Business service management method and platform based on digital twinning and AI

Abstract

The invention provides a business service management method and platform based on digital twinning and AI. The method comprises the following steps: constructing a crop digital twinborn model based on crop growth environment data and crop growth state data; performing correlation analysis on the equipment operation historical data and the meteorological historical data, and generating a growth fluctuation signal corresponding to a future preset time period based on a correlation analysis result; on the basis of the growth fluctuation signal, performing simulation analysis on the growth vigor of the crop in a future preset time period by using a crop digital twinborn model, and generating a crop growth vigor prediction result; and inputting a crop growth prediction result into the AI optimization model to obtain a planting management optimization scheme. According to the invention, the digital twinborn model is applied to crop growth prediction and management optimization, so that the planting effect of crops can be improved, and the yield and quality can be improved.

Description

Technical Field

[0001] This invention relates to the field of agricultural platform technology, and more specifically, to a business service management method and platform based on digital twins and AI. Background Technology

[0002] In traditional agricultural production management, farmers mostly rely on their long-accumulated experience to formulate planting and management plans. However, experience-based judgments often lack precision and scientific rigor, making it difficult to fully and accurately grasp the complex changes in the farmland environment and the subtle differences in the crop growth process. This results in low resource utilization efficiency and makes it difficult to effectively guarantee the yield and quality of agricultural products.

[0003] With the rapid development of information technology, digital twin technology has gradually emerged and been applied in the agricultural field. Some agricultural management platforms have begun to attempt to build virtual models based on digital twins to digitally simulate crop growth environments (such as soil moisture, temperature, and light intensity) and crop growth status. However, current applications of digital twin technology mainly focus on the digital simulation of crops, while research on using digital twin technology to assist in crop planting and management is significantly insufficient.

[0004] This application is a further application of digital twin technology to assist crop planting management based on digital twin technology, thereby improving crop yield and quality. Summary of the Invention

[0005] To address the technical problems existing in the background art, the present invention provides a business service management method, platform, electronic device, computer storage medium, and computer program product based on digital twins and AI.

[0006] The first aspect of this invention provides a business service management method based on digital twins and AI, comprising the following steps: By collecting crop growth environment data and crop growth status data in real time through sensing devices deployed in the planting greenhouse, a crop digital twin model is constructed based on the crop growth environment data and crop growth status data. Acquire historical planting data for the greenhouse, including equipment operation history data and meteorological history data. Perform correlation analysis on the equipment operation history data and meteorological history data, and generate growth fluctuation signals corresponding to the future preset time period based on the correlation analysis results. Based on the growth fluctuation signal, the crop digital twin model is used to simulate and analyze the crop growth in a future preset period and generate crop growth prediction results. The crop growth prediction results are input into the AI ​​optimization model to obtain an optimized planting management plan; wherein, the AI ​​optimization model is embedded with a crop growth-management measure association algorithm obtained through pre-training.

[0007] A second aspect of the present invention provides a business service management platform based on digital twins and AI, comprising: The modeling unit receives real-time crop growth environment data and crop growth status data collected by sensing devices deployed in the planting greenhouse, and constructs a crop digital twin model based on the crop growth environment data and the crop growth status data. The fluctuation signal generation unit acquires historical planting data of the greenhouse, including historical equipment operation data and historical meteorological data, performs correlation analysis on the historical equipment operation data and historical meteorological data, and generates a growth fluctuation signal corresponding to a preset future time period based on the correlation analysis results. The growth prediction unit, based on the growth fluctuation signal, uses the crop digital twin model to simulate and analyze the crop growth in a future preset period and generates crop growth prediction results. The management optimization unit inputs the crop growth prediction results into the AI ​​optimization model to obtain an optimized planting management plan; wherein, the AI ​​optimization model is embedded with a crop growth-management measure association algorithm obtained through pre-training.

[0008] A third aspect of the present invention provides an electronic device comprising: a memory storing executable program code; a processor coupled to the memory; the processor invoking the executable program code stored in the memory to perform the method as described in any of the preceding claims.

[0009] A fourth aspect of the present invention provides a computer storage medium storing a computer program that, when executed by a processor, performs the method described in any of the preceding claims.

[0010] A fifth aspect of the invention provides a computer program product comprising program code tools for performing the steps of the method as described in any of the preceding claims.

[0011] This invention collects environmental and growth data in real time using multiple types of sensing devices, constructs a precisely mapped digital twin model through preprocessing, and generates growth fluctuation signals by combining historical data correlation analysis, which greatly improves the accuracy and foresight of crop growth prediction and provides a scientific basis for decision-making. At the same time, the AI ​​optimization model relies on pre-trained algorithms to generate targeted management plans based on prediction results, realizing precise allocation of resources such as water and fertilizer and improving utilization efficiency. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart illustrating a business service management method based on digital twins and AI disclosed in an embodiment of the present invention.

[0014] Figure 2 This is a schematic diagram of the system architecture of the agricultural business service management solution disclosed in the embodiments of the present invention.

[0015] Figure 3 This is a schematic diagram of the structure of a business service management platform based on digital twins and AI disclosed in an embodiment of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, exemplary embodiments will be described in detail below, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0017] like Figure 1 As shown in the figure, a business service management method based on digital twin and AI according to an embodiment of the present invention includes the following steps: 100. Real-time data collection of crop growth environment and crop growth status is achieved through sensing devices deployed in the planting greenhouse, and a crop digital twin model is constructed based on the crop growth environment and crop growth status data.

