A method for dynamic optimization and control of environmental parameters for potato seedling grading

Abstract

This invention discloses a method for dynamic optimization and control of environmental parameters for graded potato seedlings, relating to the field of agricultural information technology. Based on IoT sensors deployed in the target area, data on stem diameter, chlorophyll content, and plant height of potato seedlings are collected. K-means clustering is used to classify the seedlings into three categories: weak, medium, and strong. The duration of the initial photoperiod is set. Gradient photoperiod experiments are conducted on each category of potato seedlings, using stem elongation rate and chlorophyll content as constraints. This invention overcomes the limitations of traditional fixed photoperiod control by using graded photoperiod modeling and dynamic optimization via ant colony optimization. It sets the initial photoperiod range based on the physiological differences of seedlings, and constructs a photoperiod-growth rate response surface model by combining stem elongation rate and chlorophyll content constraints. This quantifies the impact of different photoperiod combinations on seedling growth, ensuring precise matching between photoperiod and seedling needs, avoiding excessive vegetative growth or premature senescence, and improving the stability of growth rhythm and resource utilization efficiency.

Description

Technical Field

[0001] This invention relates to the field of agricultural information technology, specifically to a method for dynamic optimization and control of environmental parameters for potato seedling grading. Background Technology

[0002] Potatoes are a typical short-day crop, and their growth is regulated by photoperiod. Short days can induce tuber differentiation, while long days can delay or inhibit tuber formation, leading to excessive vegetative growth. Traditional open-field cultivation relies on seasonal changes in light, resulting in a mismatch between the tuber formation period and market demand. However, modern greenhouses, plant factories, and other enclosed environments provide a technological basis for light regulation. By supplementing light, shading, or LED spectral control, the limitations of natural light can be overcome.

[0003] Traditional methods for controlling light duration often employ fixed photoperiods, neglecting the physiological differences among seedlings after grading. This can easily lead to excessive growth or premature aging in some seedlings due to photoperiod mismatch. Furthermore, PID control suffers from slow response speeds, making it difficult to ensure the compatibility of light conditions with the control scheme for seedling growth, which can easily disrupt the tuber formation rhythm. Therefore, the problem this invention aims to solve is how to analyze seedling status, model photoperiods for different grades, and combine this with ant colony algorithm for ant colony path planning to analyze the promoting effect of different photoperiod combinations on seedlings of different grades, so as to dynamically adjust the photoperiod duration. To this end, a dynamic optimization and control method for environmental parameters of potato seedling grading is proposed. Summary of the Invention

[0004] The purpose of this invention is to provide a method for dynamically optimizing and controlling environmental parameters for potato seedling grading, so as to solve the problems mentioned in the background art.

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

[0006] A method for dynamically optimizing and controlling environmental parameters for potato seedling grading includes the following steps:

[0007] Step 1: Based on the IoT sensors deployed in the target area, collect data on stem diameter, chlorophyll content and plant height of potato seedlings. Use K-means clustering to divide the seedlings into three categories: weak seedlings, medium seedlings and strong seedlings. Set the duration of the initial photoperiod (10-12h for weak seedlings, 12-14h for medium seedlings and 14-16h for strong seedlings).

[0008] Step 2: Using stem elongation rate and chlorophyll content as constraints, a gradient photoperiod experiment was conducted on each type of potato seedling to analyze the existing photoperiod combinations. A photoperiod-growth rate response surface model was established using multiple regression analysis to preliminarily determine the appropriate photoperiod range.

[0009] Step 3: Based on the photoperiod-growth rate response surface model, the ant colony algorithm parameters and pheromone matrix are initialized, and the light intensity is introduced through heuristic factors to optimize the path selection weights.

[0010] Step 4: Each ant in the ant colony algorithm selects a photoperiod path by combining pheromone concentration and heuristic factor probability, and performs photoperiod combination screening.

[0011] Step 5: Collect stem diameter and chlorophyll content data at fixed intervals using IoT sensors to analyze the uniformity of seedling growth under each photoperiod path;

[0012] Step 6: Update the pheromone every 24 hours based on the photoperiod path evaluation results, eliminate paths that cause excessive growth or premature aging, and after 10 iterations, lock in the optimal photoperiod-graded seedling matching scheme and optimize and control the light parameters.

[0013] A further improvement to the technical solution of the present invention is that: in step 1, the process of setting the duration of the initial optical period is as follows:

[0014] By deploying IoT sensors in the target area, physiological data of potato seedlings, including stem diameter, chlorophyll content and plant height, are collected synchronously. After data cleaning and standardization, the K-means clustering algorithm is used to classify the seedlings into three categories: weak seedlings, medium seedlings and strong seedlings based on the Euclidean distance of the physiological indicators.

[0015] Based on the seedling grading results and considering the growth requirements of potato seedlings, initial photoperiod duration ranges were set for three types of seedlings: 10-12h for weak seedlings, 12-14h for medium seedlings, and 14-16h for strong seedlings. This ensures that seedlings in different growth stages receive differentiated light strategies and avoids uneven growth caused by uniform control.

[0016] A further improvement to the technical solution of the present invention is that: in step 2, the process of initially determining a suitable optical period range is as follows:

[0017] Using stem elongation rate (daily average growth) and chlorophyll content (SPAD value) as constraints, potato seedlings were grouped according to the previous grading results (weak seedlings / medium seedlings / strong seedlings), and photoperiod gradients of 10-16 h were set for each group with a step size of 2 h. Each group was repeated 3 times and cultured continuously for 14 days under controlled conditions. Stem elongation rate (cm / d) and chlorophyll content (SPAD) were monitored simultaneously to ensure the spatiotemporal consistency of the data.

[0018] Based on experimental data, photoperiod combinations that meet the constraints were selected. Photoperiod and seedling type (dummy variable encoding) were used as independent variables, and stem elongation rate and chlorophyll content were used as dependent variables. Multiple linear regression was used to fit the response surface model, and a photoperiod-growth rate response surface model was constructed to quantify the nonlinear effect of photoperiod on growth indicators.

[0019] The accuracy of the model was evaluated through cross-validation (R²). 2 >0.85, RMSE<10%), to verify the model's prediction accuracy, output the nonlinear relationship between photoperiod and growth rate, and preliminarily determine the appropriate photoperiod range.

[0020] A further improvement to the technical solution of this invention lies in the following process: the construction of the photoperiod-growth rate response surface model and the preliminary determination of a suitable photoperiod range are as follows:

[0021] Simultaneously acquire experimental data on photoperiod and seedling category (categorical variables, encoded as dummy variables: weak seedlings = 001, medium seedlings = 010, strong seedlings = 100), as well as the corresponding stem elongation rate and chlorophyll content. Remove outliers to ensure data integrity. Perform Z-score standardization on continuous variables (photoperiod, stem elongation rate, chlorophyll content) to eliminate dimensional differences. Categorical variables (seedling category) retain the dummy encoding form and are directly included in the data. Integrate the acquired data to obtain a comprehensive dataset and divide it into training and test sets in a 7:3 ratio.

[0022] Using photoperiod and seedling type (dummy variable encoding) as independent variables, and stem elongation rate and chlorophyll content as dependent variables, a photoperiod-growth rate response surface model was constructed using multiple linear regression. The least squares method was used to fit the model, and the regression coefficients and their significance were calculated. Significant variables were screened (p<0.05).

