Data management method for garden seedlings

Abstract

The invention relates to a data management method for garden seedlings. By deploying a high-precision sensor network, physiological feature data and environmental parameters of garden seedlings are collected; carrying out data dimension reduction by utilizing a feature engineering technology, extracting key parameters and establishing a dynamic association model; updating the model in real time by adopting a self-adaptive algorithm so as to realize dynamic association between nursery stock growth characteristics and environment variables; classifying different types of abnormities through an abnormal mode clustering algorithm and triggering intelligent early warning; performing real-time optimization on irrigation, fertilization and pruning management operation based on a dynamic data model and a geographic space analysis result by utilizing a reinforcement learning algorithm, and dynamically adjusting an optimization strategy through a closed-loop feedback mechanism so as to continuously improve a management effect; inputting the dynamic data model, the spatial analysis result and the optimal management scheme into a digital twinborn system, constructing a virtual model of the garden nursery stock, and realizing real-time synchronization of the physical nursery stock and the digital twinborn body; and datamation management of the whole life cycle of the nursery stock is realized.

Description

Technical Field

[0001] The invention relates to the field of garden data management, and in particular to a data management method for garden seedlings. Background Art

[0002] At present, the existing technology, such as the cloud computing-based intelligent management method and system for garden data disclosed in Chinese invention CN202310781307.9, has achieved intelligent management and cost optimization of garden data to a certain extent, but there are still many shortcomings and disadvantages. First of all, the core of the invention is to collect data by laying out sensors, and use the estimated detection value and the actual detection value for verification to optimize the number of sensor layouts. However, this method mainly focuses on the economic layout and cost reduction of sensors, ignoring the more critical dynamic and multi-dimensional data association analysis in garden seedling data management. For example, in actual garden management, seedling growth is affected by a variety of environmental factors such as soil moisture, light intensity, and temperature, and these effects have complex correlations in time and space. The prior art only calculates the detection value based on the detection data of a single verification area, lacks in-depth mining of the correlation between multiple regions and multiple factors, and cannot meet the dynamic monitoring and precise control requirements of the growth state of seedlings. In addition, the invention method judges the accuracy of the estimated detection value by setting a verification threshold, but this threshold is usually based on experience or fixed settings, and it is difficult to adapt to dynamically changing environmental conditions. For example, seasonal changes, sudden extreme weather, etc. will increase the volatility of environmental parameters, thereby increasing the deviation between the estimated detection value and the actual detection value. In this case, the fixed threshold cannot be adjusted dynamically, which may cause the system to frequently re-deploy sensors, increasing the complexity and workload of management, which is inconsistent with the goal of reducing the complexity of the system. At the same time, although the re-deployment of sensors can improve the accuracy of the data, it fails to fundamentally solve the problem of identifying causal relationships. For example, when the soil moisture detection value of a certain verification area in the garden area is abnormal, the system will only prompt to rearrange the sensor or adjust the estimation model, but it cannot identify whether the soil moisture change is caused by other variables (such as insufficient irrigation and increased temperature), so it is impossible to provide accurate management suggestions. This result-oriented passive processing method lacks in-depth analysis of the causal relationship between variables, and it is difficult to support the precise and scientific management of garden seedlings. Furthermore, the invention does not fully utilize the value of real-time data. The calculation of its estimated detection value and verification accuracy value is more of an offline calculation of a static model, which fails to respond dynamically to real-time changing environmental data. For example, under extreme weather conditions (such as continuous high temperatures or heavy rainfall), the environmental conditions in the verification area may change rapidly, and the system only determines whether the data is accurate by verifying the threshold. This static judgment mode is difficult to capture the true growth status of the seedlings in a timely manner, which may delay management decisions. At the same time, the application of sensor data in the system architecture of the invention is more limited to data calculation and verification in a single area, and fails to achieve integrated analysis of data from different regions.For example, for large-scale garden scenes, the ecological relationships between regions (such as water transport and light shielding effects) require global multi-scale analysis, but the division of verification areas and sensor layout optimization in existing technologies are too independent, ignoring the management needs of the overall synergistic effects of the garden ecosystem.

[0003] Another shortcoming is that the invention lacks management support for the entire life cycle of seedlings. The growth of garden seedlings is a long-term dynamic process, involving multiple stages such as planting, growth, pruning, and transplanting, and each stage requires targeted management data support. However, the existing technology pays more attention to the verification of sensor data at the current time point, and does not model and predict the long-term growth trend and future state of seedlings. For example, when the growth rate of seedlings in a certain area decreases, the system can only improve the accuracy of the data by re-deploying sensors, but cannot provide predictions of the future health trends and growth potential of the seedlings, which is extremely unfavorable to the long-term planning and resource allocation of the garden. Summary of the invention

[0004] The purpose of the present invention is to provide a data management method for garden seedlings, thereby solving some of the drawbacks and shortcomings pointed out in the background technology.

[0005] The present invention solves the above-mentioned technical problems by adopting the following technical solutions: a data management method for garden seedlings, comprising: S1, high-dimensional dynamic data modeling based on plant feature extraction:

[0006] S1.1. Collect physiological characteristics data and environmental parameters of garden seedlings by deploying a high-precision sensor network;

[0007] S1.2. Use feature engineering technology to reduce data dimension, extract key parameters and establish a dynamic correlation model;

[0008] S1.3, using adaptive algorithms to update the model in real time to achieve dynamic association between seedling growth characteristics and environmental variables;

[0009] S2. Early warning mechanism for abnormal seedling status based on causal reasoning:

[0010] S2.1. Based on the high-dimensional dynamic data model, a causal network of seedling status was constructed to identify the causal relationship between different variables;

[0011] S2.2, classify different types of anomalies through abnormal pattern clustering algorithm and trigger intelligent warning, and generate response suggestions for abnormal status;

[0012] S3. Multi-scale geospatial analysis and precision planting planning:

[0013] S3.1. Collect multi-scale geospatial data of the garden, including macro-level garden layout, meso-level seedling distribution, and micro-level individual plant status;

[0014] S3.2. Formulate precise planting plans through spatial data integration and optimization algorithms, and optimize spatial resource utilization based on the ecological relationship between seedlings;

[0015] S4. Adaptive management optimization system based on dynamic data:

[0016] S4.1. Using reinforcement learning algorithms, design an intelligent optimization engine based on dynamic data models and geospatial analysis results to optimize irrigation, fertilization, and pruning management operations in real time, and dynamically adjust the optimization strategy through a closed-loop feedback mechanism to continuously improve management results;

[0017] S5. Digital twin technology for seedlings throughout their life cycle:

[0018] S5.1. Input the dynamic data model, spatial analysis results and optimization management plan into the digital twin system to build a virtual model of garden seedlings and achieve real-time synchronization between physical seedlings and digital twins;

[0019] S5.2. Use virtual experiments to simulate the impact of different management plans on seedling growth, and apply them to actual operations after optimization to achieve digital management of the entire life cycle of seedlings.