[0018] like Figure 2 As shown, the solution of the present invention is applied to an agricultural business service management platform, which is a cloud platform. It is connected to the management servers of multiple planting greenhouses to receive sensor data uploaded by each management server and transmit the corresponding planting management plan or planting management optimization plan to the corresponding management server.

[0019] The greenhouse is equipped with various types of sensors, such as soil sensors to collect environmental data like soil moisture and nutrient content, and image and infrared sensors to capture data on crop growth status, including plant height, leaf area, and color. These sensors operate in real time at a preset sampling frequency to ensure the timeliness and continuity of the data.

[0020] The management platform constructs a digital twin model of crops based on collected environmental and status data, enabling precise mapping between the physical greenhouse and the virtual model. The established digital twin model integrates fundamental information such as the spatial structure parameters of the greenhouse and the growth characteristics of crop varieties, simulating crop growth patterns in a data-driven manner. For example, the digital twin model can simulate root water absorption based on changes in soil moisture and calculate photosynthetic efficiency based on light intensity data, ultimately achieving dynamic simulation of crop growth status and serving as a virtual carrier for subsequent growth prediction. Understandably, before constructing the crop digital twin model, data preprocessing is necessary, including data cleaning (removing outliers and filling missing values) and standardization (unifying data format and units).

[0021] 200. Obtain historical planting data for the greenhouse, including historical equipment operation data and historical meteorological data. Perform correlation analysis on the historical equipment operation data and historical meteorological data, and generate growth fluctuation signals corresponding to the future preset time period based on the correlation analysis results.

[0022] First, retrieve the historical planting data for the corresponding greenhouse from the database. The historical equipment operation data includes records of operating parameters of greenhouse temperature control equipment, irrigation equipment, ventilation equipment, etc. (such as heating time, irrigation flow rate, fan speed, etc.); the historical meteorological data includes records of changes in environmental parameters such as historical temperature, humidity, light duration, and CO2 concentration in the greenhouse.

[0023] When conducting correlation analysis on the two types of historical data mentioned above, statistical methods (such as Pearson correlation coefficient and partial least squares regression) are used to uncover the influence patterns of equipment operating parameters and meteorological parameters on crop growth. For example, the correlation between the operating time of heating equipment in winter and nighttime temperature fluctuations, as well as the relationship between temperature fluctuations and crop growth rates, are analyzed. Based on the correlation analysis results, key fluctuation factors affecting crop growth (such as sudden temperature changes caused by equipment failure and seasonal changes in light intensity) are identified, and growth fluctuation signals for future preset periods (such as the next 7 days and 15 days) are generated using time series prediction algorithms. Thus, this growth fluctuation signal quantifies the possible deviations in crop growth caused by various factors, providing more realistic interference variables for subsequent simulation analysis.

[0024] 300. Based on the growth fluctuation signal, the crop digital twin model is used to simulate and analyze the crop growth in a future preset period to generate crop growth prediction results.

[0025] During the simulation analysis, the growth fluctuation signal generated in step 200 is used as input parameters to import into the crop digital twin model. The crop digital twin model, combined with the current crop growth status data, simulates the crop growth process in a fluctuating environment within a preset future time period. The simulation is conducted using a dynamic iterative approach, updating the model parameters at regular time intervals (e.g., 12 hours, 24 hours). Environmental variables are adjusted based on the fluctuation signals (e.g., simulating insufficient water due to irrigation equipment failure, and changes in light intensity due to extreme weather), and the impact of these changes on crop physiological processes (e.g., cell division, nutrient absorption, fruit development, etc.) is calculated.

[0026] The generated crop growth prediction results include multi-dimensional indicators, such as plant height growth curves, leaf number change trends, and fruit maturity predictions. For example, the model can predict that if there are 3 days of weak light fluctuations in the next 10 days, the average daily growth of crop height will decrease from the normal 2cm to 1.5cm, and the fruit enlargement rate will slow down by 15%. These prediction results can be presented in the form of visual charts or structured data.

[0027] 400, The crop growth prediction results are input into the AI ​​optimization model to obtain the planting management optimization plan; wherein, the AI ​​optimization model is embedded with a crop growth-management measure association algorithm obtained through pre-training.

[0028] The AI ​​optimization model employs a deep learning architecture (such as convolutional neural networks and recurrent neural networks), pre-trained using large-scale historical planting data, thereby embedding a crop growth-management linkage algorithm. The training data needs to include growth data for different crop varieties and under different environmental conditions, along with records of the effectiveness of corresponding management measures (such as fertilizer application rate, irrigation frequency, and timing of pest and disease control). Sufficient training is used to learn from this data and uncover the mapping relationship between growth characteristics, management measures, and yield / quality.

[0029] After inputting the crop growth prediction results generated in step 300 into the model, the AI ​​optimization model first extracts features from the prediction results to identify key growth indicators (such as slow growth, risk of nutrient deficiency, etc.). Then, it calls a pre-trained crop growth-management measure association algorithm to match the optimal combination of management measures that can improve the current predicted growth. For example, if the prediction results show that the crop has a risk of leaf yellowing due to insufficient potassium fertilizer, the AI ​​optimization model recommends specific potassium fertilizer application rates, application times, and corresponding irrigation strategies based on historical data to ensure maximum nutrient absorption efficiency. The final generated planting management optimization plan includes specific parameters, execution sequence, and expected effects for each measure, providing growers with precise operational guidance.

