Dynamic decision-making method and system for soilless culture growth process of ananas comosus

Abstract

The application provides a dynamic decision method and system for soilless cultivation of Anoectochilus roxburghii, and relates to the field of agricultural intelligence technology.The method comprises the following steps: in step 2, based on multi-dimensional environmental data and plant physiological indexes, the time sequence physiological indexes are analyzed, and the dynamic change mode characteristics of the stage change of the physiological indexes are extracted; the shape characteristics of different time segments are matched and fused, and the current growth stage of Anoectochilus roxburghii is dynamically recognized through a preset growth stage recognition model; in step 3, according to the current growth stage of Anoectochilus roxburghii, the dynamic ideal threshold range corresponding to the current stage is obtained from the preset growth stage and environmental mapping relationship.The application realizes self-adaptive regulation and control of the soilless cultivation environment, and improves the growth quality and yield.

Description

Dynamic Decision-Making Method and System for Soilless Cultivation of Anoectochilus roxburghii Technical Field

[0001] This invention relates to the field of agricultural intelligent technology, and in particular to a dynamic decision-making method and system for the hydroponics growth process of Anoectochilus roxburghii. Background Technology

[0002] In the hydroponic cultivation of Anoectochilus roxburghii, precise control of environmental conditions is key to its growth and development, accumulation of active ingredients, and final yield and quality. Most common environmental control methods rely on preset fixed parameters. These methods are mostly based on the general laws of crop growth cycles, setting uniform environmental control thresholds and maintaining them unchanged for a certain period of time.

[0003] However, plant growth is a continuous and dynamic process, and its internal physiological state and environmental requirements change as different growth stages progress. For example, during the critical period when Anoectochilus roxburghii transitions from vegetative to reproductive growth, its requirements for light intensity and nutrient solution ratio may differ significantly from those of the seedling stage or rapid growth stage. Most existing static control strategies are unable to accurately capture and respond to these stage-specific internal physiological changes. Although stage judgment can be made through manual observation or timed sampling, this method has some lag and relies on personal experience, making it difficult to achieve real-time and objective judgment. Therefore, using fixed environmental thresholds for management may result in a mismatch between environmental conditions and the plant's actual needs at certain stages, affecting subsequent growth. Summary of the Invention

[0004] The technical problem to be solved by this invention is to provide a dynamic decision-making method and system for the hydroponic growth process of Anoectochilus roxburghii, so as to achieve adaptive regulation of the hydroponic environment and improve growth quality and yield.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] Firstly, a dynamic decision-making method for the hydroponic growth process of *Anoectochilus roxburghii*, the method comprising:

[0007] Step 1: Collect multi-dimensional environmental data and plant physiological indicators in real time during the growth of Anoectochilus roxburghii.

[0008] Step 2: Based on multidimensional environmental data and plant physiological indicators, analyze the time-series physiological indicators and extract the dynamic change pattern characteristics of the physiological indicators in stages; match and fuse the shape features of different time segments, and dynamically identify the current growth stage of Anoectochilus roxburghii through a preset growth stage identification model.

[0009] Step 3: Based on the current growth stage of Anoectochilus roxburghii, obtain the dynamic ideal threshold range of each environment corresponding to the current stage from the preset growth stage and environment mapping relationship;

[0010] Step 4: Compare the dynamic ideal threshold range of each environment with the multidimensional environmental data. By calculating the intersection interval and relative position relationship between the real-time data curve and the boundary curve of the dynamic threshold range, quantify the degree of deviation between the real-time environment and the ideal threshold range, and generate the environmental deviation index.

[0011] Step 5: Based on the environmental deviation index, construct a multidimensional evaluation vector, analyze the mutual influence relationship between the dimensions in the multidimensional evaluation vector, evaluate the contribution of each dimension to the overall control decision, and generate dynamic control correction amount.

[0012] Step 6: Integrate the dynamic adjustment correction amount with short-term climate prediction data to make a comprehensive decision, dynamically generate a control strategy for the growth environment of Anoectochilus roxburghii, and realize the dynamic control of the growth process of Anoectochilus roxburghii.

[0013] Secondly, the dynamic decision-making method and system for the hydroponics growth process of *Anoectochilus roxburghii* includes:

[0014] The data acquisition module is used to collect multi-dimensional environmental data and plant physiological indicators in real time during the growth of Anoectochilus roxburghii.

[0015] The identification module is used to analyze time-series physiological indicators based on multidimensional environmental data and plant physiological indicators, extract the dynamic change pattern features of the physiological indicators in stages, match and fuse the shape features of different time segments, and dynamically identify the current growth stage of Anoectochilus roxburghii through a preset growth stage identification model.

[0016] The acquisition module is used to obtain the dynamic ideal threshold range of each environment corresponding to the current growth stage from the preset growth stage and environment mapping relationship based on the current growth stage of Anoectochilus roxburghii.

[0017] The generation module is used to compare the dynamic ideal threshold range of each environment with multi-dimensional environmental data. By calculating the intersection interval and relative position relationship between the real-time data curve and the boundary curve of the dynamic threshold range, the deviation between the real-time environment and the ideal threshold range is quantified, and an environmental deviation index is generated.

[0018] The regulation module is used to construct a multi-dimensional evaluation vector based on the environmental deviation index, analyze the mutual influence relationship between the dimensions in the multi-dimensional evaluation vector, evaluate the contribution of each dimension to the overall regulation decision, and generate dynamic regulation correction amount.

[0019] The strategy module is used to make comprehensive decisions by combining dynamic regulation corrections with short-term climate forecast data, and dynamically generate regulation strategies for the growth environment of Anoectochilus roxburghii, thereby realizing dynamic regulation of the growth process of Anoectochilus roxburghii.

[0020] The above-described solution of the present invention has at least the following beneficial effects:

[0021] By dynamically analyzing time-series physiological indicators and matching and fusing shape characteristics of different time segments, combined with a pre-set growth stage identification model, the current growth stage is dynamically identified, capturing key nodes in the transition of growth stages to ensure that the regulation strategy is highly adapted to physiological needs. Then, based on the dynamically identified growth stage, the corresponding dynamic ideal threshold range is obtained, achieving dynamic matching between the threshold and the growth stage. By calculating the intersection interval and relative position relationship between the real-time data curve and the dynamic threshold boundary curve, the environmental deviation index is quantified, transforming environmental differences into quantifiable index indicators. This leads to the construction of a multi-dimensional evaluation vector, analyzing the mutual influence relationships between various dimensions and assessing their contribution to regulation decisions, generating dynamic regulation correction quantities, and achieving synergistic consideration of multi-dimensional environmental factors. Finally, the dynamic regulation correction quantities are integrated with short-term climate prediction data to generate dynamic regulation strategies, improving the growth rate of *Anoectochilus roxburghii*, enhancing management efficiency, and reducing resource waste and economic losses. Attached Figure Description

[0022] Figure 1 is a flowchart illustrating the dynamic decision-making method for the hydroponic growth process of Anoectochilus roxburghii provided in an embodiment of the present invention.

[0023] Figure 2 is a schematic diagram of the dynamic decision-making system for the hydroponic growth process of Anoectochilus roxburghii provided in an embodiment of the present invention. Detailed Implementation

[0024] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0025] As shown in Figure 1, an embodiment of the present invention proposes a dynamic decision-making method for the hydroponic growth process of *Anoectochilus roxburghii*, the method comprising the following steps:

[0026] Step 1: Collect multi-dimensional environmental data and plant physiological indicators in real time during the growth of Anoectochilus roxburghii.

[0027] Step 2: Based on multidimensional environmental data and plant physiological indicators, analyze the time-series physiological indicators and extract the dynamic change pattern characteristics of the physiological indicators in stages; match and fuse the shape features of different time segments, and dynamically identify the current growth stage of Anoectochilus roxburghii through a preset growth stage identification model.

[0028] Step 3: Based on the current growth stage of Anoectochilus roxburghii, obtain the dynamic ideal threshold range of each environment corresponding to the current stage from the preset growth stage and environment mapping relationship;

[0029] Step 4: Compare the dynamic ideal threshold range of each environment with the multidimensional environmental data. By calculating the intersection interval and relative position relationship between the real-time data curve and the boundary curve of the dynamic threshold range, quantify the degree of deviation between the real-time environment and the ideal threshold range, and generate the environmental deviation index.

[0030] Step 5: Based on the environmental deviation index, construct a multidimensional evaluation vector, analyze the mutual influence relationship between the dimensions in the multidimensional evaluation vector, evaluate the contribution of each dimension to the overall control decision, and generate dynamic control correction amount.

[0031] Step 6: Integrate the dynamic adjustment correction amount with short-term climate prediction data to make a comprehensive decision, dynamically generate a control strategy for the growth environment of Anoectochilus roxburghii, and realize the dynamic control of the growth process of Anoectochilus roxburghii.

[0032] In this embodiment of the invention, dynamic analysis of time-series physiological indicators and matching and fusion of shape features of different time segments are combined with a preset growth stage identification model to achieve dynamic identification of the current growth stage, capture key nodes in the transition of the growth stage, and ensure that the regulation strategy is highly adapted to physiological needs. Then, based on the dynamically identified growth stage, the corresponding dynamic ideal threshold range of the environment is obtained to achieve dynamic matching between the threshold and the growth stage. By calculating the intersection interval and relative position relationship between the real-time data curve and the dynamic threshold boundary curve, the environmental deviation index is quantified, and the environmental difference is transformed into a quantifiable index indicator. Then, a multi-dimensional evaluation vector is constructed to analyze the mutual influence relationship between the dimensions and evaluate their contribution to the regulation decision, generate a dynamic regulation correction amount, realize the synergistic consideration of multi-dimensional environmental factors, and finally make a comprehensive decision based on the dynamic regulation correction amount and short-term climate prediction data to generate a dynamic regulation strategy, improve the growth rate of Anoectochilus roxburghii, improve the management efficiency of Anoectochilus roxburghii, and reduce resource waste and economic losses.

[0033] In a preferred embodiment of the present invention, step 1 above, which involves real-time collection of multidimensional environmental data and plant physiological indicators during the growth of *Anoectochilus roxburghii*, may include:

[0034] In this embodiment of the invention, the acquisition parameters are determined, including two types of core data to be collected: multidimensional environmental data covering air temperature, air humidity, light intensity, CO2 concentration, EC value of the cultivation substrate, pH value of the substrate, nutrient solution temperature, and dissolved oxygen content of the nutrient solution; and plant physiological indicators including plant height, stem diameter, number of leaves, leaf area, chlorophyll content, root fresh weight, and leaf water content. Environmental data are acquired using high-precision digital sensors. Air temperature / humidity sensors are deployed 50cm above the planting trough, with two sensors per trough to ensure data uniformity. Light intensity sensors are installed 20cm above the plant canopy, with one sensor per trough. CO2 concentration sensors are deployed in the central ventilation area of ​​the greenhouse, with one sensor per 100㎡. Substrate EC and pH value sensors are directly inserted into the nutrient solution in the planting trough, with one set per trough. Nutrient solution temperature and dissolved oxygen sensors are installed at the outlet of the nutrient solution circulation pipe, with one set per trough. Plant physiological indicators are acquired using a manual-assisted automatic acquisition mode, equipped with a ruler, vernier caliper, leaf area meter, chlorophyll meter, and electronic balance, along with a portable... Data acquisition terminals record data; environmental data is collected every 5 minutes, with timestamps accurate to the second; plant physiological indicators are collected daily at 10:00 AM, with timestamps uniformly in UTC format to ensure consistency with environmental data timestamp standards; to address the difference in collection frequencies between the two types of data, environmental data within 2.5 minutes before and after the physiological indicator collection time is extracted, resulting in two collection points. The arithmetic mean is calculated as the synchronized environmental data to achieve timestamp alignment; the collected data is stored in real-time in CSV format, with fields including collection timestamp, sensor number, parameter name, value, unit, and data status; abnormal data is synchronously removed. When the collected value exceeds ±3 times the standard deviation of the parameter's historical 30-day valid data, it is marked as abnormal, and the value from the previous valid collection point is used to fill the gap; then, using a moving average method, three consecutive collection points form one window, corresponding to 15 minutes for data smoothing, and the arithmetic mean within the window is calculated as the intermediate time data, controlling the data fluctuation range within ±5% to ensure the validity and stability of the collected data.

[0035] In a preferred embodiment of the present invention, step 2 above, based on multidimensional environmental data and plant physiological indicators, analyzes the time-series physiological indicators and extracts the dynamic change pattern characteristics of the phased changes in the physiological indicators; matches and fuses the shape features of different time segments, and dynamically identifies the current growth stage of *Anoectochilus roxburghii* through a preset growth stage identification model, which may include:

[0036] In this embodiment of the invention, step 220 involves performing timestamp alignment processing on the multidimensional environmental data and plant physiological indicators to generate a time-series data pair of *Anoectochilus roxburghii* growth process that is strictly synchronized in time. Specifically, this includes: first, determining the basic parameters for collecting the two types of data; for multidimensional environmental data, high-precision sensors are used for collection at a fixed frequency of once every 5 minutes, with a UTC format timestamp recorded immediately after each collection, along with the sensor number, parameter value, and unit, to ensure the traceability of each piece of environmental data; for plant physiological indicators, a portable, manually assisted collection terminal is used, with a fixed daily collection frequency. The data collection was conducted at 10:00 AM. First, the collection tool was calibrated. Then, each *Anoectochilus roxburghii* plant was measured. After recording the data for each plant, the arithmetic mean of all plants in the same planting trough was taken. This mean was the sum of the individual plant values ​​divided by the number of plants to obtain the measured physiological index value for that planting trough. Simultaneously, a UTC timestamp consistent with the environmental data was linked to avoid time discrepancies. Because the collection frequencies of the two types of data differ significantly, direct correlation would result in time misalignment, requiring targeted alignment. The UTC timestamp of the physiological index at 10:00 AM daily was used as the core reference point, and the alignment time range was defined as the period preceding this reference point. 2.5 minutes (UTC timestamp minus 2 minutes and 30 seconds to 2.5 minutes after the baseline, or UTC timestamp plus 2 minutes and 30 seconds) is the time interval within which exactly two environmental data points can be collected. Since environmental data is collected every 5 minutes, it covers one collection node before and after. These two environmental data points are first validated, and values ​​marked as abnormal or exceeding ±3 times the standard deviation of the historical 30 days of valid data for that parameter are removed. If both data points are valid, the two values ​​for the same environmental parameter are added together and divided by 2 to obtain the synchronous mean value of that parameter. If only one data point is valid, that valid value is used directly. The data is used as a synchronization value; if both data are abnormal, the environmental data from the previous valid collection period is traced back to the temporary synchronization value, and the abnormal data status is marked. After the single parameter synchronization is completed, the synchronization values ​​of the 8 environmental parameters under the same reference timestamp are integrated with the measured average values ​​of the 7 physiological indicators to generate a set of synchronized time-series data pairs. Each set of data pairs contains a unique UTC timestamp, the synchronization values ​​and corresponding units of the 8 environmental parameters, the measured average values ​​and corresponding units of the 7 physiological indicators, and a data validity identifier to ensure that each set of data is completely corresponding in time, without misalignment or omission.