[0020] Furthermore, the abnormal state warning mechanism of seedlings based on causal reasoning:

[0021] Multimodal sensors are deployed to collect data on key physiological and environmental variables of seedlings, including but not limited to soil moisture, light intensity, and leaf temperature. The data are stored in the form of time series, denoted as a set X = {x1, x2, …, x n}, where x i represents the real-time observation value of the i-th variable; the data is processed using feature dimensionality reduction technology to obtain the principal component set Z = {z1, z2, …, z m}, where z i It represents the i-th principal component variable after dimensionality reduction, reflecting the comprehensive effect of the main influencing factors. Based on this, a dynamic association model M(t) is established, and its formula is:

[0022]

[0023] Where: M(t): the output of the dynamic correlation model at time t, used to describe the comprehensive correlation between seedling status and environmental variables; i (t): the value of the i-th principal component at time t; f i(t): represents the dynamic weight function of the i-th variable over time t, reflecting its influence on the model; α: the adjustment factor in the model, controlling the amplitude of the interaction term between variables; β: the decay rate of the time-varying influencing factor, used to smooth the influence of time changes; e -β·t : Exponential decay term, indicating the weakening trend of the variable over time;

[0024] Through the above formula, the model can dynamically describe the comprehensive impact of various variables on the state of seedlings and adapt to the nonlinear characteristics of time series data.

[0025] Furthermore, the abnormal state warning mechanism of seedlings based on causal reasoning:

[0026] Based on the dynamic model M(t), a causal network is constructed using causal reasoning methods. The causal network is constructed through the effect function C(x i ,x j ) quantifies the causal relationship, the formula is as follows:

[0027]

[0028] Where: C(x i ,x j ):variable x i For variable x j the strength of the causal effect; variable x i The impact rate on the dynamic model output M(t) is represented by x i The importance of g(x j ,t): variable x j The response function is used to characterize the sensitivity of the dependent variable to time t; The cumulative impact from the initial time t0 to the current time t represents the time range of the causal effect; when C(x i ,x j )>γ, where γ is the threshold of causal effect; i x j There is a significant causal effect, which in turn establishes directional connections between variables in the causal network.

[0029] Furthermore, the abnormal state warning mechanism of seedlings based on causal reasoning:

[0030] According to the dynamic results of the causal network, the abnormal deviation of each variable is calculated in real time; when the variable deviates from the normal range, it is marked as abnormal, and the abnormal pattern is classified by the abnormal clustering method; the abnormal pattern clustering is performed by the abnormal clustering function A(k), and the formula is as follows:

[0031]

[0032] Where: A(k): clustering score of the k-th abnormal pattern, used to evaluate the confidence that the data point belongs to cluster k; z i : Principal component value of abnormal data point i; z k : The principal component value of the k-th cluster center; d(z i ,z k ): point z i With cluster center z k The Euclidean distance indicates the degree to which the data point deviates from the cluster center; n k : The number of abnormal data points in cluster k;

[0033] Through the above formula, abnormal data points are clustered into different patterns, including water stress, disease, and nutrient deficiency; and specific warning signals are triggered for each pattern; combined with the causal network analysis results and clustering patterns, targeted response recommendations are generated based on the seedling management knowledge base.

[0034] Furthermore, the digital twin technology of seedlings throughout their life cycle:

[0035] Through a multimodal sensor network, the growth parameters of seedlings, including height, chlorophyll concentration, photosynthesis efficiency, and environmental parameters, including soil moisture, temperature, and light intensity, are collected in real time. The collected data are integrated into a dynamic data stream. Combined with the geographic information system GIS and spatial analysis, the distribution and growth space information of seedlings in the garden area are obtained to form a multi-dimensional dynamic data set for building a digital twin model of seedlings. The digital twin model describes the real-time state and dynamic change trend of seedlings based on their geometric structure, physiological state, and environmental interaction. The state model is defined as:

[0036]

[0037] Among them: S(t): the comprehensive state of the digital twin model at time t, representing the virtual model result of the overall growth of the seedlings; G(t): the geometric structure change model, describing the physical characteristics of the seedlings (including crown width and tree height), which is the static basis of the twin model; The integral operation of the cumulative effect, from the initial time t0 to the current time t, is used to calculate the long-term impact of the environment and variables on the growth state; E(τ): Environmental interaction model, describing the direct effect of environmental variables, including light intensity and soil moisture on the seedlings at time τ; z i (τ): the value of the i-th key variable at time τ, representing the measured value of a specific environmental factor, including temperature and light; h i (τ): weight function, reflecting the variable z i (τ) The degree of influence on the overall model state; α, β: adjustment factors used to balance nonlinear dynamic behavior, α controls the basic influence intensity of the variable, and β represents the decay rate of the time effect; e-β·τ : Time decay term, indicating that the influence of the variable on the state gradually weakens over time.

