to be 18.5%.

[0092] 2.4 Disease and pest warning and risk response

[0093] January 8th, the crop sensor collected spectral data, the reflectivity at 550 nm decreased by 16%, the CNN model identified powdery mildew (disease leaf rate 12%), the environmental temperature was 22℃, the humidity was 78%, and the warning was high risk. Push low-residue scheme: kresoxim-methyl 100 ml / acre, sprayed in the morning at 6 o'clock, avoiding the honey source period. 7 days later, the disease leaf rate was 3%, the pesticide residue detection was <0.01 mg / kg, which met the feeding standard.

[0094] January 15th, the weather forecast low temperature (-3℃ for 3 hours), the risk assessment sub-module according to Calculation, W(t') = 0.85 (low temperature), P(t') = 0.3, D(t') = 0.2, R = 6.8 (high risk). Push emergency solution: cover 3 cm thick straw insulation, low temperature after detection of no freeze injury.

[0095] 2.5 Rotation planning and data traceability

[0096] The rotation planning submodule generates a scheme: Astragalus 15 February (early pod) is turned over, with a depth of 12 cm, and the following crop of wheat is increased by 100 kg / acre of organic fertilizer, which is expected to increase wheat yield by 12%, and is rotated with rapeseed after 2 years of continuous planting. The cultivation data traceability submodule records the whole cycle data, generates a two-dimensional code, and connects to the Henan agricultural product traceability platform. Scanning can view information such as palatability detection (23% crude fiber) and pesticide residue.

[0097] 3. Data representation of running effect

[0098] Table 2: Comparison of traditional planting of northern forage Astragalus and the effect of the system of the present application

[0099] Index Traditional experience planting The system of the present application Crude protein content 15% 19% Crude fiber content 30% 23% Disease and pest loss rate 28% 4% Pesticide residue over-standard rate 12% 0% Subsequent wheat yield increase rate 6% 12%

[0100] Table 2 data comes from a comparison test of 300 acres of land. The traditional planting variety has poor palatability, with only 15% crude protein and 30% crude fiber, which cannot meet the needs of silage. Disease and pest control relies on chemical pesticides, with a loss rate of 28% and a residue exceeding standard rate of 12%, posing a high risk to forage safety. The rotation planning is not reasonable, and the following crop of wheat has an increase of only 6%. The present application selects high-protein varieties through the variety adaptation submodule, increases protein to 19% through nutrient regulation, reduces the loss rate to 4% through disease and pest warning and low-residue control, and optimizes wheat yield by 12% through rotation planning, perfectly adapting to the needs of yield, quality, and safety for large-scale planting of northern forage Astragalus.

[0101] Reference Figure 2 This figure directly shows the core value of the variety adaptation submodule in claim 3. Traditional experience selection relies on habit, and the adaptation rate is generally lower than 65%. Local conventional varieties do not meet the requirements in various scenarios. The present application constructs a three-level index system through the analytic hierarchy process, calculates the adaptation index in combination with soil, climate, and use, and achieves 85 for Minzhi No. 6 in clay green manure scenarios and 88 for Pingning No. 3 in sandy loam forage scenarios, both of which far exceed the adaptation threshold. This verifies the effect of the variety adaptation submodule in solving the blindness of traditional seed selection, and precisely matches the optimal variety for different plots, improving resistance and yield potential.

[0102] Reference Figure 3The figure clearly reflects the optimization effect of the yield prediction sub-module in claim 2. The traditional experience prediction relies on historical mean value, and the error in the seedling stage is more than 25%, which cannot guide dynamic regulation and control; the present application fuses multi-dimensional data of soil, environment, management and microorganisms through integral formula, and the error is reduced to less than 6% as the growth period advances, and is reduced to less than 3% after introducing adjacent land data correction. This verifies the accuracy of the spatio-temporal attention mechanism model, predicts the yield trend 45 days in advance, provides a reliable basis for adjusting fertilizer and irrigation, and avoids yield loss caused by lagging management.