[0037] Step 221 involves fusing the synchronous time-series data pairs of *Anoectochilus roxburghii* growth process. The environmental vector and physiological indicator vector corresponding to each aligned time point are combined to construct a structured time-series data table of *Anoectochilus roxburghii* growth status. Specifically, this includes: preprocessing all generated synchronous time-series data pairs, verifying the uniqueness of each timestamp, checking for duplicate timestamps, multiple data sets corresponding to the same UTC timestamp, missing timestamps, and no corresponding data for a particular collection day. If duplicate timestamps exist, the set with the data validity flagged as normal is retained, and the remaining sets are discarded. If timestamps are missing, the missing date and reason are noted, and the impact of the missing data on subsequent analysis is explained. (The text then repeats the steps 221-222, which is not directly related to the steps 221.) After the experiment was completed, all data pairs were sorted in ascending order by UTC timestamp, starting from the day of *Anoectochilus roxburghii* sowing, arranged sequentially with one set of data per day to form a coherent time series chain. Then, a vector combination operation was performed. For each timetamp-corresponding synchronous data pair, environmental and physiological indicator vectors were constructed. The environmental vectors were strictly constructed in a fixed order: air temperature - air humidity - light intensity - CO2 concentration - substrate EC value - substrate pH value - nutrient solution temperature - nutrient solution dissolved oxygen. Each vector element corresponds to the synchronous value and unit of one environmental parameter; for example, the air temperature element was labeled as 25.3℃, and the substrate EC value element was labeled as 1.6 mS / cm, ensuring... The parameters within the vectors are ordered uniformly and are traceable. The physiological indicator vectors are strictly constructed in a fixed order: plant height - stem diameter - number of leaves - leaf area - chlorophyll content - root fresh weight - leaf water content. Each element corresponds to the measured mean and unit of a physiological indicator. For example, the plant height element is labeled as 8.2cm, and the leaf area element is labeled as 4.5cm². After the vectors are constructed, the environmental vector and the physiological indicator vector at the same timestamp are concatenated sequentially to form a combined data containing 15 values ​​and units. Then, core fields are added to this combined data: UTC timestamp, collection date, planting trough number, data validity identifier, collection personnel, and sensor group number, ultimately constructing a 16-column structure. The data table is stored on local edge storage devices in an editable tabular format for easy retrieval and modification. It is managed in folders according to the growth cycle. After daily data collection and alignment, newly generated synchronized time-series data pairs are appended to the end of the data table in timestamp order. The append time and operator are recorded to create an operation log. After appending, the data table undergoes integrity verification, checking that all 16 columns of the day's data are complete, that numerical units are consistent, and that timestamps are consecutive, ensuring the data table's integrity and standardization. This data table's cumulative storage period covers the entire growth cycle of *Anoectochilus roxburghii*, forming a complete growth data chain from sowing to harvest.

[0038] Step 222 involves performing spatiotemporal correlation analysis on the fused time-series data table of *Anoectochilus roxburghii* growth status. This involves calculating the dynamic correlation between the environmental time-series data sequence and the time-series data sequences of various physiological indicators, and extracting key correlation features and delay pattern features characterizing the physiological response driven by environmental changes. Specifically, this includes: first, selecting analysis samples and extracting complete data for 30 consecutive days from the structured fused time-series data table as the basis for analysis, ensuring no data is missing within these 30 days. If 1-2 days of data are missing, the mean of the data from the two consecutive days is used to complete the missing data. The completed data is then labeled and interpolated, and the proportion of data with normal validity is no less than 90% to avoid affecting the analysis results due to sample data quality issues. For the selected 30-day data, the analysis is performed according to environmental parameters and physiological... The correspondence between indicators was established by constructing time-series data sequences one by one. Each environmental parameter corresponds to a time-series sequence containing 30 values, with one synchronized value per day, arranged in ascending order of timestamp. Each physiological indicator corresponds to a time-series sequence containing 30 values, with one measured mean value per day, arranged in ascending order of timestamp. A total of 56 independent environmental and physiological indicator time-series data sequence pairs were constructed (8 environmental parameters × 7 physiological indicators). Each sequence pair was labeled with the corresponding parameter name, time range, and data source. When calculating the dynamic correlation, for each sequence pair, the following steps were performed step-by-step: First, the average value of the environmental parameter sequence and the average value of the physiological indicator sequence were calculated separately. The average value of the environmental parameter sequence = the sum of the 30 synchronized values ​​of that sequence. The first step is to calculate the deviation at each of the 30 time points. The environmental parameter deviation at each time point is calculated as follows: Environmental parameter deviation = synchronous environmental parameter value at that time point - environmental parameter series average. The physiological indicator deviation at each time point is calculated as: measured physiological indicator average at that time point - physiological indicator series average. Deviation values ​​can be positive or negative, indicating whether the value at that time point is higher or lower than the series average. The specific value of each deviation is recorded. The third step is to calculate the sum of the products of the deviation values. This involves multiplying the environmental parameter deviation value by the physiological indicator deviation value at each of the 30 time points. The first step is to calculate the sum of the products of all deviation values. The second step is to calculate the sum of squares. The environmental parameter deviation values ​​at each of the 30 time points are squared (deviation value × deviation value). The sum of all squared results is then obtained. Similarly, the sum of squared deviation values ​​of physiological indicators is calculated. The two sums of squares are then multiplied to obtain the product of the sums of squares. The square root of this product is then taken. The third step is to calculate the dynamic correlation result. The sum of the deviation value products is divided by the square root of the product of the sums of squares to obtain the dynamic correlation result of the set of environmental and physiological indicator sequence pairs. The result range is marked as -1 to 1. The larger the absolute value of the result, the stronger the dynamic correlation between the two.

[0039] When extracting key correlation features, a screening threshold was first set: an absolute value greater than 0.6. The dynamic correlation results of 56 sequence pairs were compared one by one, and sequence pairs with an absolute value greater than 0.6 were selected as key correlation pairs. These pairs represent combinations of environmental factors and physiological indicators that significantly affect the physiological state of *Anoectochilus roxburghii*. For each key correlation pair, three core pieces of information were recorded in detail: first, the average correlation, obtained by summing the dynamic correlation results at 30 time points and dividing by 30; second, the maximum correlation value, selected from the dynamic correlation results at 30 time points, noting its specific magnitude and corresponding time point; and third, the correlation trend, determining whether the correlation result of the pair is positive (when environmental parameter values ​​increase, physiological indicator values ​​increase synchronously) or negative (when environmental parameter values ​​increase, physiological indicator values ​​decrease synchronously). For example, a positive correlation between air temperature and plant height indicates that when the temperature rises within a suitable range, plant height growth accelerates. Conversely, a negative correlation between substrate EC value and root fresh weight indicates that EC value is not positive. Excessive C-values ​​inhibit root growth. When extracting delayed pattern features, for each key association pair, the peak time points of the environmental parameter sequence and the physiological indicator sequence are first determined. The peak time point of the environmental parameter sequence refers to the time point with the largest value among the 30 values ​​in the sequence. If multiple values ​​have the same maximum value, the earliest occurrence time point is taken. Similarly, the peak time point of the physiological indicator sequence is the time point with the largest value among the 30 values ​​in the sequence. Then, the time difference between the two peak time points is calculated in days. If the peak time point of the environmental parameter is earlier than the peak time point of the physiological indicator, the time difference is the delay time of the physiological response driven by the environmental change. If the peak time point of the environmental parameter is later than or equal to the peak time point of the physiological indicator, it is marked as no significant delay. The delay times of all key association pairs are classified into fixed intervals: 1-2 days is short-term delay, and the physiological indicator shows a response peak 1-2 days after the environmental change; 3-5 days is medium-term delay; and 6-8 days is long-term delay. At the same time, the delay time type and specific duration of each association pair are recorded to determine the response lag pattern of different environmental factors to the physiological indicators of Anoectochilus roxburghii.

[0040] Step 223: Based on key correlation features and delay pattern features, classify and label the dynamic response patterns of *Anoectochilus roxburghii* and its environment according to a pre-set rule base to obtain the dynamic change pattern features of *Anoectochilus roxburghii* growth within the current growth cycle. Specifically, this includes: first, constructing a pre-set dynamic response pattern rule base. The rule base is built based on the growth characteristics of *Anoectochilus roxburghii*, cultivation experience, and historical data summaries to avoid subjective assumptions. The rules are clearly and unambiguously expressed, specifically including three core rules. Each rule has dual judgment conditions: the average correlation of key correlation pairs + delay time type, ensuring judgment accuracy. Rule 1 (temperature and light driven rapid growth mode) simultaneously satisfies both conditions. The conditions are as follows: First, the average correlation between the key correlation pair of air temperature and plant height is greater than 0.7, and the lag time is short; second, the average correlation between the key correlation pair of light intensity and leaf area is greater than 0.65, and the lag time is short. This model corresponds to the vigorous growth stage of *Anoectochilus roxburghii*, when temperature and light conditions are suitable, the plant's photosynthetic rate is high, and plant height and leaf area increase rapidly, commonly seen from the late seedling stage to the early seedling stage. Rule 2 (nutrient-oxygen driven robust growth model) simultaneously meets two conditions: first, the average correlation between the key correlation pair of substrate EC value and stem diameter is greater than 0.6, and the lag time is medium; second, the average correlation between the key correlation pair of nutrient solution dissolved oxygen and root fresh weight is... The correlation coefficient (C0.55) is greater than 0.55, and the delay time is in the middle stage. This pattern corresponds to the nutrient accumulation stage of *Anoectochilus roxburghii*, when the nutrient concentration and dissolved oxygen content of the nutrient solution are suitable, the root system is well developed, and the stems are thick. This is commonly seen in the middle stage of seedling formation. Rule 3 (humidity-carbon-source driven slow growth pattern) simultaneously meets two conditions: first, the average correlation between key correlation pairs of air humidity and leaf water content is greater than 0.6, and the delay time is long-term; second, the average correlation between key correlation pairs of CO2 concentration and chlorophyll content is greater than 0.5, and the delay time is long-term. This pattern corresponds to the slow growth stage of *Anoectochilus roxburghii*, when air humidity and CO2 concentration are the main influencing factors, and the plant's metabolic rate is low. Commonly seen in the budding stage or pre-harvest stage; after the rule base is constructed, the key correlation features and delay pattern features extracted in step 222 are substituted into the rule base one by one, and the matching operation is performed; during matching, the dual conditions of each group of core rules are checked one by one to determine whether they are satisfied. If only all the conditions of a certain rule are satisfied and no other rules are matched, the pattern corresponding to that rule is directly used as the core dynamic change pattern feature in the current growth cycle. At the same time, the key correlation pair information and delay pattern under this pattern are supplemented. For example, it is marked that the current mode is temperature and light driven rapid growth mode, the average correlation between air temperature and plant height is 0.72, and the short-term delay is 1 day; the average correlation between light intensity and leaf area is 0.68. Short-term delay of 2 days; if multiple rules are met simultaneously, such as both rule 1 and rule 2, the pattern is prioritized and selected as the core pattern based on the rule with the highest average correlation sum. The average correlations of the two key association pairs under that rule are added together, with the highest sum taking precedence. The patterns corresponding to the remaining matching rules are used as auxiliary features, collectively forming the dynamic change pattern features. The matching degree of each pattern, the number of conditions met, and the average correlation sum are also noted. For example, the current core pattern is a temperature-light driven rapid growth pattern, and the auxiliary pattern is a nutrient-oxygen driven robust growth pattern. The average correlation sum of temperature-light association is 1.4, and the average correlation sum of nutrient-oxygen association is 1.15. If none of the rules are met, and no dual conditions of any rule are met, it is marked as a transitional growth pattern. The characteristics and delay patterns of all current key association pairs are recorded in detail, indicating that the plant is in a transitional growth stage, and the interaction pattern between the environment and physiological indicators has not yet formed a stable pattern. Through the above classification and labeling, the interaction state between *Anoectochilus roxburghii* and the environment during the current growth cycle is characterized.