[0038] Furthermore, the digital twin technology of seedlings throughout their life cycle:

[0039] Virtual experiments were conducted in the digital twin model to simulate the dynamic effects of different schemes on seedling growth by changing environmental parameters, including temperature and light; or management strategies, including irrigation amount and fertilization frequency. The virtual experiment used a multi-objective optimization algorithm to evaluate the effects of management strategies, including maximizing growth rate and minimizing resource utilization. The optimization objective function was defined as:

[0040]

[0041] Where: P: comprehensive score of management strategy, used to quantify the effect of a strategy in virtual experiment; T = t f -t0: test time range, t0 and t f are the start and end time of the experiment respectively; R(t): growth rate at time t, indicating the increment of seedlings at that time; C(t): resource consumption at time t, including the use of water and nutrients; λ1, λ2: weight coefficients of growth rate and resource consumption respectively, λ1 reflects the preference for growth, and λ2 represents the demand for saving resources.

[0042] Furthermore, the digital twin technology of seedlings throughout their life cycle:

[0043] The optimal management strategy is selected based on the virtual test results, and adjusted in combination with the actual operating conditions. The optimization results are fed back to the physical seedling management system to form a closed-loop management process that integrates virtual and physical elements. Through the IoT platform, the real-time data of the physical seedlings is dynamically synchronized with the digital twins to ensure that the geometric structure and physiological state of the twin model are consistent with the actual state, achieving real-time adjustment and consistency correction.

[0044] Record the data of the entire process from planting to maturity of seedlings, and establish a full life cycle management file; predict future growth trends based on the digital twin model, and calculate the future state through the following prediction functions:

[0045]

[0046] Where: F(t): the predicted value of the future state after time t, indicating the comprehensive growth of the seedlings at a certain time in the future; T p : The length of the prediction time window, that is, the time range from the current moment in the future; G ′ (t): time derivative of the geometric structure model, representing the rate of change of the physical properties of the seedlings; E ′(t): time derivative of the environmental interaction model, representing the impact of changes in environmental variables on future states; k1, k2: weight coefficients of geometric changes and environmental interactions, respectively. k1 and k2 are used to balance the relative contributions of geometric structures and environmental variables in future trends.

[0047] The data management method for garden seedlings of the present invention collects real-time data through multimodal sensors, performs analysis based on causal reasoning and dynamic modeling, and combines digital twin technology to achieve virtual and physical fusion management, as well as full life cycle growth status monitoring and prediction, and has the following beneficial effects:

[0048] This method can obtain the physiological characteristics and environmental data of seedlings in real time through the sensor network, and dynamically integrate them into the data model to realize real-time monitoring of the seedling status. Through multi-scale geospatial analysis, this method can optimize the spatial layout of the garden, rationally plan the seedling planting area, and improve resource utilization; and dynamically adjust management operations such as irrigation and fertilization in combination with the optimization algorithm to reduce resource waste and improve garden management efficiency. Using reinforcement learning algorithms and multi-objective optimization models, the present invention can evaluate the effects of different management strategies in virtual experiments, and formulate the optimal management plan by weighing the maximization of growth rate and the minimization of resource utilization, thereby avoiding the inefficiency and blindness of traditional empirical decision-making. The introduction of digital twin technology realizes the real-time synchronization of physical seedlings and virtual models, and dynamically updates data through the Internet of Things platform, so that managers can intuitively grasp the current status and future trends of seedlings. The prediction results of the virtual model directly guide the actual management operations, forming a closed-loop management system. This method can record the data of the entire process of seedlings from planting to maturity, and establish a complete growth archive. Based on the prediction function of the model, managers can formulate future management strategies in advance, providing a scientific basis for the long-term planning and sustainable development of the garden. Through in-depth data mining and analysis, the present invention greatly reduces manual dependence and management costs, while improving the quality of seedling growth through precise intervention, thereby achieving higher economic benefits. The method has good adaptability and is not only suitable for garden seedling management, but can also be extended to other agricultural and forestry fields, providing innovative solutions for intelligent ecological management and sustainable development. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 The present invention is a flow chart of the data management method for garden seedlings.

[0050] Figure 2 This is a flow chart of the early warning mechanism for abnormal status of seedlings based on causal reasoning of the present invention.

[0051] Figure 3 This is a flowchart of the digital twin technology for seedlings throughout their life cycle according to the present invention. DETAILED DESCRIPTION

[0052] The specific implementation modes of the present invention will be described in detail below in conjunction with the accompanying drawings.

[0053] Combined with Figure 1 In the data management method of garden seedlings of the present invention, S1 is based on high-dimensional dynamic data modeling extracted from plant features. First, in S1.1, by deploying a high-precision sensor network, real-time collection of multimodal data is performed for key areas of garden seedlings, such as leaves, stems and roots. These data include physiological characteristic data, such as chlorophyll concentration, photosynthesis efficiency, stem diameter, and environmental parameters, such as soil moisture, light intensity, temperature, etc. The sensor network is characterized by low power consumption and high precision, and can maintain the stability and accuracy of data collection in large-scale garden scenes. Secondly, in S1.2, feature engineering technology is used to reduce the dimension of the collected high-dimensional data. Through methods such as principal component analysis (PCA) or sparse coding technology, key parameters affecting seedling growth, such as water utilization rate, photosynthetic rate change trend, etc., are extracted from the original data to reduce data complexity and retain data core information. At the same time, based on the extracted key features, a dynamic association model is established to describe the relationship between seedling growth characteristics and environmental variables. Finally, in S1.3, an adaptive algorithm is used to update the dynamic association model in real time. As time goes by and the environment changes, the model can dynamically adjust the weight parameters so that it can accurately reflect the current relationship between the growth status of the seedlings and environmental conditions. For example, when environmental parameters such as light intensity or soil moisture fluctuate, the model will adjust to the new data input to ensure the real-time and accurate prediction.