[0023] Three-dimensional response surfaces of photoperiod-growth rate and photoperiod-chlorophyll content were generated based on regression coefficients. The marginal effect of photoperiod on growth indicators including stem elongation rate and chlorophyll content was analyzed by partial derivatives. Weak seedlings were more sensitive to photoperiod, while strong seedlings showed a faster decrease in chlorophyll content as photoperiod lengthened due to light inhibition. Piecewise regression was introduced to handle the nonlinear relationship, dividing the photoperiod into two segments: <12h and ≥12h. Models were fitted to each segment separately, and the AIC values ​​of the piecewise model and the global model were compared. When ΔAIC>2, the piecewise model was selected as the optimal model. The interval corresponding to the peak value of the response surface ±0.5 standard deviation was then taken as the suitable photoperiod interval for each seedling type.

[0024] A further improvement to the technical solution of this invention lies in the following: In step 3, the process of optimizing the path selection weights is as follows:

[0025] Based on the photoperiod-growth rate response surface model, the initial ant colony algorithm parameters are set, including the number of ants (a fixed number of individuals are assigned to each type of seedling), the pheromone volatility coefficient, and the initial pheromone concentration. A three-dimensional structure is constructed according to the seedling category (weak / medium / strong) and photoperiod to form a pheromone matrix, whose dimensions correspond to the grid points of the photoperiod-growth rate response surface.

[0026] Light intensity is introduced as a heuristic factor into the ant colony algorithm. It is normalized and mapped to the [0,1] interval. Together with pheromone concentration, it forms the path selection probability formula, i.e. the transition probability. The weight of the heuristic factor is dynamically adjusted so that the guiding effect of light intensity on path selection decreases with the number of iterations, avoiding local optima, while retaining the biological constraints of the photoperiod-growth rate response surface.

[0027] A further improvement to the technical solution of the present invention is that: in step 4, the process of screening the photoperiod combination is as follows:

[0028] Each ant locates the corresponding level of the pheromone matrix based on the current seedling category (weak / medium / strong seedling). At the current position (photoperiod-seedling combination), it calculates the transition probability of all feasible neighbor grid points by combining pheromone concentration and normalized light intensity, and randomly selects the next position based on the transition probability, taking into account both deterministic guidance and random exploration, and avoiding getting stuck in a fixed path.

[0029] After the ant completes its path selection, it records the photoperiod-seedling combination it passes through, and after the ant completes one iteration, it releases pheromones along the path it has passed through.

[0030] The process involves continuous iteration, updating the pheromone matrix and evaluating path adaptability in each round. A dual termination condition is set. After the iteration terminates, the grid point with the highest pheromone concentration in the corresponding photoperiod interval for each type of seedling is extracted from the final pheromone matrix and used as its optimal photoperiod matching scheme. Weak seedlings are locked with long-day conditions to prevent premature aging, strong seedlings are determined to have short-day conditions to promote tuber growth, and medium seedlings are matched with intermediate values ​​to balance growth.

[0031] A further improvement to the technical solution of this invention lies in the following: the process for determining the optimal optical period matching scheme is as follows:

[0032] The ant colony algorithm continuously iterates and updates the three-dimensional pheromone matrix. In each iteration, the ant colony selects a path based on the current pheromone distribution and illumination heuristics, and then globally updates the three-dimensional pheromone matrix. This involves reducing the pheromone concentration of all grid points according to a preset volatility coefficient to simulate the natural evaporation process, and adding pheromone increments to the corresponding grid points based on the predicted growth rate of the ants along the path. At the same time, after each iteration, the path fitness is re-evaluated, a comprehensive fitness score is calculated, and the predicted growth rate of the photoperiod-seedling combination is correlated with the pheromone concentration to ensure that the matrix update is synchronized with the biological model. This gradually guides the search direction to converge towards a high fitness region, determines the optimal path, and improves the global search efficiency of the algorithm.

[0033] The algorithm sets up dual termination conditions to balance computational efficiency and solution quality, namely, a preset maximum number of iterations and a dynamic convergence criterion. The preset maximum number of iterations prevents infinite loops; the dynamic convergence criterion considers the algorithm to have converged to a stable solution when the concentration value of the optimal path (the grid point with the highest pheromone concentration) does not change significantly for 10 consecutive iterations.

[0034] After the algorithm terminates, the optimal photoperiod matching scheme for each type of seedling is extracted from the final pheromone matrix. Weak seedlings are locked in the long-day interval, and the apical dominance is suppressed and premature aging is prevented by extending the light. Strong seedlings are determined in the short-day interval, and the expression of tuber formation-related genes is activated to promote the allocation of nutrients to underground organs. Medium seedlings are matched with the intermediate photoperiod.

[0035] A further improvement to the technical solution of this invention lies in the following: In step 5, the process of analyzing the uniformity of seedling growth under each photoperiod path is as follows:

[0036] Using IoT sensors in the target area, stem diameter and chlorophyll content data are collected at fixed intervals of 4 hours, and corresponding photoperiod parameters are recorded simultaneously. After the data is uploaded to the edge computing node, outlier removal and missing value imputation are performed to generate a standardized time series dataset, ensuring data integrity and consistency.

[0037] Based on the clustered seedling categories, the seedlings are grouped according to photoperiod paths. The coefficient of variation (CV) of stem diameter and the uniformity index (calculated based on the entropy value of stem diameter data) of each seedling category under the same photoperiod path are calculated. The time dynamics of uniformity are analyzed by sliding window to identify the trend of the influence of photoperiod on growth stability and generate curves of uniformity changing with time under each path.

[0038] Based on the uniformity analysis results, photoperiod paths with a stem diameter variation coefficient of <15% were selected as candidate schemes and marked as highly adaptable paths.

[0039] A further improvement to the technical solution of this invention lies in the following: the process of analyzing the time dynamics of uniformity through a sliding window and generating curves showing the change of uniformity over time for each path is as follows:

[0040] Based on the seedling classification results obtained by K-means clustering, the data are grouped according to the photoperiod parameter to ensure that each group of seedlings is in the same photoperiod path. For each group of data, the coefficient of variation (CV) of stem diameter is calculated, which is the ratio of the standard deviation of stem diameter to the mean. This quantifies the dispersion of stem diameter among individual seedlings under the same photoperiod. The smaller the CV value, the higher the growth consistency. At the same time, the evenness index is introduced to evaluate growth stability. The evenness index is calculated based on the entropy value of the stem diameter data.

[0041] The sliding window method was used to analyze the time variation trend of the uniformity index. The window width was set to a fixed time interval (7 consecutive time points). The window was slid point by point with a step size of 1. The mean value of the uniformity index in each window was calculated to generate time series data. By comparing the time series curves under different photoperiod paths, the variation pattern of uniformity was identified. At the same time, the slope and fluctuation range of the uniformity index under each path were calculated to quantify the difference in trend strength and stability.

[0042] The sliding window analysis results are visualized, generating curves showing the change of uniformity index over time for each photoperiod path. The horizontal axis represents time, and the vertical axis represents the uniformity index value. Different paths are distinguished by different colors. By comparing the curves, the influence trend of photoperiod on growth stability is presented.

[0043] A further improvement to the technical solution of the present invention is that: in step 6, the optimization and control process of the illumination parameters is as follows:

[0044] Every 24 hours, based on the photoperiod path evaluation results, the stem elongation rate and chlorophyll content of seedlings under each photoperiod path are calculated. If the path results in a stem elongation rate >0.5cm / d (excessive growth) or a chlorophyll content <30SPAD (premature senescence), it is marked as a poor path and its pheromone concentration is reduced; otherwise, it is marked as a good path.