[0041] Step 224 involves dividing the temporal features related to the dynamic change pattern into multiple consecutive growth time segments according to a preset time granularity, and calculating the shape features of the change curves of each physiological indicator within each growth time segment. Specifically, this includes: first, determining the preset time granularity as 7 days. This time granularity is based on the actual collection frequency of *Anoectochilus roxburghii* physiological indicators: plant height, stem diameter, and chlorophyll content are collected once daily at 10:00 AM; root length is collected once every 7 days at the same time; and the regular change cycle of *Anoectochilus roxburghii* growth stages is set. This ensures complete coverage of the effective change range of physiological indicators while avoiding feature fragmentation due to excessively short granularity and masking of stage changes due to excessively long granularity. Subsequently, the temporal features related to the dynamic change pattern obtained in Step 223 are extracted, focusing on the complete temporal sequences of four physiological indicators: plant height, stem diameter, chlorophyll content, and root length. Taking the sowing day of *Anoectochilus roxburghii* as the starting point of the growth cycle, multiple consecutive time segments are divided into 7-day units. The growth time segments are continuous, non-overlapping, and without gaps. For example, the first segment is from day 1 to day 7 after sowing, the second segment is from day 8 to day 14, the third segment is from day 15 to day 21, and so on. If the remaining time of the current growth cycle is less than 7 days, the remaining time is used as the last complete growth time segment to ensure that no growth period is missed. Then, for each segment, the change curves of each physiological indicator are generated one by one. The date in the segment is used as the horizontal axis, such as day 1 to day 7 for the first segment. The physiological indicator measurement value of the corresponding date is used as the vertical axis. The physiological indicator values ​​collected daily in the segment are marked in the coordinate system in chronological order, and then connected by straight lines to form the continuous change curve of the physiological indicator in the segment. Since the root length is collected once every 7 days, there is only one root length data in each segment. Its change curve is represented by a horizontal straight line, and the trend of the curve is determined by comparing it with the root length value of the previous segment.

[0042] Next, the three shape characteristics of each change curve were calculated. The specific calculation method is as follows: First, the overall trend direction. Physiological index values ​​were extracted from the start and end dates of the segment. The total change was obtained by subtracting the start date value from the end date value. If the total change was positive and the total change in plant height was ≥0.5cm, stem diameter ≥0.05mm, chlorophyll content ≥2SPAD, and root length ≥1cm, it was determined to be an upward trend. If the total change was negative and the above indicators decreased by the corresponding values, it was determined to be a downward trend. If the absolute value of the total change was less than the corresponding values, such as the total change in plant height being between -0.1cm and 0.1cm... If the value is between m, it is considered a gradual trend. The specific range of the total change is recorded to aid subsequent feature analysis. Secondly, the distribution of local extreme points is analyzed. Physiological indicator values ​​for three consecutive days within a segment are compared. If the value on a given day is greater than both the previous and following day's values, and the difference meets the criteria of plant height ≥ 0.2 cm, stem diameter ≥ 0.02 mm, and chlorophyll content ≥ 1 SPAD, it is considered a local maximum point. If the value on a given day is less than both the previous and following day's values, and the difference meets the above criteria, it is considered a local minimum point. The specific number of local maximum and minimum points within each segment, and their corresponding dates, are recorded. If there are only two days of data within a segment, they are not counted. The process involves three steps: First, identifying local extreme points and marking them as having no valid extreme points. Second, determining the curvature change pattern. Each 7-day segment is divided into a first half and a second half. The total change in physiological indicators for each segment is calculated: the total change in the second half = day 7 value - day 4 value; the total change in the first half = day 3 value - day 1 value. The total change for the segment is obtained by adding the total change in the first half to the total change in the second half. Then, the proportion of the total change in the first half and the second half to the total change in the segment is calculated separately. If the proportion of the second half exceeds 70%, and the daily average change in the second half (total change in the second half ÷ 3 days) is greater than the daily average change in the first half (total change in the first half ÷ 2 days), then the curvature is considered steep. The trend is defined as a significant acceleration in the growth / decline rate in the later stages. If the difference between the proportions of the first and second halves does not exceed 20%, and the difference in the daily average changes between the two halves is small, it is judged as a flat curvature trend, meaning the growth / decline rate is stable. In other cases, such as a high proportion in the first half and a low proportion in the second half, or similar proportions in both halves but large rate fluctuations, it is judged as a curvature fluctuation trend. The specific patterns of rate changes are recorded in detail. Finally, for each growth time segment, the overall trend direction, local extreme point distribution, and curvature change pattern of plant height, stem diameter, chlorophyll content, and root length are summarized one by one to form a unique set of shape features for each segment. Each feature is labeled with specific judgment criteria.

[0043] Step 225: Compare the shape features with the pre-stored reference time segment shape features representing each typical growth stage of *Anoectochilus roxburghii*. Calculate the shape similarity between the current segment and each reference typical segment using a shape matching algorithm, generating a similarity matrix. Specifically, this includes: first, retrieving the pre-stored reference shape feature library for typical growth stages of *Anoectochilus roxburghii*. This feature library is based on a large amount of historical data from hydroponically cultivated healthy *Anoectochilus roxburghii*, identifying reference time segment shape features corresponding to four typical growth stages: seedling stage, seedling stage, flowering stage, and harvesting stage. Each typical stage corresponds to a 7-day reference segment, and the shape features of each reference segment are labeled with specific parameter ranges. To ensure consistent comparison standards, the specific reference characteristics are as follows: For the seedling stage (1-21 days after sowing), the plant height curve shows a gentle upward trend with a total change of 0.3-0.5 cm, ≤1 local extreme point, and a gentle curvature; the stem diameter curve shows a gentle upward trend with a total change of 0.03-0.05 mm, no local extreme points, and a gentle curvature; the chlorophyll content curve shows a gentle upward trend with a total change of 1-2 SPAD, no local extreme points, and a gentle curvature; the root length curve shows a gentle upward trend with a total change of 0.5-1 cm, no local extreme points, and a gentle curvature. For the mature seedling stage (22-60 days after sowing), the plant height curve shows an upward trend with a total change of 0.8-1.5 cm. The number of local extreme points is 2-3, with a steep curvature; the stem diameter curve rises, with a total change of 0.08-0.12 mm, 1-2 local extreme points, and a gentle curvature; the chlorophyll content curve rises, with a total change of 3-5 SPAD, 1 local extreme point, and a gentle curvature; the root length curve rises, with a total change of 1.5-2.5 cm, no local extreme points, and a gentle curvature; the reference characteristics for flowering period (61-80 days after sowing) are: a gentle plant height curve, with a total change ≤0.3 cm, about 2 local extreme points, and fluctuating curvature; a gently rising stem diameter curve, with a total change of 0.02-0.03 mm, no local extreme points, and a gentle curvature; chlorophyll content... The content curve decreases, with a total change of 1-2 SPAD, one local extreme point, and a gentle curvature; the root length curve is gentle, with a total change of ≤0.5cm, no local extreme points, and a gentle curvature; the reference characteristics for the harvest period (81 days after sowing to harvest) are: plant height curve is gentle, with a total change of ≤0.2cm, ≤1 local extreme point, and a gentle curvature; stem diameter curve is gentle, with a total change of ≤0.01mm, no local extreme points, and a gentle curvature; chlorophyll content curve decreases, with a total change of 2-3 SPAD, one local extreme point, and a steep curvature; root length curve decreases, with a total change of 0.3-0.5cm, no local extreme points, and a steep curvature.

[0044] Subsequently, for each current growth time segment's shape feature set, each physiological indicator was compared with the corresponding physiological indicator shape features of the reference segments at each typical stage. The shape matching algorithm was used to calculate the individual similarity, and then the overall shape similarity was obtained by summing the results. Individual similarity was calculated on a 100-point scale, with the following scoring criteria: Completely consistent trend direction (e.g., the current segment's plant height curve is rising, and the reference segment is also rising) earns 30 points; basically consistent trend direction (e.g., the current segment is rising gently, and the reference segment is rising) earns 15 points; inconsistent trend direction (e.g., the current segment is rising, and the reference segment is falling) earns 0 points; 30 points are awarded if the number of local extreme points differs from the reference segment by ≤1, and the change range corresponding to the extreme points conforms to the reference range. Two differences with roughly the same magnitude earn 10 points; three or more differences, or excessively large deviations, earn 0 points; identical curvature change patterns (e.g., steep curvature in the current case, and a similar pattern in the reference case) earn 40 points; roughly identical patterns (e.g., steep curvature in the current case, and a slightly steep pattern in the reference case) earn 20 points; inconsistent patterns (e.g., steep curvature in the current case, and fluctuating pattern in the reference case) earn 0 points; single-item similarity is the sum of the scores of the above three items, divided by 100 to convert to a percentage. For example, if the scores of the three items are 30, 30, and 40 respectively, the single-item similarity is 100%; overall shape similarity is the weighted average of the single-item similarity of each physiological indicator. The weights are determined based on the priority of each physiological indicator in representing the growth stage. Based on the growth characteristics of *Anoectochilus roxburghii*, the plant... Plant height, stem diameter, and chlorophyll content more clearly characterize changes in different stages and have higher weights. Specifically, plant height has a weight of 30%, stem diameter 30%, chlorophyll content 30%, and root length 10%. The calculation method is as follows: first, multiply the individual similarity of each physiological indicator by its corresponding weight to obtain a weighted score for each indicator; then, sum the four weighted scores to obtain the overall shape similarity. For example, if the current segment has a plant height similarity of 80%, stem diameter similarity of 70%, chlorophyll content similarity of 90%, and root length similarity of 60%, then the plant height weighted score = 80% × 30% = 24%, the stem diameter weighted score = 70% × 30% = 21%, and the chlorophyll content weighted score = ... 90% × 30% = 27%, root length weighted score = 60% × 10% = 6%, overall shape similarity = 24% + 21% + 27% + 6% = 78%; then, a similarity matrix is ​​constructed, with all current growth time segments as rows of the matrix, arranged in chronological order, such as segment 1, segment 2, segment 3, and the four typical growth stages of seedling stage, seedling stage, flowering stage, and harvest stage as columns of the matrix. The overall shape similarity of each current segment with the corresponding typical stage is filled into the corresponding position of the matrix. Each element in the matrix is ​​a percentage value of 0-100%, clearly showing the matching degree of each current segment with each typical growth stage, without missing any segment and stage comparison results.

[0045] Step 226: Based on the similarity matrix and combined with the pre-set inherent contribution of each reference typical segment in the growth stage judgment, the matching results of multiple highly similar reference segments are fused and calculated to generate a stage membership vector. Specifically, this includes: First, determining the pre-set inherent contribution of each reference typical segment. The contribution is set based on the duration proportion, physiological significance, and cultivation control priority of each growth stage of *Anoectochilus roxburghii*, ensuring that key growth stages have higher weight in the judgment. The inherent contribution of the four typical stages are: seedling stage 0.6, mature seedling stage 0.3, flowering stage 0.07, and harvesting stage 0.03. The sum of the contribution of all stages is 1 (0.6+0.3+0.07+0.03=1). Among them, the seedling stage and mature seedling stage are the core stages of *Anoectochilus roxburghii* growth, with a contribution ratio of 90%, which meets the need for accurate judgment of these two stages in actual cultivation. Then, extracting the overall shape similarity of each typical stage corresponding to each current growth time segment from the similarity matrix, and screening out high similarity reference segments. For each segment, the criteria for high similarity are: overall shape similarity ≥ 60%. If a current segment has multiple reference segments with similarity ≥ 60%, all of them are included in the high similarity set. If a current segment has no reference segments with similarity ≥ 60%, the top two reference segments with the highest similarity are selected as the high similarity set, and segments without high similarity are marked as the second highest similarity segment. If only one reference segment has similarity ≥ 60%, that segment is used as a separate high similarity set. Next, the high similarity reference segments of each current segment are fused and calculated to generate the stage membership degree of each typical stage. The specific calculation logic is as follows: First, calculate the preliminary score of each typical stage: preliminary score = overall shape similarity of the stage × inherent contribution of the stage. Second, add the preliminary scores of the four typical stages to obtain the total score. Third, calculate the stage membership degree of each typical stage: stage membership degree = preliminary score of the stage ÷ total score. The sum of the membership degrees of all typical stages must be 1.

[0046] For example, the current segment has a similarity of 75% with the seedling stage, 68% with the mature seedling stage, 45% with the flowering stage, and 32% with the harvest stage. The high similarity sets are the seedling stage and the mature seedling stage. Calculate the preliminary scores: Preliminary score for the seedling stage = 75% × 0.6 = 0.450, preliminary score for the mature seedling stage = 68% × 0.3 = 0.204, preliminary score for the flowering stage = 45% × 0.07 = 0.0315, and preliminary score for the harvest stage = 32% × 0.03 = 0.0096. Total score = 0.450 + 0.204 + 0.0315 + 0.0096 = 0.6951; Calculate the membership degree: membership degree during seedling stage = 0.450 ÷ 0.6951 ≈ 0.65, membership degree during mature seedling stage = 0.204 ÷ 0.6951 ≈ 0.29, membership degree during flowering stage = 0.0315 ÷ 0.6951 ≈ 0.05, membership degree during harvest stage = 0.0096 ÷ 0.6951 ≈ 0.01, and the adjusted total membership degree is 1 (0.65 + 0.0096). (29 + 0.05 + 0.01 = 1); For example, if the current segment only has a similarity of 62% in the seedling stage, while the similarity in other stages is <60%, then the preliminary scores are: Seedling stage = 62% × 0.6 = 0.372, Mature seedling stage = 58% × 0.3 = 0.174, Flowering stage = 50% × 0.07 = 0.035, Harvest stage = 48% × 0.03 = 0.0144; Total score = 0.372 + 0.174 + 0.035 + 0.0144 = 0.5954; Belongs to The membership degrees are calculated as follows: seedling stage ≈ 0.62, mature seedling stage ≈ 0.29, flowering stage ≈ 0.06, and harvesting stage ≈ 0.03, with a total of 1. Finally, for each current growth time segment, the membership degrees of the four typical stages are arranged in a fixed order of seedling stage, mature seedling stage, flowering stage, and harvesting stage to form a stage membership degree vector corresponding to the segment. Each element in the vector corresponds to the membership degree of a typical stage, and the calculation basis of each membership degree is also marked, which intuitively reflects the degree to which the current segment belongs to each growth stage.