[0054] S2 The causal reasoning-based warning mechanism for abnormal seedling status mainly includes two core steps. First, in S2.1, a causal network is constructed using the physiological characteristics and environmental data of seedlings through a high-dimensional dynamic data model to identify the causal relationship between different variables. The key to this step is to analyze the variables in the dynamic data model (such as soil moisture, light intensity, and leaf health index) through causal reasoning technology to clarify the causal path between the variables. For example, through the causal network, it can be inferred how "reduced soil moisture" leads to "decreased leaf photosynthesis efficiency", and then a clear causal chain can be constructed. This analysis not only relies on statistical correlation, but also combines the time series changes and intervention effects between variables to ensure the reliability and scientificity of the causal relationship. Secondly, in S2.2, the abnormal pattern clustering algorithm is used to classify the abnormal data, and the intelligent warning is triggered based on the analysis results of the causal network. The abnormal pattern clustering algorithm classifies the collected abnormal data into different abnormal types according to feature similarity, such as abnormalities caused by diseases, abnormalities caused by water stress, or abnormalities caused by insufficient nutrients. The early warning mechanism provides specific response recommendations based on the anomaly type and causal path, such as "When soil moisture is detected to be continuously decreasing and causing a decrease in leaf health index, it is recommended to increase irrigation" or "In response to the decrease in growth rate caused by insufficient light, it is recommended to adjust shading facilities."

[0055] S3 multi-scale geospatial analysis and precision planting planning include two key steps. In S3.1, first, the multi-scale geospatial data of the garden needs to be collected. These data are divided into three levels: the macro level includes the overall layout information of the entire park, such as regional terrain, roads, and water distribution; the meso level covers the distribution of seedlings in the park, such as seedling density and variety distribution in each area; the micro level focuses on the specific status of individual seedlings, such as the precise coordinates of the individual position, the canopy coverage, and the distance from surrounding plants. Through drone remote sensing, GPS positioning and high-precision imaging technology, these multi-scale data are collected and digitally stored in real time to form a geographic information database for the entire park. Then, in S3.2, a precise planting plan is formulated through the integration and optimization algorithm of spatial data. Spatial data integration unifies the macro, meso and micro data into a unified model to analyze the spatial distribution law of seedlings and potential resource utilization problems. Combined with optimization algorithms, such as spatial optimization strategies based on genetic algorithms, available space in the park is identified and the optimal planting layout is planned. In addition, the planting plan must fully consider the ecological relationship between seedlings, including competitive effects (such as shading and root competition) and synergistic effects (such as the disease protection of adjacent plants). Through these analyses and optimizations, the maximum utilization of park space resources can be achieved, such as rationally planning planting density to improve light uniformity, optimizing the layout of irrigation systems to reduce resource waste, or adjusting seedling distribution to improve overall ecological benefits.

[0056] S4 is an adaptive management optimization system based on dynamic data, which mainly realizes intelligent management optimization through reinforcement learning algorithm. In S4.1, dynamic data model and geospatial analysis results are first used as inputs of reinforcement learning algorithm. These data include the real-time growth status of seedlings (such as chlorophyll concentration, crown width change), environmental parameters (such as soil moisture, light intensity) and spatial distribution characteristics (such as seedling spacing, terrain characteristics). The reinforcement learning algorithm designs an intelligent optimization engine by setting state, action and reward mechanism. The state space represents the current overall growth status of garden seedlings; the action space covers management operations, such as adjusting irrigation frequency, optimizing fertilization, implementing pruning measures, etc.; the reward function measures the pros and cons of management operations based on seedling growth performance (such as growth rate, health index) and resource utilization efficiency (such as water and fertilizer consumption). The optimization engine gradually finds the optimization solution that can maximize rewards by continuously exploring and experimenting with different management strategies. For example, the algorithm finds that appropriately reducing irrigation volume but increasing fertilization frequency can promote rapid growth of seedlings while saving resources. Through the closed-loop feedback mechanism, after each management operation is performed, the growth response data of the seedlings is collected in real time and fed back to the optimization engine to update the policy parameters of the reinforcement learning model. This closed-loop mechanism ensures that the optimization engine can continuously adjust the management strategy according to the dynamic changes in the environment and seedling status, thereby maintaining continuous improvement in management results. For example, when seasonal changes cause soil moisture to drop, the optimization engine can dynamically adjust the irrigation frequency to maintain optimal growth conditions.

[0057] S5's digital twin technology for seedlings throughout their life cycle uses digital means to achieve full-process management of seedlings from planting to maturity. In S5.1, the dynamic data model, spatial analysis results, and optimization management schemes are first input into the digital twin system. The dynamic data model includes the physiological characteristics, environmental parameters, and time series changes of the seedlings. The spatial analysis results include the geometric position distribution and spatial characteristics of the seedlings in the park. The optimization management scheme is generated by reinforcement learning algorithms and multi-objective optimization strategies. These data together construct the digital twin of the seedlings. The digital twin can not only virtually present the geometric structure of the seedlings (such as tree height and crown width), but also dynamically reflect the physiological state of the seedlings (such as health index and growth rate) and their interactive relationship with the environment (such as light and water use efficiency). Through the Internet of Things platform, the real-time data of the physical seedlings is synchronized with the digital twin to ensure that the virtual model is consistent with the actual seedlings and reflects the changes in the physical environment and management operations in real time. In S5.2, virtual experiments are carried out through the digital twin system to simulate the dynamic effects of different management schemes on seedling growth. The test content includes management measures such as adjusting irrigation volume, optimizing fertilization strategies, and adjusting pruning plans, and is evaluated in combination with different environmental conditions (such as soil moisture changes or light intensity fluctuations). The virtual experiment uses a multi-objective optimization algorithm to analyze the effects of different management schemes, such as maximizing the growth rate and health index of seedlings while minimizing resource consumption. After the test is completed, the management strategy with the best effect is selected and applied to actual operations. In this way, the digital twin system can not only predict the results of different management schemes, but also further optimize the management strategy through real-time feedback to achieve virtual and physical closed-loop management. In the end, digital twin technology realizes the data management of the entire life cycle of seedlings, and can accurately monitor, dynamically optimize and scientifically predict each stage from planting to maturity, which greatly improves the efficiency and intelligence level of garden management, and provides reliable data support for long-term planning and resource optimization.

[0058] Embodiment 1:

[0059] Combination Figure 2 In this embodiment, sensors are deployed in a garden to collect key physiological data required for seedling growth, including soil moisture (x1), light intensity (x2) and leaf temperature (x3). It is hoped that the abnormal state warning mechanism of seedlings based on causal reasoning can monitor these data in real time, timely discover potential problems of seedlings (such as insufficient water, excessive light, etc.), and give response suggestions.