[0103] Referring to Figure 4 The chart verifies the practical value of the irrigation optimization sub-module in claim 4. The traditional irrigation is executed according to fixed amount, and the total irrigation amount during the whole growth period is 150 / acre, which is low in water use efficiency and serious in waste. The present application is based on four-dimensional decision-making of "soil moisture- growth period-texture-weather", combined with irrigation time window optimization, and the total irrigation amount of sandy soil during the whole period is only 97 / acre, and the total irrigation amount of clay soil is 105 / acre, and the water use efficiency is stable at more than 0.78. This reflects the saving effect of precise irrigation on water resources, avoids problems such as flushing in the seedling stage and water accumulation in the flowering stage, and improves water use efficiency.

[0104] Referring to Figure 5 The figure highlights the comprehensive benefits of the disease and pest intelligent early warning sub-module in claim 6. The traditional manual prevention and control relies on patrol, and the recognition lag is more than 30 minutes, the prevention and control accuracy is only 65%, the loss rate is as high as 25%, and the risk of pesticide residue is high. The present application identifies through "image-spectrum-environment" trinity, responds within 10 seconds, combines with risk grading to push biological or low residue scheme, the prevention and control accuracy reaches 95%, the loss rate is reduced to 5% and there is no residue, and the proportion of ecological prevention and control is increased to 60%. This verifies the early discovery and precise treatment ability of the system to diseases and pests, and takes into account yield protection and ecological safety.

[0105] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can make equivalent replacement or change according to the technical scheme and inventive concept of the present application within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A big data analysis-based optimization system for the cultivation and management of milkvetch, characterized in that, Includes the following modules: The field data acquisition module is equipped with soil parameter sensors, environmental sensors, and crop growth sensors. The soil sensors measure nitrogen, phosphorus, and potassium; the environmental sensors monitor temperature, humidity, light intensity, and precipitation; and the crop sensors detect plant height, leaf area index, and flowering rate. The big data processing and analysis module uses edge computing nodes and cloud servers to work together to perform data filtering, normalization and anomaly removal, and divides the cultivation areas through K-means clustering; the cloud server uses a random forest regression algorithm and a CNN model to identify 12 types of diseases and pests of milk clover. The cultivation plan generation module has a built-in database of milkvetch growth period. Combining real-time data with historical cases, it constructs a multi-objective optimization model with the goals of increasing yield, optimizing quality, and reducing costs. The intelligent control and execution module supports Modbus-RTU and MQTT protocols, and can connect to integrated water and fertilizer equipment, intelligent seeders and plant protection drones; The user interaction module is equipped with a 7-inch touch screen and a mobile APP, which supports cultivation data visualization, manual adjustment of plans, anomaly alarms, and query of historical plans; The data storage and communication module adopts edge caching and cloud-distributed database, and the communication integrates 4G / 5G and LoRa. The data is encrypted using AES-128. The system operation and maintenance module enables 24 / 7 status monitoring, tracks sensor online rate, equipment failure rate and solution execution success rate, automatically generates maintenance work orders after fault warning, and supports remote firmware upgrades and parameter backups.

2. The milkvetch cultivation and management optimization system based on big data analysis according to claim 1, characterized in that, It also includes a yield prediction submodule, which constructs a three-dimensional data system encompassing historical, real-time, and predictive data. This submodule integrates nearly five years of plot yield data, hourly soil nutrient data, environmental parameters, cultivation management records, and soil microbial activity. It employs a spatiotemporal attention mechanism to optimize the prediction model. Calculate the predicted yield of milkvetch during the pod-setting stage, where Y(t) is the predicted yield at time t. S(t') represents the basic yield of the plot, S(t') represents the comprehensive value of soil nutrients at time t', E(t') represents the environmental suitability at time t', M(t') represents the implementation rate of management measures at time t', F(t') represents the soil microbial activity value at time t', and a, b, c, and d represent the influence coefficients of soil, environment, management, and microorganisms, respectively. The prediction model updates its parameters every 15 days and corrects for deviations by combining the actual biomass measured in the field.