[0045] Based on the daily evaluation results, the pheromones of inferior paths decrease at a fixed decay rate, while the pheromones of superior paths increase according to the fitness ratio. If a path is marked as inferior for 3 consecutive days, it will be forcibly eliminated. If the pheromone content of a superior path exceeds the threshold of 80%, it will be retained first and its coverage will be expanded. After 10 iterations, only the top 20% of paths with the highest pheromone concentration will be retained to form a candidate photoperiod set.

[0046] The candidate photoperiod set is grouped according to seedling category. The photoperiod distribution of each group path is statistically analyzed. The photoperiod range with the highest pheromone concentration is locked as the optimal matching scheme. The optimal matching scheme is input into the illumination control system to dynamically adjust the photoperiod.

[0047] Due to the adoption of the above technical solution, the technical progress achieved by this invention compared to the prior art is as follows:

[0048] 1. This invention provides a method for dynamic optimization and control of environmental parameters for potato seedling grading. By using graded photoperiod modeling and ant colony algorithm for dynamic optimization, it overcomes the limitations of traditional fixed photoperiod control. Based on the physiological differences of seedlings, an initial photoperiod range is set. Combined with the constraints of stem elongation rate and chlorophyll content, a photoperiod-growth rate response surface model is constructed to accurately quantify the impact of different photoperiod combinations on seedling growth. This ensures that the photoperiod is precisely matched with the needs of seedlings, avoids excessive growth or premature aging, and improves the stability of growth rhythm and resource utilization efficiency.

[0049] 2. This invention provides a method for dynamic optimization and control of environmental parameters for potato seedling grading. It uses the photoperiod path optimization mechanism of ant colony algorithm, uses the pheromone matrix to record the promoting effect of each photoperiod combination on seedling growth, and combines light intensity as a heuristic factor to dynamically adjust the path selection weight. It adopts a dynamic balance search strategy to avoid local optima and ensures that the algorithm converges efficiently to the global optimal photoperiod scheme within the iteration range. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0051] Figure 1 This is a schematic diagram illustrating the workflow of a dynamic optimization and control method for environmental parameters in potato seedling grading according to the present invention.

[0052] Figure 2 This is a schematic diagram of the process for dynamically optimizing and controlling environmental parameters in a potato seedling grading method according to the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] Example 1, such as Figure 1 , Figure 2As shown, this invention provides a method for dynamically optimizing and controlling environmental parameters for potato seedling grading, comprising the following steps:

[0055] Step 1: Based on IoT sensors deployed in the target area, collect data on stem diameter, chlorophyll content, and plant height of potato seedlings. Use K-means clustering to classify the seedlings into three categories: weak, medium, and strong. Set the initial photoperiod duration (10-12 hours for weak seedlings, 12-14 hours for medium seedlings, and 14-16 hours for strong seedlings). Simultaneously collect physiological data on potato seedlings, including stem diameter, chlorophyll content, and plant height, using IoT sensors deployed in the target area. After data cleaning and standardization... The K-means clustering algorithm was used to classify seedlings into three categories: weak, medium, and strong seedlings, based on the Euclidean distance of physiological indicators. According to the seedling classification results and the growth requirements of potato seedlings, the initial photoperiod duration ranges for the three categories of seedlings were set respectively. The initial photoperiod duration ranges for weak, medium, and strong seedlings were 10-12h, 12-14h, and 14-16h, respectively, to ensure that seedlings in different growth stages receive differentiated light strategies and avoid uneven growth caused by uniform control.

[0056] The specific work involves: synchronously collecting key physiological data of potato seedlings through a high-precision IoT sensor network deployed in the target planting area, including stem diameter (reflecting the plant's nutrient reserves and mechanical support capacity), chlorophyll content (characterizing photosynthetic efficiency and health status), and plant height (reflecting vertical growth rate and population density). The collected data is wirelessly transmitted to edge computing nodes for data cleaning, removing outliers caused by sensor malfunctions or environmental interference. Z-score normalization is then used to normalize the multi-dimensional data, eliminating dimensional differences. The standardized data is then used for clustering. To provide a unified input and avoid classification bias caused by different index magnitudes, IoT sensors were deployed in a 5m×5m grid in the target planting area (i.e., the potato experimental field). Stem diameter monitoring used laser rangefinders (accuracy ±0.01mm), installed 3cm from the base of the plant, calculating stem diameter using time-of-reflection to avoid mechanical damage. Chlorophyll content monitoring used a multispectral SPAD instrument to non-destructively measure leaf chlorophyll content, selecting the third fully unfolded leaf from the top of each plant. Plant height monitoring deployed a ToF lidar to vertically scan the plant canopy, extracting plant height using point cloud processing algorithms and sampling. The frequency was 1 Hz. Using preprocessed stem diameter, chlorophyll content, and plant height data as feature vectors, the K-means clustering algorithm was used for unsupervised classification based on Euclidean distance. By iteratively optimizing the objectives of minimizing intra-cluster distance and maximizing inter-cluster distance, the seedlings were divided into three categories: weak, medium, and strong, corresponding to low, medium, and high growth vigor groups, respectively. The classification results were evaluated for cluster quality using silhouette coefficients to ensure that intra-cluster compactness and inter-cluster separation met expectations. After grading, the mean and standard deviation of the physiological characteristics of each seedling category were recorded as feature labels. Based on the seedling grading results, combined with the photoperiod requirements for potato growth... To determine the characteristics, differentiated initial photoperiod ranges were established for different types of seedlings. Weak seedlings, due to their weak growth vigor, require extended light duration to enhance the accumulation of photosynthetic products, and their initial photoperiod was set at 10-12 hours. Medium seedlings, in the transitional stage from vegetative to reproductive growth, were given a medium photoperiod of 12-14 hours to balance the development of the aboveground and underground parts. Strong seedlings, with their vigorous growth, may have excessive light exposure that inhibits tuber formation, so a short-day range of 14-16 hours was set to induce tuber differentiation. By setting photoperiods in a tiered manner, problems such as premature aging of weak seedlings and excessive growth of strong seedlings under uniform control were avoided, while also providing an initial benchmark for subsequent dynamic optimization.

[0057] Step 2: Using stem elongation rate and chlorophyll content as constraints, a gradient photoperiod experiment (decreasing every 2 hours) was conducted on each type of potato seedling. Existing photoperiod combinations were analyzed, and a photoperiod-growth rate response surface model was established using multiple regression analysis to preliminarily determine suitable photoperiod intervals. Using stem elongation rate (daily average growth) and chlorophyll content (SPAD value) as constraints, potato seedlings were grouped according to previous grading results (weak seedlings / medium seedlings / strong seedlings), and photoperiod gradients of 10-16 hours were set for each group, with a step size of 2 hours. Each group was repeated 3 times. The seedlings were continuously cultured for 14 days under controlled conditions, with simultaneous monitoring of stem elongation rate (cm / d) and chlorophyll content (SPAD) to ensure spatiotemporal consistency of the data. The controlled environment consisted of a temperature of 25℃ / 15℃ day / night and a humidity of 60%. Growth rate was calculated by difference of plant height measurements over three consecutive days; chlorophyll content was measured using SPAD values ​​at fixed times daily. Based on the experimental data, photoperiod combinations meeting the constraints were selected: weak seedlings were retained with stem elongation rate > 0.5 cm / d and chlorophyll > 30 SPAD; medium seedlings were retained with a rate of 0.3-0.5 cm / d and chlorophyll 25-30 SPAD; and strong seedlings were retained with a rate < 0.3 cm / d and chlorophyll < 25 SPAD. Photoperiod and seedling type (coded as dummy variables) were used as independent variables, and stem elongation rate and chlorophyll content were used as dependent variables. A multiple linear regression model was used to fit the response surface model, constructing a photoperiod-growth rate response surface model to quantify the nonlinear effect of photoperiod on growth indicators. The model accuracy was evaluated through cross-validation (R²). 2 >0.85, RMSE<10%), verify the model prediction accuracy, output the nonlinear relationship between photoperiod and growth rate, and preliminarily determine the appropriate photoperiod range;