[0060] 1. Data collection and feature dimensionality reduction

[0061] The following data are collected:

[0062] x1 = soil moisture (%)

[0063] x2 = light intensity (Lux)

[0064] x3=leaf temperature (℃)

[0065] Store this data in the form of time series, for example:

[0066] At time t0, soil moisture is x1(0) = 40%, light intensity is x2(0) = 500 Lux, and leaf temperature is x3(0) = 30°C;

[0067] At time t1, soil moisture is x1(1)=42%, light intensity is x2(1)=520Lux, and leaf temperature is x3(1)=32°C.

[0068] Then, these multidimensional data are subjected to feature dimensionality reduction to extract the main influencing factors. After setting the principal component analysis (PCA), the principal component set Z = {z1, z2, z3} is obtained, where z1 represents the factor that has the greatest impact on seedling growth. The data after dimensionality reduction is as follows:

[0069] z1(t0)=0.75

[0070] z1(t1)=0.78

[0071] 2. Establish a dynamic association model

[0072] The dynamic association model M(t) is constructed using the formula as follows:

[0073]

[0074] In this formula:

[0075] M(t) represents the growth status of seedlings at time t;

[0076] z1(t) is the principal component data after dimensionality reduction;

[0077] f1(t) is the weight function of soil moisture (x1) on seedling growth;

[0078] α and β control the magnitude of the interaction term between variables and the decay rate of time changes, respectively.

[0079] set up:

[0080] f1(t)=0.6 (the weight of the influence of soil moisture on seedling growth);

[0081] α=0.5, β=0.1 (time decay factor).

[0082] For times t0 and t1, M(t) can be calculated.

[0083] 1. Calculate M(t0):

[0084]

[0085] 2. Calculate M(t1):

[0086]

[0087] From the calculations above, it can be seen that the growth state M(t) of the seedlings is 0.3 at time t0, and 0.317 at time t1, indicating that the state of the seedlings has improved during this period, which is expected to be due to changes in soil moisture or light.

[0088] 3. Causal reasoning and effect function calculation

[0089] Based on the dynamic correlation model M(t) obtained above, the causal relationship between variables is analyzed using the causal reasoning method. The formula is:

[0090]

[0091] in:

[0092] C(x i ,x j ) represents the variable x i x j causal effect;

[0093] Represents x i The impact rate on the model output M(t);

[0094] g(x j ,t) is the variable x j The response function at time t.

[0095] The causal effect of x1 (soil moisture) on x2 (light intensity) will be calculated. Setting:

[0096] (The impact rate of soil moisture on seedling status);

[0097] g(x2,t)=0.5 (response function of light intensity).

[0098] 1. Calculate the causal effect C(x1,x2):

[0099]

[0100] Assume t1-t0 = 1 hour, then:

[0101] C(x1,x2)=0.3·1=0.3

[0102] This means that the causal effect of soil moisture on light intensity is 0.3. Since the causal effect is greater than the preset threshold γ=0.2, it is considered that there is a significant causal relationship between soil moisture and light intensity.

[0103] 4. Abnormal clustering and early warning

[0104] Through causal reasoning analysis, the causal relationship between variables is known, and now anomaly detection and early warning are started. By calculating the abnormal deviation of the data at each time point, the anomaly clustering algorithm is used to identify different types of abnormal patterns. The clustering formula is:

[0105]

[0106] in:

[0107] A(k) is the clustering score of the k-th abnormal pattern;

[0108] z i is the principal component value of the abnormal data point;

[0109] d(z i ,z k ) is point z i With cluster center z k The Euclidean distance of .

[0110] It is assumed that two types of abnormal patterns are identified: water stress and disease, and the clustering scores of each type of abnormality are:

[0111] Water stress mode: A1 = 0.85;

[0112] Disease mode: A2=0.92.

[0113] If A1>0.8 or A2>0.8, an early warning signal is triggered. For example, if the cluster score of water stress A1=0.85, the system will trigger an irrigation early warning to remind you to replenish water.

[0114] Through the above four steps, the abnormal state warning mechanism of seedlings based on causal reasoning is realized. From data collection, feature dimension reduction, dynamic model establishment, causal reasoning to abnormal pattern clustering, the whole process provides real-time monitoring and precise management solutions for garden seedlings. The model effectively improves the management efficiency of garden seedlings through dynamic association, causal reasoning, anomaly detection and other methods, timely warns of potential problems, and optimizes management strategies.

[0115] Embodiment 2:

[0116] See also Figure 3,In this embodiment, the digital twin technology collects seedling growth and ,environmental parameters through a multimodal sensor network, and combines GIS and ,spatial analysis to build a digital twin model of seedlings, realizing real-time monitoring and ,dynamic management of the growth status of seedlings.

[0117] It is assumed that multimodal sensors are deployed in a certain garden. These sensors collect the growth parameters (such as height, chlorophyll concentration, photosynthesis efficiency) and environmental parameters (such as soil moisture, temperature, and light intensity) of seedlings in real time. These data are transmitted in real time via wireless networks, and combined with geographic information systems (GIS) and spatial analysis technology, a set of dynamic data streams are generated to describe the growth and distribution of seedlings in the garden area. Based on these data, a digital twin model of seedlings can be constructed, which dynamically depicts the overall growth process of seedlings based on the geometric structure, physiological state, and environmental interaction of seedlings.

[0118] Construction of digital twin model

[0119] The core formula of the digital twin model is:

[0120]

[0121] In this formula:

[0122] S(t) represents the overall growth status of the seedlings at time t, that is, the comprehensive status of the digital twin model.

[0123] G(t) is a geometric structure change model that describes the physical characteristics of seedlings, such as tree height and crown width.

[0124] It represents the cumulative effect integral from the initial time t0 to the current time t, which is used to calculate the long-term impact of environment and variables on seedling growth.

[0125] E(τ) is an environmental interaction model that describes the direct effects of environmental variables (such as soil moisture, temperature, light intensity, etc.) on seedlings at time τ.

[0126] z i (τ) is the value of the i-th environmental variable, which represents the specific environmental factors (such as light, temperature, etc.) that affect the growth of seedlings at time τ.