[0047] Step 227: Input the stage membership vector and dynamic change pattern features into the preset growth stage identification model, and dynamically infer and obtain the specific growth stage of *Anoectochilus roxburghii* at the current time point; specifically, this includes: first, describing in detail the construction process of the preset growth stage identification model. The growth stage identification model is an inference model calibrated by rules and historical data. The construction is divided into three steps: data preparation, collecting data on the entire growth cycle of *Anoectochilus roxburghii* from the last 5 batches, including the stage membership vector, dynamic change pattern features, and temporal change trends of 4 types of physiological indicators, such as the plant height having a positive and increasing slope for 3 consecutive segments. The actual growth stages, manually labeled, are divided into training and validation sets in an 8:2 ratio. Data labeling must be complete and reliable, removing outliers caused by extreme environments. Inference rules are formulated based on the core threshold of membership vectors and supplementary rules for dynamic features, establishing a three-level inference logic. The first-level rule states that if the membership degree of a single stage is ≥0.6, it is directly classified as that stage; a membership degree ≥0.6 indicates a sufficiently high match, requiring no additional verification. The second-level rule states that if there is no stage with a membership degree ≥0.6, but both stages have membership degrees ≥0.4, the judgment is based on dynamic change pattern features, such as the difference between seedling and mature seedling stages. When the membership degrees of different growth stages are similar, an increasing slope across three consecutive segments of plant height indicates a seedling stage, while a stable slope indicates a mature seedling stage. A three-tiered judgment rule is used: for all stages with a membership degree <0.4, the rule is fine-tuned based on historical stage membership vectors and dynamic features from the same batch and period, combined with current data, and low-confidence entries are marked for further verification. The rule accuracy is tested using validation set data, thresholds are adjusted, and rules for special scenarios are added. For example, if the slope of the anoectochilin content suddenly increases during the flowering period, even if the membership degree is less than 0.6, it needs to be corrected to the flowering period based on this feature. Finally, a stable growth stage identification model is formed and stored in the system for each batch. After harvesting, the rules are optimized using new data. The stage membership vector, dynamic change pattern features, and the shape feature trends of three consecutive segments of four physiological indicators are input into the growth stage identification model to verify data integrity, ensuring there are no missing or contradictory data. If the sum of the membership vectors is 1, or if the data is abnormal, an invalid data prompt is returned, and steps 224-226 are backtracked to correct the data before re-entering. The rules are reasoned step by step. First, the first-level rules are checked. If the membership degree of a certain stage is ≥0.6, it is directly determined to be that stage. For example, in the vector (0.634, 0.366, 0, 0), the membership degree of the seedling stage is 0.634 ≥ 0.6. Initially, the stage is determined to be in the seedling stage. Verification is then performed using dynamic change pattern features. If the dynamic features corresponding to the seedling stage match the current features, the determination is confirmed. If they do not match, a secondary rule is triggered, and the stage is re-determined based on the membership degree of the mature seedling stage, thus correcting the result. If it is a tertiary stage, historical data annotation results are referenced, and a follow-up tracking mechanism is set up, shortening the collection frequency to once per day, and recalculating the membership vector for three consecutive segments. The output results and confidence level annotations are then provided. After inference, the specific growth stage is output, and the confidence level is labeled. A tertiary stage determination has a high confidence level (≥90%), a secondary stage a medium confidence level (80%-90%), and a tertiary stage a low confidence level (<80%). Low confidence results trigger manual review reminders. Cultivators revise the stage determination based on on-site observations, and the revised data is simultaneously entered into the growth stage identification model.

[0048] By refining the time granularity division and shape feature calculation logic, and combining the actual collection patterns of Anoectochilus roxburghii physiological indicators to set operational details, continuous growth data is transformed into comparative stage features, fully capturing the trends, extreme values, and curvature changes of physiological indicators, and avoiding stage misjudgments caused by insufficient feature extraction.

[0049] In a preferred embodiment of the present invention, step 3 above, which involves obtaining the dynamic ideal threshold range of each environment corresponding to the current growth stage from a preset growth stage-environment mapping relationship based on the current growth stage of *Anoectochilus roxburghii*, may include:

[0050] In this embodiment of the invention, step 330 involves using the current growth stage of *Anoectochilus roxburghii* as the unique query identifier to call and access a pre-built dynamic mapping database of growth stages and environment. Specifically, this includes: first, determining the unique query identifier standard for the current growth stage of *Anoectochilus roxburghii*, using a pure Chinese character format without special symbols, corresponding only to the four typical growth stages, disallowing abbreviations and alternative names, ensuring complete consistency with the stage names in the pre-built database; simultaneously, validating the format of the current growth stage identifier, including whether it is one of the four stages, whether there are typos, and whether it contains spaces or special characters. If the validation fails, a prompt is displayed to confirm the correct identifier format and selectable stage names, and the query is re-initiated after manual correction of the identifier; second, determining the specific information of the pre-built database, which is deployed on a local server in the soilless cultivation control center, using a hierarchical storage structure, divided into four core data tables: a growth stage table, an environmental parameter table, a dynamic threshold table, and a timeline table. These four tables are linked through stage IDs and parameter IDs to ensure the continuity of data retrieval. The database data sources include historical data from multiple batches of hydroponically cultivated Anoectochilus roxburghii over the past 5 years, orthogonal experimental data on environmental requirements at different growth stages, and practical experience summaries from senior cultivators. Threshold definitions are updated quarterly based on growth data from new cultivation batches to ensure the data aligns with actual growth needs. Subsequently, a database call and access is initiated through the internal data interface of the cultivation control system. The request carries three core pieces of information: a uniquely verified query identifier, the access device number, and the access timestamp. Upon receiving the request, the server first verifies the validity of the device number, then verifies the validity of the query identifier and timestamp. If all three verifications pass, an encrypted data connection is established, and the database call is completed. Access is denied if the device number is invalid. If the query identifier remains ambiguous after verification, a vague identifier prompt is returned, and the request is re-initiated after the problem is resolved. During the access process, an access log is recorded synchronously, including the access device number, query identifier, access time, connection duration, and data return status. The log is retained for 90 days for subsequent traceability and anomaly detection.

[0051] Step 331: In the growth stage and environment dynamic mapping database, perform a retrieval operation using the current growth stage as the key index to locate and obtain the complete set of environmental mapping relationships bound to the current growth stage. Specifically, this includes: first, using the verified unique query identifier as the key index, passing it to the database retrieval function mode. The retrieval mode prioritizes locating the growth stage table, finding the corresponding stage ID through exact matching, and then querying the growth stage and environmental parameter association table through the stage ID to obtain all environmental parameter IDs bound to that stage and their corresponding associations, ensuring coverage of the six core parameters: temperature, humidity, light intensity, CO2 concentration, nutrient solution pH, and nutrient solution EC value, with no parameter omissions. During the retrieval process, the principle of exact matching and unique results is strictly followed. If multiple associated records are found through the stage ID, the record with the latest update time is automatically selected as the valid record. If no associated record is found, the correspondence between the stage ID and the query identifier is automatically verified. If the ID is incorrect, it is corrected and the retrieval is repeated. If data is missing, the default associated data for that stage is extracted from the database's backup storage area. Based on the general environmental mapping relationship for the same variety of *Anoectochilus roxburghii*, the data is marked with "spare data" to remind users that manual verification is required later. Then, a complete set of environmental mapping relationships is obtained. Using the parameter IDs in the association table as a basis, the threshold definition information and time axis information of the corresponding parameters are extracted from the dynamic threshold table and time axis table respectively. The stage association information, threshold definition information, and time axis information of the same parameter are integrated into a set of association data. The association data of the six parameters together constitute a complete set of environmental mapping relationships. The set must clearly include the binding basis of each parameter to the current stage, the effective conditions of each parameter's threshold, the correspondence between each parameter's time axis and the total duration of the stage, and also include the last update time and update basis of the set. Finally, the set is checked for completeness and consistency. The completeness check requires confirming that each of the six parameters has corresponding association data; if any parameter data is missing, spare data is added. The consistency check requires verifying that the time axis division of each parameter is consistent with the total duration of the current stage. If inconsistent, the time axis division is automatically corrected to ensure matching with the total duration of the stage. After the check passes, the final complete set of environmental mapping relationships is formed.

[0052] Step 332 involves analyzing the complete environmental mapping relationship set, separating each independent environmental parameter item and the corresponding dynamic threshold definition information. Specifically, this includes: firstly, decomposing the complete environmental mapping relationship set, separating six independent environmental parameter items according to their type, with each parameter item treated as an independent analysis unit. The specific definition and measurement range of each parameter are determined to avoid parameter confusion. Temperature refers to the air temperature in the cultivation area, measured at the height of the plant canopy; humidity refers to the relative humidity of the air in the cultivation area, measured at the same location as the temperature; light intensity refers to the photosynthetically active radiation received by the plant leaf surface; CO2 concentration refers to the CO2 concentration in the air of the cultivation area... The CO2 content in the solution; the pH value of the nutrient solution refers to the acidity or alkalinity of the nutrient solution near the roots, and the measurement location is the outlet of the nutrient solution circulation pipeline; the EC value of the nutrient solution refers to the conductivity of the nutrient solution, reflecting the nutrient concentration, and the measurement location is the same as the pH value; subsequently, for each independent environmental parameter, the corresponding dynamic threshold definition information is further separated. This information is the core basis for the control of parameter thresholds and needs to be broken down into five core contents, which are determined one by one. First, the time axis division method is determined, and the division is done by day, week, or stage duration. At the same time, the specific time interval of each segment is marked. For example, the temperature during the seedling stage is divided into 7 days, with days 1-7 as the first segment and days 8-14 as the second segment; second, each time... The threshold base range for each segment is defined, specifying the upper and lower limits of the threshold at the beginning of each segment, such as a base temperature range of 23-24℃ for the first segment; thirdly, the threshold adjustment rules determine the adjustment method and the basis for the adjustment range; fourthly, the applicable scenarios for the threshold determine whether it is applicable to the current cultivation mode, and if not, switch to the specific threshold definition for the corresponding mode; fifthly, the adjustment trigger conditions clarify whether it is necessary to trigger threshold adjustment in conjunction with changes in plant physiological indicators, for example, separating the dynamic threshold definition information corresponding to the temperature parameters during the seedling stage, dividing the time axis into 7-day segments, with a total of 5 segments, the first 4 segments each lasting 7 days, and the 5th segment lasting 11 days, adapting to a total duration of 39 days; the base range for the first segment (days 1-7) is defined. The temperature ranges from 23-24℃ for the first five periods: 23-24℃ for the second period (8-14 days), 24-25℃ for the third period (15-21 days), 25-26℃ for the fourth period (22-28 days), and 27-27℃ for the fifth period (29-39 days). The adjustment rule is linearly increasing, with the daily increase calculated by dividing the difference between the upper and lower limits of the corresponding period by the number of days in the period. The applicable scenario is hydroponics. The adjustment trigger condition is triggered only by time. The dynamic threshold definition information for each parameter is organized into a separate structured document, annotating the parameter unit, definition source, applicable scenario, and adjustment rule details. At the same time, auxiliary information unrelated to the threshold calculation is removed, and only the core calculation basis is retained, forming a one-parameter-one-threshold definition correspondence.

[0053] Step 333: Based on the preset timeline in the dynamic threshold definition information and combined with the real-time decision time, calculate and determine the dynamic ideal threshold range corresponding to each environmental parameter item at the real-time decision time. Specifically, this includes: First, for each independent environmental parameter item, extract the timeline division standard in its dynamic threshold definition information, and combine it with the start date of the current growth stage to determine the specific calendar interval of each time segment. For example, if the start date of the current seedling stage is May 1, 2025, the first segment is May 1-May 7, and the second segment is May 8-May 14. At the same time, obtain the real-time decision time, accurate to the hour, and calculate the total number of days from the start date of the current growth stage. May 12 - May 1 = 11 days, which is the 11th day of the seedling stage. First, determine the temporal position of the real-time decision point within the current growth stage. Then, substitute the temporal position of the real-time decision point into the time axis of the corresponding parameters to determine its specific time segment. Taking the seedling stage temperature parameter mentioned above as an example, the second segment of the time axis is days 8-14, and the real-time decision point is day 11, corresponding to the second time segment. At the same time, calculate the number of days that have passed in this segment for the real-time point: 11 days - 7 days = 4 days, that is, day 4 of the second segment, and the total number of days in the segment is 7 days, from 8 to 14 days. Subsequently, according to the threshold base range and adjustment rules of this segment, calculate the dynamic ideal threshold range of the real-time decision point under different conditions. The adjustment rule for condition one is linear increase. First, calculate the daily threshold increase within this segment: daily increase = (the threshold range of this segment) / ... The threshold upper limit at the end of the segment (the upper limit of the threshold at the beginning of the segment) ÷ the total number of days in the segment. Simultaneously, the lower limit of the threshold increases by the same increment each day. Taking the second segment of the temperature parameters above as an example, the initial upper limit is 25℃, and the final upper limit is 26℃. The second segment lasts 8-14 days, corresponding to a base range of 24-25℃. The initial range of the third segment is 25-26℃, so the upper limit at the end of the second segment needs to transition to 26℃. The total number of days in the segment is 7 days. The daily increase in the upper limit = (26℃ - 25℃) ÷ 7 ≈ 0.14℃. The daily increase in the lower limit is similarly 0.14℃ (increasing from 24℃ to 25℃). Then, calculate the threshold at real time: Lower threshold = Initial lower limit of the segment + (Daily lower limit increase × Number of days elapsed in the segment) = 24℃ + (0.14℃ × 4) ≈ 24.56℃ The upper limit of the threshold = the initial upper limit of the segment + (daily increase in the upper limit × number of days elapsed within the segment) = 25℃ + (0.14℃ × 4) ≈ 25.56℃. The final dynamic ideal threshold range for the real-time temperature is 24.56-25.56℃. The adjustment rule for scenario two is a step-by-step adjustment. There is no need to calculate the daily increase; the basic threshold range of the segment at the real-time time is used as the core. Only when the real-time time is less than 2 days from the end of the segment, the upper and lower limits of the threshold are finely adjusted by 0.1-0.2 units each to achieve a smooth transition between segments. For example, if the EC value of the nutrient solution during the seedling stage is segmented in 10-day intervals, and the basic range for the second segment (days 11-20) is 1.7-1.9 mS / cm, and the real-time time is day 20, then the fine adjustment is 1.75-1.9 mS / cm.95 mS / cm is used for the next segment, 21-30 days, transitioning to 1.9-2.1 mS / cm. If the real-time time is in the middle of a segment, the basic range of 1.7-1.9 mS / cm is directly adopted. Finally, the calculated dynamic ideal threshold ranges of each parameter are physically verified to ensure they conform to the actual physical meaning of the parameters. The upper limit of the humidity threshold should not exceed 100% RH, and the lower limit should not be lower than 30% RH; values ​​below this level can easily lead to leaf water loss. The upper limit of the nutrient solution pH value should not exceed 6.8, and the lower limit should not be lower than 5.2; values ​​exceeding this range will result in water loss. The environment affects root nutrient absorption; the upper limit of light intensity should not exceed 30,000 lux to avoid scorching leaves; if the calculated result exceeds the physical range, the physical boundary value should be used as the upper or lower limit of the threshold, and the threshold correction and reason should be noted. For example, if the calculated upper limit of temperature is 28.5℃, which exceeds the physical upper limit of 28℃ during the seedling stage, it should be corrected to 28℃. After verification, the dynamic ideal threshold range of all parameters should be compiled to form a list of environmental thresholds specific to real-time decision-making, determining the threshold range, calculation basis, and whether correction has been made for each parameter, for use in environmental control.