[0058] The process of constructing the photoperiod-growth rate response surface model and initially determining the suitable photoperiod range is as follows:

[0059] Experimental data on photoperiod and seedling category (categorical variables, encoded as dummy variables: weak seedlings = 001, medium seedlings = 010, strong seedlings = 100) were acquired simultaneously, along with corresponding stem elongation rates and chlorophyll content. Outliers were removed to ensure data integrity. Continuous variables (photoperiod, stem elongation rate, and chlorophyll content) were Z-score standardized to eliminate dimensional differences. Categorical variables (seedling category) were retained in dummy encoding and directly included. The acquired data were integrated to obtain a comprehensive dataset, which was then divided into training and testing sets in a 7:3 ratio. Photoperiod and seedling category (dummy variable encoding) were used as independent variables, and stem elongation rate and chlorophyll content were used as dependent variables. A photoperiod-growth rate response surface model was constructed using multiple linear regression. The least squares method was used to fit the model, and the regression coefficients and their significance were calculated. Significant variables (p<0.05) were screened, and three-dimensional response surfaces of photoperiod-growth rate and photoperiod-chlorophyll content were generated based on regression coefficients. The marginal effect of photoperiod on growth indicators including stem elongation rate and chlorophyll content was analyzed by partial derivatives. Among them, weak seedlings were more sensitive to photoperiod, while strong seedlings experienced a faster decrease in chlorophyll content as photoperiod lengthened due to light inhibition. Piecewise regression was introduced to handle nonlinear relationships, dividing the photoperiod into two segments: <12h and ≥12h. Models were fitted to each segment separately, and the AIC values ​​of the segmented model and the global model were compared. If the AIC of the segmented model was lower than that of the global model (ΔAIC>2), the segmented model was considered more reasonable. That is, when ΔAIC>2, the segmented model was selected as the optimal model. Then, the interval corresponding to the peak value ±0.5 standard deviation of the response surface was taken as the suitable photoperiod interval for each seedling type.

[0060] The photoperiod-growth rate response surface model is divided into a stem elongation rate model and a chlorophyll content model, and the expressions are as follows:

[0061] SER=β0+β1·T+β2·W+β3·M+β4·(T×W)+β5·(T×M)+∈;

[0062] SPAD=γ0+γ1·T+γ2·W+γ3·M+γ4·(T×W)+γ5·(T×M)+∈;

[0063] In the formula, SER is the stem elongation rate, SPAD is the chlorophyll content, T is the photoperiod, W is the dummy variable for weak seedlings, M is the dummy variable for medium seedlings, β0 and γ0 are the intercept terms (theoretical values ​​when photoperiod = 0 for the strong seedling baseline group), β1 and γ1 are the marginal effect of photoperiod T on strong seedlings (baseline effect), β2 and γ2 are the intercept differences between weak seedlings and strong seedlings (offset when photoperiod = 0), β3 and γ3 are the intercept differences between medium seedlings and strong seedlings, β4 and γ4 are the interaction effect between weak seedlings and photoperiod (T×W), β5 and γ5 are the interaction effect between medium seedlings and photoperiod (T×M), and ∈ is the random error term. In the global multiple linear regression model, based on the strategy of avoiding the dummy variable trap in statistics, the dummy variable for strong seedlings is not explicitly introduced. The baseline state of strong seedlings is represented by all dummy variables being 0, i.e., W = 0 and M = 0. Strong seedlings are used as a reference category, and their effect is included in the intercept term β0 or γ0.

[0064] The marginal effect of stem elongation rate is expressed as follows:

[0065]

[0066] The marginal effect of chlorophyll content is expressed as follows:

[0067]

[0068] Explanation: For weak seedlings, for every 1 hour increase in photoperiod, the stem elongation rate changes by β1+β4, and the chlorophyll changes by γ1+γ4. If β4>0, weak seedlings are more sensitive to photoperiod (risk of excessive growth 1). For strong seedlings, for every 1 hour increase in photoperiod, the stem elongation rate changes by β1 (β1<0, inhibiting excessive growth).

[0069] The specific work involved: using stem elongation rate and chlorophyll content as constraints, seedlings were grouped according to previous grading results (weak seedlings / medium seedlings / strong seedlings). Each group had a photoperiod gradient (10h, 12h, 14h, 16h, step size 2h), repeated three times. Seedlings were cultured for 14 days under controlled conditions of 25℃ / 15℃ day / night temperature and 60% humidity. The photoperiod setpoint (accurate to 0.1h) and seedling category labels were recorded periodically using edge computing nodes. Stem elongation rate was calculated by difference based on plant height measurements over three consecutive days. Chlorophyll content was measured using SPAD values ​​at fixed times each day. Based on the biological thresholds for potato photoperiod requirements, photoperiod combinations meeting the constraints were screened. For weak seedlings, combinations with stem elongation rate > 0.5cm / d and chlorophyll > 30 SPAD were retained; for medium seedlings, combinations with a rate of 0.3-0.5cm / d and chlorophyll of 25-30 SPAD were retained; and for strong seedlings, combinations with a rate < 0. Combinations with a growth rate of 3 cm / d and chlorophyll content <25 SPAD were selected. Seedling category coding was ensured as ternary dummy variables: weak seedlings = 001, medium seedlings = 010, and strong seedlings = 100. Continuous variables including photoperiod, stem elongation rate, and chlorophyll content were Z-score standardized. Categorical variables (seedling category) maintained the dummy coding form (001 / 010 / 100) and were not standardized. The training and test sets were divided in a 7:3 ratio to ensure a balanced distribution of each seedling category in the subsets. Completeness checks were performed to ensure no missing values ​​in either the training or test sets (mean imputation was used when the missing value rate was <1%). A response surface model was constructed using multiple linear regression, with photoperiod and seedling category as independent variables and stem elongation rate and chlorophyll content as dependent variables. The nonlinear relationship between photoperiod and growth indicators was quantified, and piecewise regression (photoperiod <12h and ≥12h) was introduced. The optimal model was selected by comparing AIC values. Cross-validation required R... 2 >0.85, RMSE<10%, the optimal photoperiod range was determined by the peak value of the response surface ±0.5 standard deviation, and the optimal photoperiod range for weak seedlings, medium seedlings and strong seedlings was identified;

[0070] Step 3: Based on the photoperiod-growth rate response surface model, the ant colony algorithm parameters and pheromone matrix are initialized. Light intensity is introduced through heuristic factors to optimize path selection weights. Based on the photoperiod-growth rate response surface model, the initialized ant colony algorithm parameters are set, including the number of ants (a fixed number of individuals are allocated to each type of seedling), pheromone volatility coefficient, and initial pheromone concentration. A three-dimensional structure is constructed according to the seedling category (weak / medium / strong) and photoperiod combination to form a pheromone matrix. Its dimensions correspond to the grid points of the photoperiod-growth rate response surface. Light intensity is introduced into the ant colony algorithm as a heuristic factor. After normalization, it is mapped to the [0,1] interval and together with pheromone concentration, it constitutes the path selection probability formula, i.e., the transition probability. The weights of the heuristic factors are dynamically adjusted so that the guiding effect of light intensity on path selection decreases with the number of iterations, avoiding local optima, while retaining the biological constraints of the photoperiod-growth rate response surface.