[0030] This invention collects environmental and growth data in real time using multiple types of sensing devices, constructs a precisely mapped digital twin model through preprocessing, and generates growth fluctuation signals by combining historical data correlation analysis, which greatly improves the accuracy and foresight of crop growth prediction and provides a scientific basis for decision-making. At the same time, the AI ​​optimization model relies on pre-trained algorithms to generate targeted management plans based on prediction results, realizing precise allocation of resources such as water and fertilizer and improving utilization efficiency.

[0031] Optionally, constructing a crop digital twin model based on the crop growth environment data and the crop growth status data includes: 101. Determine the environmental sensitivity parameters of the crops currently planted in the greenhouse. The environmental sensitivity parameters include the critical threshold of soil moisture, the range of light intensity adaptation, and the response coefficient to temperature changes. Based on the environmental sensitivity parameters, determine the sensitivity level of the crops to different environmental factors, including high sensitivity, medium sensitivity, and low sensitivity. First, retrieve the environmental sensitivity parameters of the crops currently grown in the greenhouse from the crop growth characteristic database. These environmental sensitivity parameters are key indicators based on crop biological characteristics and long-term planting practices, such as the critical threshold of soil moisture (e.g., the suitable humidity range for a certain vegetable is 60%-80%, and growth will be significantly inhibited if it is below 50% or above 90%), the adaptable range of light intensity (e.g., the light saturation point of light-loving crops is 80,000 lux, and photosynthetic efficiency will decrease if it is below 30,000 lux), and the temperature change response coefficient (e.g., the degree of influence of a 1°C temperature fluctuation on the crop growth rate).

[0032] Based on these parameters, environmental factors are classified into three sensitivity levels: high, medium, and low. For example, for tropical crops, temperature can be classified as a high-sensitivity factor (temperature fluctuations exceeding 3°C severely affect growth), light intensity as a medium-sensitivity factor (moderate fluctuations only slightly affect yield), and soil pH as a low-sensitivity factor (fluctuations within a certain range have little impact on growth).

[0033] 102. A refined modeling approach is used for highly sensitive environmental factors, a conventional modeling approach is used for moderately sensitive environmental factors, and a simplified modeling approach is used for low-sensitive environmental factors. In the process of establishing the digital twin model, a refined modeling approach is adopted for highly sensitive environmental factors. Specifically, this involves increasing the data sampling frequency (e.g., from the usual 2 hours / time to 30 minutes / time), increasing the variable dimensions of the corresponding parameters in the model (e.g., collecting not only air temperature, but also leaf surface temperature and root zone temperature simultaneously), and refining the simulation logic (e.g., simulating the immediate impact of sudden temperature changes on crop transpiration and the subsequent recovery process).

[0034] For moderately sensitive environmental factors, a conventional modeling approach is adopted, maintaining a standard sampling frequency (e.g., 1-2 hours / time) and basic simulation dimensions, capturing only significant changes in environmental factors (e.g., triggering model parameter adjustments when light intensity fluctuations exceed 20%). For low-sensitivity environmental factors, a simplified modeling approach is used, reducing the sampling frequency (e.g., 12 hours / time) and simplifying simulation calculations. For example, the impact of minor fluctuations in soil nutrients on crop growth is ignored, and simulations are only performed when nutrient content exceeds safe limits.

[0035] This differentiated modeling approach ensures accurate characterization of key environmental factors while avoiding redundant calculations, thereby significantly improving the model's operational efficiency.

[0036] 103. Based on the above modeling method, the crop digital twin model is constructed.

[0037] In practice, the modeling modules for different environmental factors with varying sensitivities in step 102 need to be integrated collaboratively to establish a correlation mechanism between the modules. For example, the highly sensitive temperature module and the moderately sensitive humidity module can be linked through a crop transpiration model. Simultaneously, basic crop growth algorithms (such as photosynthesis models and dry matter accumulation models) should be incorporated to enable the model to dynamically simulate crop growth based on multi-dimensional environmental data.

[0038] For example, for the highly sensitive soil moisture module and the low sensitive soil compaction module, the crop digital twin model will prioritize responding to the impact of soil moisture changes on root growth, while using soil compaction as a background parameter and adjusting the simulation results only when it exceeds the crop's tolerance range.

[0039] The digital twin model constructed in this embodiment can accurately reflect the impact of key environmental factors on crop growth and efficiently process massive amounts of data.

[0040] Optionally, a correlation analysis is performed on historical equipment operation data and historical meteorological data, and a growth fluctuation signal corresponding to a preset future time period is generated based on the correlation analysis results, including: 201. Extract key operating parameter sequences from historical equipment operation data, including start-stop duration sequences for temperature control equipment, flow rate change sequences for irrigation equipment, and speed adjustment sequences for ventilation equipment; extract environmental parameter sequences for corresponding time periods from historical meteorological data, including temperature fluctuation sequences, humidity change sequences, and light intensity fluctuation sequences. In this step, the sequence of key operating parameters extracted from the equipment's historical operating data directly reflects the working status of the greenhouse control equipment. For example, the start-up and shutdown duration sequence of temperature control equipment reflects the duration of heating or cooling operations, the flow rate change sequence of irrigation equipment reflects the dynamic adjustment of water supply, and the speed adjustment sequence of ventilation equipment is related to the efficiency of air circulation and gas exchange within the greenhouse.

[0041] The environmental parameter sequences extracted from historical meteorological data are core environmental factors that significantly affect crop growth. For example, temperature fluctuation sequences record diurnal variations and abnormal fluctuations in greenhouse temperature, humidity variation sequences reflect the dynamic balance of air and soil moisture, and light intensity fluctuation sequences include changes in the intensity of natural light and artificial supplemental lighting.