[0054] By using standardized query identifiers, a hierarchical database structure, and multiple verification mechanisms, we ensure the accurate association between growth stage and environmental mapping data, avoiding query errors caused by identifier confusion, data tampering, and abnormal access. At the same time, the database's regular update mechanism ensures that the threshold data is aligned with the latest cultivation experience, achieving dynamic adaptation of the threshold range as the growth stage changes over time.

[0055] In a preferred embodiment of the present invention, step 4 above, which compares the dynamic ideal threshold range of each environment with multidimensional environmental data, and quantifies the degree of deviation between the real-time environment and the ideal threshold range by calculating the intersection interval and relative position relationship between the real-time data curve and the boundary curve of the dynamic threshold range, and generating an environmental deviation index, may include:

[0056] In this embodiment of the invention, step 440 involves generating, within the same time window on a unified time axis, a dynamic curve of real-time monitoring data for each type of environmental parameter, as well as upper and lower boundary curves of the corresponding dynamic ideal threshold range. Specifically, this includes: first, determining the unified time window and time axis standard. The time window is set to the past 24 hours, tracing back 24 hours from the real-time decision-making time to ensure coverage of the complete fluctuation cycle of the environmental parameter, while adapting to a data collection frequency of 5 minutes per data point. A total of 288 data points are collected over 24 hours. The time axis format has the horizontal axis starting point at the hour 24 hours prior; for example, if the real-time decision-making time is 15:00, then the time axis starting point is 15:00 of the previous day. The horizontal axis scale... The data is labeled every 10 minutes, with the vertical axis representing the numerical value and unit of the corresponding environmental parameter. All curves are generated based on this unified time axis to avoid comparison bias caused by time misalignment. Secondly, dynamic curves for real-time monitoring data of environmental parameters are generated. For six parameters—temperature, humidity, light intensity, CO2 concentration, nutrient solution pH, and nutrient solution EC value—real-time monitoring data for nearly 24 hours is extracted from the local data server, with one data point every 5 minutes, and the data integrity is verified one by one. The monitoring data for each parameter are mapped to specific minute scales on the horizontal axis in chronological order, and all data points are connected point by point with a smooth straight line to form the dynamic curve for real-time monitoring data of that parameter. The curves are clearly labeled with the parameter name, data collection frequency, and data source, such as temperature... The temperature-real-time monitoring curve is labeled. If data is missing for a certain period, it is filled in using the average of the three data points before and after the missing period, and labeled with the words "data completion" to ensure the curve is continuous without breaks. Next, the upper and lower boundary curves of the dynamic ideal threshold range are generated. Based on the real-time dynamic ideal threshold range of each parameter determined in step 333, and combined with the threshold adjustment rules, the upper and lower boundary values ​​of the corresponding parameters are calculated point by point in time, and then mapped to a unified time axis to generate curves. The specific generation logic is as follows: if it is a linear adjustment rule, the upper and lower boundary values ​​of each minute are calculated according to the daily increase in step 333. For example, if the real-time dynamic ideal threshold range of the temperature parameter is 24.56-25.56℃, at 15:00 on the 11th day of the seedling stage... If the daily increase is 0.14℃, then the increase per minute = 0.14℃ ÷ 1440 minutes ≈ 0.000097℃. Starting from the beginning of the time axis, for every minute forward, the upper boundary value increases by 0.000097℃, and the lower boundary value increases by 0.000097℃ simultaneously. The upper and lower boundary values ​​for all time points are calculated point by point, and connected by straight lines to form smooth upper and lower boundary curves. If it is a stepped adjustment rule, the upper and lower boundary values ​​within the same time segment remain fixed, generating a horizontal straight line. After entering the next segment, the horizontal straight line is switched according to the segment threshold range. The time points at the segment junctions are clearly marked. For example, the nutrient solution EC value is adjusted in a stepped manner, with the threshold range of 1.7-1.9 mS / cm for the first 10 hours and 1.75-1 mS / cm for the next 14 hours.If the value is 95 mS / cm, then the upper boundary curve consists of two horizontal straight lines at 1.9 mS / cm and 1.95 mS / cm, and the lower boundary curve consists of two horizontal straight lines at 1.7 mS / cm and 1.75 mS / cm. The transition points are marked as segmented. Finally, the consistency of the three curves is checked, verifying the reasonableness of their values ​​at the same time point. For example, if the upper boundary value is consistently greater than the lower boundary value, and the real-time curve value is within the physical range of the parameters, then if an anomaly occurs (upper boundary ≤ lower boundary), the threshold calculation process in step 333 is revisited for correction. If the real-time curve exceeds the physical range, an anomaly is marked, and the original data is retained to ensure that the three curves can be used normally for subsequent comparative analysis.

[0057] Step 441: Based on the real-time monitoring data dynamic curve, upper boundary curve, and lower boundary curve, calculate the intersection interval and excess interval between the real-time monitoring data dynamic curve and the upper boundary curve, and the intersection interval and lower boundary interval between the real-time monitoring data dynamic curve and the lower boundary curve, to obtain detailed positional relationship data of the real-time data relative to the dynamic threshold boundary. Specifically, this includes: firstly, comparing the numerical relationship of the three curves at each time point, analyzing the positional status of the real-time monitoring data at the minute granularity; if the real-time data > the upper boundary value, it is determined to be beyond the upper boundary; if the real-time data < the lower boundary value, it is determined to be below the lower boundary; if the lower boundary value ≤ the real-time data ≤ the upper boundary value, it is determined to be within the ideal range; for each time point of state change, mark it as a key node, record the specific time and the corresponding values ​​of the three curves, for example, if the real-time temperature rises from 25.4℃ (within the ideal range) to 25.6℃ (exceeding the upper boundary of 25.5℃), the key node is 10:05 on the same day, and the corresponding values ​​are 25.6℃, 25.5℃, and 2... 4.6℃; Next, calculate the intersection interval. The intersection interval refers to the time range during which the real-time monitoring curve intersects with the upper / lower boundary curve and the state changes. It is recorded in the format of start time and end time. The intersection interval with the upper boundary curve starts from the critical node where the real-time data rises from within the ideal range to exceed the upper boundary, and ends at the critical node where the real-time data falls from exceeding the upper boundary back to within the ideal range. The continuous time range formed during this period is the intersection interval. If the real-time data continues to exceed the upper boundary until the end of the time window, the intersection interval extends from the state change node to the end of the window. Similarly, the intersection interval with the lower boundary curve starts from the node where the real-time data falls from within the ideal range to below the lower boundary, and ends at the node where it rises back to within the ideal range. For example, if the real-time temperature exceeds the upper boundary at 10:05 and falls back to within the ideal range at 10:10, the intersection interval with the upper boundary is 10:05-10:10; if the real-time humidity is below the lower boundary at 6:30 and rises back to within the ideal range at 6:45, the intersection interval with the lower boundary is 6:30-6:45. Each intersecting interval needs to synchronously record the difference between the real-time data at the start and end times and the boundary value, clearly defining the numerical changes at the intersection. Then, calculate the out-of-range and below-range. The out-of-range refers to the time range during which the real-time data is continuously greater than the upper boundary value, starting from the end time of the intersecting interval with the upper boundary until the next entry into the intersecting interval or the end of the window. If the real-time data starts the time sequence window directly from the out-of-range state, the out-of-range starts from the window start point until the first entry into the intersecting interval. The below-range refers to the time range during which the real-time data is continuously less than the lower boundary value, and the calculation logic is the same as for the out-of-range. For example, if the real-time temperature enters the out-of-range state from the intersecting interval at 10:10 and re-enters the intersecting interval at 10:30, the out-of-range is 10:10-10:30; if the real-time humidity enters the intersecting interval at 6:30, enters the below-range state at 6:45, and enters the intersecting interval at 7:00, the below-range is 6:45-7:00.The duration of each interval needs to be calculated, and the maximum, minimum, and average differences between real-time data and boundary values ​​within the interval should be recorded. Finally, detailed location relationship data should be compiled, and all key nodes, intersecting intervals, intervals exceeding the limit, intervals below the limit, and corresponding numerical information should be summarized in chronological order to form a structured table. The table includes fields such as time range, location status, real-time data, boundary values, differences, and duration to ensure that the location relationship at each point in time is traceable, with no omissions or duplicates.

[0058] Step 442: Based on detailed location relationship data and the duration of each time point outside or below the interval, calculate the instantaneous deviation of the corresponding environmental parameter at each time point within the time window, as well as the cumulative deviation of the corresponding environmental parameter throughout the entire time window. Specifically, this includes: first, setting the instantaneous deviation calculation rules, calculating based on location status. In case one, the real-time data is outside the interval, excluding intersecting intervals. The instantaneous deviation is calculated as: instantaneous deviation = (real-time data - upper boundary value) × deviation coefficient of the corresponding parameter. The deviation coefficient is set based on the degree of influence of the parameter on the growth of *Anoectochilus roxburghii*. The coefficients for temperature, light intensity, and humidity are 0.2, while the coefficients for CO2 concentration, nutrient solution pH, and EC value are 0.1. These coefficients are fixed and ensure that the instantaneous deviation is non-zero. For negative values, such as when the temperature exceeds the range, the real-time data is 25.7℃, the upper boundary value is 25.5℃, the coefficient is 0.2, and the instantaneous deviation is (25.7-25.5)×0.2=0.04. In case two, when the real-time data is below the range (excluding intersecting ranges), the instantaneous deviation is (lower boundary value - real-time data)×the deviation coefficient of the corresponding parameter. For example, when the humidity is below the range, the real-time data is 63%RH, the lower boundary value is 65%RH, the coefficient is 0.2, and the instantaneous deviation is (65-63)×0.2=0.4. In case three, when the real-time data is within intersecting or ideal ranges, the instantaneous deviation is 0, requiring no additional calculation. Following the above rules, the instantaneous deviation is calculated minute by minute at each time point within the time window, and each... The calculation of time points forms the instantaneous deviation time series of the parameter, ensuring that each value is traceable. Next, the cumulative deviation is calculated, with separate calculations for deviations outside and below the specified intervals, and then summed to form the total cumulative deviation of the parameter. The specific calculation logic is as follows: First, calculate the cumulative deviation of a single out-of-interval = average instantaneous deviation of all time points within that interval × interval duration; second, calculate the cumulative deviation of a single below-interval = average instantaneous deviation of all time points within that interval × interval duration; third, calculate the total cumulative deviation of the parameter = sum of cumulative deviations outside all intervals + sum of cumulative deviations below all intervals. For example, a temperature parameter has two out-of-intervals. The first interval lasts 0.5 hours, with an average instantaneous deviation... The value is 0.04, and the cumulative deviation is 0.04 × 0.5 = 0.02; the second interval lasts for 0.3 hours, with an average value of 0.06, and the cumulative deviation is 0.06 × 0.3 = 0.018; there is no value below the interval, and the total cumulative temperature deviation is 0.02 + 0.018 = 0.038. If a parameter does not exceed or fall below the interval, the cumulative deviation is 0. After the calculation is completed, it is checked whether the cumulative deviation is non-negative. If a negative value appears, the instantaneous deviation calculation process is backtracked to correct it to ensure that the result is reasonable. Finally, the instantaneous deviation time series of each parameter is summarized, and the average value of all time points is taken as the average instantaneous deviation and total cumulative deviation of the parameter to form a parameter-average instantaneous deviation-total cumulative deviation correspondence table.