[0071] The specific work involves: Based on the photoperiod-growth rate response surface model, initializing the ant colony algorithm with core parameters including ant count, pheromone evaporation coefficient, and initial pheromone concentration. For different seedling categories (weak, medium, and strong), the number of ants is allocated in a fixed ratio to ensure that the search process for each seedling category is independent and resources are balanced. The pheromone evaporation coefficient is set to a dynamic decay mode, with a higher initial value to accelerate early exploration, gradually decreasing with each iteration to balance global search and local convergence capabilities. The initial pheromone concentration is set based on the predicted value of the response surface model, with higher initial concentrations assigned to high-growth-rate regions to guide ants to prioritize exploring potential optimal solutions. When constructing the three-dimensional pheromone matrix, photoperiod is used as the horizontal axis, seedling category as the vertical axis, and growth rate as the vertical axis. The grid points of the response surface are mapped to matrix elements, with each element storing the pheromone concentration of the corresponding photoperiod-seedling combination, forming a structured data storage framework consistent with the model dimensions. Light intensity is introduced as a heuristic factor into the path selection mechanism to influence the light... The intensity data is normalized and mapped to the [0,1] interval through linear transformation to eliminate dimensional differences and unify the numerical range. The pheromone concentration and normalized light intensity are combined in a weighted sum form to constitute the path selection probability. The weight coefficient is dynamically adjusted with the number of iterations. In the initial stage, the light intensity is given a high weight (0.7) to strengthen its guiding role on the ant path. At this time, the ants will move towards the high light area first to quickly locate the potential high growth rate area. With each iteration, the light weight is reduced by 0.05, and then stabilized at 0.3 (the pheromone weight is increased to 0.7) to avoid the algorithm from getting trapped in local optima due to over-reliance on light intensity. At the same time, according to the biological characteristics of potatoes, the effective range of photoperiod is set for different seedlings. When the ants move, they are only allowed to select grid points within this range to eliminate invalid solutions that violate physiological characteristics. After each iteration, the growth rate prediction value of each photoperiod-seedling combination is recalculated based on the latest response surface model, and the pheromone matrix is ​​updated synchronously to ensure that the algorithm always conforms to the actual growth law.

[0072] Step 4: In the ant colony algorithm, each ant selects a photoperiod path based on pheromone concentration and heuristic factor probability, filtering photoperiod combinations. Weak seedlings prioritize long-day (14-16h) paths to prevent premature aging, strong seedlings focus on short-day (10-12h) paths to induce tuber formation, and medium seedlings use a median (12-14h) for balanced growth. Each ant locates the corresponding level of the pheromone matrix based on its current seedling category (weak / medium / strong). At its current position (photoperiod-seedling combination), it calculates the transition probability of all feasible neighbor grid points based on pheromone concentration and normalized light intensity. Weak seedlings prioritize calculating the 14-16h long-day path probability, strong seedlings focus on the 10-12h short-day path, and medium seedlings evaluate the 12-14h median range. The probability weights are dynamically adjusted, with early light intensity dominating and later pheromone concentration dominating. Pheromones are the dominant factor, and the next position is randomly selected based on the transition probability. This approach balances deterministic guidance with random exploration to avoid getting stuck on a fixed path. After an ant completes its path selection, it records the photoperiod-seedling combination it has passed through. After each iteration, the ant releases pheromones along its path. The increment value is positively correlated with the predicted growth rate value corresponding to that path. Paths with high growth rates receive more pheromone increments, while inefficient paths are weakened. The process continues iterating, updating the pheromone matrix and evaluating path adaptability in each round. A dual termination condition is set. After the iteration terminates, the grid point with the highest pheromone concentration in the photoperiod interval corresponding to each type of seedling is extracted from the final pheromone matrix as its optimal photoperiod matching scheme. Weak seedlings are locked in long-day conditions to prevent premature aging, strong seedlings are determined to be in short-day conditions to promote tuber growth, and medium seedlings are matched with intermediate values ​​to balance growth.

[0073] The process of determining the optimal optical period matching scheme is as follows:

[0074] The ant colony algorithm iteratively updates a three-dimensional pheromone matrix. In each iteration, the ant colony selects a path based on the current pheromone distribution and illumination heuristics, then globally updates the pheromone matrix. This involves reducing the pheromone concentration at all grid points according to a preset evaporation coefficient, simulating a natural evaporation process. Furthermore, based on the predicted growth rate of the ants along the paths traversed, pheromone increments are added to corresponding grid points, with higher growth rate paths receiving larger increments. This creates a cumulative mechanism that strengthens high-quality paths and eliminates inefficient ones. Simultaneously, after each iteration, path fitness is reassessed, and a comprehensive fitness score is calculated. The predicted growth rate of the photoperiod-seedling combination is correlated with pheromone concentration to ensure the matrix update is synchronized with the biological model. This gradually guides the search direction towards high-fitness regions, determining the optimal path and improving the algorithm's global search efficiency. The algorithm employs dual termination conditions to balance computational efficiency and solution quality: a preset maximum number of iterations and a dynamic convergence criterion. The preset maximum number of iterations prevents infinite iterations. The algorithm iterates through dynamic convergence. When the concentration value of the optimal path (the grid point with the highest pheromone concentration) does not change significantly for 10 consecutive iterations, it is considered that the algorithm has converged to a stable solution. During the termination process, the dynamic distribution of pheromone in the photoperiod interval corresponding to each type of seedling is continuously monitored. By comparing the stability of historical iteration data with the current optimal solution, the reliability of the termination decision is ensured, effectively avoiding the locking of suboptimal solutions due to premature termination or the waste of computational resources caused by excessive iteration. After the algorithm terminates, the optimal photoperiod matching scheme for each type of seedling is extracted from the final pheromone matrix. For weak seedlings, the long-day interval is locked, and the apical dominance is suppressed and premature aging is prevented by extending the light. For strong seedlings, the short-day interval is determined, and the expression of tuber formation-related genes is activated to promote the allocation of nutrients to underground organs. For medium seedlings, the intermediate photoperiod is matched to balance the needs of vegetative growth and reproductive growth. When extracting, the seedling-specific photoperiod requirement threshold is followed, and only the grid point with the highest pheromone concentration and in line with biological constraints is selected to ensure that the scheme has both mathematical optimality and physiological rationality.

[0075] The expression for the overall adaptability score is as follows:

[0076] F op =α·τ op +β·η op ;

[0077]

[0078] In the formula, F op The overall adaptability score for the current grid point (photoperiod o - seedling category p combination), where a higher value indicates a better path, τ op η represents the pheromone concentration at grid point (o, p) in the pheromone matrix, reflecting the cumulative preference of the ant colony for this path over historical iterations. It is dynamically adjusted by the pheromone update rule. opFor illumination heuristic information, defined as the matching degree between normalized light intensity and seedling requirements, L o L is the duration of the current optical period o. p α represents the optimal photoperiod center value for seedling category p, and β represents the weighting coefficients that control the contribution ratio of pheromones and light heuristic information.