[0127] h i (τ) is a weight function, which indicates the influence of each variable on the model state.

[0128] α and β are adjustment factors used to balance the nonlinear dynamic behavior between variables, where α controls the basic influence strength of the variable and β is the decay rate of the time effect.

[0129] e -β·τIt is the time decay term, indicating that the influence of environmental factors on the growth status of seedlings gradually weakens over time.

[0130] Assume that a sensor is installed in a garden and collects the following data in real time:

[0131] The height of the seedlings is 2.5 meters at time t0 and 2.8 meters at time t1.

[0132] The soil moisture (x1) is 45% at t0 and 50% at t1.

[0133] The temperature (x2) is 22°C at t0 and 23°C at t1.

[0134] The light intensity (x3) is 500 Lux at t0 and 550 Lux at t1.

[0135] These data can be substituted into the formula for calculation. Set the weight function h1(τ) = 0.5, h2(τ) = 0.4, h3(τ) = 0.6, and the value range of the following adjustment factors:

[0136] α = 0.2 to 0.5, indicating the basic influence of environmental factors on seedling growth;

[0137] β = 0.05 to 0.1, indicating the decay rate of the effect of time.

[0138] Based on these settings, the comprehensive state S(t) in the digital twin model is calculated.

[0139] 1. Geometry structure change model G(t):

[0140] The geometric structure of the seedling (tree height and crown width) is assumed to change little during this period, so G(t) is set as a constant, representing the geometric state of the seedling at different times. For example, at t0, the tree height of the seedling is 2.5 meters and the crown width is 1.5 meters. At t1, the tree height is 2.8 meters and the crown width is 1.6 meters. For simplified calculations, G(t) is set to 2.5 meters.

[0141] 2. Environmental interaction model E(τ):

[0142] The environmental interaction model E(τ) describes the effects of different environmental factors on seedling growth. The effects of soil moisture, light intensity and temperature on seedling growth are set as follows:

[0143] The impact of soil moisture (x1) on seedlings: For every 1% increase in soil moisture, the seedling growth rate increases by 0.03 meters.

[0144] The effect of light intensity (x3) on seedlings: for every 50 Lux increase, the seedling growth rate increases by 0.02 meters.

[0145] Effect of temperature (x2) on seedlings: For every 1°C increase in temperature, the seedling growth rate increases by 0.01 meters.

[0146] Therefore, the environment interaction model E(t) is set at time t0 as:

[0147] E(t0)=0.03×(50)+0.02×(50)+0.01×(1)=1.5+1+0.01=2.51

[0148] 3. Cumulative Effect Model:

[0149] The environmental factors at the current time t1 are set as follows: z1(t1)=50 (soil moisture), z2(t1)=23 (temperature), z3(t1)=550 (light intensity), and the adjustment factors α=0.3, β=0.05.

[0150] Compute the integral part of the cumulative effect:

[0151]

[0152] Since E(τ) and z i The value of (τ) changes over time, and the cumulative effect of each environmental factor needs to be calculated. Taking soil moisture (z1) as an example, it is assumed that in the time period from t0 to t1, the soil moisture increases from 45% to 50%, and the light intensity increases from 500Lux to 550Lux. By simulating the changes in each environmental factor, we get:

[0153]

[0154] 4. Comprehensive state S(t1):

[0155] Finally, by adding up the above parts, the state of the digital twin model at time t1 is obtained:

[0156]

[0157] S(t1)=2.8+2.51+1.45=6.76

[0158] Through these steps, the digital twin model was used to describe the comprehensive growth status of the seedlings. Through real-time sensor data, environmental interaction models, and integral calculations of cumulative effects, a dynamic seedling digital twin model was successfully constructed. This method can accurately capture real-time changes in the growth process of seedlings and provide decision support for garden management.

[0159] In the digital twin technology of garden seedlings, virtual experiments are used to evaluate the dynamic impact of different management strategies on seedling growth, which can simulate the long-term effects of environmental changes and management interventions on seedlings, thereby optimizing management plans.

[0160] The digital twin model is set to optimize the management of a certain seedling. The goal is to explore the impact of different environmental parameters (such as temperature and light) and management strategies (such as irrigation amount and fertilization frequency) on the growth of seedlings through virtual experiments, and evaluate the management effect on this basis. The optimization objective function is set as:

[0161]

[0162] in:

[0163] P is the comprehensive score of the management strategy, which is used to quantify the effectiveness of the strategy in the virtual experiment.

[0164] T=t f -t0 is the total time period of the experiment, which means the time from the start to the end of the experiment.

[0165] R(t) is the growth rate at time t, which represents the increment of seedlings, that is, the rate of change of growth indicators such as tree height, crown width or leaf area of ​​seedlings.

[0166] C(t) is the resource consumption at time t, indicating the usage of water, nutrients and other resources by the seedlings at that moment.

[0167] λ1 and λ2 are the weight coefficients of growth rate and resource consumption, respectively. λ1 reflects the preference for growth, while λ2 represents the demand for resource conservation.

[0168] A virtual experiment was set up with two management strategies for seedlings:

[0169] Management strategy A: Increase light intensity and increase irrigation appropriately, while keeping fertilizer application frequency unchanged.

[0170] Management Strategy B: Lower temperatures and reduce irrigation, and apply fertilizer more frequently.

[0171] First, set the following initial conditions:

[0172] Start time t0 = 0, end time t f =10 days.

[0173] The growth rate R(t) varies with light intensity and temperature during the experiment. The growth rates under the two strategies are:

[0174] For strategy A, the growth rate varied over time, with an initial rate of 0.2 m / day, increasing to 0.3 m / day after 10 days.

[0175] For strategy B, the growth rate varied over time, with an initial rate of 0.15 m / day, increasing to 0.25 m / day after 10 days.

[0176] Resource consumption C(t) represents the water and nutrient consumption of seedlings under different management strategies, set:

[0177] For strategy A, resource consumption was 2.5 L / day in the first 5 days and 3 L / day in the next 5 days.