[0059] Step 443: Integrate the instantaneous and cumulative deviations of all environmental parameters, and generate an environmental deviation index through weighting and normalization. Specifically, this includes: first, setting weighting coefficients for each environmental parameter. These weighting coefficients are determined based on the priority of the parameter's impact on the growth and development of *Anoectochilus roxburghii* and the accumulation of active ingredients, ensuring that the sum of all weighting coefficients is 1. Specifically, the weighting coefficients are assigned as follows: temperature 0.3, humidity 0.2, light intensity 0.2, CO2 concentration 0.1, nutrient solution pH 0.1, and nutrient solution EC value 0.1. These weighting coefficients have been verified through multiple batches of cultivation trials and are suitable for the entire growth cycle of hydroponically grown *Anoectochilus roxburghii*, remaining constant. Next, weighting is performed, and the weighted average instantaneous deviation and cumulative deviation are calculated separately. The weighted cumulative deviation and weighted average instantaneous deviation are calculated as follows: Weighted average instantaneous deviation = (Average instantaneous deviation of temperature × 0.3) + (Average instantaneous deviation of humidity × 0.2) + (Average instantaneous deviation of light intensity × 0.2) + (Average instantaneous deviation of CO2 concentration × 0.1) + (Average instantaneous deviation of pH value × 0.1) + (Average instantaneous deviation of EC value × 0.1); Weighted total cumulative deviation = (Total cumulative deviation of temperature × 0.3) + (Total cumulative deviation of humidity × 0.2) + (Total cumulative deviation of light intensity × 0.2) + (Total cumulative deviation of CO2 concentration × 0.1) + (Total cumulative deviation of pH value × 0.1) + (Total cumulative deviation of EC value × 0.1); For example, the average temperature deviation... The instantaneous deviation is 0.03, and the total cumulative deviation is 0.038; the average humidity is 0.02, and the total cumulative deviation is 0.025; the average light intensity is 0.01, and the total cumulative deviation is 0.012; the average and total cumulative deviation of the other parameters are both 0. Therefore, the weighted average instantaneous deviation = (0.03 × 0.3) + (0.02 × 0.2) + (0.01 × 0.2) = 0.009 + 0.004 + 0.002 = 0.015; the weighted total cumulative deviation = (0.038 × 0.3) + (0.025 × 0.2) + (0.012 × 0.2) = 0.0114 + 0.005 + 0.0024 = 0.0188. Then, normalization is performed to adjust the weighted average instantaneous deviation. The weighted total cumulative deviation is mapped to the 0-1 interval to eliminate differences in numerical range. The normalization logic is as follows: first, determine the maximum possible value of each weighted value, and calculate based on the physical range of the parameter and the deviation coefficient. For example, the maximum weighted average instantaneous deviation of temperature is 0.3 × (30-28) × 0.2 = 0.12, and the sum of the maximum weighted values ​​of all parameters is 0.5. Then, divide the actual weighted value by the maximum possible value to obtain the normalized value. For example, if the weighted average instantaneous deviation is 0.015 and the maximum possible value is 0.5, the normalized value is 0.015 ÷ 0.5 = 0.03; if the weighted total cumulative deviation is 0.0188 and the maximum possible value is 0.5, the normalized value is 0.0188 ÷ 0.5 ≈ 0.0376. Finally, an environmental deviation index is generated, which is then fused with the normalized weighted average instantaneous deviation and the weighted cumulative deviation at a 1:1 ratio. That is, the environmental deviation index = (normalized instantaneous deviation + normalized cumulative deviation) ÷ 2. The index ranges from 0 to 1, where 0 represents that the real-time environment perfectly matches the ideal threshold range, and the closer the value is to 1, the more severe the deviation. For example, in the above example, the index = (0.03 + 0.0376) ÷ 2 ≈ 0.0338, which is marked as a slight deviation. After generating the index, the weighting coefficients, normalization basis, and fusion ratio are recorded simultaneously to form an environmental deviation index report, determining the overall environmental deviation and the main contributing parameters.

[0060] By unifying the timeline and time window, three precisely corresponding dynamic curves are generated. At the same time, boundary curves are generated by combining threshold adjustment rules to ensure that the deviation from the standard is objectively reflected in terms of the real-time growth requirements of *Anoectochilus roxburghii* and the actual impact of the deviation on the growth of *Anoectochilus roxburghii*.

[0061] In a preferred embodiment of the present invention, step 5 above, which involves constructing a multidimensional evaluation vector based on the environmental deviation index, analyzing the interrelationships between the dimensions of the multidimensional evaluation vector, evaluating the contribution of each dimension to the overall control decision, and generating a dynamic control correction amount, may include:

[0062] In this embodiment of the invention, step 550 involves constructing a multi-dimensional evaluation vector for each environmental parameter, based on the deviation degree and environmental deviation index, including the dimensions of deviation magnitude, deviation direction, and deviation duration. Specifically, this includes: first, determining the core components of the multi-dimensional evaluation vector; each environmental parameter corresponds to a three-dimensional vector, with the three dimensions being deviation magnitude, deviation direction, and deviation duration, respectively. The dimension order is fixed, and all dimensions are represented using standardized numerical values ​​to ensure horizontal comparison of vectors for different parameters. Simultaneously, the parameter deviation data from step 442 and the environmental deviation index from step 443 are linked as core data support for vector construction, ensuring that the vector data is consistent with the previous deviation analysis. The definitions and numerical values ​​of the three dimensions are determined one by one. Based on the source and calculation logic, the six environmental parameters are all constructed into vectors according to a unified rule, with deviation magnitude as the dimension, comprehensively reflecting the severity of the parameter's deviation from the ideal threshold. The deviation magnitude is calculated by combining the average instantaneous deviation and the total cumulative deviation, with weights of 0.6 and 0.4 respectively. The instantaneous deviation reflects the real-time impact and has a higher weight. The calculation logic is: Deviation Magnitude = (Average Instantaneous Deviation × 0.6) + (Total Cumulative Deviation × 0.4). The result is rounded to four decimal places; a larger value indicates a more severe deviation. For example, with an average instantaneous temperature deviation of 0.0040 and a total cumulative deviation of 0.0005, the deviation magnitude = (0.0040 × 0.6) + (0.0005 × 0.4) = 0.0024 + 0.0002 = 0.0026. If the parameter has no deviation, the deviation magnitude is 0. The deviation direction dimension characterizes the specific trend of parameter deviation, using standardized numerical labels, and is divided into only three cases: first, the parameter is in a state beyond the upper boundary, labeled as +1 (the parameter needs to be adjusted down); second, the parameter is in a state below the lower boundary, labeled as -1 (the parameter needs to be adjusted up); third, the parameter has no deviation or is in the ideal range, labeled as 0 (no direction adjustment is needed). The direction determination is directly based on the interval statistical results of step 441. If the parameter is both beyond and below the interval, the direction corresponding to the interval with the longer duration is taken as the standard. If the durations are the same, the direction of the interval with the more severe deviation is taken as the standard, and bidirectional deviation is marked. For example, if the temperature exceeds the interval for 0.5 hours and is below the interval for 0.3 hours, then... The deviation is labeled +1, representing the cumulative duration dimension of the quantified parameter deviation. The sum of all durations exceeding and falling below the range for that parameter in step 441 is multiplied by 0.1 to convert it into a standardized value, and the result is rounded to four decimal places. If the parameter has no deviation range, the duration dimension value is 0. For example, if the temperature exceeds the range for a total of 0.8 hours and falls below the range for 0.3 hours, the total duration is 1.1 hours, and the duration dimension = 1.1 × 0.1 = 0.1100. After construction, the multidimensional evaluation vector for each parameter is represented in the format of deviation magnitude, deviation direction, and deviation duration, with the synchronous labeling vector construction basis, such as the temperature vector (0.0026, +1, 0.1100), based on an average instantaneous deviation of 0.The six parameters—0040, total cumulative deviation of 0.0005, out-of-direction deviation, and total deviation duration of 1.1 hours—were each used to construct a vector. These vectors were then aggregated to form a vector set. This set was verified against previous deviation data and interval statistical results to ensure that the vector values ​​were consistent with the original data and that there were no logical contradictions.

[0063] Step 551: Based on the multidimensional evaluation vector, using a pre-defined correlation analysis mechanism, analyze the mutual influence and coupling relationships between different dimensions in the vector. Specifically, this includes: first, determining the core of the pre-defined correlation analysis mechanism, which is based on the physiological characteristics of hydroponically cultivated Anoectochilus roxburghii and summarizing historical cultivation data from multiple batches. It analyzes the mutual influence between dimensions and parameters using only defined correlation rules, divided into two analysis directions: correlation within the same parameter dimension and coupling between cross-parameter dimensions. This ensures that the correlation relationship aligns with actual growth needs. For correlation analysis within the same parameter dimension, for a single parameter's three-dimensional vector, analyze the mutual influence of deviation magnitude, direction, and duration. The core rules include: firstly, deviation... The duration is positively correlated with the magnitude of the deviation; that is, the longer the duration, the greater the deviation tends to be, requiring the duration to be noted to amplify the deviation cumulatively. Secondly, the direction of deviation determines the trend of the duration's influence on the magnitude of the deviation. When the deviation exceeds the direction, the longer the duration, the greater the deviation tends to be, while when it is below the direction, the opposite is true, requiring the direction to be noted as the dominant influence trend. Thirdly, when there is no deviation direction (0), both the magnitude of the deviation and the duration are 0, with no correlation. For example, the temperature vector (0.0026, +1, 0.1100) is analyzed as having a deviation duration of 0.11 hours (after standardization) when the deviation exceeds the direction, further amplifying the magnitude of the deviation; the two are positively correlated. Cross-parameter dimension coupling analysis, targeting different... The parameter vector is used to analyze the dimensional coupling relationship between core and auxiliary parameters, focusing on parameter combinations that significantly affect the growth of *Anoectochilus roxburghii*, determining the coupling rules and degree of influence. Temperature and humidity are coupled; the magnitude of temperature deviation is negatively correlated with the magnitude of humidity deviation. When the temperature exceeds the upper boundary (direction +1), the humidity tends to fall below the lower boundary (direction -1), and for every 0.001 increase in temperature deviation, the humidity deviation may increase by 0.0008. It needs to be noted that an increase in temperature leads to a decrease in humidity, with a coupling coefficient of 0.8. Conversely, when the temperature falls below the lower boundary, the humidity tends to exceed the upper boundary, and the coupling logic remains consistent. Light intensity and CO2 concentration are also coupled; the magnitude of light intensity deviation is related to the magnitude of CO2 concentration deviation. There is a positive correlation between light intensity and EC value. When the light intensity exceeds the upper boundary, the CO2 concentration needs to be increased simultaneously; otherwise, the photosynthetic efficiency will decrease. For every 0.001 increase in light intensity deviation, the CO2 concentration needs to be adjusted by 0.0005. It is noted that increased light intensity requires simultaneous CO2 supplementation, with a coupling coefficient of 0.5. When the light intensity is below the lower boundary, the CO2 concentration needs to be reduced accordingly to avoid excessive concentration inhibiting growth. The pH value of the nutrient solution is coupled with the EC value. The magnitude of pH deviation will affect the effective absorption efficiency of the EC value. When the pH exceeds the upper boundary (becomes alkaline), the actual effectiveness of the EC value decreases, and the magnitude of the EC value deviation needs to be appropriately increased by 0.0003 (corresponding to the standardized value). It is noted that alkaline pH reduces EC effectiveness, with a coupling adjustment coefficient of 0.3. When the pH is below the lower boundary (acidic), the effectiveness of the EC value increases, and the deviation of the EC value needs to be appropriately reduced; the coupling logic is the opposite. After the analysis, compile a list of correlations and coupling relationships, summarizing them by parameter combination, correlation type, direction of influence, coupling coefficient, and correlation basis fields. Label the physiological effects of *Anoectochilus roxburghii* corresponding to each relationship to ensure that the relationship resolution is traceable and verifiable.

[0064] Step 552: Based on the mutual influence and coupling relationships between dimensions, and combined with accumulated regulatory effect data, determine the relative importance of deviations in each environmental dimension on the overall growth of *Anoectochilus roxburghii*. Specifically, this includes: first, determining the evaluation indicators and data sources. The evaluation indicators focus on the core characteristics of *Anoectochilus roxburghii* growth, and the data sources are the cumulative regulatory effect data from the hydroponics of the same variety of *Anoectochilus roxburghii* in the last five batches, ensuring that the importance determination aligns with actual cultivation feedback rather than theoretical derivation. Second, setting an importance scoring standard using a 10-point scale. The scoring dimensions include the weight of single-dimensional influence, the weight of coupling relationship influence, and the weight of historical regulatory effect, with weights of 0.4, 0.3, and 0.3 respectively. The higher the total score, the higher the relative importance. The specific scoring logic is as follows: Single-dimensional influence scores are based on the direct impact of parameter deviations on growth indicators. Temperature and light intensity directly affect photosynthesis, with a maximum score of 10; humidity affects transpiration, with a maximum score of 9; nutrient solution pH and EC values ​​affect root nutrient absorption, with a maximum score of 8; CO2 concentration assists photosynthesis, with a maximum score of 7. The actual score = maximum score × the magnitude of the parameter deviation (after standardization). A deviation of 0 results in 0 points. For example, a temperature deviation of 0.0026 results in a single-dimensional score of 10 × 0.0026 = 0.026. Coupling relationships influence the weighted scores, based on the coupling coefficient from step 551. A higher coupling coefficient results in a higher score. The calculation logic is: Score = Coupling Coefficient. The coefficient is multiplied by the single-dimensional score of the corresponding associated parameter. If multiple coupling relationships exist for the parameter, the maximum score is taken. For example, if the coupling coefficient between temperature and humidity is 0.8 and the single-dimensional score for humidity is 0.020, the temperature coupling score = 0.8 × 0.020 = 0.016. The historical control effect weight score is based on the improvement in growth indicators after parameter adjustment in historical data. The greater the improvement, the higher the score. The calculation logic is: Score = Historical average improvement (standardized) × Parameter deviation duration (standardized). The historical average improvement is calculated from the last 5 batches of data. For example, if the historical improvement in temperature is 0.2 and the deviation duration is 0.1100, the historical score = 0.2 × 0.1100 = 0.022; Then, calculate the relative importance score of each parameter. The total score = (single-dimensional score × 0.4) + (coupling score × 0.3) + (historical score × 0.3). The result is rounded to four decimal places and sorted by score from high to low to classify importance levels: Level 1 (score ≥ 0.02, core influencing parameter), Level 2 (0.01-0.02, important influencing parameter), and Level 3 (< 0.01, minor influencing parameter). For example, the total score for temperature = (0.026 × 0.4) + (0.016 × 0.3) + (0.022 × 0.3) = 0.0104 + 0.0048 + 0.0066 = 0.0218, which is determined to be a Level 1 important parameter; the total score for CO2 concentration is 0.008 is classified as a Level 3 minor parameter. Finally, the rationality of the importance level is verified, and the level is adjusted based on the coupling relationship. If a parameter is a strongly coupled auxiliary parameter to a core parameter, its level can be increased by one level; if a parameter has no coupling relationship and its historical improvement effect is poor, its level can be decreased by one level. After verification, a list of parameter names, relative importance scores, levels, core impact growth indicators, and adjustment priorities is generated. The adjustment priority is consistent with the importance level, with Level 1 parameters adjusted first and Level 3 parameters adjusted last.