3. The milkvetch cultivation and management optimization system based on big data analysis according to claim 1, characterized in that, It also includes a variety matching submodule, which constructs a feature database covering 20 mainstream varieties. The detailed parameters include flood tolerance, cold tolerance, branching ability, fresh grass yield, growth period, stress resistance, and utilization value. Combined with field measurement data: soil texture, organic matter content in the 0-20cm soil layer, average annual temperature, annual precipitation, planting purpose, and crop rotation pattern, a three-level evaluation index system is constructed using the analytic hierarchy process (AHP) to calculate the variety matching index and simultaneously generate supporting planting parameters: suitable sowing period, sowing density, sowing method, and base fertilizer application rate. Mixed sowing suggestions are also provided for complex plots.

4. The milkvetch cultivation and management optimization system based on big data analysis according to claim 1, characterized in that, It also includes an irrigation optimization submodule, which makes dynamic decisions based on four-dimensional data: soil moisture, growth stage, soil texture, and meteorological conditions. It refines irrigation strategies based on the growth stage characteristics of milkvetch: light sprinkler irrigation is initiated when soil moisture is below 60% during the seedling stage; moderate irrigation is initiated when soil moisture is below 70% during the branching stage; precision drip irrigation is initiated when soil moisture is below 75% during the flowering stage; and intermittent irrigation is initiated when soil moisture is below 65% during the pod-setting stage. It integrates meteorological forecast data for the next 72 hours, and if the predicted precipitation is ≥10mm, irrigation is delayed by 24-48 hours; if it is ≥20mm, irrigation is cancelled; if accompanied by strong winds, irrigation is initiated earlier and the irrigation amount is reduced by 10%. It also optimizes the irrigation time window.

5. The milkvetch cultivation and management optimization system based on big data analysis according to claim 1, characterized in that, It also includes a nutrient requirement calculation submodule, which, based on the nutrient absorption patterns of milkvetch and soil fertility characteristics, refines the key requirements according to the growth stage: phosphorus and molybdenum fertilizer are emphasized during the seedling stage; nitrogen, phosphorus, and potassium are considered during the branching stage; a balanced nitrogen, phosphorus, and potassium is needed during the flowering stage; and potassium and phosphorus are strengthened during the pod-setting stage. Calculate the nutrient requirements for each growth stage, among which This represents the nutrient requirements for the i-th reproductive stage. Let be the nutrient requirement coefficient for the i-th reproductive stage. The target biomass for the i-th reproductive stage. The nutrient loss rate during the i-th reproductive period. This represents the readily available nutrient content in the soil. To improve nutrient utilization efficiency, the appropriate fertilizer types for formula fertilizer formulation are as follows: 60% organic fertilizer and 40% chemical fertilizer during the seedling stage, and 40% organic fertilizer and 60% chemical fertilizer during the flowering stage. Combined with optimized fertilization methods, phosphorus and potassium fertilizers are mainly applied as basal fertilizers, nitrogen fertilizers are mainly applied as top dressings, and micronutrients are applied as foliar sprays.

6. The milkvetch cultivation and management optimization system based on big data analysis according to claim 1, characterized in that, It also includes a sub-module for intelligent early warning of pests and diseases. This sub-module constructs a three-in-one recognition system of image, spectrum and environment. The crop sensor collects leaf images and spectral data every 2 hours, and the CNN model integrates image texture features and spectral features. A risk correlation model was established by integrating hourly environmental data: the risk of powdery mildew increases when the temperature is 20-28℃, the humidity is above 75%, and there are 3 consecutive days of cloudy and rainy weather. When the temperature is 15-25℃ and the soil moisture is ≥80% after rainfall, aphids are prone to outbreaks; insufficient sunlight for 5 consecutive days increases the risk of rust; four risk levels are set: diseased leaf rate <5% is low risk; 5%-15% is medium risk; 15%-30% is considered high risk; >30% is considered emergency risk; establish a closed-loop feedback mechanism for prevention and control effectiveness, and recheck the incidence of pests and diseases 7 days after application. If the incidence decreases by less than 50%, adjust the type and dosage of pesticide.