[0042] 202. The sliding window algorithm is used to align the key operating parameter sequence and the environmental parameter sequence in time, calculate the Pearson correlation coefficient between the equipment operating parameters and the environmental parameters in different window intervals, and screen out strongly correlated parameter pairs whose absolute values ​​of the correlation coefficient are greater than a preset threshold. In this step, a sliding window algorithm (e.g., setting the window size to 1 hour) is used to align the key operating parameter sequences and environmental parameter sequences over time, ensuring that the analysis focuses on the parameter changes within the same time period. For example, the start-up and shutdown durations of temperature control equipment within a certain 1-hour period are matched with the temperature fluctuation data for the corresponding time period to avoid analytical biases caused by time misalignment.

[0043] The correlation strength between equipment operating parameters and environmental parameters is quantified by calculating the Pearson correlation coefficient ([-1,1]) of parameters within different time window intervals. Strongly correlated parameter pairs are screened. For example, the correlation coefficient between "heating equipment start-up and shutdown time" and "temperature fluctuation amplitude" is 0.75, which is greater than the preset threshold of 0.6. Therefore, the two are determined to be strongly correlated, indicating that the operation of the heating equipment directly affects temperature changes. Parameter pairs with low correlation (such as ventilation equipment speed and soil moisture) are excluded to reduce redundant analysis.

[0044] 203. Perform time-series feature extraction on strongly correlated parameter pairs to identify environmental parameter mutation events caused by abnormal equipment operation, and record the time node, impact magnitude and duration of the event; perform feature mining on each of the environmental parameter mutation events to form the occurrence pattern of environmental parameter mutation events; In this step, when extracting time-series features from strongly correlated parameter pairs, abnormal thresholds (such as a sudden drop in temperature of 5°C within 1 hour) are set to identify sudden environmental parameter events caused by abnormal equipment operation, such as a sudden drop in temperature caused by a malfunction in a temperature control device or a sudden increase in humidity caused by abnormal irrigation pump flow. At the same time, the key characteristics of the event are recorded in detail, including but not limited to the time node (such as noon in summer), the magnitude of the impact (such as a temperature deviation of 3°C from the normal value), and the duration (such as 2 hours).

[0045] When performing feature mining on a large number of environmental parameter abrupt change events, statistical analysis and data mining algorithms (such as cluster analysis and association rule mining) are used to extract the patterns of event occurrence. For example, the analysis found that the event of "temperature abrupt change caused by heating equipment failure at night in winter" occurred 30% more frequently in greenhouses with equipment that had been in use for more than 3 years, with an average duration of 1.5 hours and a maximum temperature drop of 4°C.

[0046] Understandably, these patterns quantify the conditions for occurrence, the degree of impact, and the distribution characteristics of abrupt changes in environmental parameters.

[0047] 204. Based on the occurrence patterns of abrupt environmental parameter events and basic meteorological forecast data for a predetermined future period, the growth fluctuation signal, which includes the amplitude, duration, and probability of environmental parameter fluctuations, is generated using an LSTM time series prediction model; wherein, high-probability fluctuation events are marked as key attention items in the growth fluctuation signal.

[0048] In this step, the patterns of historical abrupt events (such as seasonal distribution and correlation with equipment age) and basic meteorological forecast data for a preset future period (such as outdoor temperature and sunshine trends for the next 7 days) are used as inputs and imported into an LSTM time series forecasting model. This model predicts potential future fluctuations in environmental parameters by learning the temporal dependencies in historical data.

[0049] The generated growth fluctuation signal contains three key three-dimensional information: fluctuation amplitude (e.g., temperature may fluctuate by ±2℃), duration (e.g., the fluctuation is expected to last 3 hours), and probability of occurrence (e.g., 60%). High-probability fluctuation events (e.g., a sudden increase in humidity due to irrigation equipment malfunction with a probability >70%) are marked as key concerns, highlighting their potential impact on crop growth. This growth fluctuation signal provides realistic disturbance variables for the simulation analysis of the digital twin model, making growth prediction more forward-looking and accurate.

[0050] Optionally, based on the growth fluctuation signal, the crop digital twin model is used to simulate and analyze the crop growth in a future preset period to generate crop growth prediction results, including: 301. Perform structured analysis on the growth fluctuation signal, extract the high-probability fluctuation events marked as key concerns, and determine the environmental parameter type, fluctuation amplitude, expected duration and occurrence time window of each high-probability fluctuation event; In this step, the growth fluctuation signals are structured and analyzed to extract high-probability fluctuation events marked as key concerns from the signal data. These high-probability fluctuation events are key variables that may significantly affect crop growth within a predetermined time period. Their core characteristics include: type of environmental parameters (such as specific influencing factors like temperature, humidity, and light), fluctuation amplitude (such as a sudden temperature increase of 3°C or a sudden humidity decrease of 10%), expected duration (such as the fluctuation lasting for 4 hours), and occurrence time window (such as 9:00-13:00 the next day).

[0051] For example, if the growth fluctuation signal is marked as "summer noon light intensity suddenly increases by 50%, lasts for 2 hours, with a probability of 85%" as a key focus item, then the light type, 50% fluctuation amplitude, 2-hour duration and specific time period of this high-probability fluctuation event need to be extracted as event parameters for subsequent simulation.