[0065] Step 553 involves comprehensively calculating the current multidimensional assessment vector, the inter-dimensional interactions and coupling relationships, and the relative importance, to generate the adjustment direction and magnitude of each environmental parameter used to correct environmental deviations, i.e., the dynamic control correction amount. Specifically, this includes: first, determining the adjustment direction, directly based on the deviation direction dimension of the multidimensional assessment vector, combined with the importance level and coupling relationship for fine-tuning. The core rule is that the adjustment direction for first- and second-level parameters is directly determined according to the deviation direction; for third-level parameters, if they are strongly coupled auxiliary parameters to the core parameters, the direction needs to be adjusted according to the coupling relationship; if they are independent parameters, the direction can be temporarily left unchanged. For example, if the temperature vector direction is +1, the adjustment direction is downward; if the humidity vector direction is -1, the adjustment direction is upward; if the CO2 concentration vector direction is 0, the adjustment direction is... The result remains unchanged. Next, the adjustment range is calculated, taking into account the deviation magnitude, importance score, and coupling coefficient. This is done in three steps: basic adjustment range = parameter deviation magnitude × relative importance score × 100, with the result rounded to four decimal places. When the deviation magnitude is 0, the basic range is 0. For example, if the temperature deviation magnitude is 0.0026 and the importance score is 0.0218, the basic range = 0.0026 × 0.0218 × 100 ≈ 0.005668. The coupling adjustment range is adjusted based on the coupling coefficient from step 551. The calculation logic is: coupling range = basic range × coupling coefficient × importance score ratio of the corresponding core parameter (core parameter score ÷ sum of the two scores). If there are multiple coupling relationships between parameters, the maximum coupling range is taken.For example, the humidity and temperature coupling coefficient is 0.8, the basic humidity amplitude is 0.0045, the temperature importance score is 0.0218, the humidity score is 0.018, and the proportion = 0.0218 ÷ (0.0218 + 0.018) ≈ 0.548. The coupling amplitude = 0.0045 × 0.8 × 0.548 ≈ 0.001973. The final adjustment amplitude = basic adjustment amplitude + coupling adjustment amplitude. If the final amplitude is negative, the absolute value is taken. If the parameter is a three-level independent parameter, the final amplitude = basic amplitude × 0.5. For example, the final amplitude of temperature = 0.005668 + 0 (no coupling adjustment required) ≈ 0.005668; the final amplitude of humidity = 0.0045 + 0.001973 ≈ 0.006473. Then, the standardized final amplitude is converted into actual parameter units, and the conversion coefficient is determined by combining the physical range of the parameter and the tolerance range of *Anoectochilus roxburghii*. The temperature conversion coefficient is... 1. The conversion coefficients are: humidity 10, light intensity 100, CO2 concentration 50, nutrient solution pH 0.1, and EC 0.01. The actual amplitude is calculated as: final standardized amplitude × conversion coefficient. The results are rounded to a reasonable number of decimal places. For example, if the temperature standardized amplitude is 0.005668, the actual amplitude is 0.005668 × 1 ≈ 0.01℃; if the humidity standardized amplitude is 0.006473, the actual amplitude is 0.006473 × 10 ≈ 0.06%RH. Finally, a dynamic control correction list is generated, summarizing parameters by name, adjustment direction, standardized amplitude, actual amplitude, adjustment priority, coupled parameters, and adjustment basis. The adjustment limits for each parameter are marked. If the actual amplitude exceeds the limit, the adjustment is made according to the maximum limit amplitude, and the amplitude is marked as limited. At the same time, the change in deviation after adjustment is predicted to provide a basis for verifying the control effect.

[0066] By constructing a multidimensional evaluation vector, the magnitude, direction, and duration of parameter deviations are quantified and integrated to ensure that the deviation state of each parameter can be fully captured, avoiding the control deviation caused by single-dimensional evaluation, and achieving accurate dynamic adaptation of environmental parameters.

[0067] In a preferred embodiment of the present invention, step 6 above, which integrates the dynamic adjustment correction amount with short-term climate prediction data to make a comprehensive decision and dynamically generate a control strategy for the growth environment of *Anoectochilus roxburghii*, thereby realizing the dynamic control of the growth process of *Anoectochilus roxburghii*, may include:

[0068] In this embodiment of the invention, step 660 involves fusing and analyzing the dynamic adjustment correction amount with climate prediction data within a preset short-term time window. Based on the adjustment direction and magnitude indicated in the correction amount and the changing trend indicated by the prediction data, the trajectory of comprehensive environmental parameter changes in the near future is obtained. Specifically, this includes: first, determining the preset short-term time window and data source. The time window is set to the next 4 hours, a duration that is suitable for both the accuracy of short-term climate prediction and the response cycle of environmental control equipment. The climate prediction data comes from the joint prediction of local agricultural meteorological stations and micro-meteorological sensors in the cultivation area, updated hourly, covering the predicted values, changing trends, and rate of change of six major environmental parameters. The reliability of the forecast is marked by time, and the dynamic adjustment correction amount uses the list data generated in step 553, including the adjustment direction, actual magnitude, priority, and coupling relationship of each parameter, to ensure that the fused data is consistent with the previous control logic. Secondly, fusion analysis rules are established, and fusion calculations are carried out on a parameter-by-parameter and hourly basis. The core rules combine the growth requirements of *Anoectochilus roxburghii* with the parameter coupling relationship to ensure that the comprehensive trajectory closely matches actual environmental changes. For single-parameter fusion calculations, the hourly comprehensive parameter value = hourly climate forecast value + predicted trend change amount - dynamic adjustment correction magnitude, where the predicted trend change amount = predicted trend rate × 1 hour. For example, if the current dynamic adjustment correction amount for temperature is a decrease of 0.01℃, the next hour... The predicted temperature is 25.57℃, with a trend of increasing by 0.2℃. Therefore, the comprehensive temperature value = 25.57℃ + 0.2℃ - 0.01℃ = 25.76℃. The predicted temperature for the next 2 hours is 25.77℃, with a stable trend. The comprehensive value = 25.77℃ - 0.01℃ = 25.76℃. Coupling parameters are synergistically fused. Based on the coupling relationship in step 551, the comprehensive value of the associated parameters is adjusted synchronously. For example, the temperature and humidity coupling coefficient is 0.8. When the temperature decreases by 0.01℃, the humidity needs to be adjusted synchronously by the humidity equivalent corresponding to 0.008℃. Therefore, the comprehensive humidity value = predicted value + trend change + (temperature correction magnitude × coupling coefficient × humidity conversion coefficient). The humidity conversion coefficient is 10. Outliers... If the merged composite value exceeds the physical range of the parameters, it is corrected according to the physical boundary value. At the same time, the fusion amplitude of the associated parameters is adjusted, and the correction of the fusion value and the reason are marked. Next, the trajectory of the change of the comprehensive environmental parameters is generated. The six parameters after fusion each hour are arranged in chronological order and connected with smooth straight lines to form the trajectory of the single parameter change. The trajectories of the six parameters are summarized into a comprehensive trajectory chart. The trajectory chart is marked with key information, such as the predicted value, fusion value, correction amplitude, trend change, and coupling adjustment amount for each hour. The reliability of the trajectory is also marked. Finally, the continuity of the trajectory is verified to ensure that the change of the composite value in adjacent hours is stable and to avoid sudden changes that may affect the growth of Anoectochilus roxburghii. If the fluctuation is too large, the trend change amount or correction amplitude is corrected by backtracking the fusion rules.

[0069] Step 661 compares the trajectory of changes in comprehensive environmental parameters with the ideal threshold ranges for each environmental parameter preset for the growth of *Anoectochilus roxburghii*, and determines whether each environmental parameter enters the ideal threshold range at the end of the short-term time window. Specifically, this includes: first, determining the comparison benchmark and range. The comparison benchmark is the dynamic ideal threshold range at the real-time decision moment determined in step 333 and the hourly dynamic thresholds for the next 4 hours, ensuring a complete match between the comparison benchmark and the trajectory's time granularity; the comparison range covers the hourly comprehensive values ​​of the six parameters, focusing on verifying the parameter status at the end of the 4-hour window, while also considering deviations in the intermediate periods, establishing a three-level comparison judgment standard, comparing each parameter individually, and determining compliance. If the comprehensive value in the 4th hour falls within the corresponding hourly dynamic ideal threshold range, and the comprehensive value in the last hour (3rd-4th hour) remains within the ideal range without exceeding / falling below it, it is determined that it can automatically enter the ideal range; if the comprehensive value in the intermediate period (1st-3rd hour) falls within the corresponding hourly dynamic ideal threshold range, it is determined that it can automatically enter the ideal range; If there is a slight deviation, but the value remains stable and meets the standard in the 4th hour, it is still considered to meet the standard. If the value falls within the ideal range in the 4th hour, but fluctuates in the last hour or the deviation is close to the critical value, it is considered to enter the observation state and requires close monitoring of trajectory changes without further adjustments. If the value does not meet the standard, and the value exceeds / falls below the dynamic ideal threshold range in the 4th hour, or shows a continuous deviation trend in the last hour, it is considered to be unable to automatically enter the ideal range. The reasons need to be analyzed to provide a basis for equipment command calculation. After the comparison is completed, a comparison result list is generated, which is summarized by parameter name, hourly comprehensive value, corresponding hourly ideal threshold, 4th hour status, judgment result, and reason for non-compliance. For non-compliant parameters, the deviation direction, deviation magnitude, and trend change rate are additionally marked to clarify the gap that needs to be made up by equipment adjustment. For compliant parameters, the current correction amount is marked to ensure that resources are preferentially allocated to non-compliant parameters.

[0070] Step 662: If the judgment result is that it cannot enter automatically, calculate the environmental equipment control instruction sequence according to the dynamic regulation correction amount, combined with the response characteristics of the environmental regulation equipment and the preset energy consumption constraint conditions. Specifically, it includes: First, determine the list of environmental regulation equipment and characteristic parameters, sort out the supporting regulation equipment in the cultivation area. The corresponding six major parameters are temperature, humidity, light intensity, CO2 concentration, nutrient solution pH value, and nutrient solution EC value. All equipment is marked with response delay time, operating power range, and fault backup plan to ensure that the instruction calculation fits the actual capabilities of the equipment. Second, determine the preset energy consumption constraint conditions, which are divided into upper limits of energy consumption for each time period, upper limits of energy consumption for single equipment, and energy consumption priorities. The constraint conditions are determined through cultivation cost accounting and power load tests, and can be fine-tuned according to actual cultivation requirements. Calculate the instruction sequence according to the logic of first calculating the basic instructions, then correcting according to the characteristics, and finally optimizing according to the energy consumption constraints. For the calculation of basic instructions, based on the actual amplitude of the dynamic regulation correction amount and the unqualified gap, determine the basic operating parameters of the equipment. For example, if the temperature is unqualified and needs to be lowered by an additional 0.1°C, and the air conditioner's response speed is 5 minutes / °C, then the basic operating time = (0.1°C ÷ 1°C) × 5 minutes = 0.5 minutes, and the operating power is set according to full load. If the humidity needs to be increased by an additional 0.5%RH, and the humidifier is 10 minutes / 1%RH, then the basic operating time = (0.5%RH ÷ 1%RH) × 10 minutes = 5 minutes. For equipment characteristic correction, correct the operating parameters in combination with the equipment response delay and stabilization time. For example, if the air conditioner has a startup delay of 2 minutes, the starting time of the operation needs to be advanced by 2 minutes; after the humidifier operates, it takes 3 minutes to stabilize the humidity, so it needs to end the operation 3 minutes before the target time to ensure that the humidity reaches the standard at the 4th hour. At the same time, consider the coupling relationship to correct the instructions. For example, when the temperature is lowered by 0.1°C, the humidity is likely to rise by 0.08%RH, and the operating time of the humidifier can be appropriately shortened by 0.8 minutes. For energy consumption constraint optimization, calculate the estimated energy consumption corresponding to the basic instructions. If it exceeds the constraint, make fine-tuning. For example, during the peak period, the basic energy consumption of the air conditioner is 0.017 degrees (0.5 minutes) + the humidifier is 0.05 degrees (5 minutes) = 0.067 degrees, which does not exceed the upper limit, so the instructions are retained; if the superimposed energy consumption of multiple equipment exceeds the upper limit, cut the operating time of the secondary equipment according to the energy consumption priority, and at the same time re-calculate the correction amplitude to ensure that the compliance target is not affected. Finally, generate the control instruction sequence, arrange the equipment instructions in chronological order, and each instruction includes the equipment number, operating action, operating parameters, starting time, ending time, estimated energy consumption, associated parameters, and backup plan; the sequence is sorted minute by minute in a 4-hour window to ensure no instruction conflicts, and the estimated total energy consumption per hour is marked synchronously to check whether it meets the constraint conditions.