[0178] For strategy B, resource consumption was 3 L / day in the first 5 days and 2 L / day in the next 5 days.

[0179] 1. Calculate the comprehensive score P:

[0180] First, the growth rate R(t) and resource consumption C(t) are brought into the objective function for integration. Since the value ranges of λ1 and λ2 are 0.5 to 1.0, λ1 = 0.8 and λ2 = 0.4 are set to reflect the priority of growth and resource conservation, respectively.

[0181] For management strategy A, calculate the integral term:

[0182]

[0183] in:

[0184] R A (t) 0.2 m / day from day 0 to day 5 and 0.3 m / day for the next 5 days;

[0185] C A (t) 2.5 L / day on days 0 to 5, and 3 L / day on days 5.

[0186] Points calculation:

[0187]

[0188] For management strategy B, a similar calculation is performed:

[0189]

[0190] in:

[0191] R B (t) 0.15 m / day from day 0 to day 5 and 0.25 m / day for the next 5 days;

[0192] C B (t) 3 L / day on days 0 to 5, and 2 L / day on days 5.

[0193] Points calculation:

[0194]

[0195] Through virtual experiments, the effects of two management strategies A and B were evaluated. The calculation results showed that the comprehensive score P of management strategy A was A =-0.9 and the overall score of management strategy B B =-0.84. Although both are negative values, management strategy A is slightly superior in maximizing growth rate. By adjusting the weight coefficients λ1 and λ2, and refining resource consumption and management strategies, the scheme can be further optimized to achieve a balance between seedling growth and resource utilization.

[0196] In the digital twin management technology of the whole life cycle of garden seedlings, the optimal management strategy is screened through virtual experiments, and adjusted in combination with actual operating conditions to form a closed-loop management process, which enables the two-way integration of virtual and physical. Assuming that the seedlings of a certain garden are being optimized by the digital twin management system, the following is the implementation process and calculation example of this scheme. First, the results of the virtual test of the seedlings show that the best management strategy is to increase the amount of irrigation and appropriately increase the frequency of fertilization. However, considering the conditions in actual operation, such as soil type and climate change, these management strategies need to be further adjusted. Then, through the IoT platform, the growth data of the physical seedlings, such as soil moisture, temperature and other environmental parameters, are collected in real time, and these data are fed back to the digital twin model to ensure that the geometry, physiological state and environmental interaction of the model are consistent with the actual seedlings. By continuously synchronizing the virtual and physical systems, the real-time and accuracy of seedling management are ensured. Assuming that a certain seedling is selected for prediction of future growth status, the following prediction function is used:

[0197]

[0198] in:

[0199] F(t) is the predicted value of the future state after time t, indicating the comprehensive growth of the seedlings at a certain moment in the future;

[0200] T p is the length of the prediction time window, which represents the future time range from the current moment. For example, if we select T p =30 days;

[0201] G′(t) is the time derivative of the geometric structure model, which represents the rate of change of the physical characteristics of the seedlings (such as tree height, crown width, etc.);

[0202] E′(t) is the time derivative of the environmental interaction model, which represents the impact of changes in environmental variables on future states (such as soil moisture, light, etc.);

[0203] k1 and k2 are the weight coefficients of the geometric structure and environmental interaction, respectively, and the value range is set to 1 and k2 = 0.4, reflecting the relative importance of geometric changes and environmental interactions in future growth trends.

[0204] The current growth state of the seedlings is set as: tree height 3.0 meters, crown width 2.5 meters, and environmental parameters (temperature 22°C, soil moisture 30%) are the current actual conditions. The current geometric structure derivative G′(t) of the model is set to 0.1 meters / day, indicating that the tree height increment of the seedlings is 0.1 meters every day, and the environmental interaction derivative E′(t) is set to 0.05 meters / day, indicating the incremental impact of the environment on the seedlings.

[0205] The prediction time window is selected as 30 days, that is, t p = 30 days, and then the future state prediction is calculated.

[0206] 1. Influence of geometric structure change: The time derivative of geometric structure change is set to G′(t) = 0.1 m / day.

[0207]

[0208] 2. Impact of environmental interaction: Set the time derivative of environmental interaction E′(t) = 0.05 m / day.

[0209]

[0210] 3. Comprehensive prediction value: Substitute the above results into the future state prediction formula:

[0211]

[0212] F(t)=3.0+1.8+0.6=5.4 m

[0213] According to the calculation results, the height of the seedling is predicted to reach 5.4 meters in 30 days. At this time, the IoT platform will collect the current growth data in real time and adjust the actual management strategy according to the prediction results, such as increasing the frequency of irrigation or applying the right amount of fertilizer to ensure that the growth of the seedling meets the expected state in the digital twin model.

[0214] Through this process of prediction and real-time adjustment, the growth process of seedlings is precisely controlled, and the virtual model is effectively combined with actual management to ensure scientific and real-time management. Each step in this process is supported by data and calculations, making the full life cycle management of garden seedlings more efficient and sustainable.

[0215] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A data management method for garden seedlings, characterized in that The following steps are involved: S1. High-dimensional dynamic data modeling based on plant feature extraction: S1.

1. Collect physiological characteristics data and environmental parameters of garden seedlings by deploying a high-precision sensor network; S1.

2. Use feature engineering technology to reduce data dimension, extract key parameters and establish a dynamic correlation model; S1.3, using adaptive algorithms to update the model in real time to achieve dynamic association between seedling growth characteristics and environmental variables; S2. Early warning mechanism for abnormal seedling status based on causal reasoning: S2.

1. Based on the high-dimensional dynamic data model, a causal network of seedling status was constructed to identify the causal relationship between different variables; S2.2, classify different types of anomalies through abnormal pattern clustering algorithm and trigger intelligent warning, and generate response suggestions for abnormal status; S3. Multi-scale geospatial analysis and precision planting planning: S3.

1. Collect multi-scale geospatial data of the garden, including macro-level garden layout, meso-level seedling distribution, and micro-level individual plant status; S3.

2. Formulate planting plans through spatial data integration and optimization algorithms, and optimize spatial resource utilization based on the ecological relationship between seedlings; S4. Adaptive management optimization based on dynamic data: S4.