[0052] 302. Based on the preset crop sensitivity parameter thresholds in the crop digital twin model, the impact level of high-probability fluctuation events is assessed and classified into mild impact, moderate impact, and severe impact. In this step, based on the preset crop sensitivity parameter thresholds in the crop digital twin model (such as the suitable temperature range of a crop being 20-28℃ and the tolerance limit being 15-32℃), the extracted high-probability fluctuation events are classified into levels: (1) Mild impact refers to fluctuations within the normal tolerance range of crops (such as a short-term temperature fluctuation to 29℃, which does not exceed the 32℃ limit) and has no significant interference with growth; (2) Moderate impact refers to fluctuations that exceed the normal range but do not reach the critical value and last for a short period of time, such as the temperature rising to 33°C and lasting for 12 hours (not exceeding 24 hours), which may lead to a temporary slowdown in growth; (3) Severe impact refers to fluctuations that significantly exceed the tolerance limit and last for a long time, such as when the temperature drops to 14°C and lasts for more than 24 hours, which may cause physiological damage to crops.

[0053] 303. During the simulation of the digital twin model, the environmental parameter change data in the growth fluctuation signal are imported sequentially according to the time window, and a refined simulation strategy is adopted for high-probability fluctuation events. In this step, during the operation of the digital twin model, the environmental parameter change data in the growth fluctuation signal are imported into the digital twin model segment by segment according to the time window sequence of high-probability fluctuation events to simulate the dynamic changes of the real environment.

[0054] Meanwhile, a differentiated and refined simulation strategy is adopted for high-probability fluctuation events, as follows: For severe impact events, the simulation time step is shortened from the conventional 12 hours / time to 1 hour / time, tracking the instantaneous changes in crop physiological indicators (such as leaf stomatal conductance and root water absorption rate) in real time; for moderate impact events, the standard time step of 6 hours / time is maintained, focusing on monitoring changes in key crop growth nodes (such as pollination efficiency during flowering and fruit enlargement rate); and mild impact events are integrated into the regular simulation process without additional adjustment of the time step. For example, for a "severe low-temperature event lasting 24 hours," the digital twin model updates the low-temperature environmental parameters every hour to simulate the dynamic process of decreased crop cell activity and weakened photosynthesis.

[0055] 304. After the simulation, the basic growth prediction data under normal environmental conditions and the deviation data under the influence of high-probability fluctuation events are integrated to generate crop growth prediction results that include plant height growth curve, biomass accumulation trend, key growth period node offset and risk warning indicators; among them, the risk warning indicators correspond to the growth abnormalities that may be caused by severe impact events.

[0056] In this step, after the simulation is completed, basic growth prediction data under normal environmental conditions (without significant fluctuations) (such as normal daily plant height growth and biomass accumulation rate) are retrieved and fused with deviation data under the influence of high-probability fluctuation events (such as a 0.5 cm / day reduction in plant height growth and a 2-day delay in the growth period due to severe low temperatures). The final crop growth prediction results contain multi-dimensional information: the plant height growth curve visually shows the trend of the impact of fluctuation events on height; the biomass accumulation trend reflects the changes in dry matter accumulation; the offset of key growth period nodes (such as the delay of flowering from day 30 to day 32) quantifies the interference of events on the growth cycle; and risk warning indicators (such as "low temperature stress may lead to a 15% yield reduction") provide clear warnings for severe impact events.

[0057] The above crop growth prediction results not only reflect the normal growth trend of crops, but also highlight the potential impact of abnormal fluctuations, thus providing more decision-making basis for subsequent AI optimization models.

[0058] Optionally, inputting the crop growth prediction results into the AI ​​optimization model to derive an optimized planting management plan includes: 401. The crop growth forecast results are subjected to feature quantification processing to extract core growth indicators and risk characteristics, including the slope deviation value of the plant height growth curve, the deviation degree of biomass accumulation trend, the number of days of deviation of key growth period nodes, and the growth abnormality type corresponding to the risk warning indicator. In this step, feature extraction is performed on the crop growth prediction results to extract core growth indicators that quantify deviations and reflect growth dynamics. These include: the slope deviation of the plant height growth curve, used to measure the difference between the actual growth rate and the standard rate (e.g., a slope 0.2 cm / day lower than the baseline indicates slow growth); the deviation of the biomass accumulation trend, calculated by comparing the actual accumulation with the expected value (e.g., a deviation of 15% indicates insufficient biomass); and the number of days of shift at key growth stages, used to quantify the advancement or delay of the growth cycle (e.g., a flowering period shift of +3 days indicates a delayed growth period). Simultaneously, the types of growth anomalies corresponding to risk warning indicators are extracted, such as "low temperature stress leading to growth stagnation" and "insufficient light causing excessive vegetative growth," which serve as risk features.

[0059] 402. The extracted feature data is compared with the preset crop yield benchmark value and quality index threshold, and the difference quantification parameters corresponding to each growth index are calculated. In this step, the core growth indicators and risk characteristics extracted in step 401 are compared with preset targets to calculate the gap quantification parameters. The preset targets include a crop yield baseline (e.g., a target yield of 5000 kg / mu) and quality indicator thresholds (e.g., fruit sugar content ≥12%, vitamin C content ≥20 mg / 100g). When calculating the gap quantification parameters, the deficiencies can be presented intuitively using absolute values ​​or percentages, such as "predicted yield is 8% lower than the target value," "predicted soluble sugar content of fruit is 1.2% lower than the quality standard," and "flowering period deviation leads to a 5-day delay in harvest."