[0079] The specific work involves each ant locating to the corresponding level of a three-dimensional pheromone matrix based on its seedling category (weak, medium, or strong). This matrix stores the pheromone concentration for each combination, with photoperiod as the horizontal axis, seedling category as the vertical axis, and growth rate as the vertical axis. The ant evaluates the transition probability of all feasible neighboring grid points at its current location. The calculation combines pheromone concentration and normalized light intensity, with weights dynamically adjusted. In the initial iteration phase, light intensity has a higher weight (0.7) to guide the ant to prioritize exploring areas with suitable light. Later, as the number of iterations increases... The weighting gradually shifts towards pheromone concentration (ultimately accounting for 0.7), strengthening the cumulative preference for high-quality paths. For weak seedlings, the probability calculation focuses on long-day paths (14-16 hours); for strong seedlings, it focuses on short-day paths (10-12 hours); and for medium-sized seedlings, it evaluates the intermediate range (12-14 hours). The next position is selected randomly using probability, balancing deterministic guidance and random exploration to avoid getting trapped in local optima. After ants complete path selection, the photoperiod-seedling combination sequence they traverse is recorded. After each iteration, all ants release pheromones along their paths, with the increment value corresponding to the photoperiod of the path. The predicted growth rate of the seedling combination is positively correlated, meaning that high-growth-rate paths receive higher pheromone increments, while inefficient paths are weakened due to volatility. The pheromone matrix is ​​dynamically updated with each iteration, and global pheromones decay according to a fixed volatility coefficient. Simultaneously, the newly added pheromones in this iteration are superimposed, forming an "accumulation-elimination" mechanism. By strengthening high-value paths and weakening inefficient paths, the ant colony is gradually guided to converge toward the potential optimal solution. After each iteration, the adaptability of each path is reassessed to ensure that the pheromone distribution remains synchronized with the growth rate prediction model, so that the pheromone distribution always reflects the latest search results. The ant colony algorithm continues to iterate until the termination condition is met, reaching the preset maximum number of iterations, or the pheromone concentration of the optimal path does not change significantly for 10 consecutive iterations. After termination, the highest concentration grid point in the photoperiod interval corresponding to each type of seedling is extracted from the final pheromone matrix as the optimal solution. Among them, weak seedlings are locked with a long-day scheme to inhibit premature aging by extending the light exposure; strong seedlings are determined with a short-day scheme to promote tuber enlargement; and medium seedlings are matched with an intermediate value to balance vegetative and reproductive growth. The photoperiod requirement thresholds of different seedlings are followed to ensure biological rationality.

[0080] Step 5: Collect stem diameter and chlorophyll content data at fixed intervals (every 4 hours) using IoT sensors to analyze the uniformity of seedling growth under each photoperiod path;

[0081] Step 6: Update the pheromone every 24 hours based on the photoperiod path evaluation results, eliminate paths that lead to excessive growth (stem elongation rate > 0.5 cm / d) or premature aging (chlorophyll content < 30 SPAD), and after 10 iterations, lock in the optimal photoperiod-graded seedling matching scheme to optimize and regulate light parameters.

[0082] Example 2, as Figure 1 , Figure 2 As shown, based on Example 1, the present invention provides a technical solution: Preferably, in step 5, the process of analyzing the seedling growth uniformity under each photoperiod path is as follows:

[0083] Using IoT sensors in the target area, stem diameter and chlorophyll content data are collected at fixed intervals of 4 hours, and corresponding photoperiod parameters are recorded synchronously. After the data is uploaded to the edge computing node, outlier removal and missing value imputation are performed to generate a standardized time series dataset to ensure data integrity and consistency. Based on the clustered seedling categories, the seedlings are grouped according to photoperiod paths. The coefficient of variation (CV) of stem diameter and the uniformity index (calculated based on the entropy value of stem diameter data) of each seedling category under the same photoperiod path are calculated. The time dynamics of uniformity are analyzed by sliding window to identify the trend of the influence of photoperiod on growth stability and generate curves of uniformity changing with time under each path. Based on the uniformity analysis results, photoperiod paths with a stem diameter CV of <15% are selected as candidate schemes and marked as high-adaptability paths.

[0084] The process of analyzing the time dynamics of uniformity using a sliding window to generate curves showing the change of uniformity over time for each path is as follows:

[0085] Based on the seedling classification results obtained from K-means clustering, the data were grouped according to photoperiod parameters to ensure that each group of seedlings was on the same photoperiod path. For each group, the coefficient of variation (CV) of stem diameter was calculated, which is the ratio of the standard deviation of stem diameter to the mean, quantifying the dispersion of stem diameter among individual seedlings under the same photoperiod. The smaller the CV value, the higher the growth consistency. Simultaneously, an evenness index was introduced to assess growth stability. The evenness index is calculated based on the entropy value of the stem diameter data. Specifically, a stem diameter frequency distribution histogram was constructed, dividing the stem diameter data into 10 equally wide intervals. The probability distribution of each interval was calculated, and the entropy value (i.e., the evenness index) was calculated using the Shannon entropy formula. The lower the entropy value, the more concentrated the data distribution and the higher the evenness. The coefficient of variation of stem diameter and the evenness index characterize growth stability from the perspectives of dispersion and distribution concentration, respectively. The sliding window method was used to analyze the time trend of the evenness index. The window width was set to a fixed time interval (7 consecutive time points), and the window was slid point by point with a step size of 1. The evenness index was calculated for each window. The mean of the uniformity index within the apex is used to generate time-series data. By comparing the time-series curves under different photoperiod paths, the change pattern of uniformity is identified. If the uniformity index of a certain path increases over time, it indicates that photoperiod promotes enhanced growth stability; if the curve fluctuates sharply or continues to decline, it reflects that photoperiod inhibits stability. At the same time, the slope and fluctuation range of the uniformity index under each path are calculated to quantify the trend strength and stability differences. The sliding window analysis results are visualized to generate curves showing the change of the uniformity index over time under each photoperiod path. The horizontal axis represents time, and the vertical axis represents the uniformity index value. Different paths are distinguished by different colors. By comparing the curves, the influence trend of photoperiod on growth stability is presented. Furthermore, the reliability of the dynamic trend is verified by combining the static analysis results of the stem diameter variation coefficient: if the stem diameter variation coefficient of a certain path is low and the uniformity curve shows an upward trend, it is confirmed that the photoperiod is a high-fitness path; if the stem diameter variation coefficient is high and the curve fluctuates greatly, it is marked as a low-fitness path.

[0086] The expression for the uniformity index is as follows:

[0087]

[0088] In the formula, H is the evenness index, the larger the value, the more uniform the population growth; n is the sample size, the number of seedlings under the same photoperiod path; i is the sample number, used to traverse all individual seedlings; p i x represents the relative proportion of the stem diameter of the i-th seedling, reflecting the weight of the stem diameter of a single seedling relative to the total population. i Let be the stem diameter measurement of the i-th seedling, which is normalized to the interval [0,1]. ln(·) is the natural logarithm (base e), used to quantify the dispersion of the distribution (the essence of entropy).