1. Based on dynamic data models and geospatial analysis results, design an intelligent optimization engine to optimize irrigation, fertilization, and pruning management operations in real time, and dynamically adjust the optimization strategy through a closed-loop feedback mechanism; S5. Digital twin technology for seedlings throughout their life cycle: S5.

1. Input the dynamic data model, spatial analysis results and optimization management plan into the digital twin system, build a virtual model of garden seedlings, and control the real-time synchronization between physical seedlings and digital twins; S5.

2. Use virtual experiments to simulate the impact of different management plans on seedling growth, and apply them to actual operations after optimization to achieve digital management of the entire life cycle of seedlings.

2. The data management method for garden seedlings according to claim 1, characterized in that The seedling abnormal state early warning mechanism method based on causal reasoning: Multimodal sensors are deployed to collect data on key physiological and environmental variables of seedlings, including soil moisture, light intensity, and leaf temperature. The data are stored in the form of time series, denoted as a set X = {x1, x2, …, x n }, where x i represents the real-time observation value of the i-th variable; the data is processed using feature dimensionality reduction technology to obtain the principal component set Z = {z1, z2, …, z m }, where z i It represents the i-th principal component variable after dimensionality reduction, reflecting the comprehensive effect of the main influencing factors; The dynamic association model M(t) is established, and its formula is: Where: M(t) is the output of the dynamic correlation model at time t, which is used to describe the comprehensive correlation between seedling status and environmental variables; i (t) is the value of the i-th principal component at time t; f i (t) is the dynamic weight function representing the i-th variable over time t; α is the adjustment factor in the model, controlling the amplitude of the interaction term between variables; β is the decay rate of the time-varying influence factor, which is used to smooth the influence of time variation; e -β·t : Exponential decay term, indicating the weakening trend of the variable over time.

3. The data management method for garden seedlings according to claim 2 is characterized in that The seedling abnormal state early warning mechanism method based on causal reasoning: Based on the dynamic model M(t), a causal network is constructed using causal reasoning methods. The causal network is constructed through the effect function C(x i ,x j ) quantifies the causal relationship, the formula is as follows: Where: C(x i ,x j ):variable x i For variable x j the strength of the causal effect; variable x i The impact rate on the dynamic model output M(t) is represented by x i The importance of g(x j ,t): variable x j The response function is used to characterize the sensitivity of the dependent variable to time t; The cumulative impact from the initial time t0 to the current time t represents the time range of the causal effect; when C(x i ,x j )>γ, where γ is the threshold of causal effect; i x j There is a significant causal effect, which in turn establishes directional connections between variables in the causal network.

4. The data management method for garden seedlings according to claim 3 is characterized in that The seedling abnormal state warning mechanism based on causal reasoning: According to the dynamic results of the causal network, the abnormal deviation of each variable is calculated in real time; When a variable deviates from the normal range, it is marked as an anomaly, and an anomaly clustering method is used to classify the abnormal pattern; abnormal data points are clustered into different patterns, including water stress, disease, and nutrient deficiency; and specific warning signals are triggered for each pattern.

5. The data management method for garden seedlings according to claim 1, characterized in that The digital twin technology method of seedlings throughout the life cycle: Through a multimodal sensor network, the growth parameters of seedlings, including height, chlorophyll concentration, photosynthesis efficiency, and environmental parameters, including soil moisture, temperature, and light intensity, are collected in real time. The collected data are integrated into a dynamic data stream. Combined with the geographic information system GIS and spatial analysis, the distribution and growth space information of seedlings in the garden area are obtained to form a multi-dimensional dynamic data set for building a digital twin model of seedlings. The digital twin model describes the real-time state and dynamic change trend of seedlings based on their geometric structure, physiological state, and environmental interaction. The state model is defined as: Among them: S(t) is the comprehensive state of the digital twin model at time t, which represents the virtual model result of the overall growth of the seedlings; G(t) is the geometric structure change model, which describes the physical characteristics of the seedlings; is the integral operation of cumulative effect, from the initial time t0 to the current time t, used to calculate the long-term impact of environment and variables on growth status; E(τ) is the environmental interaction model, describing the direct effect of environmental variables, including light intensity and soil moisture on seedlings at time τ; z i (τ) is the value of the i-th key variable at time τ, which represents the measured value of a specific environmental factor, including temperature and light; h i (τ) is the weight function, reflecting the variable z i (τ) The degree of influence on the overall model state; α, β: adjustment factors used to balance nonlinear dynamic behavior, α controls the basic influence intensity of the variable, and β represents the decay rate of the time effect; e -β·τ is the time decay term, which means that the influence of the variable on the state gradually weakens over time.

6. The data management method for garden seedlings according to claim 5, characterized in that The digital twin technology method for seedlings throughout their life cycle: conduct virtual experiments in a digital twin model to simulate the dynamic effects of different schemes on seedling growth by changing environmental parameters, including temperature and light; or management strategies, including irrigation amount and fertilization frequency.

7. The data management method for garden seedlings according to claim 6, characterized in that The digital twin technology method of seedlings throughout the life cycle: Screen management strategies based on virtual test results, adjust them based on actual operating conditions, and feed the optimization results back to the physical seedling management system to form a closed-loop management process that integrates virtual and physical elements. Through the IoT platform, the real-time data of physical seedlings is dynamically synchronized with the digital twins; Record the data of the entire process from planting to maturity of seedlings, and establish a full life cycle management file; predict future growth trends based on the digital twin model, and calculate the future state through the following prediction functions: Where: F(t) is the predicted value of the future state after time t, indicating the comprehensive growth of the seedlings at a certain time in the future; T p is the length of the prediction time window, that is, the time range from the current moment in the future; G′(t) is the time derivative of the geometric structure model, which represents the rate of change of the physical properties of the seedlings; E′(t) is the time derivative of the environmental interaction model, which represents the impact of changes in environmental variables on future states; k1 and k2 are the weight coefficients of geometric changes and environmental interactions, respectively. k1 and k2 are used to balance the relative contributions of geometric structures and environmental variables in future trends.

Images