[0060] 403. The algorithm for crop growth-management measures association embedded in the AI ​​optimization model is invoked. Based on the gap quantification parameter, the historical optimization case library is retrieved, and successful management cases similar to the current growth characteristics are matched to generate an initial management measure combination, including fertilization plan, irrigation strategy, environmental control parameters and growth period intervention measures. In this step, the crop growth-management measure association algorithm embedded in the AI ​​optimization model is invoked. This algorithm learns the mapping relationship between "growth characteristics-management measures-effects" based on a large amount of historical data. Based on the gap quantification parameters in step 402, similar scenarios (such as "low temperature leading to less than 10% biomass") are retrieved from the historical optimization case library, and successful management measures that have been verified in practice are matched to generate an initial management measure combination. The combination covers multi-dimensional operations: the fertilization plan specifies the adjustment of the nitrogen, phosphorus and potassium ratio (such as increasing potassium fertilizer by 20%), and the frequency of foliar fertilizer spraying (once every 3 days); the irrigation strategy includes drip irrigation flow rate adjustment (increasing from 2L / h to 2.5L / h), and irrigation time optimization (avoiding midday high temperature); the environmental control parameters are set to supplemental light intensity (increased to 600μmol / m² / s) and CO2 concentration threshold (maintained at 800ppm); the growth period intervention measures specify the timing of spraying growth regulators (7 days before flowering).

[0061] 404. Through deep reinforcement learning, the initial management measure combination is dynamically optimized, the growth correction effect after the implementation of different measures is simulated, the cost-effectiveness of the measure implementation cost and expected benefits is calculated, and the measure combination with the best cost-effectiveness is selected. In this step, a deep reinforcement learning algorithm is used to dynamically optimize the initial combination of management measures. That is, by simulating the implementation effects of different combinations of measures multiple times, the growth correction trend is predicted. For example, the deviation in biomass accumulation can be reduced to 5% after fertilization adjustment.

[0062] At the same time, the implementation costs of the measures are calculated, including fertilizer procurement costs, equipment energy consumption costs, and labor operation costs. Combined with expected benefits (such as increased income from increased yield and premium from improved quality), the cost-benefit ratio is evaluated. For example, the input and output of two options, "increasing the duration of supplemental lighting by 2 hours" and "adjusting the fertilization ratio", are compared to select the combination of measures with the highest cost-effectiveness, ensuring that the improvement in growth is maximized under the premise of low cost.

[0063] 405. Generate a planting management optimization plan that includes a timeline of action plans corresponding to the optimal combination of measures, key parameter adjustment thresholds, expected growth improvement curves, and emergency response plans; wherein, the emergency response plan sets trigger conditions for severe impact events in the risk warning indicators.

[0064] In this step, the final optimized planting management plan includes actionable guidelines that can be directly implemented. Specifically: the implementation schedule clearly defines the daily / weekly operations (e.g., spraying foliar fertilizer at 9:00 AM on Monday, adjusting irrigation flow on Wednesday); key parameter adjustment thresholds are set with automated trigger conditions (e.g., automatically starting irrigation when soil moisture is below 60%, turning on heating equipment when temperature is below 15℃); and the expected growth improvement curve visualizes the implementation effect (e.g., the deviation in plant height growth slope is reduced to within 2% after 10 days). For severe impact events identified in the risk warning indicators, the emergency response plan sets trigger conditions, such as "automatically starting temporary heating measures to maintain the temperature above 18℃ when the temperature is below 15℃ for two consecutive days," ensuring rapid response in abnormal situations.

[0065] The above-mentioned optimized planting management plan can take into account both routine management and emergency response, providing growers with comprehensive and precise operational guidance.

[0066] like Figure 3 As shown in the figure, this invention also discloses a business service management platform 100 based on digital twins and AI, comprising: Modeling unit 101 receives real-time crop growth environment data and crop growth status data collected by sensing devices deployed in the planting greenhouse, and constructs a crop digital twin model based on the crop growth environment data and the crop growth status data. The fluctuation signal generation unit 102 acquires historical planting data of the greenhouse, including historical equipment operation data and historical meteorological data, performs correlation analysis on the historical equipment operation data and historical meteorological data, and generates a growth fluctuation signal corresponding to a future preset time period based on the correlation analysis results. The growth prediction unit 103, based on the growth fluctuation signal, uses the crop digital twin model to simulate and analyze the crop growth in a future preset period and generates crop growth prediction results. The management optimization unit 104 inputs the crop growth prediction results into the AI ​​optimization model to obtain a planting management optimization plan; wherein, the AI ​​optimization model is embedded with a crop growth-management measure association algorithm obtained through pre-training.

[0067] Optionally, the modeling unit 101 specifically: Determine the environmental sensitivity parameters of the crops currently grown in the greenhouse. These parameters include the critical threshold of soil moisture, the range of light intensity tolerance, and the response coefficient to temperature changes. Based on these parameters, determine the sensitivity level of the crops to different environmental factors, including high sensitivity, medium sensitivity, and low sensitivity. A refined modeling approach is used for highly sensitive environmental factors, a conventional modeling approach is used for moderately sensitive environmental factors, and a simplified modeling approach is used for low-sensitive environmental factors. The crop digital twin model was constructed based on the above modeling method.

[0068] Optionally, the wave signal generation unit 102 specifically: Extract key operating parameter sequences from historical equipment operation data, including start-up and shutdown duration sequences for temperature control equipment, flow rate change sequences for irrigation equipment, and speed adjustment sequences for ventilation equipment; extract environmental parameter sequences for corresponding time periods from historical meteorological data, including temperature fluctuation sequences, humidity change sequences, and light intensity fluctuation sequences. The sliding window algorithm is used to align the key operating parameter sequence and the environmental parameter sequence in time. The Pearson correlation coefficient between the equipment operating parameters and the environmental parameters in different window intervals is calculated, and strongly correlated parameter pairs with an absolute value of the correlation coefficient greater than a preset threshold are selected. Temporal features are extracted from strongly correlated parameter pairs to identify environmental parameter mutation events caused by equipment malfunctions, and the time nodes, impact magnitudes, and durations of these events are recorded. Feature mining is then performed on each of these environmental parameter mutation events to identify the occurrence patterns of these events. Based on the occurrence patterns of abrupt environmental parameter events and basic meteorological forecast data for a predetermined future period, an LSTM time series prediction model is used to generate the growth fluctuation signal, which includes the amplitude, duration, and probability of environmental parameter fluctuations. Among these, high-probability fluctuation events are marked as key focus items in the growth fluctuation signal.