[0089] The specific work involves: using an IoT sensor network deployed in the target area, synchronously collecting stem diameter, chlorophyll content, and corresponding photoperiod parameters at fixed intervals of 4 hours to form a multi-dimensional raw data stream. After the data is uploaded to the edge computing node, outlier detection and removal are performed. A statistical method based on the 3σ criterion is used to identify and filter outliers caused by sensor malfunctions or environmental interference. For missing values, a time-series linear interpolation method is used to fill the gaps using the mean of data from adjacent time points to ensure data continuity. Stem diameter and chlorophyll content are normalized to eliminate dimensional differences, while photoperiod parameters are retained according to actual values. Finally, a standardized time-series dataset is generated, containing five dimensions: timestamp, seedling category, stem diameter, chlorophyll content, and photoperiod. Based on the standardized time-series dataset, seedling categories are divided according to clustering results, and further grouped according to photoperiod paths. For each type of seedling under the same photoperiod path, the coefficient of variation (CV) of stem diameter was calculated to quantify the degree of growth dispersion. At the same time, an evenness index (calculated based on the entropy value of stem diameter data) was introduced to evaluate the growth consistency. The evenness index was dynamically analyzed through a sliding window (with a window width of 7 time points) to generate a time series curve and analyze the trend of the influence of photoperiod on growth stability. Based on the results of the dynamic analysis of evenness, photoperiod paths with a coefficient of variation (CV) of stem diameter below 15% were selected as candidate schemes, indicating that the seedlings have low growth dispersion and high stability under this photoperiod. The evenness index of the candidate paths was further tested for time trends to eliminate short-term fluctuations and confirm the long-term stability advantage. Finally, photoperiod paths that meet the conditions were marked as high-adaptability paths, and a path-seedling category mapping table was generated to clarify the optimal photoperiod range for each type of seedling.

[0090] In step 6, the process of optimizing and controlling the illumination parameters is as follows:

[0091] Every 24 hours, based on the photoperiod path evaluation results, the stem elongation rate and chlorophyll content of seedlings under each photoperiod path are calculated. If a path results in a stem elongation rate >0.5 cm / d (excessive growth) or a chlorophyll content <30 SPAD (premature senescence), it is marked as a poor path and its pheromone concentration is reduced; conversely, it is marked as a good path and its pheromone concentration is maintained or enhanced. The initial pheromone concentration is evenly distributed according to the photoperiod to ensure full coverage. Based on the daily evaluation results, the pheromone concentration of poor paths decreases at a fixed decay rate, while the pheromone concentration of good paths increases according to the fitness ratio. If a certain path... If a path is marked as poor quality for three consecutive days, it will be forcibly eliminated. If the pheromone content of a high-quality path exceeds the threshold of 80%, it will be retained and its coverage expanded. After 10 iterations, only the top 20% of paths with the highest pheromone concentration will be retained to form a candidate photoperiod set. The candidate photoperiod set will be grouped according to seedling category, and the photoperiod distribution of paths within each group will be statistically analyzed. The photoperiod range with the highest pheromone concentration will be locked as the optimal matching scheme. The optimal matching scheme will be input into the illumination control system to dynamically adjust the photoperiod and ensure that the deviation between the actual parameters and the optimal matching scheme is less than 5%, thereby achieving precise control.

[0092] The specific work involves: Calculating the stem elongation rate and chlorophyll content of seedlings based on 24-hour data collected by an IoT sensor network, grouped by photoperiod path. Path quality is determined using dynamic thresholds: if the stem elongation rate exceeds 0.5 cm / d (indicating excessive growth) or the chlorophyll content is below 30 SPAD (indicating premature aging), the path is marked as substandard; otherwise, it is marked as high-quality. Initially, the pheromone concentration of all photoperiod paths is initialized according to a uniform distribution principle to ensure coverage of the entire photoperiod range and avoid local optima due to initial bias. Based on daily evaluation results, the pheromone concentration of substandard paths decreases at a fixed decay rate (20%), while the pheromone concentration of high-quality paths increases proportionally to fitness. Fitness is defined as the product of the reciprocal of the stem elongation rate and the chlorophyll content, forming a dynamic adjustment mechanism of survival of the fittest. If a path is marked as substandard for three consecutive days, it is forcibly eliminated to prevent further damage. Ineffective resource occupation: If the pheromone concentration of high-quality paths exceeds the threshold of 80%, their photoperiod coverage is prioritized and expanded to enhance their competitiveness. Through 10 iterations, only the top 20% of paths with the highest pheromone concentration are retained to form a candidate photoperiod set. Positive feedback of pheromone concentration strengthens high-quality paths, while negative feedback suppresses low-quality paths, gradually approaching the global optimum. The candidate photoperiod set is grouped by seedling type (weak, medium, and strong), and the photoperiod distribution of each group is statistically analyzed. The continuous photoperiod range with the highest pheromone concentration is locked as the optimal matching scheme. The optimal matching scheme is input into the lighting control system. By dynamically adjusting the switching duration of LED lights or the opening angle of the shading net, the deviation between the actual photoperiod and the optimal scheme is ensured to be less than 5%, achieving precise environmental control, balancing seedling growth rate and health status, avoiding excessive growth or premature aging, and improving crop yield and quality consistency.

[0093] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for dynamic optimization and control of environmental parameters for potato seedling grading, characterized in that, Includes the following steps: Step 1: Based on the IoT sensors deployed in the target area, collect data on stem diameter, chlorophyll content and plant height of potato seedlings, and use K-means clustering to divide the seedlings into three categories: weak seedlings, medium seedlings and strong seedlings, and set the duration of the initial photoperiod. Step 2: Using stem elongation rate and chlorophyll content as constraints, a gradient photoperiod experiment was conducted on each type of potato seedling to establish a photoperiod-growth rate response surface model and preliminarily determine the suitable photoperiod range. The specific process is as follows: Using stem elongation rate and chlorophyll content as constraints, potato seedlings were grouped according to the previous grading results, and photoperiod gradients of 10-16 h were set for each group with a step size of 2 h. Each group was repeated 3 times and cultured continuously for 14 days under controlled conditions, while the stem elongation rate and chlorophyll content were monitored simultaneously. Based on experimental data, photoperiod combinations that meet the constraints were selected. Photoperiod and seedling type were used as independent variables, and stem elongation rate and chlorophyll content were used as dependent variables. Multiple linear regression was used to fit the response surface model, and a photoperiod-growth rate response surface model was constructed to quantify the nonlinear effect of photoperiod on growth indicators. The accuracy of the model was evaluated through cross-validation, the prediction accuracy of the model was verified, the nonlinear relationship between photoperiod and growth rate was output, and the appropriate photoperiod range was initially determined. Step 3: Based on the photoperiod-growth rate response surface model, the ant colony algorithm parameters and pheromone matrix are initialized, and the light intensity is introduced through heuristic factors to optimize the path selection weights. Step 4: The ant colony algorithm combines pheromone concentration and heuristic factor probability to select photoperiod paths and screen photoperiod combinations. Step 5: Collect stem diameter and chlorophyll content data at fixed intervals using IoT sensors to analyze the uniformity of seedling growth under each photoperiod path. The specific process is as follows: Using IoT sensors in the target area, stem diameter and chlorophyll content data are collected at fixed intervals of 4 hours, and corresponding photoperiod parameters are recorded synchronously. After the data is uploaded to the edge computing node, outlier removal and missing value imputation are performed to generate a standardized time series dataset. Based on the clustered seedling categories, the seedlings are grouped according to photoperiod paths. The coefficient of variation of stem diameter and the uniformity index of each seedling category under the same photoperiod path are calculated. The time dynamics of uniformity are analyzed by sliding window, and curves of uniformity change over time under each path are generated. The uniformity index is calculated based on the entropy value of stem diameter data. Based on the uniformity analysis results, photoperiod paths with a stem diameter variation coefficient of <15% were selected as candidate schemes and marked as highly adaptable paths; Step 6: Update the pheromone every 24 hours based on the photoperiod path evaluation results, eliminate paths that cause excessive growth or premature aging, and after 10 iterations, lock in the optimal photoperiod-graded seedling matching scheme and optimize and control the light parameters.