[0071] Step 663 involves sending the control command sequence to the corresponding actuators to drive the equipment to perform corresponding actions, thereby achieving dynamic control of the growth process of *Anoectochilus roxburghii*. Specifically, this includes: First, establishing a command issuance process and verification mechanism. Commands are issued through the encrypted communication interface within the cultivation control system. Before issuance, the identity of the actuator and the equipment status are verified, and commands are only issued to equipment in normal standby mode. If equipment malfunctions, backup equipment is immediately activated, and the equipment number and operating parameters in the command are simultaneously corrected, marking the equipment switch. Commands are issued device-by-device and in batches, with core parameter devices prioritized and secondary parameter devices delayed by one minute to avoid power load fluctuations caused by simultaneous activation of multiple devices. Second, execution feedback and real-time monitoring are implemented. After receiving the command, the equipment immediately returns a success / failure feedback. If reception fails, the command is reissued twice. If it still fails, an alarm is triggered, notifying the cultivation personnel for manual intervention. During command execution, the operating status is collected in real time through equipment sensors, and parameters are collected through environmental sensors. The system monitors changes in parameters and compares actual changes with the overall trajectory. If the deviation exceeds the allowable range, additional instructions are immediately added to dynamically correct the trajectory. If a device experiences a sudden malfunction during operation, the system automatically pauses subsequent instructions for that device, activates a backup device, recalculates supplementary instructions, analyzes the cause of the malfunction, and records the malfunction time, impact range, and handling measures. If there is no backup device, a shutdown instruction is immediately issued to the associated device, triggering manual intervention reminders to ensure that the growth environment of *Anoectochilus roxburghii* does not experience extreme fluctuations. Finally, the system archives and reviews the execution of instructions. After the 4-hour window ends, all control instructions are stopped or the minimum operating state is maintained. The actual operating data and final environmental parameters of each device are collected and compared with the preset trajectory and ideal thresholds to review the control effect. If the parameters meet the standards, the instruction sequence and energy consumption data are recorded as a reference for similar scenarios in the future. If the standards are still not met, the cause is analyzed, the fusion rules and instruction calculation logic are corrected, and the dynamic control correction amount is updated. At the same time, the execution log is archived and saved for 90 days for easy traceability and optimization.

[0072] By integrating short-term climate forecasts with dynamic regulation corrections and combining parameter coupling relationships to generate a comprehensive trajectory, deviations caused by blind regulation are avoided, thus ensuring the stability of the growth environment for Anoectochilus roxburghii.

[0073] As shown in Figure 2, embodiments of the present invention also provide a dynamic decision-making system for the hydroponic growth process of *Anoectochilus roxburghii*, including:

[0074] The data acquisition module is used to collect multi-dimensional environmental data and plant physiological indicators in real time during the growth of Anoectochilus roxburghii.

[0075] The identification module is used to analyze time-series physiological indicators based on multidimensional environmental data and plant physiological indicators, extract the dynamic change pattern features of the physiological indicators in stages, match and fuse the shape features of different time segments, and dynamically identify the current growth stage of Anoectochilus roxburghii through a preset growth stage identification model.

[0076] The acquisition module is used to obtain the dynamic ideal threshold range of each environment corresponding to the current growth stage from the preset growth stage and environment mapping relationship based on the current growth stage of Anoectochilus roxburghii.

[0077] The generation module is used to compare the dynamic ideal threshold range of each environment with multi-dimensional environmental data. By calculating the intersection interval and relative position relationship between the real-time data curve and the boundary curve of the dynamic threshold range, the deviation between the real-time environment and the ideal threshold range is quantified, and an environmental deviation index is generated.

[0078] The regulation module is used to construct a multi-dimensional evaluation vector based on the environmental deviation index, analyze the mutual influence relationship between the dimensions in the multi-dimensional evaluation vector, evaluate the contribution of each dimension to the overall regulation decision, and generate dynamic regulation correction amount.

[0079] The strategy module is used to make comprehensive decisions by combining dynamic regulation corrections with short-term climate forecast data, and dynamically generate regulation strategies for the growth environment of Anoectochilus roxburghii, thereby realizing dynamic regulation of the growth process of Anoectochilus roxburghii.

[0080] It should be noted that this system is a system corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0081] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A dynamic decision-making method for the growth process of *Anoectochilus roxburghii* in hydroponics, characterized in that... The method includes: Step 1, real-time acquisition of multidimensional environmental data and plant physiological indicators during the growth of *Anoectochilus roxburghii*; Step 2, based on the multidimensional environmental data and plant physiological indicators, analysis of time-series physiological indicators, extraction of dynamic change pattern features of physiological indicator stage changes; matching and fusing shape features of different time segments, and dynamically identifying the current growth stage of *Anoectochilus roxburghii* through a preset growth stage identification model; specifically including: performing timestamp alignment processing on multidimensional environmental data and plant physiological indicators to generate synchronous time-series data pairs of *Anoectochilus roxburghii* growth process that are strictly synchronized in time; fusing the synchronous time-series data pairs of *Anoectochilus roxburghii* growth process, and combining the environmental vector and physiological indicator vector corresponding to each aligned time point to construct a structured *Anoectochilus roxburghii*. A time-series data table of growth status was created. Spatiotemporal correlation analysis was performed on this table to calculate the dynamic correlation between environmental time-series data and the time-series data of various physiological indicators. Key correlation features and delay pattern features characterizing the physiological response driven by environmental changes were extracted. Based on these key correlation features and delay pattern features, the dynamic response patterns of *Anoectochilus roxburghii* interacting with the environment were classified and labeled according to a pre-defined rule base, yielding the dynamic change pattern features of *Anoectochilus roxburghii* growth within the current growth cycle. The time-series features related to the dynamic change pattern features were divided into multiple consecutive growth time segments according to a pre-defined time granularity, and the shape features of the change curves of each physiological indicator within each growth time segment were calculated. These shape features were then compared with pre-stored data. The shape features of reference time segments representing each typical growth stage of *Anoectochilus roxburghii* are compared. A shape matching algorithm is used to calculate the shape similarity between the current segment and each reference typical segment, generating a similarity matrix. Based on the similarity matrix and combined with the pre-defined inherent contribution of each reference typical segment in growth stage judgment, the matching results of multiple reference segments with similarity greater than or equal to 60% are fused to generate a stage membership vector. The stage membership vector and dynamic change pattern features are input into a pre-defined growth stage recognition model to dynamically infer and obtain the specific growth stage of *Anoectochilus roxburghii* at the current time point. Step 3: Based on the current growth stage of *Anoectochilus roxburghii*, the corresponding environmental dynamic features are obtained from the pre-defined growth stage and environment mapping relationship. Step 4: Compare the dynamic ideal threshold range of each environment with multidimensional environmental data. By calculating the intersection interval and relative position relationship between the real-time data curve and the boundary curve of the dynamic threshold range, quantify the deviation between the real-time environment and the ideal threshold range, and generate an environmental deviation index. Step 5: Based on the environmental deviation index, construct a multidimensional evaluation vector, analyze the mutual influence relationship between the dimensions in the multidimensional evaluation vector, and evaluate the contribution of each dimension to the overall control decision, generating a dynamic control correction amount. Step 6: Combine the dynamic control correction amount with the climate prediction data for the next 4 hours to make a comprehensive decision, dynamically generate a control strategy for the growth environment of Anoectochilus roxburghii, and realize the dynamic control of the growth process of Anoectochilus roxburghii.

2. The dynamic decision-making method for the hydroponic growth process of *Anoectochilus roxburghii* according to claim 1, characterized in that, The shape features include the overall trend direction of the fragment, the distribution of local extreme points, and the curvature variation pattern.

3. The dynamic decision-making method for the hydroponic growth process of *Anoectochilus roxburghii* according to claim 2, characterized in that, Based on the current growth stage of *Anoectochilus roxburghii*, the ideal dynamic threshold ranges for each environment corresponding to the current stage are obtained from the preset growth stage-environment mapping relationship. This includes: using the current growth stage of *Anoectochilus roxburghii* as the unique query identifier, calling and accessing the pre-built growth stage-environment dynamic mapping database; performing a retrieval operation in the growth stage-environment dynamic mapping database with the current growth stage as the key index, locating and obtaining the complete set of environment mapping relationships bound to the current growth stage; analyzing the complete set of environment mapping relationships, separating each independent environmental parameter item and the dynamic threshold definition information corresponding to each environmental parameter item; and calculating and determining the ideal dynamic threshold range for each environmental parameter item at the real-time decision time based on the preset time axis in the dynamic threshold definition information and the real-time decision time.

4. The dynamic decision-making method for the hydroponic growth process of *Anoectochilus roxburghii* according to claim 3, characterized in that, The upper and lower boundary values ​​of the dynamic ideal threshold range are calculated in real time based on a preset ideal growth dynamic change relationship that matches the current growth stage and time axis.

5. The dynamic decision-making method for the hydroponic growth process of *Anoectochilus roxburghii* according to claim 4, characterized in that, By comparing the dynamic ideal threshold range of each environment with multidimensional environmental data, and calculating the intersection interval and relative positional relationship between the real-time data curve and the boundary curve of the dynamic threshold range, the deviation between the real-time environment and the ideal threshold range is quantified, and an environmental deviation index is generated. This includes: for each type of environmental parameter, within the same time series window on a unified time axis, generating the real-time monitoring data dynamic curve of the environmental parameter, as well as the upper and lower boundary curves of the corresponding dynamic ideal threshold range; based on the real-time monitoring data dynamic curve, the upper boundary curve, and the lower boundary curve, calculating the intersection interval and the excess interval between the real-time monitoring data dynamic curve and the upper boundary curve, and the intersection interval and the lower boundary curve, obtaining detailed positional relationship data of the real-time data relative to the dynamic threshold boundary; based on the detailed positional relationship data, combined with the duration of each excess or lower interval, calculating the instantaneous deviation of the corresponding environmental parameter at each time point within the time series window, and the cumulative deviation of the corresponding environmental parameter throughout the entire time series window; integrating the instantaneous and cumulative deviations of all environmental parameters, and generating the environmental deviation index through weighting and normalization.

6. The dynamic decision-making method for the hydroponic growth process of *Anoectochilus roxburghii* according to claim 5, characterized in that, Based on the environmental deviation index, a multidimensional evaluation vector is constructed. The interrelationships between the dimensions of the multidimensional evaluation vector are analyzed, and the contribution of each dimension to the overall control decision is evaluated. Dynamic control correction quantities are generated, including: constructing a multidimensional evaluation vector for each environmental parameter based on the deviation degree and environmental deviation index, including the deviation magnitude, deviation direction, and deviation duration; using a pre-set correlation analysis mechanism based on the multidimensional evaluation vector, analyzing the interrelationships and coupling relationships between different dimensions in the vector; determining the relative importance of each environmental dimension deviation to the overall growth of *Anoectochilus roxburghii* based on the inter-dimensional interrelationships and coupling relationships, combined with accumulated control effect data; and comprehensively calculating the current multidimensional evaluation vector, the inter-dimensional interrelationships and coupling relationships, and the relative importance to generate the adjustment direction and magnitude of each environmental parameter for correcting environmental deviations, i.e., the dynamic control correction quantities.

7. The dynamic decision-making method for the hydroponic growth process of *Anoectochilus roxburghii* according to claim 6, characterized in that, This method integrates dynamic adjustment corrections with short-term climate prediction data to generate dynamic control strategies for the growth environment of *Anoectochilus roxburghii*, achieving dynamic control of its growth process. This includes: fusing and analyzing dynamic adjustment corrections with climate prediction data within a preset short-term time window; obtaining the trajectory of comprehensive environmental parameter changes in the short term based on the adjustment direction and magnitude indicated by the corrections and the changing trends indicated by the prediction data; comparing this trajectory with the ideal threshold ranges for each environmental parameter preset for *Anoectochilus roxburghii* growth to determine whether each environmental parameter enters the ideal threshold range at the end of the short-term time window; if the result indicates that it cannot automatically enter, calculating the environmental equipment control command sequence based on the dynamic adjustment corrections, combined with the response characteristics of the environmental control equipment and preset energy consumption constraints; and issuing the control command sequence to the corresponding actuators to drive the equipment to perform corresponding actions, thereby achieving dynamic control of the *Anoectochilus roxburghii* growth process.

8. A dynamic decision-making system for the hydroponic growth process of *Anoectochilus roxburghii*, wherein the system implements the method as described in any one of claims 1 to 7, characterized in that... include: The data acquisition module is used to collect multidimensional environmental data and plant physiological indicators during the growth of Anoectochilus roxburghii in real time; the identification module is used to analyze the time-series physiological indicators based on the multidimensional environmental data and plant physiological indicators, and extract the dynamic change pattern characteristics of the physiological indicators in stages. The shape features of different time segments are matched and fused, and the current growth stage of Anoectochilus roxburghii is dynamically identified through a preset growth stage identification model; the acquisition module is used to obtain the dynamic ideal threshold range of each environment corresponding to the current growth stage from the preset growth stage and environment mapping relationship based on the current growth stage of Anoectochilus roxburghii. The generation module is used to compare the dynamic ideal threshold range of each environment with multi-dimensional environmental data. By calculating the intersection interval and relative position relationship between the real-time data curve and the boundary curve of the dynamic threshold range, the deviation between the real-time environment and the ideal threshold range is quantified, and an environmental deviation index is generated. The regulation module is used to construct a multi-dimensional evaluation vector based on the environmental deviation index, analyze the mutual influence relationship between the dimensions in the multi-dimensional evaluation vector, evaluate the contribution of each dimension to the overall regulation decision, and generate dynamic regulation correction amount; the strategy module is used to make comprehensive decisions by combining the dynamic regulation correction amount with short-term climate prediction data, dynamically generate regulation strategies for the growth environment of Anoectochilus roxburghii, and realize dynamic regulation of the growth process of Anoectochilus roxburghii.

Images