[0069] Optionally, the growth prediction unit 103 specifically: The growth fluctuation signal is structured and analyzed to extract high-probability fluctuation events marked as key concerns, and the environmental parameter type, fluctuation amplitude, expected duration and occurrence time window of each high-probability fluctuation event are determined. Based on the preset crop sensitivity parameter thresholds in the crop digital twin model, the impact level of high-probability fluctuation events is assessed and classified into mild impact, moderate impact, and severe impact. During the digital twin model simulation, environmental parameter change data from the growth fluctuation signal are imported sequentially according to the time window, and a refined simulation strategy is adopted for high-probability fluctuation events. After the simulation, the basic growth prediction data under normal environmental conditions and the deviation data under the influence of high-probability fluctuation events are integrated to generate crop growth prediction results that include plant height growth curves, biomass accumulation trends, key growth period offsets, and risk warning indicators; among them, the risk warning indicators correspond to growth abnormalities that may be caused by severe impact events.

[0070] Optionally, the management optimization unit 104 specifically: The crop growth forecast results are subjected to feature quantification processing to extract core growth indicators and risk characteristics, including the slope deviation of the plant height growth curve, the deviation of the biomass accumulation trend, the number of days of deviation of key growth period nodes, and the type of growth abnormality corresponding to the risk warning indicator. The extracted feature data is compared with the preset crop yield benchmark and quality index thresholds to calculate the gap quantification parameters corresponding to each growth index. The algorithm for associating crop growth and management measures embedded in the AI ​​optimization model is invoked. Based on the gap quantification parameter, the historical optimization case library is retrieved, and successful management cases similar to the current growth characteristics are matched to generate an initial management measure combination, including fertilization plan, irrigation strategy, environmental control parameters and growth period intervention measures. By using deep reinforcement learning to dynamically optimize the initial management measure combination, the growth correction effect after the implementation of different measures is simulated, the cost-effectiveness of the measure implementation cost and expected benefits is calculated, and the measure combination with the best cost-effectiveness is selected. Generate a planting management optimization plan that includes a timeline of action plans corresponding to the optimal combination of measures, key parameter adjustment thresholds, expected growth improvement curves, and emergency response plans; among which, the emergency response plan sets trigger conditions for severe impact events in the risk warning indicators.

[0071] This invention also discloses an electronic device, comprising: a memory storing executable program code; a processor coupled to the memory; the processor calling the executable program code stored in the memory to execute the method described in Embodiment 1.

[0072] This invention also discloses a computer storage medium storing a computer program, which is executed by a processor as described in Embodiment 1.

[0073] This invention also discloses a computer program product comprising program code tools for performing the steps of the methods described in any of the preceding claims.

[0074] Those skilled in the art will understand that this application is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made by those skilled in the art without departing from the scope of protection of this application. Therefore, although this application has been described in detail through the above embodiments, this application is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of this application, and the scope of this application is determined by the scope of the appended claims.

Claims

1. A business service management method based on digital twins and AI, characterized in that, Includes the following steps: By collecting crop growth environment data and crop growth status data in real time through sensing devices deployed in the planting greenhouse, a crop digital twin model is constructed based on the crop growth environment data and crop growth status data. Acquire historical planting data for the greenhouse, including equipment operation history data and meteorological history data. Perform correlation analysis on the equipment operation history data and meteorological history data, and generate growth fluctuation signals corresponding to the future preset time period based on the correlation analysis results. Based on the growth fluctuation signal, the crop digital twin model is used to simulate and analyze the crop growth in a future preset period and generate crop growth prediction results. The crop growth prediction results are input into the AI ​​optimization model to obtain an optimized planting management plan; wherein, the AI ​​optimization model is embedded with a crop growth-management measure association algorithm obtained through pre-training.

2. The business service management method based on digital twin and AI according to claim 1, characterized in that: A crop digital twin model is constructed based on the crop growth environment data and the crop growth status data, including: Determine the environmental sensitivity parameters of the crops currently grown in the greenhouse. These parameters include the critical threshold of soil moisture, the range of light intensity tolerance, and the response coefficient to temperature changes. Based on these parameters, determine the sensitivity level of the crops to different environmental factors, including high sensitivity, medium sensitivity, and low sensitivity. A refined modeling approach is used for highly sensitive environmental factors, a conventional modeling approach is used for moderately sensitive environmental factors, and a simplified modeling approach is used for low-sensitive environmental factors. The crop digital twin model was constructed based on the above modeling method.