2. The method for dynamic optimization and control of environmental parameters for potato seedling grading according to claim 1, characterized in that: In step 1, the process of setting the duration of the initial optical period is as follows: By deploying IoT sensors in the target area, physiological data of potato seedlings, including stem diameter, chlorophyll content and plant height, are collected synchronously. After data cleaning and standardization, the K-means clustering algorithm is used to classify the seedlings into three categories: weak seedlings, medium seedlings and strong seedlings based on the Euclidean distance of the physiological indicators. Based on the seedling grading results and considering the growth requirements of potato seedlings, the initial photoperiod duration ranges for three types of seedlings were set: 10-12h for weak seedlings, 12-14h for medium seedlings, and 14-16h for strong seedlings.

3. The method for dynamic optimization and control of environmental parameters for potato seedling grading according to claim 1, characterized in that: The process of constructing the photoperiod-growth rate response surface model and initially determining the appropriate photoperiod range is as follows: Simultaneously acquire experimental data on photoperiod and seedling type, as well as corresponding stem elongation rate and chlorophyll content, remove outliers, standardize continuous variables with Z-score, keep categorical variables in virtual encoding form and directly incorporate them into use, integrate the acquired data to obtain a comprehensive dataset, and divide the training set and test set in a 7:3 ratio; Using photoperiod and seedling type as independent variables, and stem elongation rate and chlorophyll content as dependent variables, a photoperiod-growth rate response surface model was constructed using multiple linear regression. The least squares method was used to fit the model, and the regression coefficients and their significance were calculated to screen significant variables. Three-dimensional response surfaces of photoperiod-growth rate and photoperiod-chlorophyll content were generated based on regression coefficients. The marginal effect of photoperiod on growth indicators including stem elongation rate and chlorophyll content was analyzed by partial derivatives. Piecewise regression was introduced to handle nonlinear relationships, dividing the photoperiod into two segments: <12h and ≥12h. Models were fitted to each segment separately, and the AIC values ​​of the segmented model and the global model were compared. When ΔAIC>2, the segmented model was selected as the optimal model. Then, the interval corresponding to the peak value of the response surface ±0.5 standard deviation was taken as the suitable photoperiod interval for each seedling.

4. The method for dynamic optimization and control of environmental parameters for potato seedling grading according to claim 2, characterized in that: In step 3, the process of optimizing path selection weights is as follows: Based on the photoperiod-growth rate response surface model, the initial ant colony algorithm parameters are set, including the number of ants, the pheromone volatility coefficient and the initial pheromone concentration. A three-dimensional structure is constructed according to the combination of seedling type and photoperiod to form a pheromone matrix, whose dimensions correspond to the grid points of the photoperiod-growth rate response surface. Light intensity is introduced as a heuristic factor into the ant colony algorithm. It is normalized and mapped to the [0,1] interval. Together with pheromone concentration, it forms the path selection probability formula, i.e. the transition probability. The weight of the heuristic factor is dynamically adjusted so that the guiding effect of light intensity on path selection decreases with the number of iterations.

5. The method for dynamic optimization and control of environmental parameters for potato seedling grading according to claim 4, characterized in that: In step 4, the process of screening photoperiod combinations is as follows: Each ant locates the corresponding level of the pheromone matrix based on the current seedling category, calculates the transition probability of all feasible neighbor grid points at the current position by combining the pheromone concentration and normalized light intensity, and randomly selects the next position based on the transition probability. After the ant completes its path selection, it records the photoperiod-seedling combination it passes through, and after the ant completes one iteration, it releases pheromones along the path it has passed through. The process is iterated continuously, with the pheromone matrix updated and path adaptability evaluated in each round. A dual termination condition is set, and after the iteration terminates, the grid point with the highest pheromone concentration in the corresponding photoperiod interval for each type of seedling is extracted from the final pheromone matrix as its optimal photoperiod matching scheme.

6. The method for dynamic optimization and control of environmental parameters for potato seedling grading according to claim 5, characterized in that: The process for determining the optimal optical period matching scheme is as follows: The ant colony algorithm continuously updates the three-dimensional pheromone matrix through iteration. In each iteration, after the ant colony completes path selection based on the current pheromone distribution and illumination heuristic information, it performs a global update of the three-dimensional pheromone matrix. In this process, the pheromone concentration of all grid points is reduced according to a preset volatility coefficient to simulate the natural volatility process. Based on the predicted growth rate of the ants along the path, the corresponding grid points are added with pheromone increments. At the same time, after each iteration, the path fitness is re-evaluated, the comprehensive fitness score is calculated, and the predicted growth rate of the photoperiod-seedling combination is correlated with the pheromone concentration to gradually guide the search direction to converge toward the high fitness region and determine the optimal path. The algorithm has two termination conditions: a preset maximum number of iterations and a dynamic convergence criterion. The preset maximum number of iterations prevents infinite loops. The dynamic convergence criterion considers the algorithm to have converged to a stable solution when the concentration value of the optimal path does not change significantly for 10 consecutive iterations. After the algorithm terminates, the optimal photoperiod matching scheme for each type of seedling is extracted from the final pheromone matrix. Weak seedlings are locked in the long-day interval, strong seedlings are determined in the short-day interval, and medium seedlings are matched with the intermediate photoperiod.

7. The method for dynamic optimization and control of environmental parameters for potato seedling grading according to claim 1, characterized in that: The process of analyzing the time dynamics of uniformity through a sliding window and generating curves showing the change of uniformity over time for each path is as follows: Based on the seedling classification results obtained by K-means clustering, the data are grouped according to the photoperiod parameter. For each group of data, the coefficient of variation of stem diameter, i.e. the ratio of the standard deviation of stem diameter to the mean, is calculated to quantify the dispersion of stem diameter among individual seedlings under the same photoperiod. At the same time, the evenness index is introduced to evaluate growth stability. The sliding window method was used to analyze the time variation trend of the uniformity index. The window width was set to a fixed time interval, the mean of the uniformity index within each window was calculated, and time series data was generated. By comparing the time series curves under different photoperiod paths, the variation pattern of uniformity was identified. At the same time, the slope and fluctuation range of the uniformity index under each path were calculated to quantify the differences in trend strength and stability. The sliding window analysis results are visualized, generating curves showing the change of uniformity index over time for each photoperiod path. The horizontal axis represents time, and the vertical axis represents the uniformity index value. Different paths are distinguished by different colors. By comparing the curves, the influence trend of photoperiod on growth stability is presented.

8. The method for dynamic optimization and control of environmental parameters for potato seedling grading according to claim 1, characterized in that: In step 6, the process of optimizing and controlling the illumination parameters is as follows: Every 24 hours, based on the photoperiod path evaluation results, the stem elongation rate and chlorophyll content of seedlings under each photoperiod path are calculated. If a path results in a stem elongation rate >0.5cm / d or a chlorophyll content <30SPAD, it is marked as a poor path; otherwise, it is marked as a good path. Based on the daily evaluation results, the pheromones of inferior paths decrease at a fixed decay rate, while the pheromones of superior paths increase according to the fitness ratio. If a path is marked as inferior for 3 consecutive days, it is forcibly eliminated. If the pheromone content of a superior path exceeds the threshold of 80%, it is retained and its coverage is expanded. After 10 iterations, only the top 20% of paths with the highest pheromone concentration are retained to form a candidate photoperiod set. The candidate photoperiod set is grouped according to seedling category. The photoperiod distribution of each group path is statistically analyzed. The photoperiod range with the highest pheromone concentration is locked as the optimal matching scheme. The optimal matching scheme is input into the illumination control system to dynamically adjust the photoperiod.

Images