7. The milkvetch cultivation and management optimization system based on big data analysis according to claim 1, characterized in that, It also includes a crop rotation planning submodule, which combines the growth period of milkvetch with the annual variation of soil nutrients in the plot; if the previous crop is indica rice, the recommended milkvetch variety is Yijiangzi; if the previous crop is rapeseed, the recommended variety is Pingning No. 3; optimization for the needs of the subsequent crop: if the subsequent crop is japonica rice, apply an additional 5 kg / mu of silicon fertilizer after the milkvetch is plowed in; if the subsequent crop is wheat, delay the plowing period to the early pod-setting stage; add a crop rotation year optimization strategy, after planting milkvetch for 3 consecutive years, it is recommended to rotate with rapeseed for 1 year; calculate the comprehensive benefits of crop rotation: soil organic matter increase, subsequent crop yield increase, fertilizer reduction rate, and carbon sequestration.

8. The milkvetch cultivation and management optimization system based on big data analysis according to claim 1, characterized in that, It also includes a cultivation risk assessment submodule, which constructs a three-dimensional risk assessment system based on meteorology, biology, and soil. Calculate the comprehensive risk value for the entire growth period, where R is the comprehensive risk value, W(t') is the meteorological disaster risk at time t', P(t') is the pest and disease risk at time t', D(t') is the soil degradation risk at time t', and h, e, and f are the weighting coefficients for meteorology, pests and diseases, and soil, respectively. Refine the calculation of each risk factor: the risk of low-temperature freezing damage is calculated based on the lowest temperature in the next 48 hours; the risk of rainstorm flooding is calculated based on the plot slope and precipitation; and the drought risk is calculated based on the number of consecutive days with soil moisture <50%. Targeted emergency plans are pushed out: in high-risk situations, cover with straw for insulation before low temperatures, dig temporary drainage ditches for flood prevention before rainstorms, and activate emergency drip irrigation during droughts; in medium-risk situations, increase the frequency of pest and disease inspections and spray foliar fertilizer to enhance stress resistance; in low-risk situations, maintain routine management and monitor risk changes; simultaneously generate an emergency supplies list, indicating procurement channels and usage methods.

9. The milkvetch cultivation and management optimization system based on big data analysis according to claim 1, characterized in that, It also includes a smart agricultural timing reminder sub-module, which is based on the milkvetch growth period accumulated temperature model and real-time growth data, and pushes tiered reminders 3 days in advance via mobile APP: the basic reminder includes the node name and the best time to operate; Detailed reminders include key operational points, required equipment and materials; agricultural timing can be dynamically adjusted, and in case of extreme weather, the seedling period will be automatically extended to 37 days, the topdressing period will be postponed accordingly and the amount of phosphate fertilizer will be increased by 10%, and the turning and compaction period will be adjusted accordingly; an operational quality monitoring mechanism has been added, and the sowing uniformity will be detected by drone aerial photography and ground sampling after sowing, and the changes in soil nutrients will be detected after fertilization.

10. The milkvetch cultivation and management optimization system based on big data analysis according to claim 1, characterized in that, It also includes a cultivation data traceability submodule, which constructs a three-dimensional traceability system of plot, time, and operation, automatically collects and associates data throughout the entire growth period: basic plot information, sowing information, field management records, environmental data, growth monitoring data, and harvest data; all data are accompanied by a unique plot identifier and timestamp, and core data is stored using blockchain technology; It supports multi-channel traceability queries; generates a visual traceability map, displaying key nodes in the growth period in a timeline format, and allows users to view corresponding data, images, and operation records by clicking on nodes; and connects to the National Agricultural Product Quality and Safety Traceability Management Information Platform.

Images