3. The business service management method based on digital twins and AI according to claim 1, characterized in that: Correlation analysis is performed on historical equipment operation data and historical meteorological data. Based on the correlation analysis results, growth fluctuation signals corresponding to a preset future time period are generated, including: Extract key operating parameter sequences from historical equipment operation data, including start-up and shutdown duration sequences for temperature control equipment, flow rate change sequences for irrigation equipment, and speed adjustment sequences for ventilation equipment; extract environmental parameter sequences for corresponding time periods from historical meteorological data, including temperature fluctuation sequences, humidity change sequences, and light intensity fluctuation sequences. The sliding window algorithm is used to align the key operating parameter sequence and the environmental parameter sequence in time. The Pearson correlation coefficient between the equipment operating parameters and the environmental parameters in different window intervals is calculated, and strongly correlated parameter pairs with an absolute value of the correlation coefficient greater than a preset threshold are selected. Temporal features are extracted from strongly correlated parameter pairs to identify environmental parameter mutation events caused by equipment malfunctions, and the time nodes, impact magnitudes, and durations of these events are recorded. Feature mining is then performed on each of these environmental parameter mutation events to identify the occurrence patterns of these events. Based on the occurrence patterns of abrupt environmental parameter events and basic meteorological forecast data for a predetermined future period, an LSTM time series prediction model is used to generate the growth fluctuation signal, which includes the amplitude, duration, and probability of environmental parameter fluctuations. Among these, high-probability fluctuation events are marked as key focus items in the growth fluctuation signal.

4. The business service management method based on digital twin and AI according to claim 3, characterized in that: Based on the growth fluctuation signal, the crop digital twin model is used to simulate and analyze the crop growth in a future preset period, generating crop growth prediction results, including: The growth fluctuation signal is structured and analyzed to extract high-probability fluctuation events marked as key concerns, and the environmental parameter type, fluctuation amplitude, expected duration and occurrence time window of each high-probability fluctuation event are determined. Based on the preset crop sensitivity parameter thresholds in the crop digital twin model, the impact level of high-probability fluctuation events is assessed and classified into mild impact, moderate impact, and severe impact. During the digital twin model simulation, environmental parameter change data from the growth fluctuation signal are imported sequentially according to the time window, and a refined simulation strategy is adopted for high-probability fluctuation events. After the simulation, the basic growth prediction data under normal environmental conditions and the deviation data under the influence of high-probability fluctuation events are integrated to generate crop growth prediction results that include plant height growth curves, biomass accumulation trends, key growth period offsets, and risk warning indicators; among them, the risk warning indicators correspond to growth abnormalities that may be caused by severe impact events.

5. A business service management method based on digital twins and AI according to claim 4, characterized in that: The crop growth prediction results are input into the AI ​​optimization model to derive an optimized planting management plan, including: The crop growth forecast results are subjected to feature quantification processing to extract core growth indicators and risk characteristics, including the slope deviation of the plant height growth curve, the deviation of the biomass accumulation trend, the number of days of deviation of key growth period nodes, and the type of growth abnormality corresponding to the risk warning indicator. The extracted feature data is compared with the preset crop yield benchmark and quality index thresholds to calculate the gap quantification parameters corresponding to each growth index. The algorithm for associating crop growth and management measures embedded in the AI ​​optimization model is invoked. Based on the gap quantification parameter, the historical optimization case library is retrieved, and successful management cases similar to the current growth characteristics are matched to generate an initial management measure combination, including fertilization plan, irrigation strategy, environmental control parameters and growth period intervention measures. By using deep reinforcement learning to dynamically optimize the initial management measure combination, the growth correction effect after the implementation of different measures is simulated, the cost-effectiveness of the measure implementation cost and expected benefits is calculated, and the measure combination with the best cost-effectiveness is selected. Generate a planting management optimization plan that includes a timeline of action plans corresponding to the optimal combination of measures, key parameter adjustment thresholds, expected growth improvement curves, and emergency response plans; among which, the emergency response plan sets trigger conditions for severe impact events in the risk warning indicators.

6. A business service management platform based on digital twins and AI, characterized in that, include: The modeling unit receives real-time crop growth environment data and crop growth status data collected by sensing devices deployed in the planting greenhouse, and constructs a crop digital twin model based on the crop growth environment data and the crop growth status data. The fluctuation signal generation unit acquires historical planting data of the greenhouse, including historical equipment operation data and historical meteorological data, performs correlation analysis on the historical equipment operation data and historical meteorological data, and generates a growth fluctuation signal corresponding to a preset future time period based on the correlation analysis results. The growth prediction unit, based on the growth fluctuation signal, uses the crop digital twin model to simulate and analyze the crop growth in a future preset period and generates crop growth prediction results. The management optimization unit inputs the crop growth prediction results into the AI ​​optimization model to obtain an optimized planting management plan; wherein, the AI ​​optimization model is embedded with a crop growth-management measure association algorithm obtained through pre-training.

7. A business service management platform based on digital twins and AI according to claim 6, characterized in that: The modeling unit, specifically: Determine the environmental sensitivity parameters of the crops currently grown in the greenhouse. These parameters include the critical threshold of soil moisture, the range of light intensity tolerance, and the response coefficient to temperature changes. Based on these parameters, determine the sensitivity level of the crops to different environmental factors, including high sensitivity, medium sensitivity, and low sensitivity. A refined modeling approach is used for highly sensitive environmental factors, a conventional modeling approach is used for moderately sensitive environmental factors, and a simplified modeling approach is used for low-sensitive environmental factors. The crop digital twin model was constructed based on the above modeling method.

8. An electronic device, comprising: Memory containing executable program code; A processor coupled to the memory; characterized in that: the processor calls the executable program code stored in the memory to perform the method as described in any one of claims 1-5.

9. A computer storage medium storing a computer program, characterized in that: The computer program is executed by the processor to perform the method as described in any one of claims 1-5.

10. A computer program product, characterized in that: The computer program product includes program code tools for performing the steps of the method as described in any one of claims 1 to 5.

Images